See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/265845269

45 years recovery boiler co-operation in Finland - International Conference,

Sibelius hall, Lahti, June 3 - 5, 2009,

Book · June 2009

DOI: 10.13140/2.1.4273.7923

CITATION READS

1 1,036

23 authors, including:

Esa K. Vakkilainen Mikko Markus Hupa

Lappeenranta – Lahti University of Technology LUT Åbo Akademi University
200 PUBLICATIONS   1,906 CITATIONS    489 PUBLICATIONS   12,280 CITATIONS   

SEE PROFILE SEE PROFILE

Markus Engblom Anders Brink

Åbo Akademi University Åbo Akademi University
28 PUBLICATIONS   124 CITATIONS    105 PUBLICATIONS   1,908 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Recovery boiler fouling View project

GlaCerCo project View project

All content following this page was uploaded by Esa K. Vakkilainen on 20 September 2014.

The user has requested enhancement of the downloaded file.

 SUOMEN SOODAKATTILAYHDISTYS
FINNISH RECOVERY BOILER COMMITTEE

45 YEARS RECOVERY BOILER

CO-OPERATION IN FINLAND

PROCEEDINGS

INTERNATIONAL RECOVERY BOILER CONFERENCE

Sibelius Hall, Lahti, June 3 - 5, 2009

1
 SUOMEN SOODAKATTILAYHDISTYS
FINNISH RECOVERY BOILER COMMITTEE

45th ANNIVERSARY INTERNATIONAL

RECOVERY BOILER CONFERENCE

Sibelius Hall, Lahti, June 3 – 5, 2009

CONFERENCE CHAIRMAN

Mikko Hupa, Åbo Akademi University

PROGRAMME COMMITTEE
Klaus Niemelä, KCL, Chairman
Mikko Hupa, Åbo Akademi University
Keijo Salmenoja, Oy Metsä-Botnia Ab
Esa Vakkilainen, Lappeenranta University of Technology
Sebastian Kankkonen, Pöyry Industry Oy
Outi Pisto, Pöyry Industry Oy
Eija Turunen, Pöyry Industry Oy
Markus Nieminen, Pöyry Industry Oy

CONFERENCE MANAGEMENT
Finnish Recovery Boiler Committee
c/o Pöyry Industry Oy
Outi Pisto
Eija Turunen
Markus Nieminen

Finnish Recovery Boiler Committee

c/o Pöyry Industry Oy
P.O. Box 4, Jaakonkatu 3
FI-01621 Vantaa
FINLAND

Telephone + 358 10 3311

Telefax + 358 10 33 21163
www.soodakattilayhdistys.fi

© Copyright Finnish Recovery Boiler Committee

All rights Reserved

Cover photos:
Scenery of new recovery boiler at UPM-Kymmene Oyj, Kymi, Kuusankoski
Sibelius Hall, Lahti

2
CONTENTS

Welcoming words 5 Keijo Salmenoja

Boundaries of recovery boiler 7 Esa Vakkilainen

development

How well do we understand recovery 17 Mikko Hupa, Markus Engblom,

furnace processes? Anders Brink and Christian Mueller

Challenges to manufacture worlds 31 Kari Mäkelä

largest boilers

Benchmarking a modern recovery 39 Mikael Ahlroth

boiler

Black liquor particle and droplet 57 Warren Roberts, Hong Lu, Elvin Ip
reactions: Experimental and model and Larry L. Baxter
results

Kinetics of kraft black liquor 101 Sergey Smorodin

combustion

Getting rid of potassium and chloride 109 Heikki Jaakkola

Alternate materials for recovery boiler 117 James R Keiser, Joseph R Kish and
superheater tubes Douglas L Singbeil

Economizer leak study 131 Thomas J. Grant, Thomas M. Grace

and Jack Clement

The changing scope of black liquor 133 Kevin Whitty

gasification

Biorefineries for co-production of 145 W. J. Frederick, Jr

renewable transportation fuels

Composition of carryover in recovery 157 Honghi Tran and Asghar Khalaj-

boilers Zadeh

3
4
WELCOMING WORDS

KEIJO SALMENOJA
OY METSÄ-BOTNIA AB,
BUSINESS DEVELOPMENT

Year 2009 will be remembered for two reasons. ancier of the project. The SOTU 2-project was a
Firstly, 2009 will probably be one of the most dif- big effort to the Committee, but it became a great
ficult years in the history of the pulping industry. success. SOTU 2 concentrated mainly on the pos-
Hard times are globally present and several pulp sibilities to increase power production from the
mills have been permanently closed. All the com- recovery boilers. Another major Tekes-funded
panies have experienced diminished revenues and project (SKYREC) started in the first quarter of
have begun cost-cutting measures. Secondly, year 2008. This project will further deepen the under-
2009 will mark the 45th year for the activities of standing needed in the design and operation of
the Finnish Recovery Boiler Committee. To cel- the high-efficiency recovery boilers in the future.
ebrate the occasion this International seminar has SKYREC-project is now in a good pace and has
been arranged in Lahti, Finland. also raised a lot of international interest toward
the Finnish Recovery Boiler Committee and its
The Finnish Recovery Boiler Committee has also activities.
seen the effects of the global decline. Three pulp
mills have recently been permanently shut down The pulping industry has faced both lows and
in Finland, which means that the committee has highs in the course of history, but has always
also lost three active members. However, the bad survived. One of the main functions of the Finn-
news always comes with the good news. One of ish Recovery Boiler Committee is to distribute
the recent good news is definitely the start-up of knowledge of recovery equipment and operation,
a brand new recovery line at the UPM-Kymmene which will help the industry to develop highly ef-
Kuusankoski mill in 2008. The investment in a ficient production units with high availability.
new recovery line proves that pulp will be pro-
duced in Finland also in the future. Despite this transient decline, I hope we all look
ahead and have a strong faith in future. At least the
The Finnish Recovery Boiler Committee has, since Finnish Recovery Boiler Committee believes on
its birth in 1964, been active in promoting safe, a prosperous and active future. With these words
economic and environmentally friendly operation I’d like to congratulate the Recovery Boiler Com-
of recovery boilers and closely related processes. mittee of its 45 year milestone and welcome you
The Committee publishes different guidelines, all to this anniversary conference.
recommends best practices and arranges meetings
and conferences. Most of the active work is done
in five subcommittees, which cover the most im-
portant areas around the recovery boiler.

During the recent years, the Finnish Recovery

Boiler Committee has taken a more active role
to manage large research projects. The first ma-
jor publicly funded project (SOTU 2) was started
in 2003. The Finnish Funding Agency for Tech-
nology and Innovation (Tekes) was the main fin-

5
6
BOUNDARIES OF RECOVERY BOILER
DEVELOPMENT
ESA K. VAKKILAINEN
LAPPEENRANTA UNIVERSITY OF TECH-
NOLOGY, DEPARTMENT OF ENERGY
TECHNOLOGY,
LAPPEENRANTA, FINLAND

Introduction

Pulp mills and recovery boilers are becoming design criterion for the recovery boiler has been
larger and larger. From humble beginnings the high availability. This evolution has culminated
recovery boiler now reaches for the title of larg- with units that are among the largest biofuel boil-
est biomass fired boiler in the world. Globally ers in the world. Currently the focus is sifting
the pulp and paper industry is in recession. But towards more challenging designs that maximize
several new large mills are being built around the electricity production.
world. Even larger mills are in the works.
From old to new
New expectations have been imposed on chemi- “There is no stopping for progress” applies also
cal recovery department operations. Cutting down to recovery boilers. The recovery departments
air emissions is not enough. Pulp and paper mills built today are bigger, more efficient and geared
need to maximize their bioenergy potential and to produce more electricity than ever. This can
minimize their electricity consumption to produce be easily seen when we compare the first recov-
green electricity for sale outside the mill. ery boilers to last recovery boiler started in Fin-
land at UPM-Kymmene Kymi mill in May 2008.
Recovery boilers, which have always produced
bioenergy, have evolved significantly in the past
70 years. For pulp mills the significance of elec- First Recovery Boilers
tricity generation from the recovery boiler has The modern recovery boiler has a few strong ide-
historically been secondary. The most important as that have remained unchanged until today. It
was the first recovery equipment type where all
processes occurred in a single vessel. The drying,
combustion and subsequent reactions of black liq-
uor all occur inside a cooled furnace. This is the
main idea in Tomlinson’s work.

Secondly the combustion is aided by spraying

the black liquor into small droplets. Controlling
process by directing spray proved easy. Spraying
was used in early rotary furnaces and with some
success adapted to stationary furnace by H. K.
Moore. Thirdly one can control the char bed by
having primary air level at char bed surface and
more levels above. Multiple level air system was
introduced by C. L. Wagner.

Figure 1. First Tomlinson kraft recovery boiler

from Babcock & Wilcox in 1929 [1].

7
 Recovery boiler also improved the smelt removal.
It is removed directly from the furnace through
smelt spouts into a dissolving tank. Some of the
first recovery units employed the use of Cottrell’s
electrostatic precipitator for dust recovery.

Babcock & Wilcox was founded in 1867 and

gained early fame with its water tube boilers. It
built and put into service the first black liquor re-
covery boiler in the world in 1929 [1]. This was
soon followed by a unit with completely water
cooled furnace at Windsor Mills in 1934. After
reverberatory and rotating furnaces the recovery
boiler was on its way.[2]

The second early pioneer, Combustion Engineer-

ing based its recovery boiler design on the pio-
neering work of William M. Cary, who in 1926
designed three furnaces to operate with direct liq-
uor spraying and on work by Adolph W. Waern
and his recovery units.
Recovery boiler were soon licensed and produced
in Scandinavia and Japan. These boilers were
built by local manufacturers from drawings and
with instructions from licensors. One of the early
Scandinavian Tomlinson units employed a 8.0 m Figure 2. Latest Finnish recovery boiler Kymi
high furnace that had 2,8*4,1 m furnace bottom with Korsnäs boiler on left [6].
which expanded to 4,0*4,1 m at superheater en-
trance [3]. This unit stopped production for every
weekend. In the beginning economizers had to arrangement.
be water washed twice every day, but after in-
stallation of shot sootblowing in the late 1940s The higher main steam outlet temperature requires
the economizers could be cleaned at the regular more heat to be added in the superheating section.
weekend stop. The construction utilized was very Therefore the furnace outlet gas temperature has
successful. The 160 t/day unit at Korsnäs oper- increased. The alternative is to significantly in-
ated still almost 50 years later.[4] Edling [5] states crease superheating surface and decrease boiler
in 1937 that more than 20 units had already been bank inlet flue gas. If boiler bank inlet gas temper-
built of which 10 in Scandinavia. ature is reduced the average temperature differ-
ence between flue gas and steam is also decreased.
Newest capacity This reduces heat transfer and substantially more
The current recovery boiler, Figure 2, can be much superheating surface is needed. This approach has
larger than the previous ones. Boilers with over been abandoned because of increased cost. With
200 square meter bottom area have been bought. increasing dry solids content the furnace exit tem-
Largest recovery boilers are challenging circulat- perature can safely increase without fear of corro-
ing fluidized boilers for the title of largest bio-fuel sion caused by carryover.
fired boiler.
New capacity development
The superheater arrangement is designed for opti- Figure 3 shows the firing capacity of over 1000
mum heat transfer with extra protection to furnace recovery boilers that have been built in the world
radiation. Mill closure and decreased emissions to date.[10] This capacity has grown about 3 %
mean higher chloride and potassium contents in per year since 1970 when the maximum was 2041
black liquor. Almost all superheaters are placed tds/d. If this rate goes on, then the first 10 000
behind the bullnose to minimize the direct radia- tds/d recovery boiler will start operating in 2024.
tive heat transfer from the furnace. Increasing su- Finnish Recovery Boiler Committee will then cel-
perheating demand with increasing pressure de- ebrate its 60th operating year.
creases the need for boiler bank and water screen

8
Table 1. Comparison of early Scandinavian boiler with Kymi boiler.

Current black liquor processing capacity in Fin- In the last ten years (1999 – 2008) about one third
land is about 40 000 tds/d. Finnish Recovery Boil- of the installed capacity has been at green field
er Committee will celebrate its 100th operating mills. Most of the new recovery projects are at es-
year in 2064. Hopefully we can then show again tablished sites, which offer much more competi-
a new big recovery boiler. This boiler will be able tive infrastructure. New recovery capacity is built
to process all black liquor in Finland. The furnace especially in Asia and South-America, where pulp
loading for new boilers is now at 3400 kW/m2 and production capacity is strongly expanding, Figure
is growing about 20 kW/m2 per year. So the fur- 4. Europe and North-America account for less
nace loadings will be 3700 kW/m2 and 4500 kW/ than half of total new recovery capacity. We are
m2 respectively for 2024 and 2064. If the black facing an era where hardwood based pulp mills
liquor design HHV of 14 MJ/kgds will remain un- are favored over softwood pulp mills. About half
changed, then the new boilers will have bottom of new mills are built to produce eucalyptus pulp.
areas of 440 and 1200 m2 respectively for 2024 Recovery departments of eucalyptus mills are
and 2064. The corresponding furnace dimensions now an established technology [11].
will be 21*21*52 m and 34*35*79 m. A lot of the North-American and European recov-

6000

5000

4000
Capacity, tds/d

3000

2000

1000

0
1930 1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010
Delivery year

Figure 3. Black liquor firing capacity of recovery boilers built in the world versus their years of start-
up.[10]

9
 1%
1%

Europe
28 % 31 %

Asia
N-America
S-America
Oceania
Africa

12 %

27 %

Figure 4. New recovery capacity by region 1999 - 2008.

ery capacity is old. Average boiler age is about 30 Higher black liquor dry solids
years, Figure 3. Recovery boiler technical life ex- The modern kraft recovery process is the result
pectancy is around 30 - 40 years [12]. This means of decades of continued small improvements of
that in USA and Sweden within next ten years the basic technology. One important development
investment in older mills is needed. The alterna- in recovery boiler operation is “high-solids firing”
tive is the continuing closure of pulp mills. If we which refers to burning black liquor at solids con-
assume the age distribution to be the same then tents greater than 75%. The practice significantly
about 50 % of the recovery boiler capacity in USA increases steam generation and improves com-
and Sweden need to be rebuilt by 2024. By 2064 bustion stability, resulting in lower TRS and SO2
none of the current recovery boilers are in use, but emissions, and less boiler fouling and plugging. It
black liquor gasification will possibly still be 10 also provides for more capacity in existing units.
years from realization. In Scandinavia the new units are currently pur-
chased based on 85% dry solids content at as-fired
liquor.
Recoveryboilerageascapacityoftotalcapacity,%.

100 %
USA
90 %
Sweden
80 % Finland
Brazil
70 %

60 %

50 %

40 %

30 %

20 %

10 %

0%
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

Figure 5. Age of recovery boilers in USA, Sweden, Finland and Brazil.

10
Maximum black liquor dry solids content has al- erate from black liquor combustion depends on
ways been limited by the ability of available evap- the pressure and temperature of the steam that the
oration technology to handle the highly viscous boiler produces. Traditionally, recovery boilers
concentrated liquor. As evaporator technology has have operated at moderate steam pressures (≤80
evolved, the final black liquor dry solids content bar) and temperatures (≤480 oC) due to concerns
has increased. Dry solids content of virgin black over superheater corrosion and high costs of mate-
liquor (without ash recycle) is shown in Figure 6 rials for construction. Of the last 20 recovery boil-
as a function of start-up year of the boiler. ers about 80 % have chosen higher pressures and
temperatures. Some have done this without resort-
When looking at the virgin black liquor dry solids ing to more expensive materials. Today there is
we note that on average dry solids content has in- a number of recovery boilers operating at >100
creased. This is especially true for latest very large bar and >500°C. The lead in these developments
recovery boilers. Design dry solids for green field has been taken in Japan [13,14] and Scandinavia
mills have been either 80 or 85 % dry solids. 80 [15,16]. But high efficiency boilers are now being
% dry solids have been in use in Asia and South built also in North and South America. Because of
America (hardwood liquors). 85 % has been in corrosion issues potassium and chlorine removal
use in e.g. Finland (softwood liquors). equipment are definitely needed especially in high
pressure and temperature applications.
The maximum black liquor content has increased
by about 10 percentage points in about 30 years. Modern recovery boiler
2004 was the year when first new evaporators were The state-of-the-art recovery boiler is much larger
bought for 85 % dry solids. If this trend continues, than the previous designs. Boilers with over 250
then when Finnish Recovery Boiler Committee square meter hearth area have been bought. The
will celebrate its 60th operating year in 2024 the largest recovery boilers are challenging circulat-
black liquor is respectable 95 % dry solids. Black ing fluidized bed boilers for the title of largest bio-
liquor will be totally dry (100 % dry solids) in fuel fired steam generators in the world.
40 years (2049). When Finnish Recovery Boiler
Committee will celebrate its 100th operating year The largest operating recovery boiler in the world
in 2064 the black liquor is 105 % dry solids. The is presently the Metso Power unit started up in
first stage of combustion of black liquor will then November 2004 at the Jinhai Pulp & Paper mill
consist of black liquor droplet drawing water va- in Hainan, China, Figure 7. The unit has a furnace
por from surroundings. This kind of behavior will floor area of 268 m2, a nominal firing capacity of
necessitate new combustion and firing research. 5000 tds/d (6000 tds/d maximum) and can pro-
duce 204 kg/s steam rated at 84 bar and 480°C
The amount of power a recovery boiler can gen- [19].

90

85
Maximum

80
Virgindrysolids,%

75
Average
70

65

60

55

50
1975 1980 1985 1990 1995 2000 2005 2010
Deliveryyear

Figure 6. Virgin black liquor dry solids as a function of start-up year of the recovery boiler.

11
 Figure 7. The world’s largest recovery boiler at Jinhai Pulp &Paper, Hainan.

Recovery boiler design changes slowly. There are combustion air preheaters in order to produce
however some features that the boilers bought to- high-pressure/temperature steam. In some units
day have in common. The state of the art recovery the combustion air temperature is up to 190 °C
boiler design will be that of a single drum, high and feedwater tank operates close to low pressure
solids firing unit with multiple levels of air sup- steam temperature. Many new greenfield pulp
plied by separate fans, as shown in Figure 7. It mills are equipped with a multiple extraction back
will have a greater superheater surface area and pressure turbine combined with a condensing tur-
be equipped with feedwater preheaters and/or bine to maximize power generation. A recent new
feature is heat recovery after the electrostatic pre-
cipitators [20].

The recovery boiler has adopted a position of

waste destruction unit. What can be destroyed
by burning will likely be burned in it. Half of the
new boilers are burning sludge from biological ef-
fluent treatment and some even filtrate from bark
presses. Almost all new boilers burn weak NCG
gases and dissolving tank vent gases. A majority
of new units burn strong NCG gases with a dedi-
cated burner in the furnace.

Figure 8. State of the art recovery boiler [20].

12
The state-of-the-art recovery boiler is environ- Changing recovery boiler main
mentally friendly. It has multiple air levels above
the liquor gun level for efficient NOX control.
parameters
Main design parameters of the recovery boiler af-
Low emissions of TRS and SO2 with NCG firing
fect the steam and electricity generation. Signifi-
are typically achieved. The trend has been to de-
cantly more power generation can be achieved if
crease the particulate emission by applying larger
higher main steam pressure and temperature are
and more efficient electrostatic precipitators. Use
chosen. Average main parameters of new recov-
of scrubbers to recover heat is becoming rare.
ery capacity by region 1997 - 2006 are shown in
Table 2.
Sootblowers, which are used to control the accu-
mulation of deposits on tube surfaces in recovery
The trend in recent years has been definitely in
boilers, will be equipped with fully-expanded
favor of increased temperatures and pressures.
nozzles, variable lance speeds and intelligent
Newest Scandinavian and North-American lines,
sootblowing systems to maximize deposit remov-
have chosen main steam values in excess of 80
al efficiency while minimizing sootblowing steam
bar and 480 °C [22,23] and Table 2. It can be seen
usage. It remains to be seen whether the new re-
that in e.g. China and Indonesia there is still re-
covery boilers will use low pressure steam after
sistance to high main steam parameters. This is
the steam turbine (10 to 15 bars) for sootblowing
understandable when opportunities to sell excess
[21], instead of the conventional high pressure
electricity do not yet exist.
steam (20 to 24 bars).

Table 2. Average main parameters of new recovery capacity by region 1999 – 2008.

Region Capacity Steam temp. Steam press.

O
tds/d C bar
Africa 630 472 78
Asia 2720 474 78
Europe 2350 487 88
North-America 1780 488 90
Oceania 670 460 64
South-America 2460 480 84

550

Maximum
500
Steamtemperature,oC

450

400

350

300

250
1937 1942 1947 1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 2002 2007 2012
Delivery year

Figure 9. Development of recovery boiler main steam temperature.

13
 150

125
Maximum

100
Steam pressure, bar

75

50

25

0
1937 1942 1947 1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 2002 2007
Delivery year
Figure 10. Development of recovery boiler main steam pressure.
The maximum and average steam temperatures gasification to produce synthesis gas as a raw
have increased, Figure 9. The average steam tem- material for making higher value chemicals for
perature is now about 480 oC (Figure 9 and Table sale. This is sometimes referred to as the biore-
2). The average steam temperature has increased finery concept. It is more probable that plants
about 1.5 oC per year. By 2024 the average steam using direct biomass gasification will reach wide
temperature will be over 500 oC. The conserva- commercialization first. In any case integration of
tive steam temperature 540 oC used in PCF boilers these plants to pulp mills will provide new chal-
will be reached by 2049. This further confirms lenges and opportunities.
that technologically recovery boilers are 30 - 40
lagging the power boilers. The recovery boiler, as a producer of green en-
Steam pressure needs to correspond to steam tem- ergy, requires new materials and attention to feed-
perature. The high end of recovery boilers uses water quality. The role of chlorine and potassium
pressures of about 90 bar (Figure 10 and Table 2). removal in preventing superheater corrosion must
The average steam pressure has increased about be established. The recovery rating should be
0.8 oC per year. By 2024 the average main steam changed to tons of as fired black liquor because
pressure will be over 110 bar. The maximum us- virgin black liquor is a non measurable quantity
able natural circulation pressure will be reached in modern installations (due to complex recycling
by 2049. So after just 40 years the recovery boil- arrangements).
ers need to be of once-through design. The steam
temperatures are limited by the ash properties. So Conclusions
either new materials are needed or the ash chemis- New recovery capacity is being built at consid-
try needs to be redesigned with high temperatures erable pace. Large number of recovery boilers in
in mind. North America and Sweden are approaching the
end of their technical lifetime. If we are to reverse
Future trends and challenges the trend of pulp mill closure in North America,
Back liquor gasification still struggles to be ac- further investment to recovery is needed.
knowledged as a potential replacement for recov-
ery boilers. The driving force for these alternative The size of equipment in newest recovery depart-
recovery technology developments has been the ments continues to grow. Recovery boiler capacity
potential for much greater electricity production. has surpassed 6000 tds/d (13 million lb ds/day).
New high efficiency recovery boilers have considera-
bly reduced this gap. As many material and gas clean- The main future challenge is to respond to chang-
up issues remain to be resolved, the first large-scale ing business environment. Closer integration with
gasification plant still appears to be years away. biofuels production, higher bioelectricity produc-
tion and less CO2 production are just few of the
Attention is now shifting to using black liquor changes required.

14
References Steam. Kawasaki Technical Review, No. 131, Oc-
1. Steam its generation and use. 1992, 40th edi- tober 1996, pp. 243-248.
tion, Stultz, Steven C. and Kitto, John B. eds., 929 14. Kaneko, S., Takatuka, H., Arakawa, Y. and
p. ISBN 0-9634570-0-4. Tanaka R., Mitsubishi-CBC High Efficiency
2. Jones, Andrew K., 2004, 70 years of advances Chemical Recovery Boiler Power Generation
in recovery boiler design. 40th Anniversary In- System. 31st Pulp and Paper Annual Meeting,
ternational Recovery Boiler Conference, Finnish ABTCP, Sao Paolo, Brazil, October 19-23 1998,
Recovery Boiler Committee, Haikko Manor, Por- pp. 693-704.
voo, May 12-14, 2004, pp. 15 – 20. 15. Raukola, Antti T., Ruohola, Tuomo and
3. Pettersson, Bertil, 1983, Korsnäs sodapan- Hakulinen, Aki, Increasing power generation
nor under 40 år. (40 years of recovery boilers at with black liquor recovery boiler. Proceedings
Korsnäs) In Swedish, Sodahuskonferensen ‘83, 2002 TAPPI Fall Technical Conference, Septem-
ÅF-IPK, Stockholm, pp. 57 – 68. ber 8 - 11, San Diego, Ca, 11 p.
4. Sandquist, Kent, 1987, Operational experience 16. Vakkilainen, Esa K., Kankkonen, Sebastian,
of composite tubing in recovery boiler furnaces. and Suutela, Jukka, Advanced efficiency options
Götaverken Energy Systems, technical paper TP- - increasing electricity generating potential from
2-87, 15 p. pulp mills. Proceedings of 2004 International
5. Edling, Gustaf, 1937, Några synpunkter beträf- Chemical Recovery Conference. 6-10 June, Char-
fande sodahus av olika system. (Some observa- leston, South Carolina, 12 p.
tions concerning different types of recovery) In 17. Stern, Randy, Anängen, Fredrik and Wern-
Swedish, Presented in the Swedish Paper and qvist, Anders, Evaporation System Solutions
Pulp engineer’s winter meeting 6th march 1937, - Energy and the Environment. Proceedings of
in Sodahuset 1930 – 1945, Ångpannaföreningen, Tappi 2006 Engineering, Pulping, & Environ-
Stockholm, pp. 27 - 36 p. mental Conference, November 5 - 8, Atlanta,
6. Vakkilainen, Esa K., Kraft recovery boilers – Georgia, 17 p.
Principles and practice. Suomen Soodakattilayh- 18. Rauscher, John W., Kaila, Jarmo and Jaakko-
distys r.y., Valopaino Oy, Helsinki, Finland, 2005, la, Heikki, State of the Art Evaporation System.
246 p. ISBN 952-91-8603-7. Proceedings of Tappi 2006 Engineering, Pulping,
7. Vannérus, Torbjörn, Svensson, Gustav and & Environmental Conference, November 5 - 8,
Forslund, Folke, 1948, Ångpannor. (Steam gen- Atlanta, Georgia.
erators) In Swedish, Chapter 19, in Maskinteknik, 19. Haaga, K., and Laitinen, M., Experiences of
Ingenjörshadboken 2, Nordisk Rotogravyr, Stock- the World’s Biggest Recovery Boiler at Jinhai
holm, Sweden, pp. 937 – 1072. Pulp & Paper Co. in Yang Pu, Hainan, China. 7th
8. Tikka, Matti, UPM:n Kymin sellutehtaan uusi Colloquium on Black Liquor Combustion and
talteenottolinja. (New recovery line at UPM Gasification, 31.7-2.8.2006, Jyväskylä, Finland.
Kymi) In Finnish, Presented at Soodakattilapäivä 20. Westberg, Åke, Optimering av elproductio-
2008, Tampere. nen vid SCA Östrands massafabrik. Värme- och
9. Aikio, Anne, Kymin uusi soodakattila. (New Kraftföreningen, Panndagarna 2007, 7 -8 Febru-
recovery boiler at Kymi) In Finnish, Presented at ary, Uppsala, 14 p.
Soodakattilapäivä 2008, Tampere. 21. Kaliazine, A., Cormack, D. E. and Tran,
10. Vakkilainen, Esa K., 2007, Modern recovery H.N., Feasibility of Using Low Pressure Steam
practices and state-of-the-art technology. Pro- for Sootblowing. Pulp & Paper Canada, 107(4):
ceedings of 2007 International Chemical Recov- T80- 84.
ery Conference May 29 - June 1, 2007, Quebec 22. Vakkilainen, Esa and Holm, Ralf, Firing very
City, QC, Canada, 6p. high dry solids in recovery boilers. Proceedings
11. Tran, Honghi and Vakkilainen, Esa K., Ad- of Tappi Engineering Conference, September 17-
vances in the kraft chemical recovery process. In- 21, 2000, Atlanta.
ternational colloquium on Eucalyptus Pulp, March 23. Wallén, Jonas, Viiala, Juhani and Björkström,
4-7, 2007, Belo Horizonte, MG Brazil. 7p. Ingvar, Operational performance of Billerud
12. Vakkilainen, Esa K., Recovery Boiler Re- Gruvon’s new recovery boiler with environmen-
placement Options. 7th Colloquium on Black Liq- tal friendly solutions. Proceedings 2002 TAPPI
uor Combustion and Gasification, 31.7-2.8.2006, Fall Technical Conference, September 8 - 11, San
Jyväskylä, Finland, 22 p. Diego, Ca, 10 p.
13. Matsuda, T., Masuda, T. and Suemitsu, N.,
Outline of the World’s Largest Soda Recovery
Boiler with High Pressure and High Temperature

15
16
HOW WELL DO WE UNDERSTAND RECOVERY
FURNACE PROCESSES?
MIKKO HUPA, MARKUS ENGBLOM, ANDERS BRINK,
CHRISTIAN MUELLER
ÅBO AKADEMI PROCESS CHEMISTRY CENTRE
TURKU, FINLAND

Introduction

Black Liquor Recovery Furnace is one of the most This paper discusses some of the recent advances
complex large scale industrial reactors around. On in the CFD-based modeling of recovery furnaces.
one hand it takes care of recovery of the cooking The emphasis is on some special features of the
chemicals sodium and sulfur in a chemically suit- black liquor recovery furnace as compared with
able form – as a mixture of alkali carbonate and conventional furnaces in which CFD based mod-
sulfide – to be tapped from the furnace floor as a eling has been applied. While the material pre-
smelt. On the other hand it is a reactor for clean sented focuses mainly on the modeling carried out
combustion of the organic part of the liquor to at Åbo Akademi, significant modeling research in
generate heat and power in a steam boiler cycle. other laboratories is carried out [1-7].

Black liquor recovery furnace processes are com- There are many physical and chemical phenom-
plex, but systematic research has improved our ena which need to be described in CFD-based
understanding of these processes considerably. modeling of recovery boilers. Many of these phe-
Different approaches have been applied to learn nomena are similar to other types of combustion
more about the recovery boiler. Various types of furnaces, and can be directly applied to recovery
measurements are carried out in operating boilers boilers as well. These include the turbulent gas
to explore the state and composition of the fuel flow, radiative and convective heat transfer, par-
gases in different parts of the furnace. Ingenious ticle flight in the gas flow etc. For black liquor,
laboratory measurements have given valuable in- the most important specific submodels include
formation of the details of the properties and be- description of the liquor droplet conversion, some
havior of burning liquor droplets, chemical reac- chemical reactions in the gas phase, and the proc-
tions in the furnace gases etc. ess of the char bed formation and oxidation. The
models interact with each other, meaning that as
Mathematical modeling based on CFD has be- the droplets fly inside the furnace they exchange
come an excellent tool to really test our overall energy and mass with the gas phase, and the prod-
understanding of the furnace processes. The CFD- ucts from droplet conversion react further in the
based model gives us a possibility to combine our gas phase. In addition, the material landing on the
detailed understanding of the many furnace phe- char bed and the local gas environment affect char
nomena into an overall description of the behavior bed burning.
of a full scale furnace.
Droplet Combustion Model -
CFD has been applied to recovery boilers for
nearly two decades. CFD is originally a general The Key to Furnace Modelling
methodology for modeling fluid flow. Therefore, Black liquor is a unique fuel: it contains large
application of it to study air systems and gas flows amounts of inorganic matter, its char is highly re-
inside the recovery furnace was a natural early active due to the high loading of alkali, and its
development. More efficient computers, user- burning is connected with considerable swelling
friendly programs, and more detailed models have which varies from liquor to liquor. Laboratory
contributed to the increased use of CFD in more studies using single droplets have provided in-
comprehensive studies of recovery boilers. sight into the combustion behavior of black liq-

17
 uor, including liquor-to-liquor differences [8,9]. Although there are uncertainties, these simula-
Based on this knowledge, a general droplet burn- tions show some of the positive aspects of CFD-
ing model has been developed [10]. This droplet based modeling. The first is data visualization,
model can be made liquor specific by extracting i.e., the results can be presented graphically as if
information from laboratory experiments. It can we were looking into the furnace by using suit-
then be used in full furnace simulations to study able colors to indicate e.g. temperature or compo-
the effect of liquor-to-liquor differences on in-fur- sition of the furnace gases. This is a good way to
nace combustion. get an overall picture of what takes place inside
the recovery boiler. The second is that modeling
Such a study [10] was recently carried out by sim- gives data from within the furnace. The numerical
ulating combustion of two liquors: a High Swell- information in Figures 3 and 4 could perhaps be
ing and a Low Swelling liquor. The liquors were estimated qualitatively, but would be impossible
burned in two recovery boilers: the boilers A and to extract from an operating recovery boiler by
B. Some of the results of this study are presented any experimental techniques.
in Figures 1-4. With the high swelling liquor, com-
bustion takes place to a greater extent in-flight and In addition, the models are general and thus valid
higher up in the furnace as with the low swelling for any sizes of the furnaces. Thus, the impact of
liquor. In addition, droplet combustion occurs in furnace size on the combustion inside it can be ef-
different regions of the two furnaces (Figures 3 fectively studied. It is quite clear that the impres-
and 4). The recovery boilers had considerably dif- sive increase in the size of the furnaces in the lat-
ferent capacities: 450 t DS/day and 3600 t DS/day, est boiler deliveries would not have been possible
for A and B, respectively. In furnace B, the larger without the support of CFD based modeling.
of the two, droplet combustion takes place to a
greater extent in-flight.

Figure 1, Influence of liquor swelling on temperature in Furnace A [10].

18
Figure 2, Influence of liquor swelling on temperature in Furnace B [10].

Figure 3, Influence of liquor swelling (Low Swelling Liquor LSL; High Swelling Liquor HSL) on the occur-
rence of the stages of combustion in different regions of Furnace A (in-flight F; on furnace wall W, on the
char bed B [10].

Figure 4, Influence of liquor swelling (Low Swelling Liquor LSL; High Swelling Liquor HSL) on the occur-
rence of the stages of combustion in different regions of Furnace B (in-flight F; on furnace wall W, on the
char bed B [10].

19
 Nitrogen Oxides – of the nitrogen in the liquor goes to char; 35 % is
released as ammonia, NH3; and another 35 % as
Still a Challenge molecular nitrogen, N2 during the devolatilization
Black liquor differs from all other fuels in its ni-
(or pyrolysis) stage of the droplet conversion. The
trogen chemistry. The special feature is that part
release of ammonia can be assumed proportional
of the organic nitrogen in the liquor after com-
to the rate of carbon released during pyrolysis.
pleted char oxidation may stay in the smelt resi-
During char combustion of the droplets, nitrogen
due as inorganic salt, sodium cyanate. This ”smelt
is assumed to be released as NO if the local en-
nitrogen” is a source of large amounts of nitrogen
vironment is oxidizing, whereas no nitrogen is
species – mostly ammonia – found in the various
released if the conditions are reducing. The latter
water solutions of the chemicals recovery cycle of
situation results in an increase in the N/C ratio in
the pulp mill. Today there is relatively good un-
the char residue. If the final burnout of char car-
derstanding of the fate of nitrogen in the recovery
bon takes place under oxidizing conditions, the
cycle (Figure 5) [11-15].
remaining nitrogen is released as NO.
One main source of NO emissions from the recov-
Nitrogen chemistry in the gas phase is calculated
ery boiler is the part of the fuel nitrogen which is
using a reaction mechanism involving 21 species
released during devolatilization [14,15]; an over-
and 52 elementary reactions, including pathways
view of the reactions of black liquor nitrogen is
also for thermal NO formation. Figure 7 presents
shown in Figure 6 [13,14]. The importance of the
simulated NO concentrations in a recovery fur-
molecular nitrogen in the air as a source for NO
nace.
(the so called thermal NO mechanism) is probably
negligible (see below).
One of the findings of the study [16] was that ther-
mal NO does not contribute to the NO emission
Modeling the NO formation in the recovery fur-
from the recovery boiler. Although once again
nace requires a description of the release of the
there are uncertainties, in this respect the simu-
liquor nitrogen from the burning droplet. This
lation results are consistent with the current un-
can be made based on laboratory droplet burning
derstanding of NO formation in recovery boilers.
studies. In the latest version of the droplet model
However, the models require further development
[16], the following assumptions are included. As
to enable them to describe the full complexity of
a first approximation it can be assumed that 30 %
nitrogen chemistry in the recovery furnace.

Figure 5, Fate of nitrogen in the Kraft recovery cycle [11].

20
Figure 6, Overview of the reactions of black liquor nitrogen in the recovery furnace [13,14].

Figure 7, Simulated NO in a recovery furnace [16].

21
 According to the simulations [16], the conditions the total NO emissions when liquor is sprayed
in the furnace favor oxidation of NH3 to NO. The into the furnace as larger droplets.
amount of NH3 released and the final NO at the
furnace outlet matched almost perfectly, thus sug- Liquor Spraying –
gesting that essentially all NH3 is oxidized to NO
in the furnace process. The impact of droplet size The Most Important Input
on NO emissions from the furnace was investigat- Introduction of the fuel into the furnace as a rela-
ed by varying the average droplet size in the black tively coarse spray is a special feature of black
liquor spray. In the original coupled simulation liquor combustion. Operational experience shows
of the furnace, an average droplet size of 6.3 mm the furnace combustion process to be highly sen-
was used. To avoid excessive computation time, sitive to differences in spraying. Extensive spray
the investigation of the impact of droplet size was characterization has been carried out at the Helsin-
carried out in post-processing mode: the veloc- ki University of Technology [17,18]. In addition,
ity and temperature profiles in the furnace were modeling of the actual droplet formation process
frozen, and the impact of droplet size on these in the liquor nozzle is ongoing [19]. However,
variables was neglected. Instead, for each average such models are not yet available as submodels
droplet size, the calculation of droplet trajectories for a full simulation of the furnace. Therefore, in
took into account the release of nitrogen species the meantime, the droplet formation process is
from the droplets (Figure 8). In the droplet mod- neglected, and the black liquor spray is described
el, all volatile matter is released to the gas phase in terms of droplet boundary conditions for direc-
upon impact on a solid surface, i.e., the furnace tion, velocity, and mass flow rate.
walls or the char bed. For this reason, all volatile
nitrogen is also released to the gas phase, Figure These initial droplet parameters are essential for
8: the NH3 release rate is independent of droplet performing accurate furnace simulations, as was
size. However, nitrogen is released from droplets learned from a recent study on different spray
as NO only if char burnout takes place in-flight models [20]. In this study three models were
and under oxidizing conditions. The decrease in tested, whose features are illustrated in Figure 9.
NO release from droplets with increasing droplet In the first model, droplets were introduced into
size (Figure 8) occurs because a higher share of the furnace using a small number of discrete in-
droplets hit a furnace wall or the char bed before jections; in the second, droplets were introduced
char burnout. This result suggests a decrease in into the furnace using a continuous distribution

Figure 8, Nitrogen release as a function of average droplet size [16].

22
within a horizontal opening angle of 120°; and in carried out by Helsinki University of Technology
the third, droplets were introduced into the fur- showed that the liquor sheet formed at the liquor
nace using a continuous distribution within both gun nozzle is not two-dimensional but in fact has
a horizontal (120°) and a vertical (10°) opening a vertical opening angle. Some simulation results
angle. The third model was included in the study are presented in Figures 10 and 11.
because the spray characterization measurements

Figure 9, Features of different spray models: 1) particles introduced in 11 discrete directions; 2) continuous
distribution within a horizontal opening angle (120°); and 3) vertical opening angle (10°) also considered
[20].

Figure 10, Furnace temperature distribution using three spray models: 1) particles introduced in 11 discrete
directions; 2) continuous distribution within a horizontal opening angle (120°); and 3) vertical opening
angle (10°) also considered [20].

23
 Large droplets are not greatly influenced by tur- [21]. In this model carbon conversion, bed tem-
bulence inside the furnace. Therefore the in-flight perature, and net carbon accumulation are calcu-
combustion of these droplets is affected mainly lated locally based on mass and energy balances
by their boundary conditions at the wall, i.e., the on the bed surface. The bed model is fully coupled
spray model. The effect on furnace temperature with the droplets and gas phase above the bed.
is shown in Figure 10. The differences are best Thus, input of char carbon and inorganic salts is
seen in the lower furnace. The main differences calculated from the droplets landing on the bed. In
in the temperature fields are between the discrete addition, the local gas velocity and composition
injection model (Model 1) and models using con- affect carbon conversion. The model predicts the
tinuous distributions within opening angles (Mod- contribution of the various carbon conversion re-
els 2 and 3). However, when both the horizontal actions involved in char bed burning. Some simu-
and the vertical opening angle were considered lation results are presented in Figure 12 [21].
(Model 3), the peak temperatures were higher up
in the furnace than in Model 2, in which only the The carbon conversion rate seldom matches the
horizontal opening angle was considered. rate of carbon input to the bed. Instead, net car-
bon accumulation is calculated as the difference
The choice of spray model also influenced the between carbon arriving with droplets and carbon
droplet landing pattern on the char bed (Figure conversion; in other words, carbon can either ac-
11). Again the main difference is between contin- cumulate or be depleted locally (Figure 13).
uous distributions (Models 2 and 3) and discrete
injections (Model 1), but including the vertical The net carbon accumulation can be interpreted as
opening angle (Model 3) resulted in more effec- the parameter which determines the bed shape: the
tive spreading of the droplets onto the bed. The bed grows in areas where the net accumulation is
droplet landing pattern is important for the simu- positive, and bed depletion occurs in regions with
lation of bed shape, as will be shown later. negative net accumulation. Based on net carbon
accumulation, changes in the char bed shape can
Char Bed – be calculated as a function of time [22]. Starting
from an initial shape, the bed shape is updated in
Can it also be Modelled? small steps. For each new bed shape, the furnace
The char bed is an additional special feature of model, including droplets, gas phase and the bed,
black liquor recovery boilers. A model has been is recalculated to obtain the new net carbon ac-
developed for describing char bed burning, in- cumulation.
cluding mechanisms for carbon conversion by
oxygen, carbon dioxide, water vapor, and sulfate

Figure 11, Droplet landing patterns on the char bed using three spray models: 1) particles introduced in
11 discrete directions; 2) continuous distribution within a horizontal opening angle (120°); and 3) vertical
opening angle (10°) also considered [20].

24
Figure 12, Carbon conversion calculated using the char bed model coupled to a furnace simulation [21].

Figure 13, Net carbon accumulation calculated using the char bed model [21, 22].

25
 The model was used to study the impact of the Although the primary value of the bed shape mod-
liquor spray initial droplet size on the bed shape el is in testing the impact of different parameters
[22]. Figure 14 presents some simulation results. on the bed shape behavior, an interesting detail of
Starting from a flat bed, different bed shapes were the simulations discussed here is that they were
obtained depending on the droplet size. Three av- carried out on a recovery boiler for which images
erage droplet sizes were used: 2 mm, 3.1 mm, and of the actual char bed are available. This made it
6.3 mm. In the 2 mm case, the bed underwent de- possible to compare the simulated and the actual
pletion; with the two larger droplet sizes, the bed bed shapes. The bed shape resulting from an av-
grew. The bed grew at a higher rate with an aver- erage droplet size of 3.1 mm was found to be in
age droplet size of 6.3 mm than with a droplet size good agreement with the actual bed shape (Figure
of 3.1 mm. The beds did not reach a steady-state 16).
shape, but continued to be depleted or to grow af-
ter the times shown in Figure 14.
Where are we going?
It was suggested earlier that the choice of spray CFD-based models for droplet, gas phase, and
model affects the bed shape. The model with both char bed burning describe what can be considered
a horizontal and a vertical opening angle (Spray the main characteristics of black liquor furnace
Model 3) was used to calculate the bed shapes processes. The modeling approach is based on
shown in Figure 14. For comparison, Figure 15 fundamental theories and concepts. It builds on
shows a bed shape simulated with the spray model established expressions for heat and mass trans-
in which droplets are introduced into the furnace fer, energy and mass balances, and it takes into
as 11 discrete injections (Spray Model 1) [20]. the consideration differences in the local environ-
The uneven distribution of droplets onto the bed ment. In addition, the submodels are coupled with
results in a bed shape which is less smooth than each other, thus allowing the different phenomena
those in Figure 14. to happen simultaneously and to interact. Conse-

Figure 14, Calculated bed shapes using different average droplet sizes [22].

26
Figure 15, Calculated bed shape using a spray model in which particles are introduced in 11 discrete direc-
tions (Spray Model 1) [20].

Figure 16, Real and simulated bed shapes. Pictures were taken through the same opening in the furnace
wall and show the bed on both sides of one mound. The view of the simulated bed is from the same direction,
and the portions of the simulated bed corresponding to those seen in the upper figures are indicated with
colored boxes [22].

quently, CFD models appear to give a reasonable input boundary conditions, especially the spray
description of combustion in black liquor recovery characteristics and droplet size distribution. The
boilers. At present, their exact prediction capabil- challenge for accurate simulations is to have rep-
ity is uncertain because of the lack of quantified resentative descriptions of these parameters.
in-furnace data for model validation.
The accuracy of the models is important if the
The issue of boundary conditions arises in con- simulations are to be used, for example, in proc-
nection with model validation. The boundary con- ess optimization. However, even without having
ditions in a recovery furnace simulation include established the absolute accuracy, response of the
the characteristics of all inflows to the furnace, models to changes in model parameters can re-
as well as the characteristics of the furnace walls, veal new insights into black liquor combustion,
e.g., wall temperatures. The simulated combus- confirm previous understanding, or indicate ar-
tion is a result of both the boundary conditions eas where additional fundamental knowledge is
and the processes inside the furnace. It has been needed. Although the full complexity of the re-
shown that the models are sensitive to several covery boilers processes is not yet described by

27
 the models, CFD has developed into a research Acknowledgements
tool, which can be used to study many details in A number of people at Åbo Akademi have contrib-
the recovery furnace and to increase our under- uted to the results. Thanks go especially to Rainer
standing of the overall furnace processes. Backman, Maritta Kymäläinen, Mikael Forssén,
Niko DeMartini, Nici Bergroth, and Kaj Eklund.
The question remains of how CFD-based mod-
eling could develop further. Incorporating more Colleagues in other research groups have sup-
detailed descriptions of the chemistry will extend ported many of the studies here. Thanks go es-
the models. However, this requires a sound un- pecially to Kari Saviharju, Lauri Pakarinen, Matti
derstanding of the chemistry, and there will be Ylitalo, Pasi Miikkulainen, Mika Järvinen, Ari
a continued need for fundamental experimental Kankkunen, Carl-Johan Fogelholm, and Andrew
data. Areas where our understanding does not yet Jones. Further, the long-term support and col-
allow for quantitative modeling include: (i) fume laboration with Honghi Tran and Jim Frederick is
formation and chemistry, (ii) fate and emissions acknowledged.
of the gaseous sulfur species (SO2 and H2S), (iii)
bed chemistry and the degree of reduction of the This paper is based on the research conducted
smelt. In all these areas the main chemical mecha- at the Åbo Akademi Process Chemistry Centre
nisms are identified, but determining their quanti- within the framework of the Centre of Excellence
tative dependence on the various furnace process program of the Academy of Finland.
conditions requires further work.
The work is also a part of the Åbo Akademi col-
Perhaps most important is obtaining compre- laboration with the companies Andritz, Metso
hensive and quantified in-furnace data. Valida- Power, Vattenfall, Metsä-Botnia, Foster Wheeler,
tion would not only establish the accuracy of the UPM-Kymmene, Clyde Bergemann and Interna-
current models, but it could make it possible to tional Paper with the support from the Finnish
identify in which respects the models still fail to Technology Agency, Tekes.
describe reality. This knowledge could assist in
directing further studies and model development. The fruitful long-term collaboration with the
Finnish Recovery Boiler Committee is acknowl-
edged.

28
References Hupa, M, The fate of nitrogen in the chemical re-
1. Sutinen, Jari, Karvinen, Reijo and Frederick, covery process in a kraft pulp mill. Part II. Am-
Wm. James, Jr., A chemical reaction engineering monia formation in green liquor. Journal of Pulp
and transport model of kraft char bed burning. and Paper Science, 2001, 27(3), 75-81.
Industrial & Engineering Chemistry Research, 13. Kymäläinen, M., Forssén, M., Jansson and
2002, 41(6), 1477-1483. M., Hupa, M, The fate of nitrogen in the chemi-
2. Jones, A.K., A model of the kraft recovery fur- cal recovery process in a kraft pulp mill. Part IV:
nace. Ph.D Thesis, The Institute of Paper Chemis- smelt nitrogen and its formation in black liquor
try, Appleton, Wisconsin, January, 1989. combustion. Journal of Pulp and Paper Science,
3. Sumnicht, D.A., A computer model of a kraft 2002, 28(5), 151-158.
char bed. Ph.D Thesis, The Institute of Paper 14. Forssén, Mikael, Kilpinen, Pia and Hupa, Mik-
Chemistry, Appleton, Wisconsin, January, 1989. ko, NOx reduction in black liquor combustion-re-
4. Wessel, Richard A., Parker, Kevin L. and Ver- action mechanisms reveal novel operational strat-
rill, Christopher L. Three-dimensional kraft recov- egy options. Tappi Journal, 2000, 83(6), 56.
ery furnace model: implementation and results of 15. Kymäläinen, Maritta, Forssén, Mikael, Kilpin-
improved black liquor combustion models. Tappi en, Pia and Hupa, Mikko, Nitrogen oxide forma-
Journal, 1997, 80(10), 207-220. tion in black liquor single droplet combustion.
5. Ohran, Benjamin J., Choi, Michael M., Kaer, Nordic Pulp & Paper Research Journal, 2001,
Soren and Baxter, Larry L. Comprehensive, three- 16(4), 346-354.
dimensional CFD model of a reacting char bed. 16. Brink, A., Engblom, M. and Hupa, M.,
International Chemical Recovery Conference; ”Nitrogen oxide emission formation in a black
11th International Symposium on Corrosion in the liquor boiler. Tappi Journal, 2008, 7(11), 28-32.
Pulp and Paper Industry, Charleston, SC, United 17. Kankkunen, A. and Miikkulainen, P., IFRF
States, June 6-10 and June 7-11, 2004 (2004), Combustion Journal, Article Number 200308,
686-697. 2003.
6. Saviharju, K., Pakarinen, L., Kyttälä, J., Jukola, 18. Miikkulainen, P. Spray formation of high dry
P., Viherkanto, K., Näkki, I. and Hämäläinen, M., solids black liquor in recovery boiler furnaces”
Three dimensional char bed imaging for numeri- PhD thesis, Laboratory of Energy Engineering
cal simulation feedback. International Chemical and Environmental Protection, Helsinki Univer-
Recovery Conference, TAPPI PRESS, 2007. sity of Technology, 2006.
7. Yuan, J., Xiao, Z., Salcudean, M., Singh, P.M. 19. Levesque, D., Fard, M.P. and Morrison, S.,
and Gorog, P., Numerical study on mid-furnace BLSpray: Understanding the Effect of Black
corrosion in a kraft recovery boiler. International Liquor Properties and Splash Plate Nozzle Con-
Chemical Recovery Conference, TAPPI PRESS, figuration on Spray Characteristics. International
2007. Chemical Recovery Conference, TAPPI PRESS,
8. Hupa, M., Solin, P. and Hyöty, P., Combustion 2004.
behavior of black liquor droplets. Journal of Pulp 20. Li, B., Brink, A., Engblom, M., Mueller, C.,
and Paper Science, Vol 13(2), 1987. Hupa, M., Kankkunen, A., Miikkulainen, P. and
9. Forssen, M.; Järvinen, M.,”Liquor-to-liquor Fogelholm, C-J., Spray models for CFD of black
differences in combustion and gasification proc- liquor recovery furnaces. 15th IFRF Members
esses: simultaneous measurements of swelling Conference, Pisa, Italy, 2007.
and CO2, CO, SO2 and NO formation reveals 21. Bergroth, N., Mueller, C. and Hupa, M. CFD
new data for mathematical models. International based modelling of recovery boiler char beds.
Chemical Recovery Conference, TAPPI PRESS, International Chemical Recovery Conference,
2001. TAPPI PRESS, 2004.
10. Mueller, C; Eklund, K; Forssen, M; Hupa, 22. Engblom, M., Mueller, C., Brink, A., Hupa,
M., Influence of liquor -to- liquor differences on M. and Jones, A., Toward predicting the char bed
recovery furnace processes - a CFD study. Inter- shape in kraft recovery boilers. Tappi Journal,
national Chemical Recovery Conference, TAPPI 7(10), pp 12-16, 2008.
PRESS, 2004.
11. Kymäläinen, M., Forssén, M., Hupa, M, The
fate of nitrogen in the chemical recovery proc-
ess in a Kraft pulp mill. Part I. A general view.
Journal of Pulp and Paper Science, 1999, 25(12),
410-417.
12. Kymäläinen, M., Forssén, M., DeMartini, N.,

29
30
CHALLENGES TO MANUFACTURE WORLD’S
LARGEST RECOVERY BOILERS?
KARI MÄKELÄ, METSO POWER
TAMPERE, FINLAND

Introduction

Boiler production is unique manufacturing proc- there are no “moving” components in the pressure
ess. The recovery boiler components made from parts.
tubes, pipes and membrane bars are manufactured
at workshops, transported to site and erected to Welded products have typically an image of hand-
finally be operated as one unit. Manufacturing is work at least combined together with a big flex-
partly serial production, if the component manu- ible size. Although the aim is towards mechanized
facture is viewed as detailed work. However, or automated production process the site welding
boiler manufacturing can also be classified as in- is extremely difficult to alter from manual weld-
dividual production. ing to a more productive type.
The size of the modern recovery boiler has in-
Largest recovery boilers have their own typical creased as shown in the figure 1 and the industrial
features, not only the dimensions and weight but drivers are mainly economical. This applies as
also their own design criteria. So recovery boiler well to boiler manufacture.
pressure parts can be categorized as a welded
product, not as a machine in respect of manu- The product itself has developed. The demand
facturing. This approach gives a good possibility of higher pressure and temperature leads to use
to understand the role of tolerances and weld- of more alloyed materials and or increased wall
ing shrinkages related to final product. Basically thicknesses of the final component. Manufac-

Figure 1. Relative size and capacity increase of recovery boilers.

31
 turing is currently becoming more regulated. components, dimensions and weights. One of the
This results in detailed instructions, procedures giant 7000 t ds/day recovery boilers currently un-
and specifications as well as more educated and der construction has the following dimensions;
trained labor. The requirements for manufacturing Furnace floor size is 16,5 x 18,0 m and the height
and quality have changed during the last decades. of the furnace is about 66 m. The total length of
The recovery boiler manufacturing technology is the pressure part tubing is more than 460.000
today regulated by two major codes/standards: meters. Huge weight affects also the way this
The American Society for Mechanical Engineers boiler is manufactured. The total weight of the gi-
(ASME) Section 1 as well as European Standard ant recovery boiler pressure parts is about 5200
EN 12952, mainly parts 5 and 6. These standards tons and the total operation hanging load reaches
give very detailed instructions on how to produce 11000 tons.
boiler components. Also boilers to be built outside
USA and Europe are typically based on the previ- Handling and transportation
ously mentioned standards. Handling of base material such as very long tubes
can be difficult due to the flexibility. The length
Health, safety and environmental issues in manu- of the economizer tubes is more than 24 m. Thus
facturing have become increasingly important. even the transportation from the tube mill to the
Not only during the new boiler manufacturing but component manufacturing workshop is challeng-
also during the service of these huge giants must ing and therefore tubes with butt welds are accept-
number of requirements be met. The environment ed today. The economizer tubes are fin welded and
at site is clearly different from 60 years ago. This the butt welds need to be ground flush.
can be realized easily when looking at figures 2
and 3 Special attention is paid to lifting at site. The large
recovery boiler components such as superheater
Boiler Size and Manufacturing and economizer elements are very flexible and the
Dimensions and Weights buckling of the components during the lifting is
Boiler dimensions have naturally an effect to avoided using special lifting frames, several lift-
productivity as well as they from a challenge to ing points or at least two cranes. Workshop lifting
manufacturing. Product cost reduction favors big capacity can be limiting the weight of the com-

Figure 2. Attitudes on the HSE issues have Figure 3. 7000 tds/day recovery boiler under
changed, photo from a boiler erection site, year construction, 2009.
1951.

32
ponents to be handled as one piece. Also road ufacturing parameters have a very narrow varia-
transportation from the workshop to the site has tion. The transportation and the lifting of the drum
to be well planned. The weight is often causing require heaviest cranes and vehicles.
problems at the site due to need of bigger lifting
capacity cranes. These are often much more ex-
pensive. Details of the components
Welding is always related to joint type, although
The drums for the large recovery are still manu- the trend is to larger component arrays. Recovery
factured at the workshop. Only a few international boiler manufacturing either small or big has three
manufacturers are able to produce drums with big major joint types: butt welds, nozzle welds and
outer diameter together with heavy wall thickness membrane bar to tube fillet welds as shown in the
more than 100 mm. Manufacturing procedures of figure 6. The welder must deal with all of these
the drum are more detailed and the allowed man- joint types. The major difference is the location of
the joint in a large recovery boiler. It is crucial to
have welders who are not terrified of heights.

Bending of tubes is similar in small and big recov-

ery boilers. As more steam is produced in big re-
covery boiler, bigger tube outside diameters must
be used. This means heavier wall thicknesses and
or stronger materials and therefore the need for
heavier bending machines. Heating of the tube
or pipe can be used to reduce the needed forces
during the bending. Induction heating is typically
used for big outer diameter pipes.

Figure 4. Economizer tube bundles as received at

the boiler manufacturing workshop.

Figure 5. 20 m long 132 ton steam drum on the way to the site for 6000 tds/day boiler.

33
 Figure 6. Three main joint types: butt weld, nozzle weld (T-joint) and fillet weld.

Influence of the Product Design

Material Grades sign. Manufacturing and assembly order is well

Composite tubes were first time used in the recov- planned as shown in the figure 7.
ery boiler lower furnace in 1971. The outer layer
was standard AISI 304 quality. Stainless steel
requires cleaner working environment especially Fatigue
during welding. Carbon steel contamination has Long tubes although they are connected to each
to be avoided at all cost. During late 80’s special other are extremely flexible. Mechanical fatigue
Ni-based alloys such as 825 and 625 composite is one of the essential mechanical design criteria.
tube outer layer for lower part of the furnace were Risk to fatigue failure can be reduced by smoother
introduced. The impurity level at the welding weld contours to the base material. The fatigue
joint area has to be minimized in the workshop lifetime can be increased if the angle θ shown in
or at site. Specially designed electric motors have the figure 6 is larger. This means skilful welders
to be used near the cleaning area in order to avoid or more costly additional grinding.
explosion if solvents as acetone is used.
Creep
One membrane wall for the large recovery boiler The highest material temperatures are at the creep
can have more than 200 tubes and 400 double fil- range in large recovery boilers. Creep loads origi-
let welds. The tube tolerances and the welding nate from internal pressure as well as from exter-
shrinkages have an effect to the final wall dimen- nal mechanical load. Usage of high alloyed creep
sions. Special methods have to be used to achieve resistant steels leads to preheating and post weld
the proper size of the wall. In addition different heat treatments with exactly specified procedures.
composite tube and membrane grades have dif- Welding parameters and welding consumables are
ferent thermal expansion coefficients and shrink- selected to meet the creep requirement.
ages. The need of a specified grade depends on the
temperature, environment, client and manufactur-
ers standards. Almost every boiler furnace has its Corrosion
own material combination. Therefore experience Possible heating surface corrosion has an effect
is the key to achieving right dimensions. to manufacturing too. The interpass temperature
between weld passes is limited for stainless steels,
Design cooling to proper level leads to waiting times.
There are no extra spaces in a huge recovery Corrosion can be reduced at superheater area by
boiler. The trend is towards very compact de- proper welding consumable selection as well as
right heat input during the welding. Carbon steel

34
Figure 7. Compact layout is one of the key factors affecting to assembly order.

contamination of the stainless steels has to be large recovery boiler components. The accuracy
avoided in the final product. It is difficult or at requirement is also increased due to length of the
least not productive to protect large areas from components as well as the possible mechanized
carbon steel dust or sparks. welding

Workshop and Site Manufacturing Preheating and post weld heat

Bigger components need larger areas to be manu-
factured in the workshop. Product design require- treatments (PWHT)
ments are reflected to workshop size and vice ver- Preheating is carried out before welding and it is
sa the workshop space dimensions give guidelines local, only located near the weld joints. The dif-
to process and mechanical design. The change of ference between small and large recovery of the
the working area is dramatic as can be seen in fig- preheating temperature control comes from the
ure 8. The workshops today look nearly like of- wall thicknesses and material grades used. Post
fices of older days and enough space is reserved weld heat treatment is more challenging. Furnace
to produce the large components. sizes are limiting the PWHT to be carried out as
one component. Solution for this is carrying out
the PWHT in several stages or local. Overlapping
Pre-treatments of the PWHT area is required if the component
Ferritic tubes are typically grid blasted in order is heat treated in stages. Overlapping area means
to move the oxide layer as well as the impurities longer exposed heat treatment times for the area.
away. This is done to increase the productivity of The heat treatment parameters cap where to oper-
the welding as well as to raise the quality level ate is narrow and it results to more sophisticated
control processes related to time temperature sum
given to the work piece
Machining
Drilling the holes and groove manufacturing re-
quires bigger and heavier machining units for

35
 Figure 8. Tampella Workshops probably from 1940 and 1970.

Forming lenging not only because work is done at elevated

Forming is mainly bending, also reducing the level. Huge amount of scaffolding is needed and
size of the tubes are needed. Naturally, plate work the prevention from the effects of wind and rain
forming is concentrating to the drum; heads and is required. The roots of the welds are welded
shell. Bigger components together with high with TIG welding process requiring shielding gas.
strength materials need stronger equipment to car- Root gas is also needed for highly alloyed materi-
ry out the forming and even more the hot forming. als and good shielding for root area is difficult to
Temperature control during the hot forming of the achieve due to long distances from open place and
thicker material needs sophisticated temperature the root area of the welding joint.
measurements as well as skill. One of the key
factors in bending is ratio of bending radius (R)
divided by outer diameter (D). Smaller R/D is re-
Labour
Welders have been qualified already for 40 years
quired to pack the superheater into smaller space.
in boiler manufacturing. It is important for a
This means either hot bending or cold bending
welders manufacturing large size recovery boiler
with more detailed and specified bending param-
to understand the procedures. The information of
eters. More knowledge related to bending param-
manufacturing methods, parameters and rules are
eters such as compression forces during bending
issued in these documents.
is often required to perform superheater bending
especially using composite tubes and solid stain-
All these modern giant recovery boilers are lo-
less tubes
cated outside of Scandinavia. Naturally, skilled
workers can be found in Asia especially in South-
Welding America. Transfering the manufacturing require-
The amount of welding is dramatically increased ments needs also very experienced manufacturing
in the giant recovery boiler manufacturing. There- supervisors. It is not enough to transfer the techni-
fore, the trend is towards mechanization and auto- cal and quality information but also to understand
mated production, in order to keep the production the local culture and way of working.
cost low. Economizers can have tube to fin fillet
about 350 km and the length of the composite fil- Quality
let weld in the furnace can be roughly about 100 The basic quality requirements come from the
km. used code and standards. In addition, manufac-
tures have their own rules based on their design
The challenges to mechanize or automate the and customers have mill standards. In order to
manufacturing process are the dimensions of the control that the work is carried out according to
components, tolerances of the base tube material contract and rules authorized inspectors are used,
and welding shrinkages. The weld joints cannot third party inspectors from client side and natu-
be located before and during the manufacturing rally manufactures own personnel are involved.
process with enough accuracy in order to place
the welding torch exactly to the right position. To be able to manufacture large recovery boiler
Site welding mechanization is even more chal- manufacturer shall have a quality assurance sys-

36
tem, which will be continuously revised and
checked in the audits Same inspection methods
are used for inspection small and large recovery
boiler, only increased amount of isotope is used
instead of x-ray inspection. Traceability of the
welds is very typical for the joints in recovery
boilers: what manufacturing procedures have
been used, records how the manufacturing have
been carried out and who has done the work not
neglecting the material traceability.

Non-conformance reports are used as a tool to

record components or their parts if they do not
fulfill all the requirements stated before com-
mencing of the manufacturing.

Conclusion
The main difference between large recovery
boiler manufacturing and small is basically the di-
mension and the weight of the components. This
means more challenges during the component
transportation from the workshop to the site. Lift-
ing the components must be planned with more
detail.

The product design, bigger amount of steam flow

and higher steam parameters have lead to thicker
materials as well as using more alloyed materials.
Manufacturing has to take into account flexibility
of the mechanical design and finalize the welds
with smooth contours to avoid fatigue. Large re-
covery boiler welder at site will be one of the last
handwork professions although productivity and
therefore mechanization and automated produc-
tion is the target.

Manufacturing is not the limiting factor produc-

ing even bigger recovery boilers but then the pro-
ductivity may decrease. The quality of manufac-
turing has been increased resulting on operation
that is more reliable. Big recovery boiler has been
a forerunner showing the quality requirements
and if small size recovery boiler is manufactured
in the future it will adopt the same standards and
requirements as big recovery boilers today even if
that detail of attention is not needed.

37
38
BENCHMARKING A MODERN RECOVERY
BOILER
MIKAEL AHLROTH
THE SWEDISH-NORWEGIAN RECOVERY
BOILER COMMITTEE
SE-16 999 STOCKHOLM, SWEDEN

Introduction

The challenge today’s recovery boiler owners In the period 2005-2008, three new high pressure
face is how to ensure high boiler availability and recovery boilers were taken into operation (Stora
energy efficiency and low emissions without jeop- Enso Skoghall 107 bar/500°C, SCA Östrand 106
ardizing the safety issues. The recovery boilers in bar/515°C and SCA Obbola 110 bar/505°C). In
Sweden are today fired with increasingly higher Skoghall’s case, a new evaporation plant was also
dry solids black liquor containing relatively high taken into operation, enabling high liquor dry-
amounts of non-process elements. ness, 80% DS content or more. These three boil-
ers are expected to be in operation until the year
In 2006, the Swedish-Norwegian Recovery Boiler 2040 or longer.
Committee (SNRBC) initiated a three-year proc-
ess study regarding the state of the art recovery In the future even higher steam data and/or liquor
boiler technology called “Förutsättningar för dryness can be predicted. A Japanese recovery
framtidens sodapanna” (the Opportunities of fu- boiler is reported to operate at 133 bar/515°C steam
ture recovery boilers). The study aimed to find data. The UPM Wisaforest and Metsä-Botnia Kemi
those factors that will be crucial for the future mills are reported to burn 85% DS liquor. The
recovery boiler development in respect to safely, evaporation plant manufacturers have demonstrat-
low emissions and energy efficiency. ed that evaporation to 90% liquor DS is feasible.

Background The starting point for the process that led to this
The recovery boiler technology has developed study was the 40th anniversary of the SNRBC in
considerably during the last 15 years in Sweden. 2005. A workshop was held at the anniversary to
The liquor dryness has increased from 70% DS plan the activities for the committee for the next
content to 80% DS content or more. The boilers three years. The members expressed an almost
are bigger and have larger cross sectional area unanimous desire for a new process study on re-
than before. New air systems have been imple- covery boiler combustion and operation.
mented which enable more flexibility in liquor
combustion. On top of this, more process streams These targets are not expected to materialise with-
are led to the evaporation plant and the recovery out effort. As the members stated at the workshop,
boiler than before, increasing the NPE load. The there are difficulties in recovery boiler operation
emission regulations have been tightened and es- today. Furthermore, the operation of modern
pecially the NOX – regulations limit the degrees of state-of the-art boilers can be more difficult. The
freedom in liquor combustion. operation of future, more advanced recovery boil-
ers, can be even more difficult to handle.
The recovery boiler park of the SNRBC was some
30 years old in 2005. Several smaller boilers were Already in 2005 an inquiry was sent to Stora Enso
then – and are even more so today – close to the Skoghall mill regarding the possibility to conduct
end of their technical service life. There are also mill trials at the then brand new high pressure re-
several mills with more than one recovery boiler, covery boiler. The inquiry was well received and
while the international trend is one mill – one re- Skoghall agreed to participate in the project, if
covery boiler. external funding could be secured.

39
 The Board of the SNRBC decided in 2006 to The total cost for the project was calculated to
initiate a three-year process study on state-of-the- be 2.3 MSEK in 2006. The SNRBC allocated 1
art recovery boilers based on the opinion of the MSEK for the three year period. Additional fund-
members from the workshop in 2005. The study ing was received from Åforsk, the ÅF Research
was called ”Förutsättningar för framtidens soda- Foundation, and Värmeforsk, the Thermal Engi-
panna” (The opportunities of future recovery boil- neering Research Institute, 650 000 SEK each.
ers) or just ”3FS” in Swedish. See Table 1 below for details.

This study tried to answer to the questions that The members of the SNRBC have also contrib-
have been raised operating recovery boilers with uted with manpower in-kind to the study. Addi-
increasingly higher dry solids black liquor and tionally, the mill personnel at Stora Enso Skoghall
steam data, higher NPE content, new air sys- mill helped the project with manpower and analy-
tems and bigger recovery boilers. The members ses worth several 100 000 SEK in-kind.
of SNRBC can benefit from these answers, espe-
cially if they plan to apply for two-year operation The project was officially initiated in April 2006.
between internal and external inspection. Another During 2006 and most of 2007, the mill trials were
motive for the study was to collect and analyse planned and preparations made. The mill trials
the process information from one modern recov- were delayed a few months and were completed
ery boiler and make these results available to all November 26th to December 1st, 2007. Results and
members. analyses were completed during 2008. The final
report was issued to Åforsk and Värmeforsk in the
Two similar studies were conducted by the second week of 2009.
SNRBC earlier. In 1987-1988 a study called
”Förbränningsteknisk undersökning i sodahus” The project was reported to the Board and mem-
(A combustion technology investigation of a re- bers of the SNRBC during the course of the work.
covery boiler) was undertaken. In 1989-1991 an- A report was issued to the two external financers
other study called ”Förbränning och korrosion i Värmeforsk and Åforsk. The project was also re-
sodapannor” (Combustion and corrosion in recov- ported at the Swedish and Finnish Recovery Boil-
ery boilers) was completed. These studies were er conferences and the AF&PA’s Recovery Boiler
funded by the SNRBC with the aide of public and conference.
private research foundations.
The 1987-1988 study was presented at the Inter-
national Chemical Recovery Conference 1989
The Project [1] and in Pulp&Paper Canada [2]. The later
A separate project organisation was set up by the 1989-1991 study was presented at the Interna-
SNRBC to manage the project. A project team tional Chemical Recovery Conference 1992 [3],
was formed, headed by Per Utterström, Korsnäs Pulp&Paper Canada [4] and at the Corrosion Con-
Gävle. The team included members from ÅF and ference in 2001 [5].
Stora Enso. Two reference groups were connected
to the project team, with representatives from
SCA, Metso Power, M-Real and Södra Cell.

Table 1, 3FS budget and contributors.

40
Subprojects conditions at the tertiary air level in the boiler,
The SNRBC project team decided to split the where the unburned gases from the lower part of
project into four subprojects. the boiler are oxidised. The reason for this was to
find if the combustion of the dissolver off-gases in
this air level were beneficial from a NOX emission
Subproject Process technology point of view.
In this subproject, all process measurements and
sampling was done in the Skoghall recovery
Åbo Akademi University was subcontracted
boiler. Originally, the subproject was planned to
to take smelt samples from the smelt spouts in
be executed in September 2007. This was pushed
Skoghall and analyse these and the smelt sam-
forward to November due to a digester plant ca-
ples taken from inside the smelt bed for nitrogen
pacity increase. This allowed for a higher recov-
(cyanate) compounds. The Skoghall mill was
ery load.
conducting a millwide nitrogen assessment dur-
ing the period that the 3FS project was executed.
First a so called reference run was made. During
Some information from that study was utilized in
this 24 hrs test run, the boiler was run at maximum
this study.
available liquor load at normal liquor dryness,
80% DS. After this test run, the liquor dryness
was planned to be raised to 83% DS or more. The Subproject Energy efficiency
evaporation plant was pushed to the limit, during The idea with this subproject was to make a tech-
which the liquor dryness was targeted to be 83% nical potential study of how to improve the ener-
DS or more. This elevated liquor dryness was kept gy efficiency of a state-of-the-art recovery boiler
for two consecutive days, during which the actual similar to Skoghall’s boiler by using advanced
project measurements and sampling were made. simulation tools. A model of Skoghall’s recovery
boiler was set up and the boiler performance was
The following samples were taken and measure- calculated for 80%, 83% and 90% liquor dryness.
ments done: After this, the boiler was “fitted” with intermedi-
ate steam superheating, feed water preheating and
• Liquor samples for dry solids and a flue gas cooler. The impact of these technologies
composition analysis. was simulated one by one and in combination to
• Ash samples from the boiler bank and give an insight into how much the energy efficien-
economiser areas and from the electrostatic cy could be improved.
precipitator for composition analysis.
• Some 300 boiler instruments were tagged
(one minute averages) for a total of four days. Measurements in Skoghall
• Dissolver off-gases and dissolver off-gas The measurement campaign was successfully
condensates for various analyses. executed from November 26th to December 1st,
• HVLC gases and condensates for various 2007. The sampling and measurement points are
analyses. shown in Figure 1.
• Smelt from inside the smelt bed with a special
probe designed by ÅF. A number of analyses During Monday the 26th the last preparations were
were carried out. made. One GC was installed in a sampling booth
located in the Skoghall boiler stack to measure the
nitrogen compounds. Another GC was prepared
Subproject Material science in Skoghall’s lab to analyse the sulphur com-
This subproject handled all the material science
pounds. Sampling bottles and cans were distrib-
issues. Åbo Akademi University was subcontract-
uted. Skoghall’s dayshift and lab personnel were
ed to measure the fouling build-up and composi-
given a briefing on the final measurement plan.
tion in the superheater area of the Skoghall boiler.
Åbo Akademi University conducted a functional-
These measurements were made simultaneously
ity test of the fouling probes that afternoon.
with the process technology subproject.
The reference run at the planned 80% liquor dry-
Subproject NOX ness and 1900 tDS/24 h liquor load was executed
The purpose for this subproject was to investigate on Tuesday 27th. All samples were successfully
the nitrogen content in the process flows going in extracted, except the gas samples from the lower
and out from the recovery boiler. Additionally, an furnace. In this location the sampling tubes melt-
important part was to investigate the combustion ed. Smelt samples were also successfully extracted

41
 Figure 1, Sampling and measurement points in Skoghall. Sampling points for medium heavy liquor, HVLC
gases and HVLC condensate not shown.

from inside the smelt bed. About a dozen samples Institute for analysis. Åbo Akademi University
were taken. A few images of the measurement and analysed the cyanate content of the smelt samples
sampling activities are shown in Figure 2. and the chemical composition of the fouling col-
lected in the superheater area.
During Wednesday afternoon, the evaporation
plant was tuned up to start increasing the liquor In 2008 it became known that there had been a
dryness. The target dryness was set to 83.5% DS. leak of medium pressure steam in the primary air
Raising the liquor dry solids content was surpris- preheater. It is not known how this leak affected
ingly easy. the measurements.

The long anticipated 3FS-test run was executed

from Thursday November 29th to Saturday morn- Results and discussion
ing December 1st. Skoghall’s recovery boiler
was run at around 83.5% DS and 1900 tDS/24h.
General recovery boiler
Like two days earlier, all planned samples were performance
collected. After the measurement campaign was The results from the liquor analyses are shown
completed, liquor, smelt, condensate and boiler in Figure 3 below. Note that it is very difficult to
ash were sent to the Swedish Technical Research determine the dry solids content for black liquor

42
Figure 2, Measurement and sampling activities in Skoghall. Clockwise from top left: nitrogen compound
analyser in flue gas stack – sampling point for dissolver vent gas condensate – fouling probe in place – tem-
perature measurement of smelt bed through primary air port with radiation pyrometer.

90
Reference run High solids run
85

80

75

70

65

60

55

50

45
2007-11-27 2007-11-27 2007-11-28 2007-11-28 2007-11-29 2007-11-29 2007-11-30 2007-11-30 2007-12-01 2007-12-01 2007-12-02
00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00 12:00 00:00

Figure 3, Dry solids content of as-fired liquor during the reference and 3FS firing. The lower curve indicates
the dry solids content of medium heavy liquor prior to the ash mixing and the superconcentrator.

with extremely high dryness. dryness was met. The mill instrumentation reg-
istered 83.5% to almost 84% liquor dry solids
Table 2 below shows the composition of the as- content during the high solids firing. On average,
fired liquor during the three test days. the dry solids content was raised from 82% DS to
It is apparent that the target set regarding liquor 85% DS. The composition remained very stable

43
 during the entire test period. for potassium.
2. The carbonate level is somewhat higher dur-
A large number of boiler ash samples were col- ing the high dryness test firing than during
lected. The results are compiled below, Table 3. the reference run. This is as expected and
points to that the furnace has been hotter dur-
The following conclusions can be made ing the high dryness test.
1. The chloride level is relatively low. There is
very little difference between the normal and The trends for boiler feed water flow, steam flow
the high dry solids liquor. The same is true and as-fired liquor flow coincide during the refer-

Table 2, The composition of as-fired black liquor during the entire period (collected sample during one day,
mass-% of dry sample).

Table 3, The composition of the boiler ash, mass-% average.

44
ence run and the high solids test firing. This indi- from the DCS. The “furnace” temperature is in
cates that the boiler has been running evenly with this case the temperature at the tertiary air port
no unexpected swings. The trend is illustrated for level. What can be seen is that the average tem-
the second day of high solids firing in Figure 4. peratures between the reference run and the high
solids firing test does not exhibit any major differ-
The average temperatures in the upper part of the ence in the chosen locations. Note that the average
furnace and the boiler are shown in Figure 5. The temperatures during the high solids test are equal
data is compiled from the trend data downloaded or even lower than during the reference run.

B o iler feed water an d s team flo w High s olids tes t run Day 2, 2007-12-01
As -fired liq u o r to
t/h B oiler feed w ater flow S team flow As -fried liquor m 3 /h
th e b o iler
280 80

B o iler feed water flo w

260

75

240

S team flo w
220 70

200
As -fired liq u o r to th e b o iler
65

180

160 60

140

55

120

100 50

08:02 10:00
2007-11-30 10:40 14:00 17:20 20:40 00:00 03:20 06:40 2007-12-01

Figure 4, Second day of high dry solids firing. Flow of feedwater, steam and as-fired liquor.
F lue ga s tem pera tures
ºC
1200

1000

800
R e fe re nc e run

H igh s olids D a y 1

600
H igh s olids D a y 2

400

200

0
F urnace S P H left S P H right B oiler bank left B oiler bank E co 2 left E co 2 right B efore precip B efore precip A fter precip S tack
right left right

L oc a tion
Figure 5, Flue gas average temperatures in various parts of the recovery boiler.

45
 Figure 6, Measurement points for the deposit measurement probes.

re fe re ns
10
hö g T S
9

0
M P1 M P2
M ä tp u n k t

Figure 7, Deposit growth rate (mm/h) in sampling points MP1 and MP2 (Reference run dark grey, high
solids firing light grey).

46
Fouling, carry-over and ISP’s An earlier study of deposits in Swedish recovery
Åbo Akademi University measured the fouling boilers was made by Åbo Akademi University,
with a special fouling probe. The probe was in- Figure 9. The study was financed by Värmeforsk,
serted in the superheater area of the Skoghall re- [6]. Deposit measurements in a total of six recov-
covery boiler, Figure 6. ery boilers were made. Two boilers were meas-
ured twice, hence eight results (A-H) in the fig-
The deposit growth rate was higher on the wind ure. These six boilers represent normal Swedish
side of the probe. The following growth rate was technical status. None is even close to the liquor
observed for sampling points MP1 and MP2 just dryness and steam data of the Skoghall recovery
above the bull nose, Figure 7. boiler.

The deposit growth rate for sampling points MP3 Skoghall’s three collected results (reference, high
through MP6 was below 1 mm/h. The principal solids firing day 1 and 2) are placed to the right
composition of the deposits at each location is as in Figure 9.
follows:
The conclusions with respect to Skoghall are as
MP1: carry over and ISP follows:
MP2: almost exclusively ISP
MP3-MP5: ISP and fume 1. In Skoghall’s deposits there is more Na but
MP6: fume the K-level is at the same level as in the de
posits of the other six boilers.
The amount of sodium and sulphate in the depos- 2. The chloride level is approximately the same
its in all sampling points are shown in Figure 8 as the Cl-level in the other six boilers during
below. the reference run and the first day of high sol-
ids firing. The elevated chloride level on the
The deposit measurements can be summarized as second day of high solids firing remains a
follows: mystery. The amount of chloride in the as-fired
liquor does not change, see Table 2. It is possible
1. The increased dry solids content does not af- that this is only a coincidence. The amount
fect the amount of deposits in any major way. of chloride in the boiler ash changes very
2. The deposit growth rate in MP2 through MP6 little between the reference run and the
is correlated to the temperature of the sam- high solids test run. Skoghall’s sulphidity is
pling point. The deposits become sticky or 32-33% which should not affect the chloride
have started sintering at 700-750°C. level.
3. The sulphate formation in the carry over takes 3. The deposits contain more carbonate during
place earlier in the superheater area at high the high solids firing. This points to that the
solids firing. furnace temperature was higher during the
test run. This was expected.

Sodiumm Sulphate
38 60
37
”Normal” DS High DS 55
36

35
50
34

33 45
32

31 40

30
35
29
28 30
1 2 3 4 5 5 6 1 2 3 4 5 5 6 1 2 3 4 5 5 6 1 2 3 4 5 5 6
Mätpunkt Mätpunkt

Figure 8, Sodium and sulphate content of the deposits in all sampling points (mass-%).

47
 100
CO3
SO4
90 Cl
K

80 Na

70

60

50

40

30

20

10

0
A B C D E F G H Ref P rov 1 P rov 2

Figure 9, Deposit measurements in six Swedish recovery boilers and the Skoghall recovery boiler
(mass-%).

No accelerated corrosion tendencies were found reduction degree in Skoghall was 95%, despite
in the superheater area during the annual outage that the carbon level in the smelt was at the lower
and inspection in the fall of 2008. limit of what was found in Mörrum. The reason
for this improved reduction degree is subject to
Other factors affecting the deposit composition is discussion. One hypothesis put forward is that a
the pulp mill sulphidity, pulp wood and processes part of the reduction reaction can take place in
used (soap firing, mixing residual acid into the fir- suspension, i.e. after the liquor droplet has dried
ing liquor, chloride enrichment). The overall con- and devolatilized.
clusion is that Skoghall’s deposits are not excep-
tional in any way, apart from the higher carbonate The lower carbon content in Skoghall’s smelt bed
content, during normal and high solids firing. became evident during the sampling. The smelt
burned with an open flame in when it was taken
out of the smelt probe in Mörrum. It is thereby
Smelt composition likely that the actual carbon content may have
Half of the smelt samples from the smelt bed been higher than what is shown in Figure 10.
were sent to the Technical Research Institute of When the smelt sampling probe was opened in
Sweden for analysis. The other half was sent to Skoghall, the smelt glowed but did not burn.
Åbo Akademi University for cyanate and carbon
analyses. Another observation made in Skoghall was the
impact of increased boiler size. Twenty years
The results are shown in Figure 10 below. ago, most boilers had sloping floors and a smaller
cross-sectional area than today. Smelt beds could
The levels of sodium and carbonate show some be quite high. This means that in Mörrum smelt
variation. On the contrary, the levels of carbon could be taken to a depth of some 50 cm into the
and potassium are relatively stable. The chloride bed. During the bed sampling in Skoghall, only
level is very low. smelt from the active surface could be taken (0-20
cm). This is due to the fact that Skoghall’s recov-
What makes Figure 10 extra interesting are the ery boiler has a cross-section of some 10 x 10 m
two extra bars to the right in the figure. These and the decanting floor gives a low smelt bed.
represent the min and max carbon levels in the
smelt samples taken from Södra Cell Mörrum’s
state-of-the-art recovery boiler in 1990. With the NOX emissions
firing techniques and air systems of that day, a During the reference run and the high solids test,
reduction degree of 90% could be reached. The a number of nitrogen compounds at several loca-

48
50
2007-11-27 10:15
Min andd max
max carb
carb on
on levels
levels in
in sme
smeltlt
45 2007-11-27 15:05 (near beddsur
surface)
face)
face) fro
fro
from
m
mmMörrum
Mörrum
Mörrum
Mörrum
2007-11-27 17:05
40 2007-11-29 10:35

2007-11-29 13:35
35
2007-11-29 16:35

2007-11-29 18:35
30
2007-11-30 11:00

Mörrum min C
25
Mörrum max C

20

15

10

Na K Cl C O3 C

Figure 10, Analysis of smelt from the smelt bed (mass-%).

tions were monitored. The focus was on the flue Combustion of dissolving tank vent
gas and the dissolving tank vent gases fired in the
tertiary air level. The levels measured in the flue
gases in the tertiary air level
Skoghall fires the dissolving tank vent gases in the
gas are shown in Table 4.
tertiary air level. The amount of gaseous ammonia
in these vent gases was measured to be 273-331
The difference between the NO- and the NO2-lev-
mg/m3n dry gas during the reference run. During
els are small for the reference run and the high
the high solids test firing, the ammonia increased
solids test. The levels of N2O and NH3 are at the
to 299-393 mg/m3n dry gas. This corresponds to
detection limit.
an increase of some 20-60 mg/m3n dry gas as a
result of the higher liquor dry solids content.
These results differ from the earlier 1987 results. A
There are few comparable measurements. Tarpey
similar high solids test firing was then undertaken
el al [7] reported a level of 250 mg/m3n dry gas in
in the then state-of-the-art recovery boiler nr 5 in
a Canadian mill in 1996.
Korsnäs. The liquor dryness was increased from
65% DS to about 76-77% DS for one day. During
The dissolving tank vent gases are fed into the
this test firing, the NO -level increased with some
tertiary air level from one side only. The aver-
60% in the flue gas. In absolute numbers from 95
age temperature at this level was already shown
ppm dry gas to 155 ppm dry gas.
in Figure 5. The temperature was also measured
by hand held radiation pyrometers a few meters
Nothing like this could be detected in Skoghall.
below the intake of the vent gases, through the
It must be noted though, that a liquor dry solids
liquor gun openings. In this location, the average
increase from 65% DS to 77% DS is much more
temperature was 1079°C during the reference run
dramatic than an increase from 82% DS to 85%
and 1091°C during the high solids test firing. It is
DS. It seems unlikely that this NOX-level increase
generally known that ammonia reduces nitrogen
in Korsnäs would be due to thermal NOX forma-
oxides to nitrogen in the temperature window of
tion, since the furnace temperatures measured
850-1000°C. If the NH3 is added at higher tem-
then were well below 1400°C.
peratures, it forms NO instead [8].

Table 4, Nitrogen compounds in the flue gas from Skoghall’s recovery boiler during the entire test firing
period (average amount in ppm dry gas).

49
 This phenomenon was observed during the high Figure 11. below shows the NO- and O2-level
solids test period. The dissolver off-gases were during the reference run. The NO-level fluctuates,
bypassed to the stack four times and were then while the O2–level is fairly constant at 3%.
substituted with combustion air. During these
periods (up to two hours at a time) a decrease in Note the effect on the NO-level when the dissolv-
the NOX-level in the boiler flue gas could be ob- er vent gases are bypassed, Figure 12 and Figure
served. The NOX-levels were also fluctuating less 13. Interestingly, the amplitude in the NO-level
during these periods. fluctuations is also less during these periods.

O2 R eference run, 2007-11-28 NO

% NO- and O2-level in the s tack flue gas ppm

4.5 95

4.0

90

3.5
O2 %

3.0 85

2.5

80

2.0

1.5 75

NO ppm
1.0

70

0.5

0.0 65

10:15 9:38
2007-11-27 13:35 16:55 20:15 23:35 02:55 06:15 2007-11-28

Figure 11, NO- and O2-content of the flue gas during the reference run. Dissolver off-gases are mixed and
burned in the tertiary air level. Note the amplitude of the fluctuations in the NO-content.

O2 High s olids firing Day 1, 2007-11-30 NO

% NO- and O2-level in the s tack flue gas ppm

5.0 95

4.5

90
4.0
O2 %

3.5
85

3.0

2.5 80

2.0

75
1.5

NO ppm
1.0
70
~15-16 ~19-21
0.5 V en t g as es V en t g as es
b yp as s ed b yp as s ed

0.0 65

07:01 10:00
10:40 14:00 17:20 20:40 00:00 03:20 06:40
2007-11-29 2007-11-30

Figure 12, NO- and O2-content of the flue gas during the first day of the high dry solids run. Dissolver off-
gases are vented to the atmosphere between 3-4 pm and 7-9 pm; otherwise these are mixed and burnt in the
tertiary air level.

50
 O2 High s olids firing Day 2, 2007-12-01
NO
% NO- and O2-level in the s tack flue gas ppm

4.5 100

O2 %
4.0

95

3.5

3.0 90

2.5

85

2.0

1.5 80

1.0 NO ppm
75

0.5 ~11-13 ~17-19

V en t g as es V en t g as es
b yp as s ed b yp as ed
0.0 70

08:02 10:00
10:40 14:00 17:20 20:40 00:00 03:20 06:40
2007-11-30 2007-12-01

Figure 13, NO- and O2-content of the flue gas during the second day of the high dry solids run. Dissolver
off-gases are vented to the atmosphere between 11 am - 1 pm and 5-7 pm, otherwise these are mixed and
burnt in the tertiary air level.

The conclusion is that the NH3 –content in the dis- ids firing. The results are shown in Figure 14. The
solver vent gases is increasing the NOX-emission amount of nitrogen as cyanate (OCN-) in the smelt
of the boiler (!). Although the dissolver vent gases samples from the spouts 1, 3, 4 and 6 are shown.
were not bypassed during the reference run there
is no doubt that this phenomenon would also be The cyanate content in the middle spouts (3, 4)
found during normal black liquor solids firing. was on average 0.032% during the reference run.
The content in the outer spouts (1, 6) was on aver-
Smelt sampling and smelt nitrogen age 0.014%. As for the first day of the high solids
firing, the amount in the middle spouts increased
content to 0.036% on average. The outer spouts contained
Åbo Akademi took smelt samples from four smelt on average 0.017% cyanate.
spouts during the reference run and the high sol-

2 7 .1 1 2 9 .1 1
0.050% at 15
at 11 at 11
0.045%
at 10 at 15
0.040%

0.035% at 15
0.030%
OCN-N
0.025%
OCN-N
0.020%

0.015%

0.010%

0.005%

0.000%
1 3 4 6 1 3 4 6 1 3 4 6 1 3 4 6 1 3 4 6 1 3 4 6
Ränna

Figure 14, Measured amounts of cyanate in the smelt spouts 1, 3, 4, and 6 during the reference run and the
first day of high dry solids firing. White boxes indicate the time when the samples were taken.

51
 ÅF took temperature measurements of the smelt in It is obvious that the balance does not add up.
the spouts during the entire test week. The average According to [9], approximately one third of the
smelt temperature was 905°C during the reference nitrogen in the black liquor should be found in
run and 884°C during the high solids firing. These the smelt going out from the boiler. The balance
measurements were difficult to perform and there around the boiler only adds up if the N2-amount
may be several sources of error in the individual in the flue gas is calculated to 35-38 kg/h. The
measurements. However, the conclusion is that distribution of gaseous nitrogen compounds is not
the smelt temperature was quite normal. according to theory with this calculated amount
of N2. The amount and share of NOX is according
The following conclusions can be drawn regard- to theory, but the N2 flow is twice the expected
ing the smelt cyanate content: amount.
1. The smelt cyanate content is lower in the out-
er spouts (1, 6) then in the middle ones (3, 4). The amounts of cyanate in the smelt are low ac-
2. The cyanate level is about 10% higher during cording to Åbo Akademi University. The carbon
the high solids firing. However, there was no content is also low. Since char = carbon plays a
clear difference between the two measur- role in the cyanate transport to the smelt bed, the
ement series. question can be raised if the low cyanate and low
carbon content in the smelt are connected some-
Åbo Akademi University also received samples of how.
the bed smelt taken from inside the boiler. These
were analysed for cyanate (and carbon), Table 5. Also note the total nitrogen content of the dis-
solver vent gases. These contribute with approxi-
Note that the amount of cyanate in the bed smelt mately 10% of the amount of nitrogen influx with
exhibits a wide variation and that the absolute the black liquor to the total nitrogen influx to the
level is close to the amounts found in the outer boiler.
smelt spouts (1, 6).
Stora Enso Skoghall made a nitrogen balance over
A nitrogen balance for the recovery boiler and the entire Skoghall mill in June 2007. The follow-
dissolver vent gas system has been set up based ing measurements were made, Table 6.
on the measured amounts of nitrogen compounds,
known process flows to the recovery boiler and In this study, the nitrogen content in the medium
calculated smelt flows in the spouts. The balance heavy liquor was measured. Thereby a straight
for the reference run is shown in Figure 15. comparison with the amounts found in Figure 15
The balance for the high solids firing test is shown and Figure 16 is not applicable. In this study the
in Figure 16 (average for both days). measured NOX emission is lower than for the ref-

Table 5, Amount of cyanate in samples taken from the smelt bed in Skoghall (mass-% of total smelt).

52
Figure 15, Nitrogen balance for the recovery boiler and dissolver vent gas system during the reference
run.

Figure 16, Nitrogen balance for the recovery boiler and dissolver vent gas system during the high dry solids
run.

53
 Table 6, Nitrogen balance for the recovery boiler and recausticizing in Skoghall.

erence run and the high solids firing. The meas- boiler feed water steam preheater, intermediate
ured amounts of N for the green liquor, white liq- steam superheater and a flue gas cooler. No energy
uor and the scrubber discharge are at the detection balance for an entire pulp mill was set up, only the
limits; thereby the results can be uncertain. impact of the recovery boiler changes to the bark
boiler and the steam turbine were considered. All
The scrubber discharge is led to the green liquor mills have different energy solutions. This was a
together with the dissolving vent gas scrubber pure technical potential study, no economical con-
condensate. Even if all this nitrogen would end siderations were included. The simulation results
up in the vent gases, it is unlikely that this extra are shown in Figure 17.
nitrogen could explain the unbalance between the
nitrogen coming in to the dissolver with the smelt Abbreviations and x-axis legend explanation:
and the nitrogen leaving the vent gas from the dis- “Basfall”: base case
solver vent gas system. “MÖH”: intermediate steam superheating
Example:
Energy efficiency “80% 190C 4 12 28 bar MÖH” = 80% dry solids;
A state-of-the-art recovery boiler similar to air preheating to 190°C; 4, 12 and 28 bar steam
Skoghall’s recovery boiler was simulated with used for BFW preheating; intermediate superheat-
ing.

Figure 17, Simulation results. The base case is Skoghall’s recovery boiler at design load. Incremental in-
crease compared to base case (%).

54
The following conclusions can be drawn: ed to a temperature zone below 1000°C. By
doing this, the NH3-content will decrease the
• There is a technical potential to increase the NOX emission.
power output of a recovery boiler – bark boiler • The amount of nitrogen in the smelt is lower
combination by use of boiler feed water steam than expected.
preheating and intermediate steam superhea- • The amount of carbon in the smelt bed is lower
ting. than previous measurements, but the reduction
• The potential is considered to be good already degree is still high.
with boiler feed water steam preheating by • The deposits in the superheater area exhibit no
using normal process steam up to steam used extraordinary composition and there is little
for soot blowing. risk of accelerated corrosion tendencies during
• Increasing the liquor dry solids contributes present firing conditions.
less to increased power output than boiler feed • The energy efficiency calculations show that,
water steam preheating and intermediate steam under certain conditions, the power output can
superheating. be increased significantly with boiler feed wa-
• The possibility to fully estimate the technical ter steam preheating, flue gas cooling and in-
potential is limited by the fact that when boiler termediate steam superheating.
feed water steam preheating and intermediate
steam superheating are used, the other heating Future work
surfaces should be adapted accordingly. This study has come up with some interesting re-
• When using boiler feed water steam preheating sults that merit further investigation:
and intermediate steam superheating, the r
covery boiler layout should be modified as fo- • Continued measurements of the ammonia
lows: content in recovery boiler dissolver vent gases
o Larger superheaters in Sweden.
o Smaller boiler bank
o A flue gas cooler must be used • In depth studies of the NOX emissions in those
o The recovery boiler water circulation system recovery boilers that fire dissolver vent gases
must be enlarged in the boiler (Billerud Gruvön, Södra Cell
• The investment cost is assumed to increase for Värö, Stora Enso Skoghall, SCA Östrand, SCA
the recovery boiler Obbola).
• The optimal solution for each case can only
be found by making an economical optimiza- • Further sampling of the smelt bed in recovery
tion based on the local conditions. boilers firing high dryness black liquor to in-
crease the knowledge of the bed chemistry and
Conclusions dynamics.
Based on the high solids test run, it is striking that
the difference between firing 82% DS liquor and • In depth studies on the reduction reaction in
85% DS liquor is not greater. The mill personnel the lower furnace and possible reaction steps
experienced the high solids test firing as a smooth taking place in suspension.
ride with the boiler. The entire project must be
considered a success. Acknowledgements
This study was partially funded by Åforsk, the ÅF
The major conclusions are: Research Foundation, and Värmeforsk, the Ther-
mal Engineering Research Institute.
• Skoghall’s recovery boiler exhibits good run
ablility and low emissions at normal and high Without the support and effort of the members
dry solids firing. of the SNRBC – and especially the personnel of
• There are indicators that the furnace and the Stora Enso Skoghall mill – this study would not
smelt bed were hotter during the high solids have been possible.
test firing than during the reference run. Their support is gratefully acknowledged.
• The ammonia content of the dissolver vent
gases is substantial and affects the total emis-
sion of NOX from the boiler, in this case by
increasing it. If dissolver vent gases are to be
fired in a recovery boiler, these should be add-

55
 References
1. Björklund, H, Warnqvist, B, Pettersson, B; In-
side a kraft recovery boiler: Combustion of (high-
sulphidity) black liquor at high dry solids. TAPPI
1989 International Chemical Recovery Confer-
ence, pp 177-182, ISBN-0-919893-68-6, Mon-
treal Canada 1989
2. Björklund, H, Warnqvist, B, Pettersson, B;
Inside a kraft recovery boiler: Combustion of
high-sulphidity black liquor at high dry solids.
Pulp&Paper Canada, Vol 92, T206, Toronto
Canada 1991
3. Warnqvist, B; Char bed and smelt properties
– mill studies, TAPPI 1992 International Chemi-
cal Recovery Conference, pp 551-556, ISBN-0-
89852-849-6, Seattle USA 1994
4. Warnqvist, B; Char bed and smelt properties –
mill studies. Pulp&Paper Canada, Vol 95, T313,
Toronto Canada 1994
5. Bruno, F; Thermo chemical aspects on chloride
corrosion in kraft recovery boilers. NACE Corro-
sion 2001 Conference, Houston Texas USA 2001
6. Forssén, M, Backman, R, Wallén, J, Hupa, M;
Flygaskans sammansättning och nedsmutsande
tendens i sodapannan. (Flue gas dust composition
and fouling tendency in recovery boilers), Värme-
forsk S6-612/S9-804, ISSN 0282-3772, Stock-
holm, Sweden 2000
7. Tarpey, T, Tran, H, Mao, X; Emissions of Gase-
ous Ammonia and Particulate Containing Ammo-
nium Compounds from a Smelt Dissolving Tank.
Journal of Pulp and Paper Science, Vol 22, No 4,
J145-J159, 1996
8. Lundberg, M; SNCR för sodapannor. (SNCR
for recovery boilers). Sodahus-konferensen 2008,
pp 143-152, ÅF, Stockholm, Sweden
9. Kymäläinen, M, Forssén, M, Hupa, M; The
fate of nitrogen in the chemical recovery proc-
ess in a kraft pulp mill. TAPPI 1998 International
Chemical Recovery Conference, pp 19-32, ISBN
0-89852-716-3, Tampa, Florida, USA, 1998

56
BLACK LIQUOR PARTICLE AND DROPLET REAC-
TIONS: EXPERIMENTAL AND MODEL RESULTS
WARREN ROBERTS, HONG LU, ELVIN IP
AND LARRY L. BAXTER
BRIGHAM YOUNG UNIVERSITY
PROVO, UT 84602

Introduction

During papermaking, a mixture of sodium hy- erate at a locally reducing stoichiometry in some
droxide (NaOH), sodium carbonate (Na2CO3) and regions, reducing oxidized inorganic components
sodium sulfide (Na2S) at elevated temperatures of black liquor to more active solubilizing agents
dissolves lignin in wood chips, releasing insolu- that eventually return to the digester, and at lo-
ble cellulose fibers in a process called digestion. cally oxidizing stoichiometries in other regions,
The lignin acts as a binder in the wood to hold oxidizing organics to produce heat and light-gas
the fibers together and constitutes about 30% of products. The primary role of the recovery boiler
the wood mass [1]. Lignin has several detrimen- involves separating the salts from the organics,
tal impacts on paper; so digestion and washing/ recovering the inorganics for recycle and heat
separation steps remove lignin from cellulose fib- from the organics – it recovers chemicals and
ers in papers with quality above newsprint. The heat. Black liquor is the second largest renewable
digester effluent stream containing mostly dis- energy source in North America and Europe, fol-
solved lignin-based organic materials and spent lowing hydropower and – depending on how one
alkali salts is called black liquor and is referred to classifies fuels – about the fifth largest fuel of all
as weak black liquor in the state that it leaves the types, distantly following coal, petroleum, natural
digester. Evaporators concentrate the weak black gas, and nuclear fuel [4].
liquor to strong liquor with 65-85 % solids con-
tent to promote its combustion [2]. The so-called Black liquor impinges on a splash plate as it enters
solids refer to the non-water portion of the liquor the boiler, forming droplets in the range of 0.5 to 5
– they form solids after complete drying of the mm in diameter. As these droplets pass through the
liquor, but essentially all of these so-called solids boiler, they undergo different stages of combustion.
are either dissolved or in suspension in the liquor The stages include drying, devolatilization, char
and in these senses are more liquid than solid in burning, and smelt (molten inorganic matter) for-
practice. In the concentrated or strong black liq- mation, with many of these processes overlapping
uor, the solids mainly comprise sodium salts and in time and space. Normally, some devolatilization
lignin-based organics in solution with some pre- and most char burning and smelt formation occur
cipitates and impurities as suspended solids and after the droplet has fallen to the bottom of the boil-
small amounts of potassium, chlorine, silica, and er onto the char bed. A fraction of the droplets be-
other impurities. These last impurities come from come entrained in the combustion gases of the re-
inorganic components and contaminants in the covery boiler and combust in suspension [2]. These
wood and collectively represent the non-process droplets generally are initially 2 mm or smaller [5].
elements. Recent trends increase the solids con- These particles and some fragments and vapors
tent of strong black liquor to 80-85% to decrease from the bed flow with the gases through the upper
the possibility of water-smelt explosions and in- sections of the boiler, where they lead to deposit
crease the throughput of the recovery boiler [3]. formation/flow constrictions or plugging and other
operational problems. The particles react dynami-
A profitable and environmentally acceptable pa- cally during flight, with large changes in tempera-
per mill recycles the valuable and otherwise envi- ture, composition, structure, and physical proper-
ronmentally aggressive inorganic salts. The lower ties. The focus of this document is on the dynamic
section of the recovery boiler and the char bed op- behavior of these particles.

57
 Distinguishing Biomass behavior. A few comments about the remaining
fuels complete this section with an emphasis on
Combustion Characteristics traditional biomass fuels, whereas the bulk of this
Coal represents the dominant low-grade, ash-
document focuses on the specific biomass fuel of
forming fuel by any measure – energy generation,
black liquor.
mass consumption, CO2 production, and econom-
ic influence. Consequently, a great many of the
Results of classical analyses of biomass and coal
experimental techniques and theoretical approach
fuels reflect many, but not all, of the important
used for many low-grade fuels, including biomass
differences in these fuels. Table 1 summarizes
broadly and black liquor are usefully considered
data from coal and biomass databases including
in the comparison to coal. This section discuss
several thousands of samples of fuels in various
biomass broadly, with most of the discussion per-
categories. As is seen, biomass differs from coal
taining to traditional biomass (forest products,
primarily in its higher oxygen, lower carbon, and
agricultural products, and residues) and to black
lower sulfur content. Herbaceous biomass (grass-
liquor. While black liquor is a form of biomass
es and straws) also contain much higher chlorine
and shares many of the characteristics discussed
as a class than do either coals or woods. Some of
here with other biomass fuels, it is unique among
these differences appear in Figure 1, where coal,
even biomass fuels in important ways. This sec-
biomass, black liquor, and a few other fuel compo-
tion is followed by another section that discusses
sitions appear on a molar H:C to O:C ratio. Other
these unique combustion characteristics.
important characteristics of biomass relative to
coal include more widely ranging (0.1%-25%)
Table 1 summarizes the chemical compositions
ash contents, lower heating values (by about half
typical of several important classes of fuels. The
on an as-delivered basis), much lower bulk energy
relationships among these fuels appear most inter-
densities (by more than an order of magnitude),
esting when plotted as atomic ratios of hydrogen
generally higher alkali content for herbaceous bi-
and oxygen to carbon (Figure 1). All individual
omass dominated by potassium, and substantially
data points in this figure represent sample results
more fibrous and less friable character. Each of
whereas the round, red, labeled data represent av-
these characteristics lends specific characteristics
erages for the samples in each classification.
to the fuels, only a few of which are discussed
Black liquor exhibits both similarities and differ- here.
ences from other fuels. It is hydrogen poor com-
pared oils and other high-grade fuels, but not as Particle Size
hydrogen poor coal. It is more oxygen rich than Traditional biomass particle sizes of commercial
coal but less oxygen rich than biomass. It is lignin interest typically range from 2-10 mm. This ex-
rich, as indicated by its position in comparison ceeds, for example, pulverized coal particle sizes
to lignin and cellulose on the figure and is obvi- and volumes by orders of magnitude. Biomass
ous from the process from which it originates. particles also exhibit highly aspherical character-
Not evident from this figure is black liquor’s ex- istics and in this section size is characterized as
tremely high inorganic content, high initial mois- the sphere-equivalent diameter, that is, the diam-
ture content, and generally fluid character. Also, eter of a sphere that has the same volume/mass of
owing mainly to high moisture, inorganic, and fuel as the biomass particle. There are several im-
oxygen contents, it has a relatively low heating plications of this large size on biomass combus-
value compared to most other fuels. While each of tion, especially when compared to traditional coal
these fuels exhibit its own interesting combustion combustion concepts. The focus here is on issues
characteristics, black liquor in many ways repre- that impact practical system operation.
sents the most extreme fuel in terms of particle

Table 1, Ultimate Analyses of Coal and Biomass Samples (compare with Table 2 for Black Liquor).
Anthracite Bituminous Subbit. Lignite Grass Straw Wood Chips Waste Wood
C 90.22 78.35 56.01 42.59 45.34 48.31 51.59 49.62
H 2.85 5.75 6.61 7.40 5.82 5.85 6.14 6.34
N 0.93 1.56 1.10 0.73 2.04 0.78 0.61 1.01
O 5.03 11.89 35.25 48.02 45.95 44.18 41.57 42.89
S 0.96 2.43 0.84 1.15 0.24 0.18 0.07 0.07
Cl 0.03 0.08 0.01 0.01 0.62 0.70 0.02 0.06

58
 2.0 Black Liquor

1.5 Cellulose
H:C Molar Ratio

Grass

Average Biomass
Wood

1.0 Peat anthracite

Lignin semianthracite
bituminous coal
subbituminous coal
lignite
Lignite
0.5 Subbituminous Coal
peat
Bituminous Coal biomass
Semianthracite black liquor
Anthracite
average values
0.0
0.0 0.2 0.4 0.6 0.8 1.0
O:C Molar Ratio
Figure 1, Comparison of biomass and coal on a molar H:C vs. O:C ratio

Devolatilization

Char burning (0.26 s later)

Figure 2, Visible-light images (left), temperature maps (center) and emissivity maps (right) of an approxi-
mately 2 mm diameter burning biomass (black liquor) particle/droplet.

59
 Figure 2 illustrates visible-light, temperature-map, verized coal combustion, strongly influence the
and emissivity-map images of an approximately 2 rates of particle heating and reaction.
mm diameter burning black liquor particle/droplet
[6, 7] during both devolatilization and char burn- Particle Shape
ing. The devolatilization image illustrates both Quite independently of particle size, the unusu-
particle (roughly spherical shape at center and ab- al shapes of biomass particles also lead to more
normal shape in lower left) and soot (bright glow complex particle behaviors than commonly occur
above and below the spherical shape) combustion. during coal combustion. Pulverized coal particles
The temperature map indicates the large range of approximate spheres, with aspect ratios rarely
temperatures existing on the particle surface, with exceeding 2. Biomass particle aspect ratios com-
surface temperatures varying by several hundred monly exceed 6 and shapes more commonly re-
degrees with a high at the most exposed regions of semble cylinders or plates/flakes than spheres. A
the particle surface (lower left) and a low at the re- sphere exhibits the smallest surface-area-to-vol-
gion most sheltered from the upward moving con- ume ratio of any shape and therefore represents
vective flow in this experiment (top of particle). an extreme case in processes where such a ratio
The soot, on the other hand, exhibits much higher is important. Large particles commonly burn at or
temperatures than those at any portion of the par- near diffusion-limited rates, which makes this ra-
ticle surface, in part because the overall particle tio of paramount importance in predicting overall
temperature is still rapidly rising in these plots conversion times. Spheres, therefore, generally
and in part because the soot particles are burning represent poor approximations for many biomass
under conditions near the peak adiabatic flame particles in combustion models.
temperature for this system. The soot clouds,
however, are relatively thin and hence have very Figure 3 illustrates the impact of particle shape
low emissivity (right diagram). on biomass combustion characteristics generally.
This figure compares the measured and predicted
During the early stages of oxidation illustrated mass loss profiles of relatively small (300 μm
here, the soot cloud is gone and the particle tem- wood) biomass particles hand-sorted into dif-
peratures are on average higher than during devol- ferent shapes. As indicated, the near-spherical
atilization, although there remain significant tem- biomass particles combust more slowly than the
perature differences on the particle surface. This higher aspect ratio particles. Since both heat and
image of the same particle is particularly instruc- mass transfer are dictated mainly by surface area
tive, however, because the particle has developed for these large particles, the lower surface-area-
a hole in the surface into which oxygen cannot to-mass ratios of spheres significantly decrease
effectively penetrate. The particle surface tem- the overall combustion rate of these particles rela-
perature at this early stage of oxidation exceeds tive to less symmetric particles with the same total
the interior temperature at least 300 kelvins, illus- mass. Detailed discussions of the model and data
trating the very large temperature gradients in the on which these figures are based appear elsewhere
particle. The emissivity map illustrates how the [8, 9].
particle emissivity varies from near unity in this
particle hole (which acts somewhat like a black Experimental constraints limited these measure-
body cavity radiator) and in regions of high char ments illustrated in Figure 3 to relatively small
concentration to as low as 0.2 in regions where the particles. At more realistic biomass particle sizes
very high sodium-salt concentrations in the char the impact on particle conversion time becomes
particle are beginning to be exposed. increasingly dramatic. Figure 4 illustrates the
predicted ratio in overall conversion time for par-
These large radial and surface variations in parti- ticles of various shapes as a function of sphere-
cle properties do not exist during pulverized coal equivalent size, with all times normalized to those
combustion. They lead to significant overlaps in of flakes. In the size range of greatest interest to
particle drying, devolatilization, and oxidation many commercial biomass conversion facilities,
processes and considerably complicate compu- the conversion times of the aspherical biomass
ter modeling of conversion processes. Addition- particles change by factors of 2-3 relative to that
ally, larger particles have large boundary layers in of the spheres with the same mass. As indicated in
which particle off gases, such as CO, react with the figure, this impact becomes increasingly less
bulk gas components, such as O2, providing sig- significant as particle size decreases, becoming
nificant thermal feedback to the particles. These relatively unimportant for pulverized coal particle
near-particle flames, which do not occur in pul- sizes (100 μm and less). The single set of data indi-

60
 1.0
1.0

0.8 0.8
Mass Loss, daf

0.6 0.6

0.4 flake-like exp. 0.4

flake-like model
0.2 cylinder-like exp.
cylinder-like model 0.2
near-spherical exp.
0.0 near-spherical model
0.0
0.0 0.1 0.2 0.3 0.4 0.5
Residence Time, s
Figure 3, Mass loss histories of sawdust particles with different shapes.

2.2
Exp. Model
Conversion Time Ratio

2.0 Near-sphere/Plate
Cylinder/Plate

1.8

1.6

1.4

1.2

1.0
0 2 4 6 8 10 12 14 16 18
Equivalent Diameter dp,eq., mm
Figure 4, Conversion time vs. particle sphere-equivalent diameter for particles with aspect ratios of 5 (par-
ticles in inert – N2 – environments with 6 percent initial moisture with gas and radiative temperatures of
1050 and 1273 K, respectively). At higher aspect ratios, the ratio of conversion time increases (rises to ap-
proximately 3 for aspect ratios of 8 – which are common in biomass fuels).

61
 1400

1200
Temperature, K

1000

800
center surface
1
600 2
exp.1
3 exp.2
400 exp.3
4 exp.4
model
0 10 20 30 40 50 60 70 80
Residence Time, s

1.0
Mass Loss, 1-m/m
0

0.8

0.6

0.4
model
run1
0.2 run2
run3
0.0 run4
0 10 20 30 40 50 60 70 80
Residence Time, s

Figure 5, Comparison of data and model predictions of an initial 9.5 mm droplet in a 1050 K inert gas
(N2)/1273 K radiative environment furnace in nitrogen (aspect ratio = 5, initial moisture 6 %, wood parti-
cle). The early temperature data in experiments 1 and 2 are biased because of conducting along thermocou-
ple leads.

62
cated in this figure are based on the measurements dicted surface and center temperatures as would
illustrated in Figure 3. Many additional data have be expected. The measured temperature overshoot
since been collected that confirm the predictions centered around 5 s residence time may arise from
of Figure 4 over a broader size range. some residual oxygen in the nominally nitrogen-
purged furnace in which these experiments oc-
Moisture and Volatiles Contents curred, or it may relate to exothermic devolatili-
The shape, size, and relatively high moisture and zation reactions that occur during the latter stages
volatiles contents of biomass particles and espe- of devolatilization.
cially black liquor droplets influence conversion
histories in practical ways. Figure 5 illustrates In addition to the water-vaporization-related tem-
predicted centerline and surface particle tem- perature plateau seen at the particle center, the
perature histories compared with measured parti- rapid volatile loss also impacts temperature rise.
cle temperatures. The interior data come from a Under the relatively mild conditions of these
thermocouple imbedded in an approximately 9.5 experiments, this impact is subtle and is largely
mm initial diameter biomass particle. Model com- responsible for the departure of the observed tem-
parisons for similar data from biomass and black perature rises from smooth exponential-like pro-
liquor particles/droplets ranging from < 1 mm to files that would exit in the absence of such blowing
> 10 mm and a detailed description of the model effects. However, under more severe conditions,
appear in the literature [8-17]. Some particles and the volatile yields and especially the combustion
especially black liquor droplets distort wildly dur- of such volatiles in the particle boundary layers
ing combustion, in which cases there no practical have pronounced impacts on particle temperature
way of knowing where the thermocouple bead is histories.
relative to the droplet center during the experi-
ments, but video records confirm that it remains These process combine to produce particle off gas
inside the droplet somewhere, so its results are compositions and particle reaction timelines that
considered interior temperatures. The data shown differ substantially from those that would be pre-
here involve less wildly contorting particles and dicted if biomass particles were treated as large,
thermocouple readings that more reliably indicate low-density, coal particles using traditional mod-
a center temperature as opposed to an interior eling techniques.
temperature.
There are additional considerations associated
The figure indicates that the center of the droplet with transformations of particle inorganic ma-
remains wet and vaporizing well after the surface terial into fly ash and its subsequent impact on
reaches devolatilization temperatures, as indi- deposition, corrosion, SCR performance and fly
cated by the low and sub-boiling predicted center ash utilization. These considerations will appear
temperature coinciding in time with surface tem- in other documents.
peratures over 700 K. Note that biomass and black
liquor particles begin devolatilization and swell- Distinguishing Black Liquor
ing at lower temperatures than do coal particles,
mostly due to the overall lower molecular weight Combustion Characteristics
of their condensed-phased organic constituents. Black liquor is unique among major low-grade fu-
Heat conduction along the thermocouple leads els in amount and composition of inorganic spe-
bias the early temperature data in Experiments 1 cies that it contains. As can be seen from Table 2,
and 2 to high values – a problem largely avoided black liquor consists of about 20% sodium. About
in Experiments3 and 4 by running the leads along half of this sodium is in the form of inorganic salts,
the axial rather than the radial direction of these and the other half is in the form of acidic func-
cylindrical particles. Larger droplets/particles and tional groups such as carboxylates and phenolates
higher heating rates (these data come from a 1050 [18]. During combustion, black liquor generates
K gas/1273 radiative environments in an inert N2 sodium-containing vapors that, in turn, form
environment) generate larger temperature gra- submicron-sized inorganic particles or aerosols.
dients still. Predicted composition plots confirm These particles, known as fume, contain sodium
this inference from the temperature, but we have carbonate (Na2CO3), and sodium sulfate (Na2SO4)
yet to devise means of generating anything other with smaller amounts of sodium chloride (NaCl)
than overall mass loss with which to confirm such and other sodium- and potassium-containing spe-
predictions. The interior temperature measured by cies. Fume formation depends strongly on furnace
the thermocouple generally lies between the pre- temperature and furnace oxygen content [2]. Field

63
 experiments indicate that the majority of fume Highlights of Previous
forms from entrained black liquor particles while
some forms from the sub-stoichiometric char bed
Investigations
This section summarizes results from investiga-
or smelt in the bottom of the boiler [19].
tions related to this topic. It also includes a critical
review of the work done by this laboratory, nota-
Table 2, Typical Black Liquor Composition [9].
bly by Le-ong Teng “Elvin” Ip, who previously
worked on this project and whose techniques are
largely adapted in this work. Ip’s work appears
in his dissertation, Comprehensive Black Liquor
Droplet Combustion Investigations [6] and else-
where [6,7,10,13,14,16]. His investigations in-
cluded experimental work on black liquor droplets
to obtain data on droplet size, shape, internal and
surface temperature, and mass loss during com-
bustion and pyrolysis. This review also highlights
work done outside this research group. A nu-
merical model of black liquor droplet combustion
quantitatively predicts similar data, based largely
Other unique characteristics of black liquor fuels
on related work conducted by Hong Lu focusing
include: (1) while highly viscous at room tem-
on biomass particles. During his investigation, Ip
perature, black liquor is a relatively easily han-
also obtained qualitative and quantitative data of
dled fluid at process temperatures (near 100 °C),
intermediate sized particle (ISP) formation. Aside
unlike most highly ash-laden fuels such as tradi-
from the investigation of ISP formation, the cur-
tional biomass and coal; (2) black liquor swells
rent document improves considerably on the data
more than any other common fuel during combus-
model pioneered by Ip and Lu.
tion, commonly experiencing density decreases of
one to two orders of magnitude; (3) black liquor
in recovery boiler applications forms (by design) Droplet Combustion Experiment
droplets of nominally 2 mm, well above pulver- Many different techniques characterize black liq-
ized coal sizes (typically 45 micron average, 100 uor combustion and pyrolysis. The technique of
micron top size) and similar to entrained-flow most importance to this work is single particle
biomass particle sizes (typically 1-6 mm); (4) combustion and pyrolysis. Hupa et al. [20] pio-
the inorganic material in black liquor participates neered this work. They studied black liquor parti-
dynamically in carbon conversion and other im- cles by suspending them on a thin platinum wire
portant aspects of liquor combustion to a much or on a thermocouple and placing them in a muffle
greater extent than the inorganic material in either furnace. By placing the black liquor droplet on a
coal or traditional biomass; and (5) black liquor thermocouple, they measured the internal drop-
composition and organic structure differ from tra- let temperature during the burn. They also used
ditional biomass and coal (discussed later in more a video camera to record the burning droplet and
detail), for example with overall oxygen content measure particle size and burning times. Frederick
that lies between that of coals and traditional bio- et al. used a two-color pyrometer to measure the
mass as a group and a hydrogen content greater surface temperature of the particle during com-
than coal and comparable to many traditional bio- bustion and pyrolysis [21]. Ip used a similar tech-
mass fuels. This document details many of these nique except that his furnace gases were not stag-
aspects of black liquor combustion. nant [16]. He also incorporated a mass balance to
measure particle mass during burning. The main
The current investigation focuses on the black liq- disadvantage to this technique is that the wire on
uor droplets entrained by flue gases that react in which the droplet rests can significantly impact
suspension in the recovery boiler. As shown later, the heat transfer to the particle.
both the inorganic content of the liquor and the
substantial swelling that occurs during combus- Molecular Beam Mass
tion dramatically affect these droplets. The inves-
tigation includes both experimental and modeling
Spectrometry
A single particle reactor [22] provides time-re-
efforts of single particle combustion. Both the
solved, simultaneous data on surface temperature,
experimental data and the model give insight into
internal temperature, mass, and composition dur-
the reaction rates and mechanisms.
ing biomass drying, devolatilization, and char

64
oxidation/gasi? cation biomass particles. Figure 1 position simultaneously during particle reaction.
schematically illustrates the experimental facility This single particle reactor produces experimen-
for the single particle combustion study. tal data, including mass loss, surface and center
temperature, as functions of time for poplar dowel
Single biomass particles suspended on a type-B particles during drying, pyrolysis, and char oxida-
or type-K thermocouple and connected to a wire- tion/gasi? cation processes.
less data logger, from PACE Scienti? c provide The distinguishing feature of the MBMS relative
internal temperature data at a speed of 20 Hz. to other mass spectrometers is its sampling port.
The data logger, thermocouple, and the biomass The MBM sampling port usually consists of a se-
particle, sit on a balance, which provides dynamic ries of two or three nozzles or skimmers. The inlet
mass loss data. For suf? ciently large particles, a ori? ces to each of these nozzles cause choked ? ow
hole the size of the thermal couple wire drilled behavior. The gases expand nearly adiabatically
through the center the particle for thermocouple through a shock wave. In well-constructed noz-
suspension provides access thermocouple access zles, this expansion causes signi? cant cooling and
to the particle interior. Mass loss data from the decreases internal energy. This cooling rapidly
balance with a resolution of 0.1 mg and tempera- quenches the gases and minimizes reactions. The
ture data from the thermocouple simultaneously decrease in internal energy causes the molecules
stream to the data acquisition system. An imaging to be more resistant to fragmentation after being
system and optical pyrometer records the physical ionized [23]. The series of ori? ces decrease the
changes and spatially resolved surface tempera- pressure of the system to the appropriate operat-
ture distribution of the reacting biomass particle. ing conditions of the mass spectrometer (10-5-10-8
Alternatively or additionally, a surface thermo- Torr). Thus the series of ori? ces allows near-real-
couple records the surface temperature at a point. time and essentially in situ sampling of combus-
A molecular beam mass spectrometer MBMS, tion or pyrolysis gases into the mass spectrometer.
specially constructed at BYU, samples off gases This series of inlet skimmers, if properly aligned,
from the particle to determine composition. In allows condensable species to remain in the gas
combination, these systems provide time-resolved phase at temperatures well below their condensa-
measurements of internal temperature, surface tion temperatures because of the absence of nu-
temperature, mass, shape, size, and off gas com- cleation surfaces [24]. Figure 6 shows a diagram

View Port*
Gas Out

Data Logger Biomass

Particle

CCD Camera

Mass Balance

Thermocouple

Heating Elements Preheater

Temperature Controller

*: View ports are in three orthogonal directions N2 Air

Figure 6, Single particle reactor schematic diagram.

65
 Figure 7, Double Skimmer Assembly Used for MBMS.

of the double skimmer assembly used at BYU. In Swelling During Drying and
this configuration, both skimmers create choked-
flow, adiabatic expansion of the gases. If the cen-
Devolatilization
Black liquor dramatically swells during drying
terlines of both are aligned, the molecules emitted
and devolatilization [16, 25]. The gases created
from the second are all traveling in nominally the
upon heating the black liquor create bubbles that
same direction at near sonic-velocities and form
expand or inflate the particle. This swelling drasti-
a beam of molecules, that is, a molecular beam
cally impacts particle heating rates, drag/trajecto-
[8, 14, 17].
ry, mass transfer, and reaction rates. Various inves-
tigations indicate what variables impact the extent
Ip’s group had significantly modified an electro-
of droplet swelling. The most important of these
spray mass spectrometer to meet the requirements
variables is reportedly the lignin to carbohydrate
of the MBMS. These modifications included
ratio. Black liquor experiences maximum swell-
changing the electro-spray ion source to an elec-
ing around a ratio of 50/50. Black liquor swell-
tron impact ion source, rewiring the mass spec-
ing decreases irreversibly when its pH decreases
trometer for the new source, adding power sup-
by adding acid [26]. Wintoku et al. proposed a
plies, cutting and welding a new flange onto the
swelling model based on first principles, yet their
mass spectrometer vacuum chamber, updating the
model was not consistent with data [27-29].
data system, and rebuilding parts of the skimmer
assembly [14].
Flame Region Surrounding a
The benefits of using such a mass spectrometer Droplet
are limited. The mass spectrometer is a semi- Flame models based on mass and energy balances
quantitative instrument because many factors in- commonly assume complete or partial combus-
fluence the mass spectra. These include operating tion, usually infinite kinetics, and commonly con-
pressure, electron energy, quadrupole tuning, and stant or average diffusivities in the flame layer
the ion source. Electron energy affects ionization [30-32]. These assumptions allow computation
efficiency and ion fragmentation. In tuning the of heat and mass fluxes from the devolatilizing
quadrupole, trade-offs between signal intensity particle. When implementing this type of model,
and spectral resolution force some compromises one must assume whether the flame is present or
in each. These variables make it difficult to get ab- not. Verrill et al. assumed that a flame was present
solute species concentration from the mass spec- when the Reynolds number was less than one
trometer, requiring difficult calibrations of the [34]. Järvinen’s model uses a flame model when
mass spectrometer at the conditions used in the the Damkohler number exceeds three [33].
experiments. However, a mass spectrometer can
measure relative change in species concentration Bryden used a different approach to solving the
versus time easily and accurately [23]. flame layer problem [35,36] The domain of the
control volume extends into the gas phase, includ-

66
ing a stagnant film boundary layer. In the outer- largest fractions. Included with Equations (1) and
most shell, the concentrations of the gas species (2) are the molar heat of reaction, ΔΗ̃ rxn , and the
and the temperature equal the bulk gas conditions. molar Gibb’s energy of reaction ΔG̃ rxn for refer-
In this approach, the flame reaction kinetics and ence.
mass transfer clearly indicate—from the high
temperature and decreased oxygen content in the Sulfur Containing Inorganic
film layer—whether a flame is present or not. One
of the major variables in this approach is the film Species
thickness, or the distance from the droplet surface Black liquor typically contains about 4% sulfur.
to the bulk gas. Mass and heat transfer decrease Of this sulfur, initially about 40% is sulfate, 35%
with increasing film layer thickness. Bryden did is thiosulfate, and 25% is organic sulfur. Thiosul-
not explain how he determined the boundary layer fate decomposes almost immediately upon heat-
thickness. ing and thus is not an active participator during
the rest of the combustion process. The organic
sulfur forms H2S and COS, but can be recaptured
Inorganic Chemistry through reactions with different fume species [40-
Inorganic species play an important role during 42]. Sulfate acts as a carbon conversion catalyst
black liquor combustion. They aid in char con- through a mechanism referred to as the sulfide/
version, capture gas-phase sulfur species, and sulfate cycle (Reactions (1) and (2)). These re-
ultimately account for much of the fouling in a actions are extremely important in the char bed
recovery boiler. Inorganic species are the major where simultaneous sulfate reduction and carbon
component of fume and intermediate sized parti- oxidation are essential [3,43]. In a reducing envi-
cle (ISP), which conglomerate and ultimately ac- ronment sulfide is more thermodynamically stable
count for the deposit formation in a boiler [37-39], than sulfate; however, in an oxidizing environ-
which depends strongly on furnace temperature ment the opposite is true.
and oxygen content [26]. Black liquor is typically
about 20% sodium and 1% potassium. Although
most investigations have concentrated on sodium Carbonates
species, it is generally assumed that potassium In an investigation on fume formation, Li and van
species follow the same mechanisms as their so- Heiningen studied the decomposition of sodium
dium equivalents. Furthermore, combinations of carbonate in black liquor under different environ-
sodium and potassium or of chlorides and sulfates ments. They observed that carbon reduces carbon-
form eutectics and other solutions that impact both ate and that this process releases sodium vapor,
deposition rates and deposit properties. Of the in- Reaction (3). They were able to obtain a kinetic
organic species, sulfates and carbonates form the rate expression in a nitrogen environment. They

(1)

(2)

(3)

67
 observed that both CO and CO2 suppress sodium Experimental Approach
carbonate decomposition, but did not acquire re- The experimental portion of this project involved
action kinetics for the decomposition reaction substantial experimental design and trouble shoot-
when these species are present [44,45]. Gairns ing. Experimental facilities used in this project
et al. found that sodium carbonate decomposi- include purchased equipment (cameras, tempera-
tion increased sharply between 800-900 ºC, but ture data logger, temperature controllers, and re-
disappeared entirely by introducing 15% CO2 at lays), newly designed and constructed equipment
900 ºC [46]. In addition to sodium release due to (single-particle furnace), rebuilt equipment, the
sodium carbonate decomposition, Frederick and last category including an MBMS built from por-
Hupa found that between 23-35% of the sodium tions of an abandoned, non-molecular-beam mass
is released during pyrolysis [47]. Others later pro- spectrometer and a spare detector. This section
posed that this sodium release is due to small par- summarizes the experimental methods and the
ticles being ejected from the droplets [48]. features of the equipment used in the project.

Motivation for this Document Experimental Materials and

The current investigation builds upon the work of Methods
the research mentioned previously in this section. Five liquors formed the suite of fuels used in this
This investigation incorporates both modeling project. Their compositions are indicated in Table
techniques as well as single droplet experiments 3. These liquor samples were all from industrial
to understand the mechanisms of black liquor paper mills.
droplet combustion. This document outlines two
types of single droplet experiments. These experi- The first two liquors, A and B, are softwood liq-
ments utilize techniques similar to those devel- uors and come from a mill in the Northwest United
oped previously, including work with a molecular States. Liquor A came with very high solids-con-
beam mass spectrometer, using thermocouples to tent (>95%), and Liquor B came with about 50%
measure internal droplet temperature during com- solids. The other three liquors, C, D, and E, were
bustion, and a pyrometry technique to measure softwood/hardwood mixed liquors and came from
surface temperature. In addition to these tech- different mills. Liquor C came with 55% solids,
niques, this project developed new techniques for Liquor D came with about 70% solids, and Liquor
surface temperature measurement, mass measure- E came with 77% solids. All liquors were concen-
ment, and data synchronization. trated to at least 70% solids content prior to use.
Two types of experiments comprise the experi-
The model developed during this investigation in- mental portion of in this project: droplet combus-
corporates various sub-models to account for the tion experiments and MBMS experiments. In the
phenomena explained in this section. This docu- droplet experiments, drying increased the black
ment explains how the model quantifies swelling liquors B and C to about 70% solids content while
of black liquor, the sulfate/sulfide cycle, carbonate the other liquors remained at the concentrations
decomposition, and the presence of a flame. By shown in Table 3.
using both experiments and modeling, this inves-
tigation reveals some of the important mechanism
of black liquor combustion. Droplet Combustion Experiment
The first type of experiment was a droplet com-

Table 3, Compositions of Black Liquors Used in the Project.

68
bustion experiment. The equipment used in this connect to a single computer for synchronization
experiment includes a furnace built at BYU (Fig- purposes. More detailed descriptions of the fur-
ure 8), three SVS Vistek CCD cameras, a Pace nace, the mass balance, and the cooling jacket, as
Scientific XR440 Pocket Logger and associated well as a novel technique of using video imaging
thermocouple (for internal temperature measure- from a single camera for pyrometry measurement,
ment), a Scientech SA 310 mass balance, and a appear in Ip’s dissertation [13,16].
Thermolyne 21100 tube furnace (used to preheat During each experiment, a 1.5 – 4 mm black liq-
the gas). The furnace is equipped with six view uor droplet suspended on a thermocouple burned
ports oriented in three orthogonal directions. in the furnace. Simultaneous collection of so
These view ports are for video and pyrometry im- many parameters complicates the experimental
aging. They are all viewed in the same plane in the procedures and makes each experiment somewhat
figure to better illustrate the furnace construction, tedious, but the procedures provide unique and
but in fact the axis of the horizontal port is perpen- comprehensive data. The temperature measure-
dicular to the page. ment recorder did not connect directly to the com-
puter (by design).
Four silicon carbide electrical resistance-heating
elements from Kanthal Globar positioned at the Typical experiments proceeded as follows: A
bottom of the furnace heat the furnace. A K-type weighed droplet was placed on an approximately
thermocouple provides feedback to a Watlow 5-10 0.01 inch diameter type K thermocouple. The
V controller and two Continental solid-state relays thermocouple was attached to the temperature
to control the 110 V potential to heating elements. logger and positioned in the balance. The thermo-
The system includes a water-cooled jacket that couple-droplet assembly rolled toward the furnace
slides into the middle of the furnace to provide a until the droplet reached the center of the furnace,
shielded environment for the particles during setup. inside the water-cooled jacket. After the mass
Two rotameters control the gas flow to the furnace, balance oscillations stopped, mass measurement
one manufactured by Cole Palmer and the other and camera measurement were initialized, and the
by Dwyer. A mass flow meter calibrates both. The cooling jacket was pulled from the furnace.
cameras, the mass balance and the pocket logger

Figure 8, Setup for the Droplet Combustion Experiment.

69
 With the cooling jacket removed, the droplet al- temperature based on a few simplifying assump-
most immediately experiences the hot gases of tions, where DN is the digital number describing
the furnace. The droplet then dries, and either the intensity of one color of light. The sensitivity
pyrolyzes in an inert environment or devolatil- of each type of sensor can be described as a func-
izes and oxidizes in an oxidizing environment. tion of wavelength by a function called the spec-
Experiments include runs in both air and nitrogen tral responsivity, S(λ). The emissive power per
environments. During each experiment, the data unit wavelength, E(λ,T), of an object is described
logger records temperature at a rate of 20 Hz; The by Plank’s Law, where E(λ,T) is the product of
mass balance recorded mass at 18-19 Hz; The the surface emissivity and the blackbody radia-
camera’s images were stored at about 15 Hz. tion per steradian. The pixel value is the integral
of the product of E(λ,T) and S(λ) times the lens
After the droplet finishes reacting, the particle and camera setting factors, Equation (4). In Equa-
mass and internal temperature data download to tion (4), τ is the lens transmission factor, D is the
a spreadsheet. The pixel counts across the parti- lens diameter, d is the distance between the object
cle’s horizontal and vertical lengths determine and the sensor, a is the pixel area, X is the mag-
the particle sizes. The known view point tube nification, and t is the camera exposure time. The
diameter serves to calibrate these measurements. technique takes advantage of multiple color bands
Because the droplet is located midway between to simplify the measurement. By taking the ratio
the two view point tubes, the scaling factor for of the pixel values for different color sensors, the
the droplet is simply the average of scaling fac- lens and camera setting factors cancel in the equa-
tors calculated for the two view point tubes. This tion, Equation (5).
document reports geometric mean diameters, that
is, the square root of the product of the horizontal The main advantages of this technique include
and vertical lengths of the particle. The surface accuracy, relatively high signal to noise ratios
temperature measurements come from two points because of the integrations, inexpensive equip-
on the particle surface. At the beginning of a run, ment, and simple implementation. One of the ma-
these data may be far from representative or ran- jor drawbacks for combustion application is that
dom, because the pyrometry technique as a lower flames that surround an object interfere with the
detection limit of approximately 600 °C. Both temperature measurement. This is not unique to
surface temperature and size data have 0.5 Hz color-band pyrometry. Two- and single-color py-
temporal resolution. rometries also may not penetrate the flame layer
to obtain a surface temperature [21].
Optical Pyrometer and Color-Band
Results and Discussion: Droplet
Method
The Color-Band Method for optical pyrometry Combustion
was developed by Lu and Ip, and a detailed expla- This section discusses the experimental results
nation of this method is given by Ip [7,13,14] Typ- of the single droplet combustion experiments.
ically, a CCD camera filters light into four colors Experiments yielded insight into interactions be-
that report to red, blue, and two green pixels. Lu tween droplet swelling, mass loss, and tempera-
and Ip developed Equation (4), which describes ture changes. Observations also show mechanisms
how the camera’s pixel value, DN, relates to the that will be important in modeling, especially a
surface char “regeneration” mechanism.

(4)

(5)

70
Simultaneous Mass Loss, Size, tilization. Devolatilization causes the most rapid
mass loss. Char burning, associated with a slower
Internal and Surface Temperature mass loss, follows devolatilization. The surface
The experimental data collected offer many in-
temperature measurements come from two points.
sights into the combustion behavior of black liq-
The cameras can only accurately determine high
uor. Figure 9 shows the results from a 45-mg black
temperatures (> 600 K). This means that during
liquor A droplet in 750 ºC air. The droplet’s size
the early stages of combustion, only temperatures
changes significantly during the combustion proc-
above this limit are reported. While some region
ess. The droplet’s radius increases rapidly during
of the surface seems nearly always to have ignited
drying and especially devolatilization. Afterwards,
and be burning at this limit, the early temperatures
the droplet shrinks as the char burns out.
are highly non-representative of the overall drop-
let surface temperature. The temperature measure-
The thermocouple temperature curve shown by
ments tend to show that although there is a large
pink squares ascends quickly at first, up to a pla-
variance in the surface temperature, the maximum
teau at about 130 ºC. Water evaporation from this
surface temperature tends to stay nearly constant
salt-saturated fuel leads to this plateau. After the
throughout most of the combustion process. Some
water evaporates from the droplet, the tempera-
of the reported surface temperatures may have
ture again increases. Initially the temperature in-
been affected by flame interference.
creases slowly; the gases that are evolved during
devolatilization decrease the heat transfer to the
Another phenomenon revealed by the surface
particle. Also, local gas evolution in the particle
temperature measurement is the beginning of
decreases heat transfer locally. Once devolatiliza-
char burning on the surface. As shown in Figure
tion is complete, the temperature increase is more
9, for a very dry black liquor droplet (95% sol-
rapid due to the combined effects of no gas evo-
ids), surface char burning starts while the internal
lution, smaller mass and larger surface area. The
temperature of the particle is still near the boiling
center temperature increases to a maximum near
point of water. This creates very large temperature
the end of char burning. This maximum tempera-
gradients inside the droplet of at least 280º C/mm.
ture measured by the thermocouple usually cor-
The surface temperature measurement is always
responds closely with the time that the smelt bead
higher than the thermocouple-measured internal
collapses onto the thermocouple. The temperature
temperature, even as the droplet reaches a pseudo-
then drops slowly to a steady-state temperature in
steady combustion regime. This is likely due to
the furnace somewhere between the local gas and
the thermocouple leads being cooled by the am-
furnace wall temperatures. This temperature de-
bient gases. Because the thermocouple leads are
crease is slow for two reasons. First, by this point
exposed to the surrounding gas, the leads conduct
the particle has shrunk to its smallest volume,
heat away from the thermocouple bead, decreas-
thus the smallest surface area. Second, exother-
ing the temperature measurement.
mic sulfide oxidation reactions continue after char
burning, further slowing the temperature change.
After the initial shrinking of the particle, the parti-
Similar mechanisms affect the mass loss. Mass
cles shown in Figure 9 and Figure 10 swell again.
loss during drying is slower than during devola-
This secondary swelling may indicate that there is

Table 4, Experimental Runs

71
 1600 70
Internal Temperature
Surface Temperature 1
1400 Surface Temperature 2 60
Mass
Diameter
1200
50
C)

4xDiameter(mm)
Temperature((oC)

1000

Mass(mg)
40
800
30
600
20
400

200 10

0 0
0 5 10 15 20 25 30
Time(sec)

Figure 9, 44.8 mg Droplet of Black Liquor A in 750 ºC Air, Gas Velocity=0.35 m/s, 95% Solids.

1600 50
Internal Temperature
Surface Temperature 1
45
1400 Surface Temperature 2
Mass
Diameter 40
1200
35
C)

4xDiameter(mm)
Temperature((oC)

1000 30
Mass(mg)

800 25
20
600
15
400
10
200
5
0 0
0 5 10 15 20
T ime (s ec)

Figure 10, 23.2 mg Droplet of Black Liquor A in 800 ºC Air, Gas Velocity=0.367 m/s, 95% Solids.

72
 1600 90
Internal Temperature
Surface Temperature
1400 80
Surface Temperature 2
Mass
1200 Diameter
70
60
C)
Temperature((oC)

5XDiameter(mm)
1000
50

Mass(mg)
800
40
600
30
400 20
200 10
0 0
0 5 10 15 20 25 30
Time(sec)

Figure 11, 59.3 mg Droplet of Black Liquor D in 715 ºC Air, Gas Velocity=0.217 m/s, 70% Solids.

residual char or black liquor solids when the smelt this may only be a result of turbulence around the
collapses onto it. The smelt reactions with the re- particle, it raises the question of how much mass
sidual char release gases, expanding the particle increase might result from sulfide oxidation dur-
once again. ing smelt oxidation. It is likely that the mass of the
particle increases during smelt oxidation.
Figure 10 and Figure 11 show similar trends to A summary of some of the main combustion char-
those described above. The droplet in Figure 11 acteristics for six runs appears in Table 4. The re-
has a higher moisture content (70% solids). This ported times for drying, devolatilization, and total
leads to a later initial char burning time compared combustion time are approximate since there was
with the droplet of Figure 10. The droplet in Fig- significant overlap of the stages of combustion.
ure 10 started char burning after only about 1 ½
seconds. There was an interesting mass increase Video Observations
at the char burning shown in Figure 10. Although Video images reveal further information on com-
bustion mechanisms. Video images during drying
and devolatilization reveal that particle conditions
are highly two- or three-dimensional. This partic-
ular furnace is bottom heated and has a constant
gas flow around the particle. During devolatiliza-
tion, a bright orange-yellow flame develops near
the droplet as shown in Figure 12.

During devolatilization, the droplet becomes very

porous. Char burning exposes these large voids.
As the char burns, the remaining smelt forms
beads that agglomerate on the droplet surface. As
char burning continues, the smelt beads grow by
agglomeration. This agglomeration seems to min-
imize the mass- and heat-transfer barrier that the
smelt may have imposed if it did not agglomerate.
This means that instead of oxidizer having to dif-

Figure 12, Yellow Flame around Droplet during Devolatilization.

73
 6.94 sec 7.08 sec 7.15 sec

Agglomerating smelt

7.27sec 7.41 sec 7.48 sec

Figure 13, Successive Images during Char Burning, 26.6 mg Droplet of Black Liquor E in 650 ºC Air, 0.32
m/s Gas Velocity, 70% Solids. (Numbers Indicate Total Elapsed Burn Time).

fuse through a thin stagnant layer of a smelt to pen- greatly aid the speed of char burning because it al-
etrate to the char; the char is constantly exposed lows surrounding gases to enter close to the center
on the surface. This has significant modeling im- of the particle. The structure of the droplet may
plications as discussed later. This mechanism for also add to the turbulent convection around the
exposing new char to the surface is called the char particle, hence aiding mass transfer.
regeneration mechanism. Figure 13 illustrates this
mechanism. It shows how a large void opens near At oxidation temperatures, the particle’s interior
the center of the droplet and eventually exposes structure melts. The droplet begins to collapse
new char on the surface. upon itself and agglomerate into one large bead.
Thus, the opening of large voids in the center of
As the droplet completes char burning, smelt ag- the particle may result in a rapid volume change,
glomeration exposes large voids in the particle such as appears in Figure 9. When the droplet col-
center formed during devolatilization. The size of lapses, the unburned char reacts with liquid so-
these voids is roughly the same order of magni- dium carbonate and sodium sulfate, causing rapid
tude as the entire particle. This mechanism may gas release.

Figure 14, Large Voids during Char Burning, Very Large Droplet of Black Liquor A, 95% Solids, in 800 ºC
Air.

74
This rapid gas release may be the cause of inter- oxygen is present in the char layer, and the sul-
mediate sized particles being ejected. Near the fate/sulfide cycle is broken because the sulfide
end of char burning, the droplet releases burning does not oxidize back to sulfate. This decrease in
intermediate sized particles (ISP) easily seen by oxidized inorganic species could be the cause of
the cameras. Although these particles occasion- decreased ISP formation in initially wet particles.
ally appear throughout most of the combustion The two pictures in Figure 15 record the same par-
process, Figure 15 shows the spectacular event ticle at the same time using identical cameras and
that sometimes characterizes the end of char burn- lenses but with different aperture settings. The left
ing. This large release of ISP is consistent with image shows only one particle leaving the droplet,
Ip’s observations. Lien et al. observed that in char but the other image shows hundreds of particles
beds ISP formation increases with bed tempera- leaving the droplet. The right image indicates the
ture [49,50]. In their experiments, increasing inlet particle is nearly exploding. The left image in-
oxygen concentration increased bed temperature. dicates only minor surface reactions occurring.
Increased oxygen on the bed could oxidize inor- (However, if the left image is manipulated, such
ganic species. Thus, their results may also point as inverting the colors as in Figure 16, more parti-
toward ISP formation from a reaction of oxidized cles can be seen.) The distinct difference between
inorganics with char. these two images illustrates how difficult ISP ejec-
tion may be to quantify using visual approaches.
Ip et al. also found that increasing the moisture Conclusions derived from visual observations that
content of the liquor decreased the ISP formation more ISP form at higher temperatures should be
[7,10,13]. Increased moisture content leads to in- highly suspect. Being able to see the ISP depends
creased auto-gasification. During gasification, no strongly on their temperature and camera set-

Figure 15, ISP Ejection After Char Burning, Two Simultaneous Images, Large Droplet of black Liquor A,
95% Solids, in 800 ºC Air.

Figure 16, ISP Ejection After Char Burning, Very Large Droplet of Black Liquor A, 95% Solids, in 800 ºC
Air (Colors Inverted).

75
 tings. Ejection at lower combustion temperatures uses the power-law scheme to solve the combined
or during the early stages of combustion would advection-diffusion problem. The term “advec-
therefore be more difficult to validate visually due tion” in this document describes flow in the radial
to the cooler temperature of the ISP. direction. Darcy’s law for flow through a porous
medium and the continuity equations quanti-
fies this flow, with the numerical solution using
Droplet Model Description and De- the SIMPLE (Semi-implicit Method for Pressure
velopment Linked Equations) algorithm [52]. The model
Modeling combustion processes entails solv- predicts transient temperature, velocity, pressure,
ing transport equations [51], Equations (6), (7), composition, reaction rate, mass- and heat-trans-
and (8), using a control-volume method [52]. fer rates, and related data as a function of one spa-
The model calculations reported here result from tial dimension. This dimension is the radius in the
modifications of a code that was originally de- case of spherical particles and the equations use r
veloped for biomass particle combustion written for the independent variable describing it.
primarily by Lu and partially by Ip [16]. Many
simplifications had to be made to the model be- The integer variable n changes with the coordinate
cause of incomplete understanding of all of the system that best describes the shape and ranges
mechanisms and kinetic parameters required by from 0 to 2. For example, a sphere corresponds
a highly detailed model as well as for computa- to a value of two for n. This enables the code to
tion time considerations. Nevertheless, even the perform calculations for other shapes
simplified model provides significantly more de- Because of swelling during black liquor combus-
tail than those commonly used to describe black tion, the internal radiation heat transfer becomes
liquor combustion. This one-dimensional, tran- important. The solid-phase thermal conductivity
sient model describes heat and mass transfer and of black liquor, λS, depends on temperature and
species continuity for arbitrary shapes. The model composition as indicated in Equation (9) [26].

(6)

(7)

(8)

(9)

76
 (10)

(11)

(12)

(13)

Temperature, T, is in Kelvins, and YS is the mass tions in the reactor.

fraction of the solids. This conductivity com- This second option is better for combustion mod-
bines with the conductivity of the gas, λg, and eling because it solves the transport equations in
radiative properties to form an effective thermal the region immediately surrounding the particle,
conductivity which, when using a diffusion ap- or the boundary layer, thus resolving the gas prop-
proximation for cubic pores and an opaque cube erties near the particle surface, including the flame
pore model, is expressed as in Equation (10) [53]. that often located there. Unlike small-particle
(pulverized-coal, for example) combustion, reac-
A value of 850 m-1 for aR, known as the Rosse- tions in particle boundary layer commonly occur
land mean absorption coefficient, was given by during black liquor combustion. The boundary
Järvinen based on a sensitivity analysis [53]. The layer thickness is approximately proportional to
pressure outside of the particle is assumed con- the particle diameter and black liquor and many
stant at ambient pressure; thus, to find velocities, biomass particles are large enough that flames
in the boundary layer only the continuity equation occur in the boundary layer and provide ther-
is solved. Equation (11) is Darcy’s Law, where µ mal feedback to the particles. Specifically, dur-
is the gas viscosity, η is the permeability, and P ing devolatilization and char burning, significant
is pressure. amounts of combustibles evolve from the particle.
If the boundary layer is very thin (relative to the
reaction distance of the gas-phase components),
Boundary Conditions and Flame these species simply diffuse from the surface and
Layer burn in the bulk gas. However, when the bound-
Symmetry boundary conditions apply at the cent- ary layer is thick, these species burn within the
er of the particle (all gradients are assumed to be boundary layer, thus reducing the oxidizer con-
zero). The model has two options for the bound- centrations at the surface. When a flame is present
ary conditions at the particle edge. The first op- during char burning, the combustion products,
tion is to use standard heat- and mass-transfer cor- such as CO2 and H2O, diffuse back to the surface
relations to calculate the Nusselt and Sherwood to gasify the char.
numbers, as shown in Equations (12) & (13), to
calculate the mass and heat transfer coefficients to The gas flow around a sphere is really at least 2-
the surface [51]. This option is best for pyrolysis dimensional, as is shown in Figure 17. This figure
calculations because it reduces calculation time. shows theoretical calculations using Fluent™ that
The second option is to extend the calculation describe typical fluid mechanics of a sphere en-
domain past the solid surface into the gas phase. trained in a gas with a different velocity. On the
This gas-phase portion is described by a stagnant- upwind side of the sphere there is a stagnation
film model. At the edge of the film, the conditions point. On the sides of the sphere, there is a more
are considered to be the same as the bulk condi- fully developed boundary layer, and on the down-
wind side of the sphere, there is a recirculation

77
 Figure 17, FLUENTTM Plot of Velocity Vectors of Air Flow around a Particle.

zone. This flow pattern creates non-uniform heat heat and mass transfer decrease.
and mass transfer around the sphere. However, Instead of describing the heat and mass transfer at
with a two-dimensional model (assuming the flow each point on the particle surface, a 1-D model can
is axisymmetric around the particle), the flame only describe the average heat and mass transfer.
zone surrounding the particle could be described The Nu and Sh correlations, shown above, work
by simply setting the boundary conditions at an well for describing the average heat and mass
arbitrarily long distance from the particle and transfer in a non-reacting boundary layer. The
solving the transport equations. method developed for predicting the film thick-
ness utilizes the accuracy of these correlations.
For a one-dimensional model, the flame zone sur-
rounding the particle does not lend itself to pre- Estimating the film thickness, involves several as-
cise modeling because flow in only one direction sumptions. The first assumption is that standard
can be considered. In this approach, a film bound- heat and mass transfer correlations in conjunc-
ary layer is used. For a one-dimensional model in tion with the stagnant-film model accurately de-
spherical coordinates, the radial direction, r, does scribe mass and heat transfer and hence can be
not capture the effect of bulk airflow around the used to predict the film thickness [51]. Associated
particle except by changing the film thickness, with this assumption are assumptions of a quasi-
which is the distance from the particle surface to steady-state process, no angular flow (only radial
the bulk conditions [51]. With the bulk conditions flow) in the film, and no gas production sources in
set at an arbitrarily long distance from the particle the film. The definition of the overall mass-trans-
(i.e. very thick film), a one-dimensional model can fer coefficient to a particle appears in Equation
only accurately predict the boundary layer charac- (14).
teristics when the bulk gas is stagnant. However,
by placing the bulk conditions closer to the parti- θm is the correction factor for advection from the
cle, the heat and the mass transfer rates accurately particle, sometimes called the blowing factor, and
describe overall particle conditions even if axial/ the subscripts 0 and L represent the conditions at
vertical flow and variations in the second dimen- the surface and in the bulk, respectively. Because
sion are ignored. The crucial variable is the film of these assumptions, the concentration profile for
thickness. By increasing this distance, the tem- inert species in the film obtains from the steady-
perature and species gradients decrease, and thus state advection-diffusion equation. Another as-

78
 (14)

(15)

(16)

(17)

(18)

sumption is that the product of density and the ity equals zero. However, by applying L’hopital’s
diffusivity is constant in the film. In the calcula- rule or by re-deriving this equation and ignoring
tions below, an average value is used denoted with the advection term, this problem can be eliminat-
an over-bar, ρgDAB . Finally, the film thickness for ed. The result of the film thickness for low veloci-
both reacting and non-reacting species is assumed ties is given in Equation (18).
the same. The differential equation for steady-
state advection-diffusion without a source term The film thickness calculated from Equations (16)
yields an exact solution when coupled with the and (17) is only valid for mass transfer problems,
overall continuity equation. Equation (15) is the and a different film thickness must be calculated
solution to this equation when solved in spherical for heat transfer. The derivation for the heat trans-
coordinates. (Note: By continuity, the product of fer film thickness is analogous to the derivation
density, velocity and the radius squared is a con- for the mass transfer film thickness. The addition-
stant in the boundary layer, i.e. ρgur2 = constant.) al assumptions are that the heat capacity and the
thermal conductivity, λ, in the film are taken to be
The derivative of Equation (15) yields an expres- the average heat capacity and the average thermal
sion for the concentration/density gradient at the conductivity, respectively.
particle surface, dYi/dr|r=r0. By substituting this
expression for dYi/dr|r=r0 into Equation (14) and The most impressive results from using this
solving for rL, the thickness of the film can be cal- method for calculating the film thickness are that
culated. For very low bulk velocities, this calcula- it preserves the analytical solution for non-react-
tion yields very large quantities for the film thick- ing, steady-state advection-diffusion problems,
ness. However, in the code, the film thickness is and that the effects of the flame inside the bound-
limited to ten times the particle diameter. ary layer can be predicted from the model. After
implementing this method for determining the
This equation has a singularity when the veloc- film thickness into the particle model, this method

79
 (19)

(20)

was tested to determine how well it worked for their original diameter during devolatilization,
a particle during pyrolysis. Because there are no which is one of its unique characteristics among
significant gas-phase reactions in the boundary low-grade fuels. The large impact this has on the
layer during pyrolysis, the droplet model predic- structure may be more apparent when considering
tions based the method described above should be the implication for particle density for a nominally
very similar to model predictions based on heat spherical particle. An increase in size by a factor of
and mass transfer correlations during the transient 5 coupled with a decrease in mass through devola-
and should converge to the same steady-state tem- tilization by, say, a factor of 2 implies a change
perature at the end. The following figure shows in particle density by a factor of 250. To account
the predicted particle surface temperature when for swelling, a swelling submodel makes the shell
each of the methods are employed in the droplet volume a function of the evaporation of moisture,
model. The predictions show that by specifying devolatilization conversion, and char conversion.
the thermal and mass transfer film thicknesses in Each shell volume swells or shrinks independ-
the manner described above, the model still ac- ently of the rest. This model does not consider
curately predicts heat and mass transfer for the the droplet to be necessarily more porous near the
simple case of pyrolysis. center as some authors contend is the case [53].
Eq. (20) summarizes the overall swelling model.
Swelling and Drying Models This model is based on that of Järvinen [53]. This
Black liquor droplets swell to between 3-5 times investigation modified it to include shrinking dur-
ing char burnout.
900

800

700
Temperature((oC)C)

600

500

400

300 T Center Correlation

T Surface Correlation
200
T Center Film Model
T Surface Film Model
100

0
0 1 2 3 4 5
Time(s)
Figure 18, Comparison of Stagnant Film Model with Standard Correlations Model for a 2-mm Droplet
Pyrolyzing in 800ºC Nitrogen.

80
 (oC)

Figure 19, Boiling Point Model Regression (Data from [26]).

approach, modified because it is difficult to define

The variable char_consumed is the integral of mole fractions of black liquor. Boiling point data
the rate of char consumption with respect to time, [26] provided a basis for an effective mole frac-
and char_produced is the integral of the rate of tion for the black liquor as a function of moisture
char production with respect to time. The ratio content. The regressed equation is actually based
of these numbers approaches unity from below, on mass fraction (solids content). The regres-
but not necessarily monotonically, as conversion sion (Figure 19) resulted in Equation (21) as an
increases. The β's can be considered to be the ra- expression for water vapor pressure as a function
tios of the volume to the initial volume for drying, of composition and temperature, which is used in
devolatilization, and smelt oxidation. traditional mass- and heat-transfer models for par-
ticles in the remainder of the code. In this equa-
In this submodel, swelling during drying is instan- tion, PH O is the partial pressure of water, and PHsatO is
2 2

taneous. Hupa et al. [54] showed that swelling the vapor pressure.
during drying is initially very fast, but then prac-
tically stops for the rest of drying. Instantaneous
swelling during drying may reasonably approxi- Pyrolysis Model
mate the real behavior. The model assumes that The pyrolysis material in black liquor solids (B)
there is a zero initial concentration of char in the devolatilizes into light gases, tar (T), and char
particle. Some basis for determining shrinking (C). The tar converts into either light gas or char.
during char consumption is required. This model Although this process is conceptually accurate, it
uses the total char production as that basis. By us- does not preserve elemental conservation. Char in
ing the total char production as the basis, as op- this model has the properties of graphite. The light
posed to the char concentration during the previ- gases are assumed to be a mixture of 1.9% H2,
ous time step, the shrinking model is more linear 39.6% CO, 20.9% CO2, 24.9% H2O, and 12.7%
with char conversion. However, due to overlap light hydrocarbons. Thermodynamic properties of
between devolatilization and char burning, the tar describe tar.
particle starts shrinking before it finishes swell-
ing; thus, the degree of swelling specified by the Because no devolatilization kinetics parameters
βB will not occur. Matching swelling with experi- were available for black liquor with the pro-
mental data requires iteration with βB. posed mechanism, parameters were regressed
from available literature data. A zero dimensional
The drying model uses a modified Raoult’s Law model was made of a single black liquor droplet

(21)

81
 Figure 20, Results of Devolatilization Regression.

during pyrolysis along with a code that performs is consistent with Järvinen’s conclusions [34].
nonlinear least squares using the Gauss-Newton Thermodynamics and Kinetics
method. Pyrolysis data from Iisa and coworkers
[42] provided benchmark kinetic information.
Least-squares analyses with this model provided Table 6, Black Liquor Initial Composition.
parameter values. The resulting model predictions
appear with the data in Figure 20.

Char Regeneration
The diffusivity model accounts for the porosity,
ε, and the tortuosity, τ; the latter is assumed to be
1.5, Equation (22). Near the end of char burning,
the smelt becomes molten. The molten smelt’s po-
rosity approaches zero and has a diffusivity orders
of magnitude lower than the gas diffusivity. How-
ever, because the smelt agglomerates into beads,
the underlying char structure is continuously ex-
posed to the surrounding gases. To account for
this mechanism, mass from the neighboring cell is
added to the outermost cell during char oxidation.
When 80% of the char has been consumed from
the outermost cell, small amounts of adjacent cell The model attempts to make all of the thermody-
are added to it to maintain the 80% char consump- namics completely consistent, meaning that each
tion until 99.9% of the particle is contained within species includes thermodynamically consistent
the outermost shell. This rate of cell combination heat of formation and heat capacity values. En-

(22)

82
Table 5, Heat Capacities and Heats of Formation.

thalpies of pure species come from heat capacity reactant)/m3s.

and heats of formation information in the Gordon-
McBride Thermodynamics Database [42]. Tar, in
this model, has the specific heat capacity of ben- Sodium Carbonate Equilibrium
zene. The heat capacities and heats of formation Model
of char (C), black liquor solids (B), and tar (T) ap- In their original investigation of sodium carbon-
pear in Table 5. All of the heats of reaction come ate decomposition, Li and Van Heiningen found
from species enthalpies. All phases (gases and that carbon dioxide and carbon monoxide stabi-
mixture) form ideal solutions. Initial species com- lize sodium carbonate. Published kinetics for the
positions come from softwood liquor data given reverse reaction have not appeared since. To ac-
by Järvinen et al. [53] as shown in Table 6. count for the stabilization of sodium carbonate by
carbon monoxide and carbon dioxide, the model
For simplicity, the condensed-phase species en- modified the original rate expression to be Equa-
thalpies varied continuously as functions of tem- tion (23).
perature. Figure 21 shows an example of the re-
sulting enthalpy as a function of temperature for where the last term represents the equilibrium
sodium carbonate. This approach ignores solid- value of sodium carbonate at the current tempera-
solid and solid-liquid phase transitions. ture and pressure. This expression prohibits the
formation/destruction of sodium carbonate from
All reaction kinetics (other than pyrolysis reac- proceeding beyond its equilibrium point.
tions) for the model come from literature. Table 7
shows the kinetics and their respective references. The equilibrium concentration of sodium carbon-
In this table, species concentration, Ci, has units ate was calculated two different ways. The first
of mols/m3; MWi is in kg/mol; ri is in kg/m3; and utilized HSC Chemistry 5.11© software, where
all rate expressions are in units of kg (of the first N2, Na(v), Na(l), C(pure carbon), CO, and CO2

(23)

83
 Figure 21, Estimated Enthalpy of Sodium Carbonate.

Table 7, Chemical kinetics.

84
85
 12
10
Na2CO3(mol)

8
6
4 '
Simplified Equilibrium
2 HSC
0
900 950 1000 1050 1100 1150 1200
Temperature((oC)
C)

Figure 22, Sodium Carbonate Equilibrium at 50 mol of Nitrogen per initial mol of Na2CO3.

were included in the equilibrium calculations. The different thermodynamic coefficients used in the
second way simply uses Gibbs energy to calculate calculations.
the equilibrium constant for the reaction shown in
Equation (24), which we call the simplified ap-
proach. The simplified approach represents a sci-
Model Limitations
A one-dimensional droplet model has several
entifically rigorous approach, but it uses the same
limitations. As explained previously, the bound-
form of thermodynamic coefficients as is used in
ary layer surrounding the particle is non-uniform.
the rest of the model and it ignores the formation
Also, the particle temperatures and composition
of a liquid phase. These two equilibrium calcula-
vary with angular and radial position. Commonly,
tions were compared and the Gibbs energy mini-
a particle’s windward edges are hotter and burn
mization equilibrium calculation was deemed ad-
out sooner than the leeward edges [13,16]. Ad-
equate, and included in the model (see Figure 22
ditionally, non-uniformity in furnace conditions
and Figure 23). The small differences arise from
leads to non-uniform particle heating.

86
 12
Simplified Equilibrium
10 HSC Chemistry
Na 2CO3(mol)

0
500 600 700 800 900 1000 1100
Temperature( C)
Figure 23, Sodium Carbonate Equilibrium at 1 mol of Nitrogen per initial mol of Na2CO3.

As shown in Figure 24 particle shapes are far from of the dominate combustion stage as well as the
spherical [13,16] and the impact of the shape on extent of conversion of the droplet. In this case,
conversion time and other properties can be dra- the model accurately predicts the stages of com-
matic. Capturing some of these complexities, such bustion; the internal temperature measurement
as the windward tendencies to burn out sooner and lies between the predicted center and surface
hotter than the leeward sides, lies within the grasp temperature during drying and devolatilization
of multi-dimensional models, but others, such as and follows a similar trend to the measured tem-
the bizarre and rapidly changing shapes, depend perature during char burning and smelt oxidation.
on unknown and possibly stochastic processes that If the end of smelt oxidation is taken to be the
may prove very difficult to theoretically capture. final bump in the temperature curves (between 13-
14 seconds), the model and the measured overall
Modeling Results and Discussion conversion times agree within less than 4 %.
This section compares model predictions with the
experimental data from the furnace and illustrates The measured and predicted surface tempera-
the effects of different assumptions on model pre- tures disagree substantially, especially in the early
dictions. stages of combustion. As mentioned previously,
the measured surface temperatures during particle
heating are limited to regions of high temperature
Comparison of Model Results with and are therefore biased toward high temperatures.
Experimental Data Ip showed in his work that large temperature vari-
This section discusses experimental data as com- ations occur on the surface of the particle (about
pared to model predictions and problems associ- 400-500 ºC during char burning and much higher
ated with both. for oxidizing vs non-oxidizing regions). The opti-
cal technique cannot record temperatures as low
Figure 25 compares predicted to measured tem- as the initial temperatures. The data suggest that,
peratures, the latter measured by the thermocou- while the average particle surface temperature
ple (internal) and the camera pyrometry (surface) may be quite low initially, there are local regions
techniques. The temperature is a good indicator of oxidation at very high temperatures that form

(24)

87
 Heating on droplet’s edges

Heating on bottom due to non-uniform

furnace temperature Bizarre but typical droplet shape
Figure 24, Illustration of non-uniform behavior of black liquor droplets

Figure 25, Temperature Profile for a 23.2 mg Droplet of BL A at 800 ºC in Air, Gas Velocity=0.37 m/s, 95% Solids.

very early in the combustion process. This spatial and increase the predicted surface temperature.
variation in surface temperature, evident in the Additionally, a one-dimensional model necessar-
particle surface temperature images, cannot be ily assumes that gases uniformly exit the particle
captured by one-dimensional models. in all directions. The images clearly do not support
such a conceptual mechanism. Apparently, there
One of the assumptions in the model is that the gas are regions of rapid particle outgassing and regions
species in the particle are in thermal equilibrium of relatively little outgassing, with significant dif-
with the surrounding particle. This assumption re- ferences in behavior on the particle surface.
sults in dramatic cooling of the char surface by
the escaping pyrolysis gases. Because of the large Figure 26 illustrates measured and predicted drop-
pore size in black liquor char, it is possible that the let mass histories. Predicted and measured droplet
path of escaping gas is not sufficiently tortuous mass follow the same trend, but exhibit about 0.3
to cause thermal equilibrium with the char. If the seconds of offset. For this run, the droplet’s initial
thermal equilibrium assumption were relaxed, it and final weights were 23.2 and 8.5 mg, respec-
would increase the predicted temperature gradient tively. The model predicted a final mass of 6.5 mg

88
Figure 26. Particle Mass for a 23.2 mg Droplet of Black Liquor A at 800 ºC in Air, Gas Velocity=0.37 m/s,
95% Solids.

9
Diameter Pred Diameter
8

7
DropletDiameter(mm)

0
0 5 10 15 20
Time(sec)

Figure 27, Particle Size for a 23.2 mg Droplet of Black Liquor A at 800 ºC in Air, Gas Velocity=0.37 m/s,
95% Solids.

for the droplet. probably causing the predicted size decreases

more rapidly than the measurements. Iterating on
Figure 27 illustrates predicted and measured par- the swelling coefficient would correct the maxi-
ticle diameters. Although the model appears to mum size prediction. Droplet swelling behavior
over-predict swelling at the beginning of drying, varies radically from sample to sample despite
the first point shown on the graph is the diame- many efforts to homogenize the material and con-
ter of the droplet based on the droplets mass. As trol conditions. Indeed, swelling produces highly
shown in previous investigations, the droplet be- aspherical particles with shapes that are difficult
gins swelling almost immediately. The maximum to capture with a single swelling parameter. Even
predicted diameter is higher than that measured, though correlations indicate average swelling be-

89
 haviors, there are large inherent stochastic varia- The following graphs show the effects of some
tions in the swelling behavior among droplets. modeling assumptions on black liquor combus-
tion. The assumptions in this investigation in-
Effect of Typical Assumptions on clude lumped capacitance (or an isothermal par-
ticle), no swelling, inert inorganic species, and no
Model Results flame. This discussion examines each assumption
Black liquor combustion involves extreme behav- sequentially to show its relative importance. Ta-
iors in many respects compared to other low-grade ble 8 indicates the assumptions relating to each
fuels. With typically 30% moisture, it contains prediction for a 2-mm particle in 1073 K air. The
similar moisture to other biomass fuels. It experi- swelling model assigned βB a value of 22. The first
ences extreme changes in size and density. With prediction has the most simplifying assumptions
a typical volume increase of 35 times its original and is the simplest model, and the last prediction
volume and an associated mass loss of 50%, the has the fewest assumptions and is the most so-
particle’s density decreases by up to two orders of phisticated and accurate.
magnitude between the initial value and the point
of maximum swelling. The considerable swelling Figure 28 shows the predicted droplet surface
during periods of rapid heating contributes to large temperature during burning under different as-
internal temperature gradients. The large quantity sumptions. Prediction 1 has the slowest overall
of inorganic species tends to increase the reactiv- conversion time. Prediction 2 shows the effect of
ity of black liquor very significantly, whereas in a flame on the particle. The flame increases the sur-
most fuels the ash can be assumed inert. face temperature during devolatilization by 70 ºC.
Table 8, Modeling Assumptions Included in Prediction.

Figure 28, Surface Temperature Predictions of a 2-mm Droplet in 750 ºC Air and 0.2% Water Vapor, Gas
Velocity .30 m/s.

90
Figure 29, Predicted Fractional Mass Loss of 2-mm Droplet in 750 ºC, Air and 2% Water Vapor, Gas Veloc-
ity .30 m/s.

The overall conversion time is almost identical for tion with carbon dioxide and water is slower than
the first two predictions, Figure 29 shows that the oxidation with oxygen, char burning decelerates.
flame accelerates devolatilization and decelerates Prediction 3 shows the effect of swelling on the
char burning, leading to largely offsetting influ- particle’s reactivity. Swelling decreases conver-
ence on overall particle conversion time. During sion time by half (from 11.5 s down to 5.5 s). Of
devolatilization, a flame forms inside the bound- all of the modeling assumptions, swelling has the
ary layer, producing higher particle heat fluxes, most dramatic effect on the conversion time. Pre-
thus accelerating devolatilization. During char diction 4 shows the effect of the isothermal as-
burning, partial oxidation of carbon monoxide sumption. Internal temperature gradients tend to
consumes some of the oxygen that would other- increase the estimated conversion time, in this
wise react at the surface. Because char gasifica- case by about 0.5 seconds.

Figure 30, Fraction Carbon and Water Remaining in a 17.3 mg Droplet of Black Liquor D in Air at 750 ºC,
Gas Velocity = 0 m/s, Solids=70%.

91
 Prediction 5 shows the impact of the inorganic the off-gases from the burning particle. If the
chemistry. The inorganic species reactions in- off-gas sample is a representative sample, then
crease particle temperature during char burning. the water peak should indicate the relative water
The inorganic species reactions also increase the evaporation rate plus the hydrogen oxidation rate.
overall conversion time. This increase in conver- In addition, if one assumes complete combustion,
sion time is somewhat counterintuitive consider- the carbon dioxide concentration indicates the
ing it involves including additional char consum- rate of carbon loss. Integration of these peaks over
ing reactions. It occurs for two reasons. First, the time yields the fractional water loss and carbon
inorganic species decrease the particle size dur- loss from the droplet.
ing char conversion. During the early part of char
conversion, sulfate reacts with char to produce Figure 30 illustrates carbon and water fractions re-
CO2 and sulfide. This char conversion decreases maining in an initially 17.3 mg (nominally 2 mm)
droplet size, thus decreasing total mass transfer droplet as measured by the MBMS compared to
(though increasing mass flux). Second, the sul- those predicted by the model. Obvious discrepan-
fate/sulfide cycle is less efficient in its oxygen use cies appear between the measurements and the
than is direct oxidation of char with O2. Although model predictions. Two major differences exist;
the sulfate donates all of its oxygen to char, the (1) There is a time shift between the model pre-
product of this reaction is CO2, which is much less dictions and the measurements by approximately
reactive than O2, and much of the carbon leaves 5-6 seconds; (2) The slopes of the curves (rates of
the droplet as CO2. In the case of direct oxidation reaction) differ significantly. The first discrepancy
of char with O2, much of the carbon leaves the arises from the residence time of gases in the reac-
particle as CO. tor and sample train prior to entering the MBMS.
The reaction rates implied by MBMS data also
Prediction 5 indicates a large temperature spike are much slower , by approximately a factor of
due to the sulfide oxidation. After char conver- 2, than those predicted and those measured by the
sion, the particle cools slower for prediction 5 be- mass balance. This could arise from gas mixing
cause of sulfide oxidation reactions that continue in the sampling train or from velocity profile/
after char burning. chromatographic influences on the gas composi-
tion as it flows through the line. The MBMS data
Comparison to MBMS Data uniquely provide gas composition data, whereas
The molecular beam mass spectrometer sampled the mass-loss data do not. It is possible that there

Figure 31, Predicted Major Inorganic Species for a 23.2 mg Droplet of Black Liquor A at 800 ºC in Air, Gas
velocity=0.37 m/s, 95% Solids.

92
are complex mechanisms leading to species mass diffusion limited rates; therefore these data pro-
loss rates implied by the data, but the accuracy vide little information on char burnout kinetics.
of the drying model when tested alone (based on They are representative of commercially signifi-
mass balance mass loss) is high, indicating that cant char particle sizes and conditions.
the MBMS results are probably compromised.
An expression for the film thickness for a one-di-
Sulfate/Sulfide Cycle mensional flame layer both predicts the presence
The sulfate/sulfide reactions oxidize char. The of a flame and accounts for species steady-state
model indicates this mechanism is the most signif- heat/mass transfer in inert environments. This ex-
icant mechanism for char conversion by an inor- pression, suitable for a the type of transient and
ganic species. Figure 31 shows the model predic- one-dimensional model developed in this docu-
tion of the major inorganic species and indicates ment, does not account for the experimentally ob-
that carbonate decomposition is minimal during served multidimensional flame structures around
combustion at the specified conditions. On the burning droplets/particles in suspension. Never-
other hand, the sulfate decreases to approximately theless, it appears to provide reasonably accurate
23% of its initial mass and then re-oxidizes after estimates of flam affects on particle behavior.
char conversion is complete. Sulfate oxidation
converts one-sixth of the char to carbon dioxide. Experimental evidence indicates that large inter-
nal temperature gradients, sometimes greater than
300 ºC/mm, form during reaction, typically dur-
Conclusions ing drying/devolatilization. The model, however,
This section summarizes the general conclusions does not always predict such large temperature
to the document, including results derived from gradients in the particle. This discrepancy pos-
the both the experimental and the modeling re- sibly arises from the model assumption that the
sults. gas and the condensed phases are in thermal equi-
librium. Because of the large pores, the path of
Experimental data from a novel single droplet re- escaping gases may not be sufficiently tortuous to
actor include simultaneous and temporal data on cause thermal equilibrium. This assumption may
internal temperature, spatially resolved surface lead to another model deficiency, that the maxi-
temperature, size, shape, images from multiple mum predicted surface temperature is consistent-
directions and mass data. These simultaneous data ly lower (never higher) than the measured surface
offer insight into black liquor reaction mecha- temperature, by approximately 50-150 ºC.
nisms and provide computer code verification and
validation information. They are especially use- Sodium carbonate reactions with char proceed
ful because of accurate furnace characterization at negligible rates below 800 ºC, largely due to
(gas flow rates and furnace temperature). These equilibrium limitations. However, sodium sulfate
mechanisms include, but are not limited to, par- reactions with char proceed under a broad range
ticle drying, swelling, devolatilization, heatup, of temperatures and gas compositions and the sul-
oxidation, and smelt formation. fate/sulfide cycle is an important mechanism in
char conversion (and smelt formation).
Particle drying proceeds by a modified Raoult’s
Law expression, where the mole fraction of water Smelt agglomeration during char burning exposes
is replaced by a mass fraction expression. Mole new char to the droplet surface. This mechanism
fractions are difficult to define when dealing with minimizes the heat- and mass-transfer resistance
black liquor and similar natural components. to the char. To account for this mechanism in the
Model predictions agreed with measurements model, neighbor nodes gradually donate mass to
within their uncertainty limits. the outer cell, thus allowing char to be present in
the surface cell throughout all of char burning.
New particle devolatilization parameters were Predictions change significantly as modeling as-
regressed and validated. Devolatilization param- sumptions change. Swelling strongly influences
eters yielded accurate amounts and rates of char overall reaction rates, decreasing overall conver-
and gas formation, as is further indicated from the sion times by approximately 50% relative to pre-
prediction accuracy of char burnout times. dicted rates without swelling. Inorganic chemis-
try, internal temperature gradients, and the flame
Char burnout time predictions and measurements boundary layer also impact overall reaction rates
are in good agreement. Most chars burn at or near by measurable amounts.

93
 Nomenclature

94
ρ ρ

95
96
References in 2005 ACERC Conference. 2005. Provo, Utah.
1. Whitty, K., Pyrolysis and Gasification Behav- 14. Ip, L.-T., A. Mackrory, L.L. Baxter, and D.R.
ior of Black Liquor under Pressurized Conditions, Tree. Comprehensive time dependent black liquor
in Chemical Engineering. 1997, Abo Akademi: single droplet experiments and predictions. 2006.
Turku. Philadelphia, PA, United States: TAPPI Press,
2. Verrill, C.L. and K.M. Nichols. Fume forma- Norcross, GA 30092, United States.
tion during black liquor droplet combustion. The 15. Lu, H., J. Scott, K. Echols, P. Foster, B. Ripa,
importance of sodium release during devolatili- R. Farr, and L.L. Baxter. Effects of Particle Shape
zation. 1992. Atlanta, GA, USA: Publ by TAPPI and Size on Black Liquor and Biomass Reactiv-
Press. ity. in Science in Thermal and Chemical Biomass
3. Jarvinen, M.P., Numerical modeling of the dry- Conversion. 2006. Victoria, British Columbia,
ing, devolatilization and char conversion proc- Canada: CPL Press.
esses of black liquor droplets. Acta Polytechnica 16. Ip, L.T., L.L. Baxter, A.J. Mackrory, and D.R.
Scandinavica, Mechanical Engineering Series, Tree, Surface temperature and time-dependent
2002(163): p. 77. measurements of black liquor droplet combus-
4. Macek, A., Research on combustion of black- tion. AICHE Journal, 2008. 54(7): p. 1926-1931.
liquor drops. Progress in Energy and Combustion 17. Lu, H., W. Robert, G. Peirce, B. Ripa, and L.L.
Science, 1999. 25(3): p. 275-304. Baxter, Comprehensive study of Biomass parti-
5. Jarvinen, M.P., R. Zevenhoven, and E.K. Vak- cle combustion. Energy & Fuels, 2008. 22(4): p.
kilainen, Auto-gasification of a biofuel. Combus- 2826-2839.
tion and Flame, 2002. 131(4): p. 357-370. 18. Wag, K.J., V.V. Reis, W.J. Frederick, and T.M.
6. Ip, L.-T., Comprehensive black liquor droplet Grace, Mathematical model for the release of in-
combustion studies, in Chemical Engineering. organic emissions during black liquor char com-
2005, Brigham Young University: Provo. bustion. Tappi Journal, 1997. 80(5): p. 135.
7. Ip, L.-T., L.L. Baxter, A.J. Mackrory, and D.R. 19. Tamminen, T., J. Kiuru, R. Kiuru, K. Janka,
Tree, Surface temperature and time-dependent and M. Hupa, Dust and flue gas chemistry dur-
measurements of black liquor droplet combus- ing rapid changes ion the operation of black liquor
tion. AICHE Journal, 2008. 54(7): p. 1926-1931. recovery boilers: Part 1 - Dust formation. Tappi
8. Lu, H., E. Ip, J. Scott, P. Foster, M. Vickers, and Journal, 2002. 1(5): p. 27-32.
L.L. Baxter, Effects of particle shape and size on 20. Hupa, M., P. Solin, and P. Hyoty, Combustion
devolatilization of biomass particle. Fuel, 2009. Behavior of Black Liquor Droplets. Journal of
9. Lu, H., L.-T. Ip, A. Mackrory, L. Werrett, J. Pulp and Paper Science, 1987. 13(2): p. 67-72.
Scott, D. Tree, and L. Baxter, Particle-Surface 21. Frederick, W.J., M. Hupa, J. Stenberg, and
Temperature Measurements with Multicolor Band R. Hernberg, Optical pyrometric measurements
Pyrometry. AIChE Journal, 2009. 55: p. 243-255. of surface temperatures during black liquor char
10. Ip, L.-T., D. Ripa, A. Thiriot, W. Roberts, D. burning and gasification. Fuel, 1994. 73(12): p.
Tree, and L. Baxter. Intermediate-sized particle 1889-1893.
(ISP) formation during black liquor droplet com- 22. Ip, L.-T., Comprehensive black liquor drop-
bustion. in 2004 International Chemical Recovery let combustion studies, in Chemical Engineering.
Conference, Jun 6-10 2004. 2004. Charleston, 2005, Brigham Young University: Provo.
SC, United States: Technical Assoc. of the Pulp 23. Evans, R.J. and T.A. Milne, Molecular Char-
and Paper Industry Press, Norcross, GA 30092, acterization of the Pyrolysis of Biomass. 1. Fun-
United States. damentals. Energy & Fuels, 1987. 1(2): p. 123-
11. Lu, H., J. Scott, B. Ripa, R. Farr, and L.L. 137.
Baxter. Effects of Particle Shape and Size on 24. Dayton, D.C. and W.J. Frederick, Direct Ob-
Black Liquor and Biomass Reactivity. in 2004 In- servation of Alkali Vapor Release During Bio-
ternational Chemical Recovery Conference. 2004. mass Combustion. Energy & Fuels, 1996. 10: p.
Charleston, South Carolina. 284-292.
12. Baxter, L.L., L.-T. Ip, H. Lu, and D.R. Tree. 25. Hupa, M. and W.J. Frederick. Combustion of
Distinguishing biomass combustion characteris- black liquor droplets. 2007. Norcross, GA 30092,
tics and their implications for sustainable energy. United States: TAPPI Press.
in 5th Asia-Pacific Conference on Combustion. 26. Adams, T.N., W.J. Frederick, T.M. Grace, M.
2005. Adelaide, South Australia, Australia. Hupa, K. Iisa, A.K. Jones, and H. Tran, eds. Kraft
13. Ip, E., A. Mackrory, L.L. Baxter, and D.R. Recovery Boilers. ed. Anonymous. 1997, Ameri-
Tree. Comprehensive Time Dependent Black Liq- can Forest & Paper Association Adams, T.N: New
uor Single Droplet Experiments and Predictions. York, NY. 381.

97
 27. Wintoko, J. and T. Richards. Swelling be- 40. Boonsongsup, L., K. Iisa, and W.J. Frederick,
havior of black liquor droplet with addition of Kinetics of the sulfation of NaCl at combustion
sodium borates or sodium titanates during drying conditions. Industrial & Engineering Chemistry
and devolatilization. 2004. Norcross, GA 30092, Research, 1997. 36: p. 4212-4216.
United States: Technical Assoc. of the Pulp and 41. Iisa, K. A Comparison of the Sulfation Rates
Paper Industry Press. of Different Fume Species in Recovery Boilers.
28. Wintoko, J., T. Richards, and H. Theliander. in 2004 International Chemical Recovery Confer-
Modeling swelling of black liquor droplet dur- ence. 2004. Charleston, South Carolina.
ing pyrolysis. 2004. Norcross, GA 30092, United 42. Iisa, K. and A.D. Mahvelat. A Model for Black
States: Technical Assoc. of the Pulp and Paper In- Liquor Devolatilization Kinetics. in 2004 Inter-
dustry Press. national Chemical Recovery Conference. 2004.
29. Wintoko, J., H. Theliander, and T. Richards, Charleston, South Carolina.
Experimental investigation of black liquor py- 43. Grace, T.M., Impact of black liquor oxidation
rolysis using single droplet TGA. Tappi Journal, on total energy recovery. South African Mechani-
2007. 6(5): p. 9-15. cal Engineer, 1977: p. 99-102.
30. Turns, S.R., An introduction to combustion: 44. Li, J. and A.R.P. van Heiningen, Mass trans-
concepts and applications. 2nd ed, ed. K.T. Kane. fer limitation in the gasification of black liquor
2000, Boston, MA: McGraw-Hill. 676. char by CO2. Journal of Pulp and Paper Science,
31. Beck, N.C. and A.N. Hayhurst, Early stages of 1986. 12(5): p. 146-151.
the combustion of pulverised coal at high temper- 45. Li, J. and A.R.P. Vanheiningen, Sodium Emis-
atures II. Measurements of the size and composi- sion During Pyrolysis and Gasification of Black
tion of the resulting char. Combustion and Flame, Liquor Char. Tappi Journal, 1990. 73(12): p. 213-
1990. 80(2): p. 170-182. 219.
32. Beck, N.C. and A.N. Hayhurst, Early stages 46. Gairns, S.A., G.J. Kubes, and A.R.P. van Hei-
of the combustion of pulverized coal at high tem- ningen, Sodium loss mechanisms and the forma-
peratures I. The kinetics of devolatilization. Com- tion of reduced sodium during the fast pyrolysis
bustion and Flame, 1990. 79(1): p. 47-74. of kraft black liquor. Pulp & Paper Canada, 1996.
33. Verrill, C.L. and R.A. Wessel, Detailed black 97(6): p. 46-51.
liquor drop combustion model for predicting fume 47. Frederick, W.J. and M.M. Hupa, Evidence of
in kraft recovery boilers. Tappi Journal, 1998. Sodium Fuming During Pyrolysis of Black Liq-
81(9): p. 139-148. uor. Tappi Journal, 1991. 74: p. 192-194.
34. Jarvinen, M., R. Zevenhoven, E. Vakkilainen, 48. Gairns, S.A., G.J. Kubes, and A.R.P. van Hei-
and M. Forssen, Black liquor devolatilization and ningen, Sodium loss mechanisms and the forma-
swelling - A detailed droplet model and experi- tion of reduced sodium during the fast pyrolysis
mental validation. Biomass and Bioenergy, 2003. of kraft black liquor. Pulp & Paper-Canada,
24(6): p. 495-509. 1996. 97(6): p. 46-51.
35. Bryden, K.M., Computational Modeling of 49. Kochesfahani, S.H., H. Tran, A.K. Jones, T.M.
Wood Combustion, in Mechanical Engineering Grace, S.J. Lien, and W. Schmidl, Particulate
Department. 1998, University of Wisconsin-Mad- formation during black liquor char bed burning.
ison: Madison. Journal of Pulp and Paper Science, 2000. 26(5):
36. Bryden, K.M., K.W. Ragland, and C.J. Rut- p. 180-187.
land, Modeling thermally thick pyrolysis of wood. 50. Lien, S.J. and C.L. Verrill. Particulate Emis-
Biomass and Bioenergy, 2002. 22(1): p. 41-53. sions from Char Bed Burning. in 2004 Interna-
37. Skrifvars, B.-J., P. Yrjas, T. Lauren, J. Kinni, tional Chemical Recovery Conference. 2004.
H. Tran, and M. Hupa, The fouling behavior of Charleston, South Carolina.
rice husk ash in fluidized-bed combustion. 2. Pi- 51. Bird, R.B., W.E. Stewart, and E.N. Lightfoot,
lot-scale and full-scale measurements. Energy Transport Phenomena, E. Series, Editor. 2002,
and Fuels, 2005. 19(4): p. 1512-1519. John Wiley & Sons, Inc: New York / Chichester /
38. Theis, M., C. Mueller, B.J. Skrifvars, M. Weinheim / Brisbane / Singapore / Toronto.
Hupa, and H.H. Tran, Deposition behaviour of 52. Patankar, S.V., Numerical Heat Transfer and
model biofuel ash in mixtures with quartz sand. Fluid Flow. Series in computational methods in
Part 1: Experimental data. Fuel, 2006. 85(14-15): mechanics and thermal sciences, ed. W.J.M.a.E.M.
p. 1970-1978. Sparrow. 1980: Taylor & Francis.
39. Tran, H.N., How Does a Kraft Recovery Boil- 53. Jarvinen, M.P., R. Zevenhoven, E.K. Vakki-
er Become Plugged. Tappi Journal, 1986. 69(11): lainen, and M. Forssen, Effective thermal con-
p. 102-106. ductivity and internal thermal radiation in burning

98
black liquor particles. Combustion Science and Glassman, A multiple-step overall kinetic mecha-
Technology, 2003. 175(5): p. 873-900. nism for the oxidation of hydrocarbons. Combus-
54. Hupa, M. and W.J. Frederick, Combustion of tion Science and Technology, 1981. 25: p. 219-
black liquor droplets. Notes - Tappi Press, 1991: 235.
p. 127.
55. Merrick, D., Mathematical models of the ther-
mal decomposition of coal. 2. Specific heats and
heats of reaction, in Fuel, E. Series, Editor. 1983.
p. 540-545.
56. Chase, M.W., Jr., C.A. Davies, J.R. Downey,
Jr., D.J. Frurip, R.A. McDonald, and A.N. Sy-
verud, JANAF Thermochemical Tables, Third
Edition. Physical and Chemical Reference Data,
1985. 14.
57. Koufopanos, C.A., N. Papayannakos, G. Mas-
chio, and A. Lucchesi, Modelling of the Pyrolysis
of Biomass Particles. Studies on Kinetics, Ther-
mal and Heat Transfer Effects. The Canadian
Journal of Chemical Engineering, 1991. 69: p.
907-915.
58. Liden, C.K., F. Berruti, and D.S. Scott, A ki-
netic model for the production of liquids from the
flash pyrolysis of biomass, in Chemical Engineer-
ing Communications, E. Series, Editor. 1988. p.
207-221.
59. Di Blasi, C., Kinetic and heat transfer control
in the slow and flash pyrolysis of solids, in Indus-
trial & Engineering Chemistry Research, E. Se-
ries, Editor. 1996, ACS, Washington, DC, USA.
p. 37-46.
60. DIPPR. Design Institute of Physical Property
Data. http://dippr.byu.edu/index.asp; Available
from: http://dippr.byu.edu/index.asp.
61. Bryden, K.M. and M.J. Hagge, Modeling the
combined impact of moisture and char shrinkage
on the pyrolysis of a biomass particle. Fuel, 2003.
82: p. 1633-1644.
62. Evans, D.H. and H.W. Emmons, Combustion
of wood charcoal. Fire Research, 1977(1): p. 57-
66.
63. Li, J. and A.R.P. van Heiningen, Kinetics of
CO2 gasification of fast pyrolysis black liquor
char. Industrial & Engineering Chemistry Re-
search, 1990. 29(9): p. 1776-1785.
64. Li, J. and A.R.P. van Heiningen, Kinetics of
gasification of black liquor char by steam. Indus-
trial & Engineering Chemistry Research, 1991.
30(7): p. 1594-1601.
65. Smoot, L.D. and P.J. Smith, Coal Combus-
tion and Gasification, ed. Anonymous. 1985, New
York: Plenum Press.
66. Di Blasi, C., Kinetics and Heat Transfer Con-
trol in the Slow and Flash Pyrolysis of Solids.
Industrial & Engineering Chemistry Research,
1996(35): p. 37-46.
67. Hautman, D.J., L. Dryer, K.P. Schug, and I.

99
100
KINETICS OF KRAFT BLACK LIQUOR
COMBUSTION
SERGEY N. SMORODIN,
SAINT PETERSBURG STATE
TECHNOLOGICAL UNIVERSITY
OF PLANT POLYMERS,
SAINT PETERSBURG, RUSSIA

Introduction

The study of kinetics of kraft liquor droplets ing kinetics has been carried out at the Depart-
combustion and development of a mathemati- ment of Industrial Heat-Power Engineering of the
cal model for recovery boiler furnace processes Saint Petersburg State Technological University
meets considerable difficulties because of high of Plant Polymers [1]. Black kraft liquor sampled
temperatures, corrosive medium, and a variety of at the recovery boiler fuel guns of the Pitkäranta
processes taking place at a furnace. Pulp Mill has been studied. Its relative moisture
content was no more than 40%.
Combustion of black kraft liquor at modern recov-
ery boiler furnaces is based on coupling of flame For experiments with dry liquor particles, the
and fuel-bed combustion and consists of three ma- moist liquor was dried for the final moisture con-
jor stages: evaporation, release of volatile organic tent to be no more than 1%. Then the dry liquor
combustibles, and coke carbon burning out. Dif- was grinded and screened by fractions 0.25, 0.5,
ferent physical-chemical processes proceed also 1.0, 1.25, 2.0, 3.0 mm. The study of moist liquor
with mineral components of the liquor. combustion was carried out with liquor particles
of relative moisture content 25%. This moisture
At the recovery boiler furnace, drying of a liquor content value was obtained after drying the ini-
droplet takes place at high heat transfer potential tial liquor at a dewatering box at 105oC. Spherical
and significant intraparticle temperature gradient. particles of diameter 1, 2, 3, 4, and 5 mm were
Under these conditions, the first drying period (the produced from this liquor.
period of constant drying rate) when moisture con-
tent on the particle surface approaches to its critical Initial chemical composition of liquor dry sub-
value and surface temperature is constant and equal stances was practically unchanged during the re-
to a wet-bulb temperature does not exist. search and is given below:
Organic components: Cd = 38%; Od = 16.45%; Hd
The period of decreasing drying rate when the rate = 4.25%; Nd = 0.35%, Std = 1.9%;
is determined by that of water diffusion to the par- Mineral components: NaOH*d = 17.9%, Na2CO3d
ticle surface runs with evaporation surface deep- = 6.04%, Na2Sd = 1.47%, Na2SO4d = 7,5%, NaOHd
ening inward the particle. = 1.67%.

Combustion and gasification kinetics of partly Combustion of individual moist liquor particles
dried and inflamed liquor particles is rather com- has been investigated with an experimental device
plicated. The combustion of volatile substances consisting of quartz tube 120 mm in inner diam-
arising from a liquor particle surface occurs with eter incorporated into an electric furnace. Liquor
simultaneous oxidation of an outer coked layer particles of diameter from 1mm to 5 mm and
and water release from an inner moist zone. moisture content 25% were fixed on junctions of
platinum to platinum-rhodium thermocouples or
Investigating black liquor on platinum electrode wires 0.2 mm in diameter.
combustion The thermocouples with liquor particles and the
The experimental study of liquor drying and burn- tube were fastened to a carriage. For inserting the

101
 carriage into the furnace rapidly, a special mecha- Temperature measurements in a droplet center
nism was used. After a preset period of time, the with the use of a thermocouple at the furnace tem-
thermocouples or wires with the particles were perature T=1273 K show that warm-up and partial
withdrawn from the reactor into a chamber cooled water evaporation takes place during four seconds
with nitrogen flow. for the particles of initial diameter 3 mm and dur-
ing 6 seconds for the particles of initial diameter
The furnace temperature was automatically con- 4 mm, with the temperature in a droplet center be-
trolled and gas medium composition was sus- ing close to that of water evaporation.
tained by operation of gas-fired injector burn-
ers. The experiments were conducted at gas (air) Volatile substances ignite when the surface tem-
temperatures 1173 K, 1273 K, and 1373 K. Air perature is close to 513 K and the duration of total
or gas combustion products were supplied to the drying of the droplets far exceeds the actual time
tube furnace at excess air coefficients 0.88, 1.0, of particles warming-up to their ignition that com-
1.2, and 1.4. prises 25-30% of the total drying duration.

Results The droplet ignition takes place when dry layer

It follows from analysis of these experimental width (δ) approximates 0.2 of the drop radius.
results that significant increase in sizes of burn- Further processes of drying, volatiles release and
ing particles occurs during their heating. For the their combustion go on simultaneously.
particles of initial diameter 1 mm and 2 mm there
was 50-fold increase in their volume and 30-40 The time of total water evaporation from a drop-
times for the particles of initial diameter 3 mm let (τe) and the temperature of outer surface of a
to 5 mm. At the furnace temperature 900 oC, a dry layer (t s) are recommended to be determined
period of particles drying and warming-up to igni- from the equations:
tion comprised 20% of the total process duration.
During the whole combustion period, a particle is
surrounded with flame. Before finishing the burn-
ing process, a drastic increase in temperature in-
side a particle and its breakdown happen, with the
smallest sparks consisting of mineral salts leaving
the particle. Based on the experimental results,
average reduced rate constants of liquor burning
have been calculated. They can be used as the
basis for thermal calculations of recovery boiler
furnaces.

The experimental residual moisture content of liq-

uor droplets vs. period of their staying at the furnace
at different temperatures is shown in Figure 1.

Figure 1, Residual moisture content (W,%) of liquor droplets vs. time at the furnace for different furnace
temperatures (droplet diameter: 1 - 2 mm; 2 - 3 mm; 3 - 4 mm; 4 - 5 mm).

102
Figure 2, Time of ignition (indicted by arrows) and total drying time of liquor droplets of given diameters
at the temperatures 1173 K, 1273 K, and 1373 K.

where q is the heat of warm-up and water evapo- volatiles released depends on temperature. At the
ration; ρ is the density of material; tg and te are temperature 1073 K, their yield comprises 67%,
respectively the gas flow temperature and the tem- at 673 K – 20%, and at 473 K – 13% providing
perature on the evaporation surface of radius re; ro the time of weighed portions calcination was the
– an initial radius of a particle; λd – the thermal same.
conductance of a dry particle; α is a coefficient of
heat transfer on particle surface, t s is the tempera- The study of pyrogenic decomposition of kraft
ture of liquor droplet surface. liquor in an inert medium at the temperatures
ranging from 423 K to 1173 K has demonstrated
The plots obtained with these equations (Figure 2) that thermal decomposition of the liquor started
are in good agreement with experimental data. at 423 K – 433 K. Volatiles release increases as
temperature increases and at the temperatures
973 K and higher losses of mineral components
Black liquor pyrolysis take place. At the temperature 523K, maximum
While studying pyrogenic decomposition of kraft velocity of black liquor volatiles release has been
liquor, the procedure based on regular weight re- observed.
cording for the samples heat treated in pure ni-
trogen flow was applied. Liquor weighed portions
were being kept at temperatures 423 K, 473 K, Organics burn-out
573 K, 673 K, 773 K, 873 K, 1073 K, and 1173 K Experimental study of liquor organics burn-out has
up to complete release of volatile substances (Fig- shown that burn-out velocity depends on the fur-
ure 3). Char residue level was also determined to- nace temperature. The burn-out velocity increased
gether with measuring volatile-matter yield. 1,7 times for particles of the same diameter as the
furnace temperature increased from 1173 K to
Volatiles release started at 423 K. The amount of 1323 K. Initial size of the particles affected also

103
 Figure 3, Volatiles release (Vd) vs. time at the furnace and furnace temperature.

the velocity of organic matter burn-out. The ve- dium sulfide level under given heat treatment con-
locity decreased with increasing initial size of the ditions. Sodium sulfate level slightly decreases as
particles. For example, the burn-out velocity dif- compared to the initial one and sodium carbonate
fered almost two times for the liquor particles of content increases markedly. Alkali carbonization
diameter 1 mm and 3 mm. Figure 4 demonstrates begins immediately with introduction of the stud-
some results of these experiments. ied weighed portion into the furnace.

From chemical analysis of mineral components of Based on the results obtained the conclusion was
both dried liquor weighed portions and individual drawn that at rapid burn-out of suspended liquor
moist particles it follows that their conversions particles in the oxidative medium sodium sulfate
are similar. When introducing weighed portions was hardly reduced to sulfide. Carbonization of
of dried liquor into furnace, chemical reactions liquor was a dominant process. Sodium hydrox-
are characterized by higher rates than those taking ide (including organically bonded one), sodium
place in moist liquor combustion. sulfide as well as salts of different organic acids
take part in chemical reactions resulted in forma-
Chemical analysis of mineral components of tion of sodium carbonate and sodium sulfate. As
moist liquor particles partly burnt-out at different a result of sodium sulfide carbonization, a part of
period of heat treatment shows the unchanged so- carbon dioxide was converted into sodium carbon-

Figure 4, Changes in sodium carbonate level and velocity of liquor organics burn-out (b, %) for the parti-
cles of diameter 3 mm and 4 mm at the temperature 1273 K and air excess α = 0.88; 1.0.

104
ate to yield hydrogen sulfide which while burning velocity for particles of different diameters at
down gave water and gaseous sulfur dioxide. varying air excess and temperature.

Sodium carbonate level at the liquor weighed por- It has been determined that the terminal velocity
tions increased drastically more than three times changed in a similar manner for particles of all di-
in an initial period of heat treatment peaking at the ameters and under all conditions investigated. The
time equal to 0.4 – 0.5 of the total combustion pe- minimum velocity value was in the range from 0.9
riod. Then the sodium carbonate level decreased m/sec. to 1.2 m/sec. for the particles of diameter
because sodium carbonate reduction and evapora- 1mm – 5 mm, with the initial terminal velocity
tion became prevailing over reactions of sodium ranging from 11 m/sec. to 12 m/sec. for original
carbonate formation. particles. The minimum value of terminal velocity
was achieved at relative burning time 0.6 – 0.7.
Chemical analysis of the smelt obtained in the
course of experiments has shown that sodium sul- Analysis of the results obtained has shown that
fate was partially reduced only at a final stage of the lesser is the diameter of a burning particle and
the process. In this case, Na2S and Na2SO4 levels the higher are air excess and furnace temperature,
in the smelt were found to comprise respectively the faster is the decrease in the terminal velocity
2.7 % - 9.3% and 2.6 % - 10.6 %. The reduction up to its minimum value.
value of sodium sulfate increased with increasing
temperature and initial size of liquor particles. The Changes in the terminal velocity of liquor droplets
maximum Na2S level (9.29%) has been found out depending on the burning time at the coefficient
in combustion of moist liquor particles of initial of air excess α = 1 and the temperature 1373 K
diameter 5 mm at α = 1 and the furnace tempera- are shown in Figure 5 (diameter of droplets: 1 – 1
ture 1273 K. mm; 2 – 2 mm; 3 – 3 mm; 4 – 4 mm; 5 – 5 mm).

Terminal velocity New air arrangement

Because of significant changes in sizes and forms A large set of experiments aimed at studying ki-
of liquor particles in the course of their combus- netics of liquor combustion, physical-chemical
tion it is difficult to calculate terminal velocity for processes in smelt and aerodynamics of kraft liq-
the burning particles. At the same time, the termi- uor particles has made it possible to work out a
nal velocity is an important parameter of liquor procedure for calculating burn-out parameters and
particles combustion. That is why the experiments liquor droplets paths at a recovery boiler furnace
have been conducted to determine the terminal [2] as well as to develop and introduce a radically

Figure 5. Changes in the terminal velocity (Wh) of liquor droplets.

105
 new scheme of liquor combustion in intercrossing (which don’t require long preparation to their ig-
jets [3]. nition) in the bottomward direction. Large-sized
droplets are in inertial motion to a central furnace
The method proposed is based on a radically new zone and then fall slowly downwards by gravity
aerodynamic scheme of liquor combustion in in- in gas upflow.
tercrossing jets under U-shaped flame motion.
The project has been implemented at the Kotlas Thus, in accordance with dispersivity of liquor
Pulp & Paper Mill where the “Tampella” recov- droplets, their separation as well as adjustment of
ery boiler has been upgraded through arranging duration of their stay at the furnace takes place in
aerodynamic pinch formed with front and rear these intersecting jets.
water-wall tubes at the secondary air level. The
secondary air was introduced into a furnace with Air supply is directed to a root of fuel jets pro-
20 adjustable nozzles (positioned on two noses of viding conditions for rapid formation of a homog-
ten each), with air jets being directed at the angle enous fuel mixture and intensive preparation of
60o to the furnace bottom. Liquor guns are locat- fuel droplets to their ignition. Inflow of combus-
ed on the front and rear walls below the level of tion products from the furnace due to injecting ac-
secondary air nozzles by 0.9 m. Tertiary air was tion of air jets contributes also to this process. In
completely removed, primary air was unchanged. so doing, vortex motion of combustion products
and fuel droplets arises in a vertical plane contrib-
Secondary air produces a continuous flat jet at the uting also to intensifying the combustion process
level of liquor guns. Injecting action of this jet and complete burning-out of combustibles.
causes inflow of liquor combustion products to air
jet root initiating vortex motion. Air jets intersect Operation of the system has demonstrated its ef-
liquor ones, entrain small-sized liquor droplets ficiency including reduction in flue dust [4].

Figure 6, Vortex furnace of recovery boiler: 1 – furnace; 2 – liquor guns; 3 – secondary air; 4 – aerody-
namic nose.

106
References
1. Evseev, O. D., Investigation of furnace process-
es in kraft liquor combustion. Author’s Abstract
of PhD Thesis; LTI PPI, Leningrad, 1982, 15 p.
2. Smorodin, S. N., Study of furnace processes
in kraft liquor combustion at a recovery boiler
vortex-type furnace aimed at increase in furnace
efficiency. Author’s Abstract of PhD Thesis; SPb
STUPP, Saint Petersburg, 2000, 24 p.
3. Certificate of Authorship No. 1617249, USSR.
Recovery Boiler Furnace/ P. A. Zhuchkov, O. D.
Evseev, E. V. Alekseev, S. N. Smorodin. Published
in 1990. Bulletin No. 48.
4. Zhuchkov, P. A., Evseev, O. D. Alekseev, E. V.
and Smorodin. S. N., Kraft liquor combustion at
a vortex-type furnace. Paper Industry. 1989, No.
2, pp. 24 – 25.

107
108
GETTING RID OF POTASSIUM AND
CHLORIDE
HEIKKI JAAKKOLA
ANDRITZ OY
TAMMASAARENKATU 1
00180 HELSINKI

Introduction

Modern Kraft mill chemical liquor cycles are ESP ash consists of valuable process chemicals,
more and more closed, causing an accumulation sodium sulfate and sodium carbonate, the value
and hence high concentration of Non Process Ele- of sodium losses has to be accounted for. The
ments (NPE) in liquor streams which is harmful sodium losses depend in addition to the selected
to the mill operation. High chloride and potassium process and chloride input to the mill also on the
content in black liquor increase the risk for corro- required chloride + potassium removal, Cl + K
sion and plugging in recovery boilers. concentration in the ash (or in black liquor). The
performances of the three processes are presented
Chloride and potassium are enriched in electro- as a function of required chloride removal rate
static precipitator ash and therefore ESP ash is and Cl-concentration in the ash.
a favorable stream to process when aiming to
control the chloride and potassium content of the Effect of chloride and potassium
whole Kraft liquor cycle. In the paper three dif-
ferent commercial chloride control methods are content on recovery boiler fouling
described and compared. Composition of dust from recovery boiler furnace
(ESP ash) varies much depending on liquor prop-
- Ash out take (dumping) erties and firing conditions including the heat load
- Leaching and especially of the concentration of the potas-
- Ash Crystallization sium and chloride in the firing black liquor. In
Table 1 is the range of major components in ESP
In all chloride removal methods also valuable ESP ash presented.
ash chemicals will be lost. Because main part of

Table 1, Typical composition range of ESP ash.

Component wt-% Compound wt-%

Na 28 - 35 Na2SO4 65 - 90
SO4 45 - 65 Na2CO3 5 - 20
CO3 5 - 15 NaCl 2 - 10
K 2 - 10 K2SO4, KCl 5 - 20
Cl 1-5

109
 2

Figure 1, Sticky temperature T15 as function of chloride and potassium content in ESP ash; Point 1. Ash Cl-
content = 8 mol-% (4 wt-%) and K = 9 mol-% (5 wt-%) => T15 = 580 ºC, Point 2. Cl = 2 mol-% (1 wt-%)
and K = 3,6 mol-% (2 wt-%) => T15 = 700 ºC.

The tendency of ash to form deposits is related In practice some safety margin for the maximum
to the smelt behavior of the ash. Smelting curve allowed temperatures has to be reserved. Gener-
depends on ash composition, generally a high ally, the following maximum chloride and potas-
chloride and potassium concentration causes low sium concentrations (Table 2) are recommended:
sticky temperature T15, therefore high Cl+K re-
quires, to avoid fouling and corrosion, low steam To make detailed estimation for “safe, non-foul-
temperature to obtain low surface temperature for ing” maximum chloride and potassium content,
the superheater tubes. boiler CFD calculations including temperature
profile and smelt curve for the ash have to be
Figure 1, is an example, of the effect of chloride done.
and potassium content on the sticky temperature
T15 (figure concentrations are given as a mole ra- Depending on construction materials risk for
tio of either Cl or K per alkali metals (Na + K)). chlorine corrosion may be the limiting factor to
maximum allowable temperature.

Table 2, Recommended maximum concentrations for Cl and K.

Steam VirginBL ESPash

Temp. / C Cl / % K/% Cl / % K/%
< 480 0.5 1.5 - 3 2.5 4-7
< 500 0.2 - 0.3 1-2 1 - 1.5 2.5 - 5
> 500 0.15 1-2 1 - 1.5 2.5 - 5

110
Chloride and Potassium sources The enrichment is usually defined as the ratio in
the dust to the content in virgin black liquor. In the
and losses in liquor cycle calculation the contents are expressed as the mo-
lar ratio of chloride and respectively of potassium
The NPE input sources and losses in liquor cycle
to the sum of sodium and potassium, enrichment
are listed in Table 3. Cl- and K-input from wood
factors for Cl and K in equations 1 and 2. Concen-
depends largely on the soil and transportation (e.g.
trations in the equations are expressed as mol-%.
sea floated). Flue gas losses are connected to SO2
emissions, the higher liquor DS the lower SO2 and
(1)
Cl emissions, with DS over 75 % the losses are
ECl = [ Cl / (Na + K)] ash / [ Cl / (Na + K)] virgin BL
practically zero.

In euca mills typical net input for chloride is 1 – 3

(2)
kg Cl /ADT and for potassium 2 – 5 kg K / ADT
EK = [ K / (Na + K)] ash / [ K / (Na + K)] virgin BL
and for the soft wood respectively 0,3 – 1 kg Cl
/ADT, 1 – 2 kg K / ADT.
Cl and K enrichment factors vary very much (Ta-
Chloride and potassium ble 4). This depends on liquor properties, potassi-
enrichment in ESP ash um and chloride concentrations, firing conditions
Chlorides and potassium compounds have higher and especially on heat load.
vapor pressures than sodium sulfate. This results
in their preferential volatilization from the smelt Rough estimate for typical enrichment factors on
bed into the ash stream. As a result, chloride and weight basis for chloride = 5 and for
potassium have higher concentrations (of DS) in potassium = 2.5. E.g. If Cl in virgin liquor = 0.5
ash than in virgin black liquor. wt-%, Cl in ash = 2.5 wt-%.

Table 3, Chloride and potassium sources and losses.

Source kgCl/ADt kgK/ADt

Wood 0.2 – 3 0.4 – 3
Make up chemicals and water 0.2 – 0.8
Waste acid 0.1 – 0.5

Losses kgCl/ADt kgK/ADt

Washing losses 0.1 – 4 0.2 – 3
Flue gases 0.01 – 0.5

Table 4, Enrichment factors.

NPE Range Typical

mol/mol mol/mol
Chloride 0.3 - 6 2.6
Potassium 1 - 2.5 1.6
111
 Chloride and potassium removal tank to a decanter centrifuge that separates solids
from liquid. The filter cake is returned to black
methods liquor mix tank and the filtrate, enriched in chlo-
Modern methods to control chloride and potas-
ride and potassium is purged to sewer.
sium content in ESP ash are:
When carbonate (CO3) concentration in the ash is
- Ash out-take (dumping)
over 5 %, the settling velocity of the slurry parti-
- Ash leaching
cles becomes very low and the ash cake remains
- Ash crystallization, Andritz Ash ReCrystal
very wet, meaning that Cl and K removal rates are
lization, ARC
decreased and sodium losses are increased. The
carbonate concentration can be decreased by add-
Ash dumping is widely used in the industry, no
ing sulfuric acid in the leaching tank, the pH is
need or low investment costs but chemical losses
decreased from 10 - 11 to 9 – 9.5. The acid will
are high. The two other methods are based on
react with carbonate forming sodium sulfate and
solubility differences between different ash com-
carbon dioxide according to the formula below:
ponents: chloride and potassium compounds are
more soluble than sulfates and carbonate.
Na2CO3 + H2SO4 => Na2SO4↓ + CO2↑ + H2O

Ash leaching The acid consumption depends on the carbonate

In leaching process (Figure 2) ESP ash is mixed content in the ash. For example: if the untreated
with hot water in ratios of 1 – 1.5 kg ash / kg wa- ash contains 10 % carbonate about 80 kg sulfuric
ter forming slurry: ash is only partially dissolved acid / ton of ash is required to decrease the carbon-
in water. Most of chloride and potassium com- ate concentration down to 5 %. The added acid is
pounds are dissolved in water while major part of not wasted; it compensates partially the need of
sodium sulfate and carbonate remain as solids in sodium sulfate make up to the liquor cycle.
the slurry. The slurry is pumped from the mixing

Figure 2, Principle of Ash Leaching.

Figure 3, Principle of Ash Crystallization.

112
Ash crystallization, ARC Comparison of chloride and potas-
In crystallization process ESP ash is first dissolved sium removal processes
completely in hot water (or condensate) in ratios Chloride and potassium removal efficiency de-
of 0,3 – 0,6 kg ash / kg water. The solution is fed pends on ash composition and is related to chemi-
to evaporative crystallizer where most of cal losses in each controlling method, part of the
sodium sulfate and carbonate is crystallized and process elements and compounds are purged with
most of chloride and potassium compounds re- Cl and K to sewer. Typical chloride removal ef-
main in the solution. The crystal slurry is pumped ficiency for leaching and crystallization processes
from the crystallizer to the centrifuge where crys- is presented in figure 4 as a function of sodium
talls and liquid are separated. The filter cake is losses.
returned to black liquor mix tank and the filtrate,
enriched in chloride and potassium is purged to Potassium removal efficiency is about the same
sewer. as that for chloride when potassium content in the
ash is less than 4 wt-%. In higher potassium con-
Metal ions in the removal process centration glaserite, K3Na(SO4)2, co-crystallizes
ESP ash water solution has pH 10 -11. Most metal with sodium sulfate and potassium removal will
ions form insoluble hydroxides in basic solutions be decreased remarkably. So when a Cl&K re-
and therefore metals in the ash are retuned back moval process is installed in an existing mill with
to liquor cycle with centrifuged cake from the high potassium content, ash has to be taken out as
removal process and only 2 – 15 % is purged to ESP dust until the potassium concentration in ESP
the sewer or otherwise: 85 – 98 % of metals are ash has decreased to about 4 %.
recovered.

Table 5, Measured metal concentrations in an ARC process.

VirginBL EspAsh CrystalsfromARC PurgefromARC

Al mg/kg ds. < 200 39.4 37.0 2.96
Cd mg/kg ds. < 0.4 0.28 0.41 < 0.05
Co mg/kg ds. <2 0.15 0.26 0.10
Pb mg/kg ds. <4 9.94 17.4 0.24
Mg mg/kg ds. 345 168 97.8 15.8

100%

80% ARC

Leaching
Clremoval

60%

40%

20%

0%
0% 10% 20% 30% 40%
Naloss

Figure 4, Chloride removal efficiency as a function of sodium losses.

113
 Figure 4 shows that the crystallization process is process alternative for the specific case can be se-
much more effective and has less chemical losses lected.
than the leaching process. Depending on the re-
quired chloride and potassium content in ash and In Table 6 is the ash re-crystallization (ARC) and
the input of Cl and K into the liquor cycle, we will leaching processes compared with respect to the
only need to treat a part of the ash in most cases, ash out-take.
to reach an acceptable concentration for Cl+K. In
such cases, the efficiency of the Cl + K removal Investment costs for ash out-take is much lower
process will, with respect to the chemical losses, than the other removal processes in table 6, but
have a less significant role to determine the proc- the annual make up chemical costs are very high.
ess alternative that is most cost effective. Depending on mill sulfur and sodium balances the
share of the make up chemicals may be different
While each pulp mill has different ash composi- from the figures shown in the table. Generally the
tion Andritz has developed a comprehensive liq- pay back time for the investment costs of ARC or
uid-solid equilibrium program by which the two leaching is less than one year compared to total
processes can be well simulated and the optimum out-take costs.

Table 6, Comparison of Cl and K removal processes.

Milldata ESPAsh
Component wt-%
Pulp production ADt/d 2 000 Na 32.5
Recovery Boiler ton DS/d 3 500 K 2.7
K removal kg/ADt 1.1 Cl 1.0
ton/d 2.3 SO4 53.8
Cl removal kg/ADt 0.4 CO3 10.0
ton/d 0.9
Process ARC Leaching Outtake
ESP ash feed ton/d 108 111 87
of ash % 39 40 31
Removal( of treated ash )
K % 80 80 100
Cl % 80 80 100
Loss
Na ( of treated ash ) % 20 30 100
ton/d 7 10 28
kg/ADt 3.5 5.2 14
SO4 ton/d 6 16 47
kg/ADt 2.9 8.1 23
Makeupchemicals(100%)
H2SO4 ton/d - 8.6 -
kg/ADt - 4.3 -
Na2SO4 ton/d 11.7 8.7 68.8
kg/ADt 5.9 4.4 34
NaOH ton/d 7.3 11.6 10.2
kg/ADt 3.7 5.8 5

114
Comparing ARC and leaching processes with the Summary
same Cl and K target for (virgin liquor / ESP ash) The liquor cycles in modern pulp mills are well
chloride and potassium concentrations: closed meaning increased NPE concentrations
also in black liquor. Investment in high energy re-
− Make up chemical costs for ARC is lower covery boiler with high power production is often
due to higher removal efficiency very feasible. Due to recent changes in cost ef-
− Electrical power consumption is about fectiveness, caused by higher energy efficiencies
same and higher energy prices, more than dozen pulp
− Heat consumption in leaching is much mills are to-day controlling their chloride and po-
lower tassium concentrations by ash leaching or crystal-
− Different secondary heat sources (green lization.
liquor flash steam, integration with black
liquor evaporator) can be utilized in ARC The selection between the two processes has to be
− Investment cost for leaching is lower done considering:

− Chloride and potassium input

− Capacity
− Ash and black liquor composition
− Target chloride and potassium concentra
tions in black liquor / ESP ash
− Removal efficiency
− Investment costs
− Mill balance (sodium / sulfur)
− Make-up chemical prices

115
116
ALTERNATE MATERIALS FOR RECOVERY
BOILER SUPERHEATER TUBES
JAMES R KEISER, OAK RIDGE NATIONAL
LABORATORY, OAK RIDGE, TENNESSEE
JOSEPH R KISH, MCMASTER UNIVERSITY,
HAMILTON, ONTARIO
DOUGLAS L SINGBEIL, FPINNOVATIONS,
VANCOUVER, BRITISH COLUMBIA

Introduction

The black liquor recovery boilers used in pulp and overcome the corrosion problems that are barri-
paper mills operating with the kraft process are ers to achieving higher efficiency performance.
critical components of those mills. These boil- This paper describes the issues associated with
ers generate steam that is utilized for power gen- recovery boiler superheater tubes and reviews
eration and process applications, and they are the industry’s experience with various materials. It
first step in recovering the chemicals used in the also provides information about the materials
pulping process. Boilers that have a lot of cor- problems with other types of boilers.
rosion, cracking, and plugging issues often oper-
ate at less than optimal conditions and generally
require more down time in order to address the Background and Experience with
maintenance issues. Consequently, the recovery Superheater Corrosion in Current
of energy from burning of the organic components Generation High Pressure
in such recovery boilers can be considerably less
efficient than desired, and significant safety issues Chemical Recovery Boilers
can also be introduced. Corrosion of superheaters is generally not an issue
in recovery boilers that are operated at a steam
Biomass-fired boilers, other than black liquor re- temperature below 450°C [1]. In current genera-
covery boilers, are used to produce steam from tion high pressure (HP) boilers, steam tempera-
materials like hog fuel, agricultural waste, solid tures are somewhat higher, usually ranging from
waste from effluent treatment, etc. The heteroge- 480 to 500°C. It is generally accepted that ac-
neity of the fuels as well as the often significant celerated corrosion of superheaters in a HP recov-
water content cause a variety of materials prob- ery boiler is mainly caused by molten salt attack
lems for boiler designers and operators. [1-5]. Molten salts are generally a good fluxing
agent effectively removing protective metal oxide
Boilers burning pulverized coal produce steam scale from an alloy surface. Accelerated corrosion
for generation of a very significant portion of the on the unprotected surface proceeds primarily by
electric power used in the United States. These oxidation, which is then followed by dissolution
boilers operate at efficiencies of about 35%, and of metal oxides in the molten salt.
they produce a significant amount of CO2 as well
as other pollutants. Operation at higher efficien- Superheater deposits consist mainly of sodium
cies would reduce the production of pollutants, sulfate (Na2SO4) and sodium carbonate (Na2CO3)
but the relatively significant sulfur and chlorine with small amounts of sodium sulfide (Na2S), so-
content of most coals cause serious materials dium chloride (NaCl) and similar potassium (K)
problems. To be able to operate at higher efficien- salts [1,6]. The first melting temperature (FMT) of
cies, stronger, more corrosion resistant materials the deposit depends on the composition of potas-
would be needed. sium, carbonate (CO32-) and sulfide (S2-), typically
ranging from 520°C to 580°C [6]. Another criti-
The need for better materials for heat transfer sur- cal factor is the proportion of liquid phase present
faces, and particularly for superheater tubes, is a at the FMT, which strongly depends on the chlo-
common issue that must be addressed in order to ride concentration [7]. Reported field studies and

117
 operating experience consistently show that the For superheaters in HP recovery boilers, a bal-
corrosion rate of superheater tubes increases rap- ance between performance, plugging and corro-
idly when the FMT of the deposit is exceeded, and sion is required. With respect to corrosion, a key
more so if the molten deposit contains sulfide or superheater design constraint is to maintain tube
reduced sulfur compounds [2,8-13]. In the past, temperatures below the first melting point of the
many superheater corrosion problems were caused deposits. Temperatures are reduced by ensuring
by molten acidic sulfate deposits containing so- relatively low flue gas velocity, ensuring relatively
dium and potassium bisulfate [(Na,K)HSO4] and high steam flow through each tube, arranging for
pyrosulfate [(Na,K)S2O7] salts, which are stable the coolest steam to flow through tubes exposed
at elevated concentrations of acidic sulfur gases to hottest flue gas, and locating the majority of the
(SOx) [14]. However, typical SOx levels produced superheater sections behind the bull nose (furnace
with current operating practices are generally not arch), shielded from direct radiation [16,17]. This
high enough to promote corrosion by the forma- design philosophy has permitted the widespread
tion of these molten acidic sulfate salts [3]. use of the less-expensive low alloy steels as su-
perheater tube materials. However, such applica-
Superheater design for low pressure (< 6,000 tion has not been without problems. Reported ex-
kPa), (LP), recovery boilers traditionally has been perience with HP superheaters fabricated entirely
governed by the need to achieve a balance be- of low alloy steel, revealed rapid corrosion at the
tween performance (steam temperature and pres- bottom bends, with measured rates ranging from
sure) and plugging. Relative to other boilers, re- 0.3-0.6 mm/y [2,8,13] to as high as 3.9 mm/yr
covery boilers have a much higher particulate flux [19]. The cause in each case was attributed to el-
through the superheater section and, consequently, evated temperatures of the lower bends. Based
tend to plug more easily [15]. Side-to-side spac- on the various resistances to heat transfer, the sur-
ing (across the bank), tube size and mechanical face temperature of a superheater tube is expected
design (including support) has varied throughout to be about 30C° to 50C° hotter, or higher if af-
the decades to accommodate overall boiler design fected by direct radiation, than the steam tempera-
[16,17]. Low alloy steels have been commonly ture inside the tube [20,21]. In addition, the tube
used for superheater tubes in LP boilers, with the temperature may be 20C° hotter on the windward
selection generally governed by similar design side than on the leeward side [12]. Note also that
constraints as for other types of boilers, namely the individual tube steam temperature can vary ±
high temperature strength, creep resistance, oxi- 23C° from the average across the steam manifold
dation resistance and economics. Table I iden- [22]. The resultant combined increase may be be-
tifies steels used along with the recommended yond the FMT of the deposit depending upon the
maximum allowable tube temperature. composition.

Table 1. Typical Superheater Tube Materials in LP/HP Kraft Recovery Boilers.

Specification Nominal Composition (wt.%) Max. T1

ASME UNS Fe Cr Ni Mo Cu C (°C)
SA-209T1A K12023 Bal. --- --- 0.44-0.65 --- 0.15-0.25 524
SA-213T11 K11597 Bal. 1.0-1.5 --- 0.44-0.65 --- 0.05-0.15 566
SA-213T22 K21590 Bal. 1.9-2.6 --- 0.87-1.13 --- 0.05-0.15 602
SA-213T91 K90901 Bal. 8.5-9.5 --- 0.85-1.05 --- 0.08-0.12 649
SA-213T304 S30400 Bal. 18.0-20.0 8.0-11.0 --- --- 0.08 Max. 760
SA-213TP347H2 S34709 Bal. 17.0-20.0 9.0-13.0 --- --- 0.04-0.10 760
SA-213TP310H3 S31009 Bal. 24.0-26.0 19.0-22.0 --- --- 0.04-0.10 816
SB-407-800H N08810 Bal. 19.0-23.0 30.0-35.0 --- 0.75 Max 0.10 Max. 816
SanicroTM 283 N08028 Bal. 27.0 31.0 3.5 1.0 0.02 Max. n/r4

1 Recommended maximum use temperature for superheater tube use in fossil-fired boilers [18].
2 Nb/Ta content of not less than eight times the carbon content and not more than 1.0 wt.%.
3 Fabricated as a co-extruded composite tube; exterior layer over an SA-213T22 core.
4 n/r = Not Reported

118
Replacing the low alloy steel lower tube bends However, there have been reported cases in which
with higher chromium-containing, more corro- these alternatives have experienced accelerated
sion-resistant alloys has been one approach used corrosion. A summary of these cases is provided
to address the accelerated corrosion observed. A in Table 2. A reported case of accelerated corro-
listing of the more typical alloys used is also pro- sion exists for each of the major alloys used as re-
vided in Table 1. The selection of these alloys was placement upgrades, with the exception of Alloy
based largely on prior field testing using air-cooled 28 lower superheater tube bends. Alloy 28 tube
probes [2,8-13]. Reported experiences of using bends have been installed in Swedish, Brazilian
these upgrades have been positive for the most and Finnish HP recovery boilers since 1994 [27].
part. A summary of preliminary experience (no An elevated tube temperature was identified as the
more than seven years service) of austenitic stain- root cause of the accelerated corrosion in each of
less steel lower tube bends installed in Finnish HP the reported cases. This conclusion is consistent
recovery boilers (480°C/8,000 kPa superheater with the findings of the air-cooled probe studies,
units) was published some time ago [23,24]. As which show that corrosion of these alternative al-
reported, the corrosion rate of the installed tube loys increases rapidly when the first melting point
bends, fabricated from Type 304, 304L, 321 and of the deposit is exceeded [2,8-13].
347, stainless steel, was around 0.1 mm/yr, with
no significant differences found between the vari- In addition to the fireside corrosion issues noted
ous grades of stainless steel used. There are also above, the reliability of superheaters in recovery
limited reports of low corrosion rates observed on boilers is also strongly affected by stress-related
Alloy 800H lower tube bends installed in Cana- factors. Given that the superheaters are suspend-
dian HP recovery boilers (482°C/6,200 kPa super- ed through the roof and hang freely in the flue gas
heater units); one based on a little more than two environment, they are susceptible to a combina-
years of service [10], and the other based on about tion of mechanical and thermal stresses, static
twenty years of service [25]. Furthermore, no cor- and/or cyclic, during start-ups, operation and
rosion was reported for Type 310 stainless steel shut-downs. As recently discussed, these prob-
weld-overlaid composite tube bends after twenty lems can result in distortion (in-plane and out-of-
four months of service when installed in a South plane) of the assembly, cracks at attachments, or
American HP recovery boiler (432°C/6,400 kPa crack-induced tube leaks [29]. Another cause for
superheater unit) [19], or for Alloy 28 co-extrud- concern is waterside stress corrosion cracking. As
ed composite tube bends after twelve months of reported in the literature, the monolithic authentic
service when installed in a Canadian LP recovery stainless steel tube bends are likely more suscep-
boiler (400°C/4,300 kPa superheater) [26]. tible [30].

Table 2. Reported Cases of Accelerated Lower Tube Bend Corrosion of Replacement Alloys.

Tube Alloy Design Steam Design Steam Tube T Deposit FMT Rate Ref.
T (°C) P (kPa) (°C) (°C) (mm/yr)
SA-213T304 480 8,000 N/R N/R N/R [11]
SA-213T304 480 6,500 N/R N/R N/R [11]
SA-213TP347H 480 8,000 N/R N/R 2.4-3.6 [24]
SA-213TP347H 480 8,000 N/R N/R 0.3-1.6 [24]
SA-213TP347H 480 8,000 N/R N/R 1.5 [24]
SA-213TP347H 482 8,620 > 590 520 1.0-2.0 [12]
SA-213TP310H/SA-213T22 454 6,200 > 590 560 1.0-2.0 [28]
SB-407-800H 482 6,200 N/R N/R 0-1.4 [25]
*N/R = Not Reported.

119
 Superheater Corrosion and studied is provided in Table 4. As the summary
shows, a wide range of alloys have been tested
Control in Next Generation HT/ in a significant number of salt mixtures, exposed
HP Recovery Boilers to a wide range of cover gas mixtures, for several
Next generation HT/HP recovery boilers, which different time periods.
are operated at a steam temperature of 500°C to
515°C and a steam pressure of 9,900 kPa to 10,900 In spite of the many differences among the testing
kPa, already exist in Japan [31-34]. Key design parameters, these studies have generated a rea-
features implemented that help control superheat- sonably consistent set of results, and thus conclu-
er tube corrosion are reported by the boiler manu- sions can be drawn regarding corrosion above the
facturer to include the installation of non-process FMT. Major findings are summarized as follows.
element (NPE), namely potassium and chlorine, Corrosion increases significantly at temperatures
removal equipment downstream of the recovery above the FMT, the extent of which increases with
boiler, use of special austenitic (25Cr-14Ni) stain- the amount of molten phase present at a given
less steels, and a large number of sootblowers. temperature. The corrosion mode is mostly local-
Reducing the potassium and chlorine content in ized (pitting and intergranular attack), consistent
the liquor cycle should help to control corrosion with a mechanism involving attack by a molten
by lowering the potassium and chlorine content chloride-containing salt, and does not require an
in the deposits that form on the superheater tubes, interaction with SO2 or HCl in an O2-containing
which in turn increases the FMT [6] and decreases gas phase. Alloys with higher chromium contents
the proportion of the liquid phase present [7], re- have a higher corrosion resistance which further
spectively. The special stainless steels were devel- increases if molybdenum and nitrogen are present
oped with input from laboratory studies, as well as additional alloying elements. Finally, the re-
as from operating experience [31,35,36]. Table 3 moval of NPE (potassium and chlorine) from the
documents the nominal compositions of the three chemical recovery process is a critical part of cor-
grades developed. The original grade YUS170 rosion control.
was used in early HT/HP recovery boilers. The
latest grade MN25R was developed to improve In contrast, the conclusions drawn from these
the strength at high temperatures and to increase studies regarding the corrosion of alloys below
the (intergranular) corrosion resistance. Experi- the FMT are more controversial. There is a con-
ence reported has been positive with these alloys, sensus that the risk of corrosion below the FMT
except when there has been contact with a high by the so-called “active-oxidation” mechanism
sulfur dioxide (SO2) concentration in the flue gas, [42,43] is low for the set of alloys studied, with
and with deposits enriched in chlorine (as chlo- the exception of the low alloy steel grade SA-
ride) [33]. There was no mention of a waterside 213T22. Active oxidation involves the forma-
stress corrosion cracking concern. It is noted that tion of Cl2/HCl gas within a chloride-containing
no independent confirmation of boiler operation deposit, which then migrates to the alloy/scale
and/or superheater tube performance has been re- interface where the partial pressure of oxygen
ported in the public domain. is low, and then reacts to form volatile metallic
chlorides. As the vapor pressure of the metal-
The recent interest in implementing the next gen- lic chloride products is considerable at the typi-
eration HT/HP recovery boilers in Finland and cal heat transfer surface temperatures involved,
North America has also initiated laboratory-based continuous sublimation occurs, and the gaseous
research and development programs designed metallic chlorides diffuse towards the scale/flue
to identify more corrosion resistant alloys from gas interface. Upon reaching regions with high
which to fabricate superheater tubes [4,37-41], in oxygen (O2) partial pressures, the gaseous metal-
addition to those conducted in Japan [31,35,36]. lic chlorides react with O2/water vapor (H2O) to
An overview of the alloys and testing conditions form solid oxides, releasing Cl2/HCl. The result-

Table 3. Superheater Tube Materials in HT/HP Recovery Boilers in Japan.

Nominal Composition (wt.%)
Grade
Fe Cr Ni Mo Mn Si N C
YUS170 Bal. 23.0-26.0 12.0-16.0 0.5-1.2 < 2.0 < 1.5 0.25-0.40 < 0.06
HR-2EL Bal. 21.0-23.0 12.5-15.5 1.0-2.0 2.5-3.5 < 0.5 0.10-0.25 < 0.02
MN25R Bal. 23.0-26.0 13.0-16.0 0.5-1.2 < 2.0 < 0.7 0.25-0.40 < 0.025

120
Table 4. Laboratory Superheater Tube Corrosion Testing for HT/HP Recovery Boilers.

ing scale is almost detached from the substrate conducted in HP recovery boilers using air-cooled
and offers little protection. Two studies [37,44] probes and laboratory studies conducted under
conclude that, below the FMT, KCl-containing controlled conditions, it remains difficult to di-
recovery boiler deposits are more corrosive than rectly compare the corrosion resistance of can-
NaCl-containing ones, whereas others report the didate tube alloys. Comparisons among the field
opposite result [45,46]. Another controversy in- test results reported are not reliable considering
volves the critical temperature at which active the variable, poorly-defined, albeit real, environ-
oxidation initiates for stainless steel alloys. Based mental conditions encountered within the vari-
on a laboratory test, corrosion of stainless steel by ous boilers under study. Comparisons among the
active oxidation was found to be moderately de- controlled laboratory test results reported are not
pendent on temperature in a range from 550°C to reliable as a consequence of the differences in
650°C, and strongly dependent on temperatures the testing parameters. Among the documented
in excess of 650°C [47]. However, a recent study lab tests, it is difficult to assess, in a quantitative
documenting the corrosion of high alloy super- manner, the relative effect of elevated tempera-
heater tubes in a waste-wood fluidized-bed boiler ture, time at temperature both below and above
shows the critical temperature may be as low as the FMT, and gas composition, for a given deposit
500°C for stainless steel [48]. composition and alloy. In addition, more informa-
tion is required on the possible influence of water
As the literature points out, there are many criti- vapor (H2O) on corrosion both above and below
cal factors that influence the corrosion resistance the FMT. This is important since sootblowing
of superheater tube alloys, both above and below can increase the local H2O concentration by 3.5
the FMT of a typical deposit in a HT/HP recov- vol.% [49]. Dissolved H2O in molten chloride-
ery boiler. In spite of the relevant field studies containing salts has been reported to enhance melt

121
 corrosivity [39,45]. Also, in recently reported ers. Rapid fluxing of the metal surface happens
corrosion tests, water vapor additions (2 vol.%) when the FMT of deposits on the tubes is exceed-
were found to increase the corrosion rate of chro- ed, and is a potential failure mode. Potassium
mium-containing, nickel-based alloys tested in and chlorine are key melting point depressants
a simulated reducing low nitrogen oxide (NOx) familiar to operators of recovery boilers, but some
coal-fired combustion environment [50,51]. This heavy metal chlorides have extremely low FMTs
finding may be relevant in HT/HP recovery boil- and these are readily formed in deposits of boilers
ers under conditions of high carryover. used to burn urban biomass. Corrosion of tubes
in many biomass boiler superheaters has been at-
tributed to attack by alkali chlorides (principally
Superheater Corrosion in KCl), most likely via active oxidation, although as
Biomass-Fired Boilers with recovery boilers, there remains considerable
Although the chemical environment in kraft re- speculation about the details of the mechanism.
covery boilers is much different than in boilers Erosion-corrosion is also a mode of failure, al-
burning other types of biomass fuel, there are though not common in superheater tubes. Silica,
some common issues, and it is instructive to ex- sand, and other hard particulates are common
amine a few of the problems encountered when contaminants brought into the boiler with the fuel.
burning biomass other than black liquor. As it These, along with residual bed media from bub-
is for black liquor recovery boilers, specifica- bling fluid bed or circulating fluidized bed boilers,
tion of superheater materials for biomass boilers can be entrained in the flue gas passing through
is challenging compared to coal and petroleum- the boiler.
fired utility boilers. Principally, this is due to
the variety of biomass available for combustion, A comprehensive body of results from labora-
along with the wide range and nature of the con- tory and field trials was published on the cor-
taminants present in the fuel. While nearly all rosion consequences of firing straw in utility
biomass fuels are vegetative in nature, the term boilers largely based on anticipated increases in
includes such diverse sources as wood and bark, superheater corrosion when utility boilers began
waste agricultural debris, peat, construction widespread use of straw as a fuel in Denmark [53-
waste, solid wastes from effluent treatment plants 56]. Relative to other biomass, straw can contain
and organic refuse diverted from urban landfills, fairly large amounts of potassium and chloride,
as well as black liquor. and it is these elements that were singled out as
contributing to rapid corrosion in straw-fired boil-
Relative to coal and petrochemical-based fuels, ers. Data for alloys ranging from 10CrMo910
biomass fuels are quite heterogeneous in nature ferritic steels to nickel-base A625 are provided
and generally have significant water content in these reports, based on both in situ probe ex-
– higher than 60 wt% in some cases. The ratio posures for as long as 3000 hrs and steam-cooled
of hydrogen to carbon is higher than for coal, and test superheater loops installed in boilers [57-58].
biomass contains more volatile components. On From a materials point of view, a key finding was
average, the distribution of elemental contami- that superheater corrosion rates were found to be
nants in biomass fuels is different than for coal negligible as long as steam temperatures are kept
and includes alkali earth metals such as sodium, below 470°C. This mirrors long-standing practice
calcium and potassium, as well as silicon, chlo- in recovery boilers, and although the chemical
rine and nitrogen. Urban biomass, in particu- environments differ, suggests some common as-
lar, may also contain unusual amounts of heavy pects of how protective oxides are broken down
metals like cadmium, lead, zinc, chromium and in service. An observation arising from the pro-
nickel. A critical difference between biomass and gram was that at higher temperatures, alloys with
other fuels is the sulfur content, or more specifi- intermediate chromium content (between about
cally the sulfur to chlorine ratio in the fuel. With a 16-18 wt% Cr) offered better overall corrosion re-
few exceptions, this ratio is generally much low- sistance than alloys with less or more chromium,
er in biomass fuels than in coal. Properties and although drawing such a conclusion is somewhat
chemical analyses of biomass fuels have been empirical, since the data comes from exposures of
documented by many authors; a very comprehen- a diverse range of commercial ferritic and auste-
sive summary for many forms of natural biomass nitic alloys that differ in many aspects other the
can be found in an NREL-issued report [52]. chromium content.

As with recovery boilers, a number of damage Maintaining a sulfur to chlorine ratio of at least
mechanisms are operative in biomass-fired boil- 2.0 and preferably over 4.0, has been suggested as

122
a means of controlling corrosion in biomass boil- in 31 months. In addition, longitudinal through-
ers by forcing sulfation of alkali chlorides before cracking of the A625 layer on the co-extruded
they condense on tube surfaces [59]. Due to con- tubes was observed after 24 months. Subsequent
straints on sulfur emissions and the need to avoid investigations found tube surface temperatures of
possible downstream dewpoint corrosion, this ap- the affected loops at the steam header to be quite
proach is likely only to be effective when the total variable, and peak values (575-618°C) measured
chlorine content of the fuel is low. Support for this at those points – not at the hot ends of the loops
measure was demonstrated in follow-on studies in – were well in excess of those predicted in the
the Danish program, which looked at the influ- boiler design. It should also be noted that despite
ence on corrosion of co-firing straw and coal in co-burning the contaminated bark with effluent
different proportions. Corrosion rates of TP347 sludge containing large amounts of sulfur, the S:
SS resulting from firing fuel mixtures containing Cl ratio in the fuel would never have reached a
10-20% straw in coal were about the same as fir- value sufficient to completely sulfate the alkali
ing coal, except at higher tube metal temperatures chlorides on formation; in fact deposit analyses
for the 20% straw/coal mixture, where formation found (Na,K)Cl to be as much as 50 wt% of ma-
of a potassium sulfate/iron sulfate eutectic melt ture deposits removed from the surface of tubes.
was suspected of causing corrosion rates to in-
crease [56]. Corrosion of Pulverized Coal-fired
In comparison to straw, wood contains much less Boiler Superheater Tubes
potassium and chloride, and field studies have As is the case with other types of boilers, coal-
demonstrated that superheater corrosion rates fired power plants in the United States are under
are not nearly as rapid in wood-fired boilers [60]. increasing pressure to improve their efficiency
Nonetheless, bark and foliage are enriched in chlo- in order to provide affordable electric power to
rine and potassium, and even small amounts in the a continually growing population. At the same
fuel can eventually lead to significant corrosion time, more stringent air quality regulations require
of superheater tubes. In such cases, mixing the significant reductions in emissions. Consequent-
woody fuel with peat (containing higher amounts ly, fossil fuel-fired power plants are subject to the
of sulphur) proved beneficial in preventing forma- same pressures as other plants - operate at higher
tion of alkali chlorides in the flue gas [61]. steam temperature and pressure to achieve higher
efficiency and reduce emissions of carbon diox-
In some coastal communities, logs are sea-floated ide and other pollutants. The target conditions for
to mills for processing and the bark and sapwood advanced ultra-supercritical (AUSC) coal-fired
from these logs can contain up to 4 wt% chloride power plants are operation at steam temperatures
on a dry basis. When used as a fuel in boilers, as high as 760°C (1400°F) and steam pressures
this contaminated bark has been demonstrated to of at least 35 MPa (5000 psi). Operation at these
cause severe corrosion [48]. This latter example conditions would significantly improve plant effi-
serves as somewhat of a precautionary note, be- ciency and reduce carbon dioxide emissions com-
cause materials selection in the superheater for the pared to conventional pulverized coal-fired power
boiler in question were made with the knowledge plants. However, operation at these conditions
that elevated levels of potassium and chlorine presents a significant challenge to boiler manufac-
were present in the fuel, and that active oxida- turers and alloy developers since improvements in
tion was a likely corrosion mechanism. T22 was high temperature strength and fireside corrosion
chosen for portions of the superheater where tube resistance would be required of the alloys used for
surface temperatures were predicted to be lower superheater tubes.
than 465°C, while SS310H was chosen for hot-
ter loops. The maximum superheater tube surface Increasing steam temperatures would greatly in-
temperature in this boiler was predicted to be crease the efficiency of coal-fired power plants;
about 505°C, well under the supposed safe limit however, such increases in temperature can have
of 550°C for SS310H based on existing literature. other implications, such as adverse effects on
Degradation of the SS310H tubes in the hottest fireside corrosion. The severe fireside wastage
loops was observed after only 18 months opera- seen in pulverized coal-fired boilers is typically
tion and the first tube failure occurred within 36 caused by a deposit-induced liquid phase corro-
months. Damaged SS310H tubes were replaced sion mechanism referred to as coal ash corrosion.
with either weld overlay A625/T22 composite The rate for coal ash corrosion generally increases
tubes or A625/T22 co-extruded tubes. These fared very rapidly with temperature up to about 735°C
little better, with severe corrosion observed with- (1355°F), after which the wastage rates tend to

123
 drop off. For the temperatures expected with ad- will operate at considerably higher temperatures
vanced steam cycles, aggressive fireside deposits, (and pressures) than are anticipated for the most
containing moderate to high concentrations of advanced recovery boilers, strength and creep re-
alkali sulfates and alkali chlorides, are expected sistance, in addition to corrosion resistance, are
to cause rapid metal wastage, especially in units important considerations in the selection of super-
burning high sulfur coal. When high-sulfur, high heater materials. Consequently, the emphasis for
alkali, and high chlorine content coals are used, superheater tube materials has been on advanced
coal ash corrosion is expected to be a widespread austenitic and nickel-based alloys which have
problem for superheater and reheater tubes, and a been developed to have much improved strength
critical problem that needs to be resolved before at the desired operating temperature.
AUSC boilers can be deployed.

Coal ash corrosion is caused by the formation of Recent Studies of Recovery Boiler
molten alkali iron trisulfates on superheater and Superheater Corrosion
reheater tube surfaces [62-66]. Reid [65] and Since superheater tubes in black liquor recovery
Kihara [67] have described the process by which boilers are not subjected to the mechanical loads
coal ash corrosion occurs as beginning with the anticipated for AUSC, there are differences in
deposition of alkali sulfates on the leading sides the materials requirements. Both types of boilers
of superheater tubes. Because of the increasing along with many biomass fired boilers have the
temperature gradient, the outermost material be- potential for subjecting the superheater tubes to
comes sticky and captures particles of fly ash. molten phases that can serve as a flux that pro-
Potassium and sodium sulfates are concentrated motes loss of the protective oxide. Two fairly
at the metal surface along the edge of the deposit recent studies have investigated the performance
band where metal temperatures are typically in of selected alloys both above and below the FMT
the range of about 595-740°C (1100-1360°F). of superheater deposits through laboratory stud-
The sulfur compounds in the ash will form SO2 ies and exposure of corrosion probes in operating
by thermal dissociation and catalysis of this SO2 recovery boilers [40,41,68].
by the Fe2O3 will produce SO3. Reaction of SO3
with metallic oxides and alkali sulfates will result These studies (summarized in Table 4) had many
in the formation of molten alkali-iron trisulfates at differences but also notable similarities in part
the metal interface through the reaction: because of efforts at collaboration between the re-
spective researchers. For instance, both projects
Fe2O3 + 3(K,Na)2SO4 + 3SO3 = evaluated six alloys, three of which were common
2(K,Na)3Fe(SO4)3 to both studies. The laboratory studies for both
projects involved exposing metallic samples to a
These molten alkali sulfates flux the protective selection of salts using exposure temperatures both
oxide species from the surface thus accelerating above and below the FMT of the salts. The Finn-
oxidation and sulfidation of the alloy. Wastage of ish study used six different salt mixtures while the
the superheater tubes is influenced by both tem- North American study used only one. On the oth-
perature and the composition, so the traditional er hand, the Finnish study used air as an oxidizing
approaches to minimizing coal ash corrosion in- cover gas while the North American study used
clude using coals with lower alkali and sulfur con- two different gases - nitrogen as a relatively inert
tent, limiting steam temperature, improving boiler gas and a reducing mixture containing hydrogen
design, using more corrosion resistant alloys and sulfide. The Finnish study was conducted at six
providing protective baffles around the tubes in temperatures ranging from 340 to 600°C while
locations that make them the most susceptible to the North American study only considered three
corrosion. temperatures over a much more narrow range -
510 to 560°C. Both studies were conducted for
As far as the selection of corrosion resistant al- relatively short times - the Finnish study lasted for
loys, the general rule is that the higher the chro- one week (168 hours) while the North American
mium content the more resistant the alloy. For study was carried out for two weeks (336 hours).
mildly corrosive coal ash environments, type Both studies considered the thickness of the cor-
347 stainless steel has been found to provide ad- rosion product layer as a measure of the extent of
equate corrosion resistance. However, for more reaction and from that estimated an annual cor-
corrosive environments, more corrosion resist- rosion rate. The North American study also con-
ant alloys with chromium content in the 22-27% sidered weight change and used elemental analy-
range would be required. Because AUSC boilers sis of cross sections to assess changes in surface

124
composition as a result of exposure to corrosive tack is clearly seen on these samples.
environments. Consequently, the two studies pro-
duced some information that was complementary
but also some that was unique. Summary
The three types of boilers discussed in this paper,
Both studies also utilized field exposures of corro- chemical recovery boilers, biomass-fired boilers
sion probes in operating recovery boilers. In both and pulverized coal-fired boilers, all have limita-
studies, air cooling was used in an effort to devel- tions on their operating temperatures and pres-
op a temperature gradient along the probe so that sures because of issues with degradation of high
there were samples exposed to temperatures both temperature components, particularly superheater
below and above the first melting temperature tubes. Corrosion by a molten phase that forms on
that was measured for deposits recovered from the tubes and fluxes the normally protective oxide
the respective boilers. Two probes were exposed is a common problem as well as corrosion by gas-
in the Finnish study; each for a little less than eous species that typically contain sulfur and/or
700 hours. The North American study success- chlorine compounds.
fully exposed only one probe, but it was exposed
for 1000 hours. Both studies utilized advanced For service below the FMT of the corrosive phase,
analysis techniques to examine samples exposed the active oxidation mechanism is apparently op-
in these probes. The Finnish studies utilized a erative, but using alloys with higher chromium
scanning electron microscope to examine cross concentrations generally provides some solution
sections of their probe samples while the North to the corrosion issue.
American study examined their cross sectioned
samples with an electron microprobe. For service above the FMT, more highly alloyed
metals are required but use of alloys with higher
An example of results from the North American chromium content does not necessarily reduce the
laboratory corrosion studies is given in Fig. 1 corrosion rate.
which shows the measured weight loss of sam-
ples as a function of temperature. The samples Additional, more extensive studies, both labora-
were embedded in a salt mixture with a FMT of tory and field, are needed to gain a better under-
525°C and a reducing cover gas containing hy- standing of the variables that affect superheater
drogen sulfide. Exposure temperatures were 510 tube corrosion and to better determine the best
and 560°C, but no T91 sample was exposed at the means to control this corrosion to ultimately per-
higher temperature. The samples were exposed mit operation of recovery boilers at higher tem-
for 336 hours, and these results show higher cor- peratures and pressures.
rosion rates above the FMT but increasing the
chromium content decreases the corrosion rate
for these conditions. Figure 2 shows the effect of
the cover gas on the calculated corrosion rate for
samples embedded in salt and exposed at 510°C
(below the FMT) for 336 hours. It is clear that the
reducing, hydrogen sulfide containing cover gas
leads to significantly higher corrosion rates for
nearly all the samples.

For samples exposed on a corrosion probe in the

superheater area of an operating recovery boiler,
the extent of attack was measured on samples
exposed above and below the FMT. The results
shown in Fig. 3 indicate that the below the FMT
the total depth of attack decreases with increas-
ing chromium content. However, above the FMT
the depth of attack does not correlate with the
chromium content. Clearly other factors play a
significant role in developing an alloy’s corrosion
resistance. Micrographs of cross-sections of the
five samples exposed at the highest temperatures
are shown in Fig. 4. The extent of subsurface at-

125
 Acknowledgements (1981).
The authors appreciate the efforts of reviewers S. 10. H.N. Tran, D.W. Reeve, D. Barham, “Local
J. Pawel and B. A. Pint. Research was sponsored Reducing Atmosphere – A Cause of Superheater
by the U.S. Department of Energy, Industrial Corrosion in Kraft Recovery Units,” Tappi Jour-
Technologies Program, under Contract DE-AC05- nal, 68(6): 102 (1985).
00OR22725 with UT-Battelle, LLC. By accepting 11. R. Backman, M. Hupa, E. Uppstu, “Fouling
the article for publication, the publisher acknowl- and Corrosion Mechanism in the Recovery Boiler
edges that the United States Government retains Superheater Area,” Tappi Journal, 70(6): 123
a non-exclusive, paid-up, irrevocable, worldwide (1987).
license to publish or reproduce the published form 12. C. Dees, J. Simonen, H. Tran, “Experience of
of this work, or allow others to do so, for United Recovery Boiler Superheater Corrosion at Wil-
States Government purposes. lamette Hawesville, 7th International Symposium
on Corrosion in the Pulp and Paper Industry, TAP-
PI Press, Atlanta, GA: 251 (1992).
References 13. W. Blakemore, M. Ellis, H. Tran, “Experi-
1. T.N. Adams, W.J. Frederick, T.M. Grace, M.
ence of Recovery Boiler Superheater Corrosion at
Hupa, K. Iisa, A.K. Jones, H. Tran, Editors, Kraft
Carter Holt Harvey Kinleith,” 11th International
Recovery Boilers, TAPPI Press, Atlanta, GA: p.
Symposium on Corrosion in the Pulp and Paper
310 (1997).
Industry,” TAPPI Press, Atlanta, GA: 217 (2004).
2. D.C. Crowe, W.C. Youngblood, “Recovery
14. R. Backman, M. Hupa, P. Hyoty, “Corrosion
Boiler Superheater Corrosion,” 9th International
due to Acidic Sulphates in Sulphate and Sodium
Symposium on Corrosion in the Pulp and Paper
Sulphite Recovery Boilers,” Pulp and Paper In-
Industry, CPPA, Montreal, QC: 225 (1998).
dustry Corrosion Problems, Volume 4, Swed-
3. F. Bruno, “Thermochemical Aspects on Chlo-
ish Corrosion Institute, Stockholm, Sweden: 76
ride Corrosion in Kraft Recovery Boilers,” Paper
(1983).
No. 01426, CORROSION/2001, NACE, Hou-
15. T.N. Adams, W.J. Frederick, T.M. Grace, M.
ston, TX (2001).
Hupa, K. Iisa, A.K. Jones, H. Tran, Editors., Kraft
4. M. Mäkipää, E, Kauppinen, T. Lind, J.
Recovery Boilers, TAPPI Press, Atlanta, GA: p.
Pyykönen, J. Jokiniemi, P. McKeough, M. Oksa,
247 (1997).
Th. Malkow, R.J. Fordham, D. Baxter, L. Koivis-
16. S.C. Stultz, J.B. Kitto, Editors, Steam/Its Gen-
to, K. Saviharju, E. Vakkilainen, “Superheater
eration and Use, 40th Edition, Babcock & Wilcox
Tube Corrosion in Recovery Boilers,” 10th Inter-
Company, Barberton, OH: 26-10 (1992).
national Symposium on Corrosion in the Pulp and
17. M.P. LeBel, “CRU Superheater Experience,”
Paper Industry, Volume 1, VTT, Espoo Finland:
Alstom Power Inc., Correspondence to Paprican,
157 (2001).
May 11, 2005.
5. J. Tuiremo, K. Salmenoja, “Control of Super-
18. S.C. Stultz, J.B. Kitto, Editors, Steam/Its Gen-
heater Corrosion in Black Liquor Recovery Boil-
eration and Use, 40th Edition, Babcock & Wilcox
ers” 10th International Symposium on Corrosion
Company, Barberton, OH: 6-21 (1992).
in the Pulp and Paper Industry, Volume 1, VTT,
19. G.Y. Lai, A. Wensley “Metallurgical Charac-
Espoo Finland: 181 (2001).
teristics and Performance Experience of Spiral
6. H. Tran, M. Gonsko, X. Mao, “Effect of Com-
Overlay tubes in Black Liquor Recovery Boilers,”
position on the First Melting Temperature of Fire-
Paper No. 138, 2005 TAPPI Engineering, Pulping
side Deposits in Recovery Boilers,” Tappi Jour-
and Environmental Conference, TAPPI Press, At-
nal, 82(9): 93 (1999).
lanta, GA (2005).
7. K. Salmenoja, “Field and Laboratory Stud-
20. J.F. La Fond, A. Verloop, A.R. Walsh, “Engi-
ies on Chlorine-Induced Superheater Corrosion
neering Analysis of Recovery Boiler Superheater
in Boilers Fired with Biofuels,” Report 00-01,
Corrosion,” Tappi Journal, 75(6): 101 (1992).
Faculty of Chemical Engineering, Abo Akademi,
21. F. Bruno, “The Significance of Superheater
Turku, Finland (2000).
Tube Corrosion on the Steam Temperature of
8. K.W. Morris, A.L. Plumley, W.R. Roczniak,
Kraft Recovery Boilers,” 8th International Sympo-
“The Effect of Chlorides on Recovery Unit Super-
sium on Corrosion in the Pulp and Paper Industry,
heater Wastage: An R&D Progress Report,” Pulp
Swedish Corrosion Institute, Stockholm, Sweden:
and Paper Industry Corrosion Problems, Volume
138 (1995).
3, NACE, Houston, TX: 47 (1980).
22. D.N. French, Metallurgical Failures in Fossil
9. D.W. Reeve, H.N. Tran, D. Barham, “Super-
Fired Boilers, John Wiley & Sons, New York, NY:
heater Fireside Deposits and Corrosion in Kraft
248 (1983).
Recovery Boilers,” Tappi Journal, 64(5): 109
23. P-E. Ahlers, A. Jaakola, “Experiences of

126
Austenitic Superheater Tube Bends in Finnish 37. M. Mäkipää, M. Oksa, L. Koivisto, “Super-
Recovery Boilers, Black Liquor Recovery Boiler heater Tube Corrosion in Changing Operation
Symposium 1982, Ekono Oy/Finnish Recovery Conditions of Recovery Boilers,” CORRO-
Boiler Committee/Finnish Energy Economy As- SION/2004, Paper No. 01424, NACE, Houston
sociation, Helsinki, Finland: D7-1 (1982). TX (2001).
24. A. Jaakola, T. Rose, “Operational Experiences 38. R. Backman, M. Huppa, J. Lagerborn, T. Lep-
of Austenitic Superheater Tube Bends in Recov- istö, “Influence of Partially Molten Alkali Depos-
ery Boilers,” Pulp and Paper Industry Corrosion its on Superheater Corrosion in Black Liquor Re-
Problems, Volume 4, Swedish Corrosion Institute, covery Boilers,” 10th International Symposium on
Stockholm, Sweden: 82 (1983). Corrosion in the Pulp and Paper Industry, Volume
25. C. Reid, “Lower Bend Superheater Tube Ex- 1, VTT, Espoo, Finland: 137 (2001).
perience in Coastal Recovery Boilers,” Acuren, 39. M. Oksa, M. Mäkipää, “Comparative Labora-
Correspondence to FPInnovations – Paprican, tory Testing of Molten Salt Attack of Superheater
November 5, 2001. Tube Materials,” 10th International Symposium on
26. J. Kish, Unpublished Data, FPInnovations Corrosion in the Pulp and Paper Industry, Volume
– Paprican, May 8 (2008). 2, VTT, Espoo, Finland: 523 (2001).
27. “Sandvik Sanicro 28 – Composite Tube for 40. J.R. Keiser, J.R. Kish, L.A. Frederick, A.W.
Steam Boiler Application,” S-12111-ENG, Sand- Willoughby, K.A. Choudhury, D.L. Singbeil, F.R.
vik Steel, Sandviken, Sweden (1999). Jetté, J.P. Gorog, “Recovery Boiler Superheater
28. B.K. Trask, “Superheater Corrosion at Scott Probe Studies,” 2007 International Chemical Re-
Maritimes Limited,” CPPA Atlantic Branch Con- covery Conference, Part 1, PAPTAC, Montreal,
ference 1992, CPPA, Montreal QC (1992). QC, p. 285 (2007).
29. J. McMillan, “Superheater Problems, Their 41. B-J. Skrifvars, L. Silvander, M. Hupa, K. Sal-
Causes and Solutions,” 2004 International Chem- menoja, E. Vakkilainen, “Laboratory Studies of
ical Recovery Conference, Volume 2, TAPPI the Corrosion of Superheater Materials Covered
Press, Atlanta, GA: 953 (2004). with Alkali Salt Mixtures,” 2007 International
30. T. Odelstam, “Material Evaluation for BLRB Chemical Recovery Conference, Part 1, PAPTAC,
Superheater Tubes,” Black Liquor Recovery Boil- Montreal, QC, p. 295 (2007).
er Symposium 1982, Ekono Oy/Finnish Recovery 42. H.J. Grabke, E. Reese, M. Spiegel, “The Ef-
Boiler Committee/Finnish Energy Economy As- fects of Chlorine, Hydrogen Chloride and Sulphur
sociation, Helsinki, Finland: D6-1 (1982). Dioxide in the Oxidation of Steels below Depos-
31. A. Fujisaki, H. Takatsuka, M. Yamamura, its,” Corrosion Science, 37: 1023 (1995).
“World’s Largest High Pressure and Temperature 43. R.W. Byers, “Fireside Slagging, Fouling and
Recovery Boiler,” 1992 International Chemical High Temperature Corrosion of Heat Transfer
Recovery Conference, TAPPI Press, Atlanta, GA: Surface due to Impurities in Steam-Raising Fu-
1 (1992). els,” Prog. Energy Combust. Sci., 22: 29 (1996).
32. A. Fujisaki, H. Takatsuka, M. Yamamura, 44. T. Nishio, H. Matsumoto, M. Shinohara, Y.
“World’s Largest High Pressure and Temperature Arkawa, “Influence of Ash Composition on the
Recovery Boiler,” Pulp & Paper Canada, 95(11): Corrosion Rate of Superheater Tubes,” TAPPI
T452 (1994). 2001 Engineering, Finishing and Converting
33. H. Tran, Y. Arakawa, “Recovery Boiler Tech- Conference, TAPPI Press: CD-ROM (2001).
nology in Japan,” TAPPI 2001 Engineering/Fin- 45. M. Mäkipää, T. Malkow, D.J. Baxter, “The
ishing and Converting Conference, TAPPI Press, Effect of Salt Composition on the Chlorine Cor-
Atlanta, GA: 49 (2001). rosion of Low Alloy Steels,” CORROSION/2001,
34. Y. Arakawa, Y. Taguchi, T. Meda, Y. Bara, Paper No. 01186, NACE, Houston, TX (2001).
“Experience with High Pressure and High Tem- 46. S.C. Cha, M. Spiegel, “Local Reactions be-
perature Recovery Boilers for Two Decades,” tween NaCl and KCl Particles and Metal Surfac-
Pulp & Paper Canada, 106(12): T269 (2005). es,” Corrosion Engineering, Science & Technol-
35. H. Matsumoto, A. Notomi, T. Nishio, “Ad- ogy, 40(3): 249 (2005).
vanced Technology for Corrosion Resistant Ma- 47. Y. Shinata, F. Takahashi, K. Hashiura, “NaCl-
terials for Recovery Boilers,” 1998 International Induced Hot Corrosion of Stainless Steels,” Mate-
Chemical Recovery Conference, TAPPI Press, rials Science and Engineering, 87: 399 (1987).
Atlanta, GA: 51 (1998). 48. J.R. Kish, C. Reid, D.L. Singbeil, R. Seguin,
36. H. Fujikawa, H. Makiura, Y. Nishiyama, “Corrosion of High Alloy Superheater Tubes
„Corrosion Behaviour of Various Steels in Black in a Coastal Biomass Power Boiler,” CORRO-
Liquor Recovery Boiler Environment,” Materials SION/2007, Paper No. 07349. NACE, Houston,
and Corrosion, 50: 154 (1999). TX (2007).

127
 49. H. Tran, N. Mapara, D. Barham, “The Ef- 58. M Montgomery, A Karlsson and O.H. Larsen,
fect of H2O on Acidic Sulfate Corrosion in Kraft Materials & Corrosion, 53, 121-131 (2002).
Recovery Boilers,” Tappi Journal, 79(11): 155 59. K. Salmenoja and K. Makela, Paper 00238,
(1996). CORROSION/2000.
50. L. Paul, G. Clark, M. Eckhardt, “Laboratory 60. A. Nafari and A. Nylund, Materials and Corro-
and Field Corrosion Performance of a High Chro- sion 55 909-920 (2004).
mium Alloy for Protection of Waterwall Tubes 61. M. Orjala, R. Ingalsuo, K. Paakkinen, J.
from Corrosion in Low NOx Coal Fired Boilers,” Hamalainen, M. Makipaa, M. Oksa, T. Malkow,
CORROSION/2006, Paper No. 06473, NACE, R.J. Fordham and D. Baxter, 10th International
Houston, TX (2006). Symposium on Corrosion in the Pulp and Paper
51. S.D. Kiser, E.B. Hinshaw, T. Orsini, “Ex- Industry, Vol 1, 117-136, (2001).
tending the Life of Fossil Fired Boiler Tubing 62. V. Viswanathan, R. Purgert, and P. Rawls,
with Cladding of Nickel Based Alloy Materials,” “Coal-Fired Power Materials, Part II”, Advanced
CORROSION/2006, Paper No. 06474, NACE, Materials & Processes, September, 2008, Vol 166,
Houston, TX (2006). No. 8, pgs. 41-45.
52. “The Behavior of Inorganic Material in Bio- 63. W. Nelson and C. Cain, Jr., “Corrosion of Su-
mass-Fired Power Boilers--Field and Laboratory perheater and Reheaters of Pulverized Coal Fired
Experiences: Volume II of Alkali Deposits Found Boilers”, Transactions of the ASME, Journal of
in Biomass Power Plants”, L.L. Baxter, T.R. Engineering for Power, July 1960, pp. 194-204.
Miles, T.R. Miles Jr., B.M. Jenkins, D.C. Dayton, 64. W.T. Reid, “Formation of Alkali Iron Trisul-
T.A. Milne, R.W. Bryers, L.L. Oden, NREL/TP- fates and Other Compounds Causing Corrosion in
433-8142, 1996]. Boilers and Gas Turbines”, Project Review July 1,
53. F.J. Frandsen, Fuel 84, 1277-1294, (2005). 1966, June 30, 1968, prepared by Battelle Memo-
54. L.A. Hansen, H.P. Nielsen, F.J. Frandsen, K. rial Institute, Columbus, OH, June, 1968.
Dam-Johansen, S. Horlyck and A. Karlsson, Fuel 65. W.T. Reid, External Corrosion and Depos-
Processing Technology 64, 189-209, (2000). its: Boilers and Gas Turbines, American Elsevier
55. M. Montgomery, A. Karlsson and O.H. Larsen, Publishing Company, New York, 1974.
Materials & Corrosion, 53, 121-131 (2002). 66. G.J. Hills, “Corrosion of Metals by Molten
56. M. Montgomery, O.H. Larsen and O. Biede, Salts”, Proceedings of the Marchwood Confer-
CORROSION/2003, paper 03356 (2003). ence: Mechanism of Corrosion by Fuel Impuri-
57. M Montgomery and A Karlsson, Materials ties, Johnson and Littler, eds., Butterworths, Lon-
and Corrosion 50, 579-584 (1999). don, 1963.

510 °C 560 °C

80

70

60

50

40

30

20

10

0
T91 347H 310H A lloy 28 A lloy 33 A lloy 693

A lloy

Figure 1. Weight loss measured after exposure to the synthetic salt mixture with the reducing cover gas (H2S-
H2-H2O-N2) for 336 hours at 510°C (below FMT) and 560°C (above FMT) [40].

128
67. S. Kihara, K. Nakagawa, A. Ohtomo, H. Aoki 68. J.R. Keiser, J. R. Kish, P.M. Singh, G.B.
and S. Ando, “Simulating Test Results for Fire- Sarma, J. Yuan, J. P. Gorog, L.A. Frederick, F.R.
side Corrosion of Superheater & Reheater Tubes Jetté, R.A. Meisner and D.L. Singbeil, Final
Operating at Advanced Steam Conditions in Coal- Technical Report, Materials for Industrial Heat
Fired Boilers”, High Temperature Corrosion in Recovery Systems, Tasks 3 and 4, Materials for
Energy Systems, TMS/AIME, M. Rothman, ed, Heat Recovery in Black Liquor Recovery Boilers,
1984, pp. 361-376. December 31, 2007

N2 AVE H2S-H2-H2O-N2

1.6

1.4

1.2

0.8

0.6

0.4

0.2

0
T91 347H 310H Alloy 28 Alloy 33

A llo y

Figure 2. Annual corrosion rate (extrapolated from 336 hour weight loss) for selected alloys immersed in the
synthetic salt mix and exposed at 510°C to N2 cover gas or to the reducing gas mixture H2S-H2-H2O-N2.

Scale Internal Total A verage T

120 700
A bove FM T Below FM T
100 600

80 500

60 400

40 300

20 200

0 100

Alloy

Figure 3. Depth of attack (measured from micrographs) for samples from the corrosion probe that was
exposed in a recovery boiler for 1000 hr. The estimated average exposure temperature is indicated, as
squares, showing whether samples were above or below the FMT [40,68]

129
 Type 347 stainless steel Type 310H stainless steel

Sanicro 28 alloy Alloy 33 weld overlay

Alloy 693

Figure 4. Micrographs from cross-sections of the five hottest samples on the corrosion probe showing the
outer surface on the “bottom” side [68].

130
ECONOMIZER LEAK STUDY

THOMAS J. GRANT, THOMAS M. GRACE

AND JACK CLEMENT
AMERICAN FOREST & PAPER ASSOCIATION

Introduction

The AF&PA Recovery Boiler Committee looked operating sources. Sorting techniques were ap-
at the number of economizer pressure part fail- plied to the data to identify the types of failures,
ures in the recovery boilers. Economizer failures their frequency and how these failures relate to
remain a significant cause of boiler downtime in the design and operating variables. It was appar-
the industry. No recovery boiler explosions have ent that economizer design and construction play
been directly attributed to economizer failures. a critical role in economizer integrity. The inves-
Yet, there is a risk of an explosion in those econo- tigators had follow-up meetings with suppliers to
mizer arrangements where water from an econo- ascertain design details and discuss the available
mizer failure could enter the furnace cavity. Not all failures information. The overall objective of the
economizer failures have been reported to BLR- study was to understand the causes of recovery
BAC. Thus the magnitude of the problem may be boiler economizer failures and to identify means
underestimated. Although the root causes of some for preventing their occurrence. Five specific ob-
economizer pressure part failures are known, the jectives were met:
information has not been widely disseminated. No - Document and evaluate the industry experi-
systematic study of economizers has been made. ence;
The AF&PA Recovery Boiler Committee realized - Categorize the types of failures that have oc-
the need for an industry-wide effort to review curred;
economizer tube failures, to collect and analyze - Determine the correlation of failures with de-
available information on economizers. Dr. Tom sign and operating factors;
Grace and Mr. Jack Clement were contracted to - Provide Guidelines for specification and con-
study this problem. struction of economizers, and
- Provide a reference for operating and mainte-
Study nance practices impacting economizers.
Data on economizer failures were collected from
operating mills, recovery boiler suppliers and Findings
BLRBAC records. A total of 346 economizers In analyzing the data, it was apparent that a
were included in the study of which 266 were number of specific types of failures occurred fre-
located in the US and 80 in Canada. Data were quently. Due to the lack of detailed information
also collected from Sweden (1998 to 2001) from in a number of incidents, it was difficult to know
41 incidents involving 12 different economizers, the extent of any given operation. It was even less
and from Finland (2000 – 2004). Because of the clear how operating practices affect stress and
brevity of these reports, they were not used in the fatigue cracking problems. The frequency and
study. nature of the economizer failures were classified
into a relatively small number of categories:
The data were reviewed, interpreted and ana- - stress or fatigue cracks
lyzed. The investigators met with suppliers and o cracks in the tube – header attachments re-
some operating companies to resolve some dis- gion
crepancies. Additional data were obtained from o cracks in the fin terminations
detailed incident reports, equipment suppliers and o cracks at other locations

131
 - handhole and radiography test plug seal weld leaks in a given area that were directly connected
leaks with previous leaks in that area. These might take
- poor quality welds, other than those above the form of a leak in a repair weld for a previ-
- external tube thinning ous leak, a leak in the seal weld of a tube plug,
- internal corrosion in thinning or a leak in a thinned area where a previous leak
- mechanical damage sprayed on adjacent tubes. Lack of access for
- leaks at rolled joints. making effective repairs and doing proper inspec-
tions around failure sites is a critical factor in this
The most widespread problem was stress and fa- regard. The most serious problem in economizer
tigue cracks near the tube-header interface. There today is cracks near the tube-to-header welds,
are indications that some of these cracks originate most commonly at the feedwater inlet headers and
on the waterside, but it is unclear as to what extent the economizer discharge headers.
water-side conditions play in this role. Although
it was difficult to prove, some operation condi- Acknowledgments
tions are clearly detrimental. A listing of operating I would like to thank Dr. Grace and Mr. Clement
variables that could potentially affect economizer for the work they have done in this project and
integrity was developed. As a result of providing helping me in the presentation of this study.
a listing or description of operating practices that
are detrimental to the economizer, it became ap-
parent that repair and maintenance practices were
also vitally involved. Thus, a stand-alone docu-
ment “Operating and Maintenance Practices Im-
pacting Economizers” was produced.

Design details and experience information were

supplied by suppliers. The investigators were able
to show some conclusions from this information
and the reports of economizer failures. For exam-
ple, economizers installed before 1990 indicated
many more leaks and about half of these leaks
were attributed to handhole seal weld leaks and
radiography hole plug leaks. Nearly 300 leaks oc-
curred in the 62 economizers installed between
1980 and 1990. In the period from 1990 to 2004
where 70 economizers were installed, there were
45 leaks.

Experience shows that economizers can be oper-

ated for over 20 years without experiencing tube
leaks if they are properly designed, constructed,
operated and maintained. Design evolution has
had a very important influence on economizer re-
liability. There has been a significant reduction in
the frequency of economizer leaks since the adop-
tion of the mini-header design by the suppliers in
1990 A very small number of economizers have
been responsible for most of the reported leaks
with the mini-header design. With the change in
design and the guidelines written to consolidate
the best experience, and the application of these
principles can result economizers in place to oper-
ate many years without unplanned outages. Weld
quality is a critical aspect of getting a reliable
economizer that will last.

It was common to see a pattern of economizer

132
THE CHANGING SCOPE OF BLACK LIQUOR
GASIFICATION
KEVIN WHITTY
THE UNIVERSITY OF UTAH

Introduction

Gasification of black liquor has long been recog- gasification development, taking into considera-
nized as an interesting alternative to the conven- tion the technical and economic motivations be-
tional recovery boiler. Proponents of black liquor hind various technologies. Black liquor gasifica-
gasification point out its many potential advantag- tion is also reevaluated in the context of today’s
es, including improved overall energy efficiency, market, which has very different priorities from
enablement of advanced pulping technologies, those of even ten years ago, with regard to techni-
opportunities for product diversification, poten- cal understanding, economics, energy production
tially lower capital costs and avoidance of the and environmental consciousness.
smelt-water explosion risk inherent with recovery
boilers. Over the past few decades, millions of
dollars have been spent developing black liquor Changing Drivers for Black Liquor
gasification technology at industrial R&D cent- Gasification Development
ers, national laboratories and universities. Yet, Development of the first pyrolysis- and gasifi-
despite seemingly strong drivers and years of de- cation-based alternative recovery processes in
velopment, only two relatively small black liquor the 1960’s and 1970’s was driven primarily by
gasifiers are in commercial operation today, and a desire for processes that were safer and more
no mills have an integrated gasification system environmentally sound than the recovery boiler.
producing biofuels or biopower. Recovery boiler explosions were an uncomfort-
ably common occurrence during that time, with
There are many reasons that black liquor gasifi- roughly 3-4 occurrences per year in North Amer-
cation technology has not achieved commercial ica [1], and likely twice that number world-wide.
success. Some are technical, of course. Black The recommended practices for safe operation
liquor is a challenging fuel and gasification, even and shutdown that are commonplace today were
for conventional fuels like coal, is still not well- not available at that time. So, there was strong
established. Advances in recovery boiler design interest in developing safer recovery systems with
have also made that technology more efficient no risk for explosion. Many proposed systems
and competitive. Financial considerations have used fluidized beds to process black liquor since
significantly limited development. Bringing a this eliminated the presence of molten smelt.
new technology to market is expensive, and few
companies have had the resources or fortitude to Improving environmental performance was anoth-
progress beyond small-scale testing. Additionally, er driver for development of alternative recovery
end users are reluctant to adopt a new, unproven systems in the 1960’s and 1970’s. In particular,
technology so key to successful operation of the there was interest in minimizing the smell associ-
mill unless they can be sure that its availability ated with kraft pulping. New gasification-based
and performance are at least on par with existing technologies focused on eliminating “malodor-
technology. Yet another challenge for black liquor ous compounds” associated with recovery boilers
gasification is shifting priorities within the pulp were proposed [2].
and paper industry and global energy markets.
In the 1980’s, gas turbine technology had ad-
This chapter explores the history of black liquor vanced to the point where coal-based integrated

133
 gasification combined cycle (IGCC) power plants through gasification of black liquor is not a new
were identified as being potentially superior to idea [3]. But changes in policy, environmental
conventional steam turbine-based plants. It was consciousness and instability in the cost and avail-
soon recognized that similar improvements in en- ability of petroleum have made production of bio-
ergy efficiency could result if black liquor were fuels from domestic, renewable, carbon-neutral
processed in a gasifier integrated into a combined- feedstocks an attractive option. This has provided
cycle power generation system. It was estimated a new driver for development and commercializa-
that a so-called BLGCC (black liquor gasifica- tion of gasification technology.
tion-combined cycle) could increase electrical
production efficiency by as much as ten percent- Clearly, since the 1960’s incentives behind de-
age points without affecting steam production. velopment of black liquor gasification technol-
Spurred by the potential energy and cost savings, ogy have shifted as markets and priorities have
a surge of interest in black liquor gasification fol- changed. Yet, the drivers for black liquor gasifi-
lowed. The period from 1985 to 1995 was the cation development are not mutually exclusive. A
most active in the history of black liquor gasifica- fully integrated black liquor and biomass gasifi-
tion technology development. cation-based mill would be able to use advanced
pulping technologies that take advantage of the
The natural separation of sodium and sulfur that sodium-sulfur split while simultaneously produc-
occurs during black liquor gasification gave rise ing biofuels and/or biopower, all the while offer-
to a new set of drivers for the technology in the ing a more energy efficient, more environmentally
1990’s. Unlike a kraft recovery boiler, where friendly alternative to the recovery boiler. The
sulfur remains bound to sodium as sulfide or sul- key is to have a viable black liquor gasification
fate, most of the sulfur in a gasifier is converted technology. The following sections review devel-
to gaseous hydrogen sulfide. The sulfur can be opment of gasification technology and summarize
recovered from the gas and used at targeted loca- the state of black liquor gasification technology
tions in the pulping process. Not only does this today.
improve control of the sodium-sulfur balance, but
it enables advanced pulping technologies such as
split-sulfidity pulping, which can improve both History of Black Liquor
pulp yield and pulp quality. In short, integrated Gasification Development
black liquor gasification creates opportunities to A recent review of alternative recovery technolo-
increase production of a mill’s primary product, gies identified over twenty serious efforts to de-
pulp, thereby improving revenues. This contin- velop black liquor gasification systems over the
ues to be a strong driver for gasification, and one past 40 years [4]. The technologies are about
which is specific to the pulp and paper industry. evenly split between low temperature fluidized
bed-based systems and high temperature systems
In recent years, the role envisioned for black liq- in which smelt leaves the gasifier in molten form.
uor gasification has changed significantly in both Notable technologies which were developed at
Europe and North America. Worldwide, there has least to the pilot stage are described below.
been a push to limit emissions of carbon dioxide,
which is seen as a key contributor to global warm- The Copeland Recovery Process
ing. In the European Union countries, policy as- The Copeland process (Fig. 1) was one of the most
sociated with this environmental consciousness successful alternative recovery systems of the past
has resulted in real economic incentive to reduce half century, with several installations worldwide.
fossil fuel-based CO2 emissions via the cap-and- The process was developed by George Copeland
trade system for carbon dioxide. Similar policies during the 1960’s and 1970’s as a safer alternative
to put a price on CO2 emissions are expected in to the recovery boiler and as an alternative to pro-
the U.S. and Canada in the near future. Renewa- vide incremental black liquor processing capacity
ble, carbon-neutral biomass-based fuels are an at- to a mill [5-8]. The process involved spraying liq-
tractive alternative, and can be produced through uor onto a wide, bubbling fluidized bed comprised
gasification of black liquor and biomass. The of inert material and black liquor inorganics and
synthesis gas resulting from gasification is rich fluidized with preheated air. The bed temperature
in hydrogen and carbon monoxide, which can be was relatively low, in the range 650-750°C. Most
catalytically converted to a variety of biofuels, installations were at non-kraft mills, although a
including methanol, dimethyl ether, and Fischer- relatively large (300 tpd) system for processing
Tropsch fuels. Producing transportation fuels kraft liquor was installed in Canada. Typically,

134
the system was operated at or near stoichiomet- The SCA-Billerud Process
ric conditions, so that it behaved as a black liquor SCA-Billerud of Sweden developed a relatively
combustor. Later development focused on operat- successful alternative recovery process in the
ing the system under sub-stoichiometric (gasifica- 1960’s and 70’s [9-11]. The process (Fig. 2) in-
tion) conditions and combusting the product gas volved spraying black liquor into the top of a re-
in either a dedicated boiler or by co-firing in an actor heated by an oil or gas burner, which caused
existing boiler. Ultimately, the Copeland proc- the liquor to dry and pyrolyze. The temperature
ess lost favor as improvements in recovery boiler was kept low enough to avoid melting the ash in
technology and a push towards higher energy ef- the liquor. The char flowed with the exiting gas
ficiencies made the process uncompetitive. through a heat recovery boiler and into a spray
scrubber, where it was captured by the liquid and
filtered to recover green liquor. The unburned
char was mixed with fresh black liquor and recy-
cled to the process. The pyrolysis gas was cooled
to condense out water and burned in a boiler to
produce steam. A 25 ton/day pilot system for
kraft liquor was operated in the early 1970’s. The
first commercial plant was installed in Sundsvall,
Sweden in 1968 and several other plants were sub-
sequently installed in Japan and the United States
at mills as large as 1000 tons pulp/day. However,
technical challenges, including buildup of ash and
char in the gas passages, as well as relatively low
efficiency compared to conventional technology
caused interest in the SCA-Billerud process to
evaporate by the mid-1970’s and no further plants
were ever constructed.

The Direct Alkali Recovery

System (DARS) Process
The DARS process was similar to the Copeland
process in that black liquor was combusted in a
Figure 1. The Copeland Recovery Process [6].
fluidized bed reactor. Unlike the Copeland proc-

Figure 2. SCA-Billerud process.

135
 ess, however, the DARS process involved addi- mid-1970’s, but no commercial system was ever
tion of iron oxide to the black liquor to achieve constructed. Eventually, the hydropyrolysis proc-
in-situ causticizing via formation and subsequent ess was deemed uncompetitive and development
leaching of of sodium ferrite [12]. The largest was terminated. Interestingly, about the same
DARS process was constructed in 1986 by As- time that St. Regis was developing hydropyroly-
sociated Pulp and Paper Mills at their Burnie, sis, Texaco Corporation was doing in-house work
Tasmania soda process pulp mill, and processed on a very similar process, which they described
roughly 150 tds/day of black liquor. Problems as- as a coking process [18,19]. Texaco was granted
sociated with fluidization and sodium ferrite dust over 15 patents relating to their process, but no
formation were encountered during the first few pilot or commercial plant was ever constructed.
years of operation. Nonetheless, the system was
operated until the mill was shut down in the late
1990’s [13,14]. VTT’s Circulating Fluidized Bed
Gasification Process
The Technical Research Center of Finland (VTT)
The St. Regis Hydropyrolysis developed a pressurized air/oxygen-blown cir-
Process culating fluidized bed gasifier between 1988 and
During the 1960’s, St. Regis Corporation devel- 1993 [20,21]. The process operated at roughly
oped a novel system for processing black liquor 650°C and at pressures to 35 atm. A 200 kW de-
(Fig. 3). The process involved heating low-solids velopment gasifier processing roughly 1.5 tds/day
black liquor to roughly 330°C at approximately of liquor was constructed at a mill in Äänekoski,
200 atm pressure to produce combustible gases, Finland and was tested for several years. Ag-
a low-ash char and an alkali salt solution similar glomeration of the bed was problematic, but was
to green liquor [15-17]. The char was washed improved by using alumina particles as bed mate-
and could be used as activated carbon or fuel for rial. No commercial plant was ever built based on
a boiler. The gases evolved during the process the VTT process, and development was ultimately
were burned to provide heat for the system. St. terminated due to technical difficulties and a chal-
Regis constructed a 16 tds/day pilot plant in the lenging economic climate.

BL
HE AT E R

HY DR O-
P Y R OL Y S IS
R E AC T OR
P R E HE AT E R /
P R ODUC T C OOL E R
B L AC K
L IQUOR
P UMP S
P R ODUC T
C OOL E R
F L UE
G AS
F L AS H T ANK
P R E S S UR E
L E T DOWN

S O2
S C R UB B E R

C HAR
F IL T E R #1
C HAR
F IL T E R
#2

Figure 3. Diagram of the St. Regis hydropyrolysis process. [17]

136
ABB’s Circulating Fluidized Bed cyclone gasifier dubbed the NSP, or “Ny Sodahus
Process” [25]. The cyclone gasifier operated in
Gasification Process a manner similar to a cyclone combustor used in
ABB developed a circulating fluidized bed gasifier
coal power plants, except that only about 90% of
similar to the VTT system during the early 1990’s
the air required for complete combustion was fed
[22,23]. A 2.5 tds/day development gasifier was
through the gasifier (Fig. 4). Additional air was
constructed at ABB’s research facility in Västerås,
fed into the main part of the boiler system. The
Sweden and ran many successful, albeit short, test
temperature at the exit of the gasifier ranged from
campaigns over the course of several years. Dur-
1100 to 1400°C. An attractive feature of this tech-
ing later years of development, ABB recognized
nology was that existing recovery boilers could be
that addition of titanate to the black liquor offered
rebuilt to have several cyclone gasifiers around the
the potential to implement in-situ causticization
perimeter. Char and smelt from the black liquor
of the black liquor. It also increased the melting
would be forced to the walls of the gasifier due to
temperature of the bed solids, allowing operation
the strong swirling action, so the exiting fuel gas
at higher temperatures [24]. ABB was able to op-
would be relatively free of particles. The molten
erate the gasifier at 680°C without problem and
slag was tapped from the low end of the cyclone
achieved good carbon conversion. ABB contin-
and fed to a dissolving tank.
ued development until 1997, when the program
was abandoned due to shifting development pri-
In the early 1980’s a 100 tds/day NSP gasifier was
orities and lack of a clear market for the gasifier in
installed at a mill in Karlsborg, Sweden. During
a reasonable time frame.
the two years the system operated, it experienced
severe technical problems such as cracking of the
The Swedish NSP Cyclone Gasifier steam-cooled cyclone shell [26]. Ultimately, the
In the early 1970’s, a team of Swedish and Finn- NSP program was terminated due to overwhelm-
ish companies, research institutions and forest ing technical challenges and lack of continued
products organizations jointly developed a unique funding.

Figure 4. NSP cyclone gasifier. Adapted from [27].

137
 The Champion-Rockwell Molten development funding ran out and development of
the Champion-Rockwell gasifier was postponed
Salt Gasification Process indefinitely.
The Champion-Rockwell molten salt gasifier
(Fig. 5) was originally developed by Rockwell
Corporation to process fuels such as coal. In the MTCI/TRI’s Fluidized Bed Steam
early 1980’s researchers at Champion Paper Cor-
poration recognized the potential for black liquor
Reforming Process
Manufacturing and Technology Conversion In-
gasification and several years of development
ternational (MTCI) began developing a bubbling
ensued [27]. The Champion-Rockwell black liq-
fluidized bed steam reformer (gasifier) in the
uor gasifier was a counter-flow, refractory-lined
mid-1980’s [30,31]. The black liquor gasification
entrained-flow reactor which contained a pool of
technology is now under license to ThermoChem
molten smelt at the bottom. Black liquor was in-
Recovery International (TRI), which continued
jected at the top and flowed downwards, undergo-
development until recently. The process (Fig. 6)
ing pyrolysis and gasification as it reacted with
involves injecting black liquor into a bubbling flu-
the upwards-flowing air. The resulting producer
idized bed in which steam is the fluidizing and re-
gas exited near the top of the gasifier and the char
acting gas. The overall steam reforming process
from the black liquor fell into the pool of smelt,
is endothermic, so it is necessary to provide heat
where conversion was completed and sulfate re-
to the system. In the MTCI/TRI system, natural
duction took place. The gasifier was hottest at
gas or syngas is combusted in a proprietary pulsed
the bottom, where the temperature was roughly
combustor and the hot exhaust gases pass through
950°C. Several tests were performed using Rock-
bundles of tubes immersed in the fluidized bed.
well’s 6 ton/day pilot gasifier [28]. Carbon con-
The overall process is energy positive and pro-
version was good and sulfate reduction in excess
duces a gas containing 65-70% hydrogen.
of 95% could be achieved. Plans were developed
for a larger, pressurized pilot gasifier [29]. But,

Figure 5. Champion-Rockwell pilot gasifier. [30]

138
 Figure 6. Conceptual MTCI system configuration for kraft liquor. [32]

In the late 1980’s, a 6 tds/day test gasifier was technology was owned by Kamyr and eventually
constructed at a recycled paper mill in Ontario, Kvaerner. The initial concept involved plasma
California and processed paper mill sludge. Kraft gasification, but the very high energy require-
black liquor from another mill was also processed ments necessary for the plasma guns turned favor
in this system during dedicated test campaigns. towards partial oxidation [34]. In the Chemrec
In the mid-1990’s, a 50 tds/day kraft black liquor process (Fig. 7), black liquor and either air or
gasification demonstration system was construct- oxygen are fed into a burner at the top of a re-
ed at Weyerhaeuser’s New Bern, North Carolina fractory-lined reactor. Only 40-50% of the total
mill [32]. This system was operated for a couple oxygen needed for combustion is fed, resulting
years, ultimately achieving 500 hours of continu- in an overall reducing (gasifying) atmosphere.
ous operation, after which it was shut down and
dismantled. In 2003, a 100 tds/day MTCI gasifier
was installed at a Norampac mill in Ontario, Can-
ada, and continues to process all the mill’s liquor.
In 2004, a 200 tds/day demonstration system hav-
ing two parallel gasifiers was started up at Geor-
gia-Pacific’s Big Island, Virginia mill [33]. This
system experienced technical problems with the
fluidized beds, pulsed heater combustors and gas
cleaning system. Ultimately, Georgia-Pacific’s
management determined that the system would
not be able to achieve the necessary availability
and the gasification system was replaced with
conventional technology.

Chemrec’s Entrained-Flow
Gasification Process
The Swedish company Chemrec began develop-
ment of its high temperature entrained-flow gasi-
fication technology in the mid-1980’s, when the Figure 7. Chemrec gasifier.

139
 The liquor pyrolyzes and is gasified at roughly of operation were associated with frequent shut-
950°C, producing a low to medium heating value downs resulting from failure of different parts
syngas and molten smelt. The products are cooled of the system, especially auxiliary systems, but
and separated in a quench system below the reac- performance and availability have improved sig-
tor. The smelt dissolves in a quench bath, forming nificantly [35]. The main challenge with the gasi-
green liquor. fier itself was refractory degradation. Refractory
lifetimes were on the order of 6-8 months and the
Chemrec has two variants of their technology, a system had to be shut down for several weeks
low pressure air-blown system intended to pro- each time the refractory had to be replaced. Over
vide incremental capacity and a high pressure the past decade, however, research has resulted in
oxygen-blown system targeted as an alternative much improved refractory materials installation
to the recovery boiler. Chemrec’s first pilot plant designs. Consequently, the most recent materials
was a 3 tds/day air-blown, atmospheric pressure used to line the gasifier have lasted more than two
system constructed in Hofors, Sweden in 1987. years and the gasifier is now a vital component
This plant operated well enough to prove the of the mill’s recovery system. Chemrec also has
technical feasibility of entrained-flow gasification a very active research program centered on their
and a commercial 75 tds/day plant was installed in 20 tds/day pressurized pilot plant in Piteå, Swe-
Frövi, Sweden in 1991. Five years later, Weyer- den and is in discussion with various clients about
haeuser constructed a 330 tds/day system in New commercial installations of this technology.
Bern, North Carolina [35]. Development of the
pressurized system began in 1994 with a 6 tds/day The other commercial gasifier in operation is a
air-blown pilot plant in Skoghall, Sweden. This 100 ton/day MTCI/TRI fluidized bed steam re-
was rebuilt to a 10 tds/day oxygen-blown sys- former at a Norampac mill in Ontario, Canada
tem in 1997, which was operated until 2000 [36]. which started up in 2003. The gasifier processes
Based on experience from these systems, a new all the black liquor from this mill, which uses a
20 tds/day pressurized oxygen-blown pilot was carbonate pulping process. The process has a
constructed in Piteå, Sweden in 2005. This sys- very simple configuration, almost ideal for this
tem has logged more than 8000 hours of operating technology, which avoids several challenges with
time and has been key to Chemrec’s technology low temperature black liquor gasification. Black
development [37]. liquor is fed to a single steam-blown fluidized
bed that is heated by pulsed heaters firing natural
gas. The raw syngas from the system is kept hot
Other Processes and fed directly into a boiler for steam produc-
In addition to the processes described above, tion. Because the syngas is not cooled, there is
many more processes were proposed and tested little opportunity for tars to condense before the
at laboratory or bench scale, but never advanced gas is combusted in the boiler. Indeed, the lack
to pilot scale. These include Paprican’s “atomiza- of a syngas cleaning system simplifies the process
tion suspension technique” (AST) process (1955- enormously and removes a potential location for
1968) [38], the University of California pyroly- operational problems. Using natural gas rather
sis-gasification-combustion process (1967-1980) than product gas from the gasification process
[39], Weyerhaeuser’s fluidized bed dry pyrolysis also improves availability of the system. The No-
process (1972-1978) [40], Tampella’s pressu- rampac gasifier continues to operate. However,
rized entrained-flow gasifier (1988-1993) [41], due largely to technical challenges and termina-
Ahlstrom’s suspension gasifier (1988-1993) [42], tion of the Georgia-Pacific Big Island demonstra-
Babcock & Wilcox’s fluidized bed gasifier (1993- tion system mentioned earlier, TRI has placed
1998) [43], Noell’s entrained-flow gasifier (1996- development of their black liquor gasifier on hold
2000) [44] and KBR’s transport gasifier (1999- and is focusing their efforts on processing more
2002) [45]. conventional biomass materials.

Current Status of Black Liquor New Opportunities for Black

Gasification Liquor Gasification
Today, there are two commercial black liquor As indicated earlier, incentives for development
gasifiers in operation. One is the air-blown, of black liquor gasification technology have var-
low pressure 330 ton/day Chemrec booster gasi- ied over the past several decades and today we are
fier processing kraft liquor at Weyerhauser’s New again witnessing a shift in thinking about black
Bern, North Carolina mill. The first few years

140
liquor gasification. The high price of oil and ing offered commercially, marking a milestone in
transportation fuels in recent years has made peo- gasification development. The intense interest in
ple think seriously about alternative fuels for the production of non-food crop-based biofuels which
first time in decades. When this occurred in the has arisen from concern over global warming and
1970’s, there was significant investment in tech- energy security has produced a strong market for
nologies for extraction of oil from oil shale and products which can be produced through black
oil (tar) sands in North America. However, today liquor gasification. This in turn has sparked the
there is a heightened environmental conscious- interest of investors. So, for the first time in the
ness regarding use of fossil fuels and the damag- history of black liquor gasification development,
ing influence of carbon dioxide and related global three key elements for successful commerciali-
warming. The only renewable source from which zation are in place: (1) a feasible, demonstrated
one can produce replacement transportation fuels technology, (2) a strong market and (3) financial
is biomass. In the U.S. and elsewhere, interest support.
and investment in corn-based ethanol exploded
in recent years. Similar growth has occurred Within the next few years we will likely see the
with biodiesel derived from natural oils, particu- first integrated biorefineries incorporating black
larly soybean oil. Unfortunately, this growth had liquor gasification for thermochemical production
the unintended consequence that the cost of the of biofuels. Undoubtedly, these first-of-a-kind
feedstock materials, corn and soybeans, spiked plants will experience technical problems, if not
to historical highs. The high cost and reduced with the gasifier then with some other part of the
availability of these crops as foodstuffs severely process. The ultimate success of black liquor gas-
affected regions of the world, particularly third- ification technology will depend on how quickly
world countries, which are struggling to feed their any technical difficulties can be overcome. If
populations. There is now significant interest in these challenges can be resolved quickly and the
production of biofuels from biomass which does biorefineries operate successfully, then the pros-
not involve displacement of food or cropland oth- pects for black liquor gasification are good. If,
erwise dedicated to food production. Production however, these first plants are plagued with insur-
of biofuels from forest-based biomass is an obvi- mountable technical problems then black liquor
ous solution. gasification’s reputation will suffer by association,
even if the problems are unrelated to the gasifier.
Among forest-based biomass, black liquor offers
the best opportunity for biofuel production in the No matter what happens over the next few years,
near term. It contains the most energy-rich com- it is certain that in the long run priorities will con-
ponent in the wood, lignin, and has little product tinue to shift, technology will continue to advance
value on its own. Black liquor is readily avail- and the economics of papermaking and energy
able, is very reactive so it can be processed at will continue to evolve. In tandem with those
relatively low temperature, and it can be pumped changes, development of black liquor gasifica-
into pressurized systems. High pressure, high tion technology will undoubtedly also continue as
temperature gasification technology is also com- long as there is black liquor available.
mercially available. In comparison, handling of
conventional biomass is more challenging and no
pressurized biomass gasifier suitable for integra-
tion into a thermochemical biofuel production fa-
cility is currently commercially available.

The Future of Black Liquor

Gasification
The future for black liquor gasification looks
more promising today than ever. Currently, it
appears that high temperature, air/oxygen-blown
entrained-flow gasification has emerged as the
technology of choice. This is due to the relatively
simple design, good conversion efficiency, low tar
production and thousands of hours of successful
operation. Entrained-flow technology is now be-

141
 References rolysis recovery process. Tappi 56(8):100 - 103
1. Recommended Good Practice. Guidelines for (1973).
Post-ESP Procedures for Black Liquor Recovery 16. Miller, R.L., The St. Regis hydropyroly-
Boilers. The Black Liquor Recovery Boiler Advi- sis process. New Recovery Processes for Black
sory Committee (2002). Liquor. Proceedings from the Gunnar Sundblad
2. Feuerstein, D. L., Thomas, J. F., and Brink, D. Seminar, 6 7 May 1976, Skövde, Sweden, pp. 36
L., Malodorous products from the combustion of - 62 (1976).
kraft black liquor. Tappi 50(6):258 (1967). 17. Watkins, J.J. and Timpe, W.G., Pyrolysis of
3. Prahacs, S., Production of methanol from the spent pulping liquors. U.S. Patent No. 4,208,245
organic materials present in spent pulping liquors, (1980).
with the simultaneous recovery of the pulping 18. Hess, H.V. and Cole, E.L., Treatment of waste
base and sulphur values present in said liquors. liquors from pulp and paper mills. U.S. Patent No.
Canadian Patent No. 733546 (1966). 3,558,426 (1971).
4. Whitty, K. and Verrill, C.L., A Historical Look 19. Hess, H.V., Franz, W.F. and Cole, E.L., Liquid
at the Development of Alternative Black Liquor phase coking of spent kraft pulping liquors. U.S.
Recovery Technologies and the Evolution of Patent No. 4,067,767 (1978).
Black Liquor Gasifier Designs. 2004 Int’l Chemi- 20. Saviharju, K., Black liquor gasification: Re-
cal Recovery Conf., 6-10 June 2004, Charleston, sults from laboratory research and test rigs. Biore-
SC, USA (2004). source Tech. 46(1-2):145 - 151 (1993).
5. Kraft recovery boiler faces new challenge. Can. 21. McKeough, P., Research on black liquor con-
Chem. Processing 51(12):54 - 56 (1967). version at the technical research center of Finland.
6. Copeland, G.G., The Copeland process fluid Bioresource Tech. 46(1-2):135 - 143 (1993).
bed system & pollution control worldwide. Proc. 22. Tanca, M.C., Dahlquist, E.G. and Flink, S.,
1969 Purdue Industrial Waste Conference, pp. CFB black liquor gasification system operating at
1017 - 1023 (1969). low pressures. U.S. Patent No. 5,425,850 (1995).
7. Evans, J.C.W., Many process applications are 23. Dahlquist, E. and Jacobs, R., Development of
appearing for fluid-bed reactors. Pulp & Paper a dry black liquor gasification process. Proceed-
49(4):82 - 85 (1975). ings 1992 Int’l Chem. Recovery Conf., Seattle,
8. Copeland, G.G. and Hanway, J.E. “Fluidized WA, USA, pp. 457 - 471 (1992).
bed oxidation of waste liquors resulting from the 24. Dahlquist, E. and Ringvall, T., Presentation of
digestion of cellulosic materials for paper making. a dry black liquor gasification process with direct
U.S. Patent No. 3,309,262 (1967). causticization developed by ABB. Olle Lindström
9. Horntvedt, E., The SCA-Billerud Recovery Symposium (1997).
Process. Proc. Symp. Recovery of Pulping Chem- 25. Magnusson, H. and Warnqvist, B., The NSP
icals, 13-17 May 1968, Helsinki, Finland (1968). project: An alternative to the conventional recov-
10. Maloney, J.D., An SCA-Billerud Kraft Chem- ery furnace. Chem. Eng. Progress 76(2):54 - 56
ical Recovery Process. Forum on Kraft Recovery (1980).
Alternatives, 29-30 April 1976, Institute of Paper 26. Lekander, K.-E., Warnqvist, B., Roos, T.
Chemistry, Appleton, Wisconsin, pp. 182 - 204 and Björklund, H., Ny SodahusProcess (NSP)
(1976). Slutrapport. (NSP Process Final Report) In Swed-
11. Horntvedt, E., SCA-Billerud pyrolysis proc- ish (1985).
ess in kraft pulping. New Recovery Processes for 27. Kelleher, E.G., Feasibility of black liquor gas-
Black Liquor. Proceedings from the Gunnar Sun- ification in combined cycle generation. Final re-
dblad Seminar, 6 7 May 1976, Skövde, Sweden, port – Phase I. U.S. DOE Report DOE/CS/40341
pp. 132 - 138 (1976). T4 (1983).
12. Covey, G.H., Development of the direct alkali 28. Kelleher, E.G., Gasification of kraft black liq-
recovery system and potential applications. Pulp uor and use of the products in combined cycle co-
Paper Can. 83(12):T350 - T354 (1982). generation. Final report – Phase II. U.S. DOE Re-
13. Maddern, K.N., Mill-scale development of the port DOE/CS/40341 T5 (DE86003647) (1985).
DARS direct causticization process. Pulp Paper 29. Kelleher, E.G., Conceptual design of a black
Can. 87(10):T395 - T399 (1986). liquor gasification pilot plant. U.S. DOE Report
14. Scott-Young, R.E. and Cukier, M., Commer- DOE/CS/40341-T6 (1987).
cial Development of the DARS Process. Proc. 30. Durai-Swamy, K., Mansour, M.N. and Warren,
1995 Int’l Chem. Recovery Conf., Toronto, Can- D.W., Pulsed combustion process for black liquor
ada (1995). gasification. U.S. DOE Report DOE/CE/40893-
15. Timpe, W.G. and Evers, W.J., The hydropy- T1 (DE92003672) (1991).

142
31. Mansour, M.N., Steedman, W.G., Durai- oxygen-blown gasification. Tappi J. 80(12):111 -
Swamy, K., Kazares, R.E. and Raman, T.V., 116 (1997).
Chemical and energy recovery from black liquor 45. Hood, K.T. and Henningsen, G.B., Black liq-
by steam reforming. Proc. 1992 Int’l Chem. Re- uor gasification… Past experience can create the
covery Conf., 7 11 June 1992, Seattle, WA, pp. roadmap of the future pathway. Presented at Pulp
473 - 475 (1992). and Paper International Congress, 23 26 October
32. PulseEnhanced™ steam reforming and recov- 2000, Sao Paulo, Brazil (2000).
ery of kraft spent liquor. Weyerhaeuser New Bern
demonstration. MTCI (1997).
33. Decarrera, R., Demonstration of black liquor
gasification at Big Island. Final Technical Re-
port. U.S. DOE Report 40850R24 for project DE-
FC22-01NT40850 (2007).
34. Kignell, J-E., Process for chemicals and en-
ergy recovery from waste liquors. U.S. Patent No.
4,808,264 (1989).
35. Brown, C. and Landälv, I., The Chemrec black
liquor recovery technology — a status report.
Proc. 2001 Int’l Chem. Recovery Conf., 11 - 14
June 2001, Whistler, Canada (2001).
36. Whitty, K. And Nilsson, A., Experience from a
high temperature, pressurized black liquor gasifi-
cation pilot plant. Proc. 2001 Int’l Chem. Recov-
ery Conf., 11 - 14 June 2001, Whistler, Canada
(2001).
37. High Temperature Black Liquor Gasifica-
tion. Synthesis report 2007-2008. Chemrec AB
(2008).
38. Prahacs, S., Pyrolytic gasification of Na –,
Ca–, and Mg– base spent pulping liquors in an
AST reactor. in Fuel Gasification, R. Gould edi-
tor, Am. Chem. Soc., pp. 230 - 252 (1967).
39. Brink, D.L., Faltico, G.W. and Thomas, J.F.,
The University of California pyrolysis-gasifica-
tion-combustion process. Forum on Kraft Recov-
ery Alternatives, 29-30 April 1976, Institute of
Paper Chemistry, Appleton, Wisconsin, pp. 148
- 181 (1976).
40. Dehaas, G.G., Hurley, P.J. and Root, D.F., Dry
pyrolysis approach to chemical recovery. Forum
on Kraft Recovery Alternatives, 29-30 April 1976,
Institute of Paper Chemistry, Appleton, Wiscon-
sin, pp. 103 - 123 (1976).
41. Salmenoja, K., Black liquor gasification:
theoretical and experimental studies. Bioresource
Tech. 46(1-2):167 - 171 (1993).
42. Saviharju, K., Black liquor gasification: Re-
sults from laboratory research and test rigs. Biore-
source Tech. 46(1-2):145 - 151 (1993).
43. High-solids black liquor firing in pulp and
paper industry kraft recovery boilers. Phase 1a:
Final report. Low temperature black liquor gasi-
fier evaluation. U.S.DOE report DOE/ER/10002-
T1 (1997).
44. Lorson, H., Schingnitz, M,. White, V. and
Dean, D.R., Black liquor recovery by pressurized,

143
144
BIOREFINERIES FOR CO-PRODUCTION OF
RENEWABLE TRANSPORTATION FUELS
W. J. FREDERICK, JR.
NATIONAL RENEWABLE ENERGY LABORATORY
GOLDEN, CO, USA

Introduction

As the 21st century proceeds, civilization faces fuels from forest biomass. We first consider al-
perhaps its greatest challenges yet: ternatives for producing transportation fuels from
• A steady population growth, moving toward 9 lignocellulosic biomass. We then consider how to
billion people or more worldwide by 2050, integrate fuels production into the forest products
• An expectation for an increasing standard of industry, and consider examples where this is be-
living in many countries, including the most ing done.
populated ones,
• An increasing appetite for raw materials, en-
ergy, and fresh water, and Options for Renewable Transpor-
• A finite or declining resource base on planet tation Fuels
earth. Many fuel and energy alternatives are being con-
sidered today for renewable transportation. Figure
The concepts of renewability and sustainability 1 shows some of the alternative routes to transpor-
have become recognized as essential if we are to tation fuels from lignocellulosic biomass. Produc-
manage these seemingly conflicting challenges, tion of biofuels from ligno-cellulosic biomass has
not only for ourselves, but for future generations not yet become a significant source of transporta-
as well. tion fuels, although the potential great. Estimates
are that 25% of EU’s transportation fuels needs
Industries with renewable resources and the ca- and 30% of the U.S’s could be met by renewable
pability to convert their raw materials sustainably fuels by 2030.
to valuable products are in a strong position to
benefit from this capability. One of the few indus- What Transportation Fuels should
tries that is well positioned to provide a range of
products, both new and traditional, from renew- be Produced?
able resources is the Forest Products Industry. Op- It can be argued that electrical vehicles will ulti-
portunities abound for new materials, chemicals, mately replace liquid fuel-powered vehicles [1,2].
fuels, and energy as well as for conventional pa- However, the time frame for electrical vehicles to
per, board, and consumer products. replace liquid fuel-powered vehicles would de-
pend on several factors including improvements
One specific opportunity is fuels from forest bio- in electric vehicle technology, phase-out of the
mass. existing transportation vehicle fleet, and the avail-
ability of electrical power to meet transportation
Currently there is enormous interest in producing needs. 25 years is likely the minimum time frame
renewable fuels to offset and eventually replace for electrical vehicles to gain a dominant position
petroleum-derived fuels for transportation [1,2,3]. if they indeed do that. In this scenario, liquid fu-
There is also a growing interest to produce renew- els must satisfy transportation needs. Which fuel
able materials from sustainably grown biomass or fuels will succeed depends on several factors,
[1,2,3]. including
• Energy density
In this presentation, we focus on transportation • Distribution system options

145
 Figure 1, Alternative routes to biomass-derived transportation fuels [7].

• Environmental impact (LCA) t/yr of biomass is available from U.S. forests, and
• Local temperature another 194 million t/a is available from agricul-
• Engine technology advances tural resources. Biomass availability was project-
ed to increase to 1.3 billion t/a by 2030, enough
to generate sufficient biofuel to replace 30% of
Availability of Biomass Resources the current petroleum-derived transportation fuel.
In Europe: In 2006, Ericsson and Nilsson as- The increase in biomass availability from 2005
sessed the potential biomass supply in 27 coun- to 2030 was expected to come entirely from an
tries in Europe[1]. They found that the potential increase in agricultural crop production includ-
supply of biomass energy amounts to up to 17.2 ing energy crops; forest residues were projected
EJ/a, about 27% of the overall energy supply in to remain constant over this time period. This
fifteen western European countries (the EU15) underestimates the potential for increasing forest
in 2001. They concluded that the biomass sup- productivity.
ply target of 5.6 EJ/a in the EU15 by 2010, set
by the European Commission in 1997, was clearly A breakdown of the 358 million t/yr of forest resi-
achievable. dues identified in the Billion Ton Biomass study
is included in Table 2. 39% of the total is either
In the United States: 1n 2005, Perlack and co- black liquor or mill residues. Most of these two
workers at Oak Ridge National Laboratory pub- potential feedstock are consumed on site by pulp
lished an extensive analysis of the availability of & paper mills and wood processing plants to pro-
cellulosic biomass in the U.S[1]. This report is duce power and process steam. The remaining
often referred to as “The Billion Ton Biomass Re- 61% are largely available although the cost of col-
port.” It concluded that, as of 2005, 368 million

Table 1, Estimated future biomass resource base for the U.S.[11]

146
Table 2, Estimated availability of biomass residue from U.S. forests by source.[11]

Table 3, Value of lumber, pulp fiber, and ethanol in the U.S. per unit of product and per tonne of wood con-
sumed in producing them.

lecting and transporting unmerchantable timber Cellulosic Biorefineries Worldwide

and harvesting residues from forests has so far A number of cellulosic biofuel installations are
limited their widespread use as biomass feedstock being constructed or planned around the world,
for fuels or power generation in the U.S. but no production-scale facilities are operating
yet. In the U.S., thirteen companies have received
The Integrated Biorefinery U.S. Department of Energy backing to install cel-
lulosic biofuel plants. Their locations are shown
Concept in Table 4. Six of these plants will use forest bio-
One of the principles of sustainability is that raw
mass as feedstock. Two of them, New Page and
materials should be utilized first for the highest
Flambeau River, both in Wisconsin, will produce
value products, with residual raw material used
both fiber and biofuels.
for other, less valuable materials. Table 3 shows
typical values for lumber, bleached softwood
A partnership between two leading U.S. compa-
pulp, and ethanol, both per unit of material of fuel
nies in energy (Chevron Corporation) and forest
and per metric ton consumed to produce them.
products (Weyerhaeuser Company), announced in
As expected, the order of value is lumber > pulp
early 2008, promises to lead to major biofuels pro-
fiber > ethanol > power. For any reasonable com-
duction capability. The new company, Catchlight
bination of prices for these four commodities, the
Energy LLC was created to develop and demon-
ranking of their value per ton of wood consumed
strate novel technologies for converting cellulose
does not change.
and lignin from a variety of sources into biofuels.
Weyerhaeuser Company plans to use its forest-
This comparison suggests that, from a global eco-
land to co-produce energy crops and trees, thus
nomic standpoint, it would always makes better
supplying biomass feedstock. Chevron brings the
economic sense to capture products from wood
capability for commercialization of those tech-
from highest value to lowest, i.e. to use high qual-
nologies and operation of the resulting biofuels
ity bole wood first for lumber or other solid wood
production plants.
products. The residue from lumber production
should be used to produce fiber. Any remaining
Two major biorefinery efforts are under way in
residues from lumber manufacture or pulping
Finland. UPM-Kymmene one of the world’s larg-
should be converted to liquid fuels such as etha-
est pulp and paper manufacturing companies, an-
nol, and then to electrical power. Please note that
nounced that they will invest in biodiesel plants,
this is not a new philosophical argument. It has
which will be built in connection with some of
been the practice of the Forest Products Industry
their pulp and paper mills. Their biodiesel tech-
for well over a century.
nology will be based on Finnish syngas technol-
ogy, where wood biomass is gasified and then
syngas is converted to biodiesel.

147
 Table 4, Locations of biofuels plants in progress in the U.S., supported in part by the U.S. Department of
Energy.

Thermochemical Conversion
Range Fuels, Inc., Soperton, GA
Lignol (KS)
Flambeau River (WI)
NewPage (WI)

Biochemical Conversion
Abengoa Bioenergy, Southwestern KS
BlueFire Ethanol, Inc., Corona, CA
Poet, Emmetsburg, IA
Mascoma (MI)
RSE (ME)
Pacific Ethanol (OR)
ICM (MO)
Altech (KY)
Verenium (LA)

NSE Biofuels Oy Ltd. will build a demonstration or low and medium pressure steam demand in
plant in Varkaus, Finland, to develop processes to evaporators through better water management in
produce renewable biofuels from wood residues. pulp mills. Major, capital intensive improvements
NSE Biofuels Oy Ltd. is a joint venture between may be required, from upgrading to high solids
Stora Enso Oyj and Neste Oil Corporation. concentrators for black liquor to installing inte-
grated combined-cycle power plants for hog fuel
In Sweden, SunPine AB is building a produc- or black liquor. To be successful in an integrated
tion plant in Piteå, Sweden to convert crude tall biorefinery where liquid fuels and/or power are
oil (CTO) either to “crude tall diesel” for subse- co-products, the entire complex must be very
quent hydroprocessing at a refinery into a renew- efficient with regard to energy management and
able diesel component or, via purification, into conservation. Finland has a significant advantage
standard EN14214 biodiesel. The plant will have in this area, marketing high efficiency recovery
a capacity of up to 100,000 m3 (26 million gal) of boilers, and having led the way to highly energy
crude tall diesel per year. Construction began in efficient kraft pulp mills.
September, 2008 and production is scheduled to
begin in September, 2009. The second task is to identify opportunities to in-
tegrate fuels or energy production with fiber pro-
Also in Sweden, CHEMREC AB continues devel- duction. This is illustrated in Figures 2 and 3. Fig-
opment of its pressurized black liquor gasification ure 2 shows how energy flows between the three
technology, with expectations to produce clean major components of a pulp and paper complex.
syngas for production of biofuels. Wood is the feedstock for both fiber and energy
production. Energy is produced from black liquor,
Renewable Fuels: How the Forest bark, and forest residues (lumped here with wood
or with residual biomass). The Combined Heat
and Forest Products Industry Can and Power (CHP) plant utilizes these fuels to pro-
Contribute and Benefit duce power for internal use and export, and steam
The Forest Products Industry can contribute to to meet process demands.
and benefit from the growth of a renewable fuels
industry in several ways. An important first task The diagram in Figure 3 illustrates an integrated
is to improve the efficiency of biomass utilization forest biorefinery that includes a biofuels plant as
in existing pulp and paper mills. Efficiency im- well as a pulp and paper mill and a CHP plant.
provements include improving biomass conver- The main difference between the diagrams in Fig-
sion efficiency in recovery boilers and hog fuel ures 2 and 3 is that, in Figure 3, some or all of the
boilers. It may include reducing steam demand at residual biomass from the pulp mill is converted
all pressure levels, for example by reducing soot to biofuels in the fuels synthesis plant in Figure
blowing steam requirements in recovery boilers 3 before it reaches the CHP plant. More biomass

148
Figure 2, Conceptual diagram of a stand-alone pulp and paper mill.

Figure 3, Conceptual diagram of an integrated forest biorefinery that includes a pulp and paper mill, a fuels
plant, and a CHP plant.

(wood) input is likely required when both fiber In this example, integration between the ethanol
and biofuels are produced. plant and pulp mill includes a common CHP plant
that provides steam to both plants, and utilization
A third task is to utilize forestlands more effec- of residual organic from the ethanol plant as fuel
tively to produce biomass for biorefinery feed- in the CHP plant. The pulp mill has a conventional
stock. One possibility is to collect more of the Tomlinson recovery boiler and steam turbine for
forest residues as feedstock, and to deliver them power generation, with process steam extraction
in a cost-effective way to the biorefinery. A sec- at three pressures. The CHP plant consists of
ond approach is to utilize forestlands to grow trees the recovery boiler, a hog fuel boiler that burns
and energy crops simultaneously, as is being un- biomass residues from the pulp mill and ethanol
dertaken by Weyerhaeuser Company in their col- plant, and the steam turbines. The process steam
laboration with Chevron Corporation. production is balanced with the total demand from
the integrated biorefinery. To balance the steam
Energy and Economic demand and production, either unneeded biomass
residue from the integrated complex is exported
Comparisons for Alternative as biofuel or additional fuel (biomass residue) is
Forest Biorefineries purchased.
Let us now compare the energy requirements
and energy output as products from two different A block diagram representing the stand-alone
types of biorefineries with those from convention- pulp mill is shown in Figure 4. The energy value
al kraft pulp mills. of each input and output stream, in millions of
Btu/hr, is indicated next to each process stream.
Integrated Ethanol Plant This particular mill produces 1398 t/d of fiber
In the first case, we compare an integrated bi- with an energy value of 737 million Btu/hr. It also
orefinery that consists of a kraft pulp mill and generates a net of 19.3 MW (66 million Btu/hr) of
integrated ethanol plant with a stand-alone pulp exportable electrical power.
mill [17]. The pulp production is the same from
both plants, but they produce different quantities An integrated biorefinery that produces ethanol
of power, and consume different amounts of bio- as well as fiber and power is shown in Figure 5.
mass and other energy supplies. The integrated bi- The fiber production rate is exactly the same as
orefinery produces ethanol from sugars extracted in Figure 4, but the input rates of wood, hog fuel,
from wood prior to pulping, fiber, and power. and natural gas have increased. The energy values

149
 Figure 4, A block diagram representing energy inputs and outputs to/from a kraft pulp mill. The energy
inputs and outputs are indicated next to each process stream.

Figure 5, A block diagram representing an integrated forest biorefinery that produces ethanol and power as
well as cellulose fiber.

Table 5, Energy balances and incremental efficiency for fuel and power generation for a biorefinery that
consists of an ethanol plant integrated with a kraft pulp mill.

150
of two of the output streams – ethanol and power The biorefinery in this example actually includes
– account for a significant part of the increase in three different options for co-production of fiber,
total energy input. fuel, and power. A major modification for these
integrated biorefinery options (but not the stand-
Table 5 contains an overall energy balance for alone pulp mill) would be to modify the chemi-
the examples in Figures 4 and 5. The high ener- cal recovery and power island in the pulp mill to
gy-value output streams in Figure 5 have a com- make it more efficient in high-level energy con-
bined energy value of 442 million Btu/hr, nearly 7 version. The most important differences are (a)
times the energy value of net power generated by the Tomlinson recovery boiler is replaced with
the pulp mill in Figure 4. This additional energy a high temperature black liquor gasifier that pro-
output is less than the increase in energy input in duces synthesis gas at elevated pressure for liquid
Figure 5 versus Figure 4. The incremental energy hydrocarbon production; (b) the hog fuel boiler is
efficiency for ethanol production plus the increase replaced with a pressurized fluidized bed gasifier
in power generated (Figure 5) is a respectable that produces synthesis gas for either power gen-
50%. By comparison, the efficiency for estimates eration (in options A and B), or for conversion to
for stand-alone, ethanol-from-wood plants are liquid hydrocarbons (in option C); and (c) a gas-
from 32% to 41% depending on the efficiency of fired power turbine is added to generate power
conversion of wood carbohydrates to fermenta- from part of the synthesis gas from the biomass
ble sugars.[12] The gains in energy efficiency gasifier and/or unconverted syngas from the fuel
come from mass and energy integration within synthesis generate power. In cases B and C, a heat
the biorefinery versus a stand-alone ethanol plant. recovery boiler and condensing turbine was used
Please see Frederick et al., 2008[12] for details of downstream of the gas turbine to produce addi-
this case study. tional power. Process steam is generated by cool-
ing the hot syngas and by recovering heat from the
exothermic reactions in the hydrocarbon synthesis
Integrated F-T Hydrocarbons Plant reactors. The process steam production in all op-
with Upgraded Power Island tions is balanced with the total demand from the
The second example is one of an integrated bi- integrated biorefinery, with biomass purchased to
orefinery that produces hydrocarbons (gasoline balance the fuel requirement. Process flow dia-
precursors) via Fischer-Tropsch (F-T) synthesis grams for each option have been published and
from synthesis gas (CO and H2) as well as pro- are reproduced in Appendix 1 of this document
ducing fiber and power. The analyses precented for convenience.
here is based on a study by Larson et al. [13,14].

Table 6, Energy balances and incremental efficiency for fuel and power generation for a biorefinery that
consists of a Fischer-Tropsch liquid hydrocarbons plant integrated with a kraft pulp mill.

151
 In all options, the fiber production is exactly the while power generation is increased by 20% when
same in the stand-alone pulp mill and each biore- compared with the pulp mill base case.
finery option, but the energy input as purchased
wood residues has increased substantially, as Table 6 also shows the incremental conversion ef-
shown in Table 6. ficiency for the additional biomass purchased for
each of the three biorefinery options. In case A, the
Also shown in Table 6 is the liquid hydrocarbon conversion efficiency is 172%. This surprisingly
and power production for each of the three bi- high incremental efficiency is mainly the result
orefinery configurations. In all three biorefinery of far more efficient energy conversion by using
cases, there is substantial liquid fuel production (a) gasification instead of combustion to convert
and increase in power production. biomass and black liquor a syngas intermediate,
and (b) using a gas combustion turbine rather
In biorefinery option A, the heat and power plant than a steam power cycle for electrical power
is configured to produce liquid hydrocarbons form production. The incremental efficiencies are also
syngas generated from black liquor only. Single- very high for options B and C, 68% and 70% re-
pass conversion of syngas to liquid fuel is well spectively. Please see the references by Larson et
below 100%, so unconverted syngas from the fuel al.[13,14] for full details of this case study.
synthesis reactor is burned in the gas turbine to
produce power. Syngas from the biomass gasifier Capital Investment
is burned to produce power and is not used for liq- In all of the options considered here, it is assumed
uid fuel production in this option. In this option, that there is an existing kraft pulp mill in good
liquid fuel production is nearly 112 MWth, and operating condition. The capital costs required to
power generation is increased by 36% beyond the convert the pulp mill to a forest biorefinery are
stand-alone pulp mill option. for new equipment for an ethanol plant or F-T
hydrocarbons synthesis plant, modification and/or
The design of biorefinery option B is optimized expansion of the chemical recovery and power is-
for power production. The input of biomass fuel land as required for each specific case, and modi-
is increased beyond the minimum required to fications to accommodate additional biomass han-
satisfy pulp mill steam demand, and the syngas dling, energy integration, etc. Table 7 includes the
produced in the biomass gasifier, along with the total incremental capital investment for each case
unconverted syngas from the fuel synthesis reac- and option considered here. More detailed capi-
tor, is used entirely for power and process steam tal cost summaries are available from Larson et
production. In option B, power generation is in- al.[13] and Frederick et al.[12]
creased by 256% when compared with the pulp
mill base option. The results in Table 7 show that a significant
capital investment is required to create a forest
The design of biorefinery option C is optimized biorefinery by integrating a fuel production plant
for liquid fuel production. Again, the input of bio- with a kraft pulp mill. The capital investments
mass fuel is increased, but the syngas produced in are not at all unreasonable, however. The capital
the biomass gasifier is used entirely for liquid fuel intensity for a fuel grade ethanol plant with cel-
production. The unconverted syngas from the fuel lulosic feedstock falls in the range of $1.50 to
synthesis reactor is used for power production. In $2.50/MWth depending upon feedstock and plant
this option, the liquid fuel production is increased size [15,16,17].
by 207% over the other two biorefinery options,

Table 7, Incremental capital costs for ethanol or F-T hydrocarbon production from an integrated forest
biorefinery.

152
Profitability
Larson and co-workers estimated that without
subsidies, F-T hydrocarbon plants would have
a discounted cash flow rate of return (DCFRR)
between 14% and 18% on invested capital. With
expected subsidies, the DCFRR would increase to
between 28% and 34%. In their study, Larson
et al.[13,14] considered these subsidies: an excise
tax credit (currently available for ethanol), an in-
vestment tax credit for gasification investments
as mandated by EPAct 2005, a 10-year produc-
tion tax credit for renewable electricity of $9 per
MWh, and a renewable energy credit (REC) of
$20/MWh.

Co-production of ethanol and fiber from southern

pine, the case study summarized here, would not
currently be profitable because of loss of cellulose
to black liquor from pulping of wood chips from
which hemicellulose had been pre-extracted with
H2SO4. For at least some other wood species, cel-
lulose is not degraded by dilute acid pre-extrac-
tion. In those cases, integrated biorefineries are
profitable.

Summary
Integrated forest biorefineries that produce trans-
portation fuels and power as well as cellulose fiber
are moving closer to becoming real options. De-
velopment of at least two of these biorefineries is
currently being funded by the U.S. Department of
Energy’s Office of Energy Efficiency and Renew-
able Energy, and several others are under devel-
opment around the world. An analysis of several
published case studies suggests that integrated
forest biorefineries would be economically vi-
able, with capital investments in the order of $280
- $500 million required for constructing and inte-
grating fuel and power plants of viable capacity
with existing, fully operational kraft pulp mills.

153
 References Assessment. Bioresource Technology 99 (2008)
1. Larson, E.D., Consonni, S., Katofsky, R.E., 5051–5057.
Iisa, K. and Frederick, W.J., A Cost-Benefit As- 13. Larson, E.D., Consonni, S., Katofsky, R.E.,
sessment of Gasification-Based Biorefining in the Iisa, K. and Frederick, W.J., A Cost-Benefit As-
Kraft Pulp and Paper Industry. Volume 1-4. Final sessment of Gasification-Based Biorefining in the
report under US Dept. of Energy contract DE- Kraft Pulp and Paper Industry, Volume 1-4, Final
FC26-04NT42260, 21 December 2006. Available report under US Dept. of Energy contract DE-
at www.osti.gov/servlets/purl/912520-8bRNAH/. FC26-04NT42260, 21 December 2006. Available
2. Rostrup-Nielsen, J.R., Making Fuels from Bio- at www.osti.gov/servlets/purl/912520-8bRNAH/.
mass. Science (3 June 2005) Vol. 308. no. 5727, 14. Larson, E. D., Consonni, S., Katofsky, R. E.,
pp. 1421 – 1422. Iisa, K. and Frederick, Jr., Gasification Based
3. Ragauskas, A.J., Nagy, M., Kim, D.H., Eckert, Biorefining at Kraft Pulp and Paper Mills in the
C.A., Hallett, J.P. and Liotta, C.L. From wood United States, Part A: Background and Assump-
to fuels: Integrating biofuels and pulp production. tions. TAPPI Journal, 7(11): 8-14, November
Industrial Biotechnology (2006), 2(1), 55-65. 2008. Part B: Results,” accepted for publication
4. Wyman, C.C. and Goodman, B.J., Biotechnol- in TAPPI J, 2008.
ogy for production of fuels, chemicals, and mate- 15. Wooley, R., Ruth, M., Sheehan, J., Ibsen, K.,
rials from biomass. 1993, Appl. Biochem. Biotech. Majdeski, H. and Galvez, A., Lignocellulosic Bi-
39-40(1): 41-59. omass to Ethanol: Process Design and Economics
5. Gielen, D.J., de Feber, M.A.P.C., Bos, A.J.M. Utilizing Co-Current Dilute Acid Prehydrolysis
and Gerlagh, T., Biomass for energy or materials? and Enzymatic Hydrolysis - Current and Futuris-
A Western European systems engineering per- tic Scenarios. NREL/TP-580-26157 (1999).
spective. Energy Policy, 2001, 29(4): 291-302. 16. McAloon, A., Taylor, F., Yee, W., Ibsen, K.
6. Ragauskas, A.J., Williams, C.K., Davison, and Wooley, R., Determining the Cost of Produc-
B.H., Britovsek, G., Cairney, J., Eckert, C.A., ing Ethanol from Corn Starch and Lignocellulosic
Frederick, W.J., Jr., Hallett, J.P., Leak, D.J., Li- Feedstocks. NREL-TP-580-28893, National Re-
otta, C.L., Mielenz, J.R., Murphy, R., Templer, R. newable Energy Laboratory, Golden, CO (Octo-
and Tschaplinski, T. The Path Forward for Biofu- ber, 2000).
els and Biomaterials. Science (2006), 311(5760), 17. Frederick, W.J. Jr., Lien, S.J., Courchene, C.,
484-489. Ragauskas, A.J. and Iisa, K., Co-production of
7. Huber, G.W., Iborra, S. and Corma, A. Synthe- ethanol and cellulose fiber from Southern Pine: A
sis of Transportation Fuels from Biomass: Chem- technical and economic assessment. Biomass &
istry, Catalysts, and Engineering. Chem. Rev. Bioenergy, 32(12):1293-1302 (2008).
2006, 106, 4044-4098.
8. Lave, L.B., Hendrickson, C.T. and McMichael,
F.C., Environmental Implications of Electric Cars.
Science 1995, 268(5213): 993 – 995.
9. McNicola, B.D., Rand, D.A.J. and Williams,
K.R., Fuel cells for road transportation purposes
— yes or no? J. Power Sources, 2001, 100(1-
2)47-59.
10. Ericsson, K. and Nilsson, L.J., Assessment of
the potential biomass supply in Europe using a
resource-focused approach. Biomass and Bioen-
ergy, 2006, 30(1):1-15.
11. Perlack R.D., Wright L.L., Turhollow A.F.,
Graham R.L., Stokes B.J. and Erbach D.C., Bi-
omass As Feedstock for a Bioenergy and Bio-
products Industry: The Technical Feasibility of a
Billion-Ton Annual Supply. Oak Ridge National
Laboratory, Oak Ridge, TN, report numbers DOE/
GO-102005-2135 and ORNL/TM-2005/66.
12. Frederick, W.J. Jr., S.J. Lien, C.E. Courch-
ene, N.A. DeMartini, A.J. Ragauskas and K.
Iisa. Production of Ethanol from Carbohydrates
from Loblolly Pine: A Technical and Economic

154
Finnish Recovery Boiler Conference, June, 2009
Appendix Process flow diagrams for three options for co- producing F-T Hydrocarbons in an inte-
grated biorefinery [13].

Pressurized, Syngas Rectisol for Once-thru

Black liquor high-temp cooling H 2S and CO 2 LP FT
gasifier & cleaning removal synthesis

O2 condensed phase S to polysulfide

liquor preparation
to causticizing
Air separation unconverted
air clean syngas
unit
biomass CRUDE F-T
O2 syngas LIQUIDS (to an
existing refinery)
Syngas Gas Turbine
Fluidized-bed
Dryer cooling CC ELECTRICITY
gasifier
& cleaning (Frame 6FA)

process steam to mill recovered process

heat
Biomass chips

Process flowsheet for Option A: biomass gasifier, gas turbine, no syngas recycle.

Pressurized, Rectisol for Once-thru

Syngas cooling
Black liquor high-temp H 2S and CO 2 LP FT
& cleaning
gasifier removal synthesis

O2 S to polysulfide
condensed phase liquor
to causticizing
Air separation unconverted
air clean
unit syngas
biomass CRUDE F-T
syngas LIQUIDS (to an
O2
existing refinery)
Fluidized-bed Syngas cooling Gas Turbine CC
Dryer ELECTRICITY
gasifier & cleaning (Frame 7FA GT)

process steam to mill

recovered process heat

Biomass chips

Process flowsheet for Option B: biomass gasifier, larger gas turbine, no syngas recycle.

Pressurized, Rectisol for H 2S

black Syngas cooling Once-thru LP
high temp and CO 2
liquor & cleaning FT synthesis
gasifier removal
oxygen condensed phase S to polysulfide
to causticizing liquor preparation
Air separation clean
air
unit biomass
syngas unconverted CRUDE F-T LIQUIDS
oxygen syngas (to an existing
refinery)
Fluidized-bed Syngas quench Gas Turbine CC
Dryer
gasifier w/ heat recovery ELECTRICITY
(Frame 6FA GT)

process steam to mill recovered

process heat
biomass

Process flowsheet for Option C: syngas from biomass gasifier to fuel synthesis, gas turbine.

155
156
COMPOSITION OF CARRYOVER IN KRAFT
RECOVERY BOILERS
HONGHI TRAN AND ASGHAR KHALAJ-ZADEH
PULP & PAPER CENTRE, UNIVERSITY OF TORONTO
TORONTO, ON, CANADA

Abstract

The composition of carryover particles is of great liquor composition, solids content and droplet
importance in determining to the ability of the size, flue gas temperature and burning conditions
particles to deposit on recovery boiler superheater in the lower furnace. As a result, the composition
tube surfaces. Results of laboratory and field stud- is usually not known, particularly at the moment
ies show that chloride (Cl) and potassium (K) con- immediately before the particles strike the tube
tents in carryover particles are linearly proportion- surface. This makes it difficult to determine the
al to, but much lower than their values in the black fouling tendency of the particles, i.e. whether the
liquor. Cl and K enrichment factors for carryover particles would deposit on the tube surface upon
are below 1 and decrease further as the particle impact.
temperature increases. Carryover contains more
sulphate and less carbonate than smelt. The sul- Much work has been done over the past decade to
phate content in carryover increases only slightly systematically study the composition of carryover
with an increase in exposure time, suggesting that particles under controlled environment using an
the majority of sulphate must have formed prior to Entrained Flow Reactor (EFR) and to examine
deposition. There is a linear correlation between important parameters that affect the composition
Cl and K enrichment factors for the as-fired black [3,4]. Field studies were also performed using
liquor, carryover, precipitator ash and smelt. The air-cooled probes to collect carryover samples in
correlation can be theoretically verified. the superheater region of recovery boilers, and to
examine how the composition is affected by liq-
Introduction uor composition and boiler operation [5,6]. This
As black liquor is sprayed and burned in recovery paper summarizes the key findings obtained from
boilers, a small portion of black liquor droplets these studies and discusses their practical implica-
are entrained in the flue gas where they undergo tions for recovery boiler operation.
drying, swelling, devolatizing and char burning,
and ultimately, form carryover particles. Due to Laboratory Studies
their relatively large particle size, 100 μm to 3 Six black liquor samples from five different kraft
mm, and high temperatures, carryover particles pulp mills were used either as-is or were mixed
tend to strike and deposit on tube surfaces in the with various amounts of precipitator dust col-
superheater region of the boiler. The ability of car- lected from the same recovery boiler as the black
ryover particles to form deposits depends greatly liquor, or were mixed with various amounts of
on the liquid content of the particles at the mo- pure KCl to obtain a wide range of Cl and K con-
ment of impact, which, in turn, is a function of tents [4]. The mixed black liquor samples were
particle temperature and composition, mainly the poured onto stainless steel trays to form a thin
chloride (Cl), potassium (K) and sulphide (S) con- sheet, which was dried in an oven controlled at
tents [1-3]. 130°C, and were then ground and sieved to make
dried particles in the size range of 300-500μm.
Since the composition of carryover continuously The dried particles were fed into the EFR which
changes during combustion, it varies greatly from was set at 800°C. A mixture of natural gas com-
boiler to boiler, and from location to location in a bustion products with a varying amount of excess
given boiler, depending on particle residence time, air flowed through the EFR with a velocity of 1.8

157
 m/s. Four different O2 concentrations were used, and hence a greater amount of Cl and K vaporized
1.7, 6.2, 9.5, and 12.3%. from the particles.

Carryover particles collected at the EFR exit were The enrichment factors for chloride (EFCl) and
analyzed for residual char, sodium (Na), potas- potassium (EFK) are typically defined respectively
sium (K), chloride (Cl), sulphide (S), sulphate as the ratios of the mole fractions of Cl and K in
(SO4), and carbonate (CO3) contents. The compo- carryover to the mole fractions of Cl and K in the
sition was expressed in mole% Cl/(Na+K) for Cl, black liquor. Thus, the slope of each line in Fig-
mole% K/(Na+K) for K, mole% CO3/(Na2+K2) for ure 1 is essentially the same as the Cl enrichment
carbonate, mole% SO4/(Na2+K2) for sulphate, and factor (EFCl) in carryover at each O2 concentra-
mole% S/(Na2+K2) for sulphide. The total sulphur tion. At 1.7%O2, for instant, the slope is 0.78 or
content was expressed as mole% Stotal /(Na2+K2) EFCl = 0.78; the chloride content in carryover was
which is the sum of sulphate and sulphide con- about 78% of that in black liquor. At 12.3%O2, the
tents, mole% (S+SO4)/(Na2+K2). slope is much lower, 0.3 or EFCl = 0.3; the chlo-
ride content in carryover was only 30% of that in
Cl and K Contents black liquor.
As shown in Figure 1, at a given O2 concentration,
the chloride content in carryover increased linear- Similarly, the slope of each line in Figure 2 is the
ly with an increase in chloride content in the black same as the K enrichment factor (EFK) in carryo-
liquor. For a given black liquor sample, however, ver at each O2 concentration. In this case, how-
burning at a higher O2 concentration resulted in ever, the decrease in K content in carryover was
carryover that had a lower Cl content. The re- much less than the decrease in Cl content (i.e. the
sults are plausible since a higher O2 concentration data are located closer to the diagonal line in Fig-
would give rise to a higher particle temperature, ure 2 than they are in Figure 1). This implies that
the enrichment factor for K in carryover is much
12 C HL OR IDE higher than that for Cl.
O2 %
10 1.7 Figure 3 shows a clear linear relationship with a
Cl/(Na+K) mole%

8 6.2 R2 value of 0.998 between EFK and EFCl values

in Carryover

9.5 for all four O2 concentrations tested in this study.

6 12.3 Extrapolating the line on both ends (broken lines)
4 shows that when EFCl = 1, EFK = 1 and when EFCl
= 0, EFK = 0.68. These two end points suggest that
2 if Cl is not deleted at all from the carryover, then
0 potassium is not deleted either. On the other hand,
0 2 4 6 8 10 12 if Cl is completely deleted from the carryover (0%
Cl/( Na+K) mole% in Black Liquor Cl), there is still 68% K left in the carryover com-
Figure 1. Correlation between chloride content in pared to that in the black liquor. The linear cor-
carryover and that in black liquor at various O2 relation between EFCl and EFK is intriguing, but it
concentrations. can be theoretically explained as will be seen later
in this paper.
12 P OT AS S IUM
O2%
10 1.7
K/(Na+K) mole%

8 6.2
in Carryover

9.5
6 12.3
EFK

0
0 2 4 6 8 10 12
K/(Na+K ) mole% in Black Liquor

Figure 2. Correlation between potassium content

in carryover and that in black liquor at various O2 Figure 3. Enrichment factors of Cl and K in car-
concentrations. ryover obtained at different O2 concentrations

158
CO3, SO4, S and Stotal Contents perature (Figure 5). The sulphate content in both
Figure 4 shows the change in carbonate, sulphate, carryover samples, however, increased with an in-
sulphide and total sulphur contents in carryover crease in gas temperature. This again confirms the
produced from the same black liquor at differ- results shown in Figure 4 that carryover formed at
ent O2 concentrations. The sulphide content de- higher temperatures retains more sulphur.
creased from 5 mole% S/(Na2+K2) at 1.5% O2 to Field Studies
zero at 12.5% O2 due to oxidation to sulphate. The
sulphate content increased from 15 mole% SO4/ In order to examine if the above laboratory re-
(Na2+K2) at 1.5%O2 to 40 mole% SO4/(Na2+K2) at sults using the EFR were applicable to conditions
12.5% O2, which is much more than the decrease in recovery boilers, field studies were performed
in sulphide content. The carbonate content de- on three operating boilers. Boiler A was the only
creased from 75 to 55 mole% CO3/(Na2+K2). The recovery boiler at a kraft mill in Eastern Canada.
extent of decrease in carbonate content is essen- Boilers B and C were two of the three recovery
tially the same as that of the total increase in total boilers at a kraft mill in Southern United States,
sulfur content. These results are consistent with burning black liquor of a similar composition.
the fact that black liquor burning at a higher tem-
perature would retain more sulphur in the char/ 1.0 C HL O R IDE
smelt residue, and would release less S to the gas
in the form of SO2. 0.8
mole% Cl/(Na+K)
100 0.6
#2
80 0.4
mole% `C`/(Na2+K2)

CO 3
60 0.2 #1
40 0.0
STotal
SO 4 600 700 800 900 1000 1100
20
Gas Tem perature ( oC )
S
0
0 3 6 9 12 15 8 P O T A S S IU M
%O2
Figure 4. Carbonate, sulphate, sulphide and total 6 #1
mole% K/(Na+K)

sulphur content in carryover obtained at different

O2 concentrations. Note: ‘C’ is either CO3, S, SO4 4 #2
or S+SO4.
2

Since the retention time of the particles in the EFR

was less than 2 seconds, the above results imply 0
600 700 800 900 1000 1100
that the formation of sulphate in carryover in an
oxidizing environment is a rapid process that oc- Gas Tem perature ( oC )
curred during combustion, and not a result of the
sulphation of carbonate after the carryover has 100 S U L P HA T E
been formed.
80
mole% SO4/(Na2+K2)

Effect of Gas Temperature 60

#2
The effect of gas temperature on Cl and K con-
tents in carryover was studied on two black liquor 40 #1
samples of different Cl and K contents. Sample
20
#1 contained low Cl and high K, while Sample #2
contained high Cl and low K. The gas temperature 0
was changed by operating the EFR at different 600 700 800 900 1000 1100
temperatures. Gas Tem perature ( oC )

As expected, Cl and K contents in both carryover Figure 5. Effect of gas temperature on Cl, K and
samples decreased with an increase in gas tem- SO4 contents in carryover obtained at 12.5% O2.

159
 Carryover Collecting Probe ing the fact that they are in the same mill. Boiler
The probe consists of two concentric stainless A had a lower Cl content and a higher K content
steel pipes. The outer pipe is 50 mm in ID and compared to Boilers B and C. These as-fired black
3 m long. The inner pipe, 25 mm in ID and 3.2 liquor concentrations were used in the calculation
mm long, is equipped with a removable carryo- of Cl and K enrichment factors in the carryover,
ver sampling tip, a thermocouple embedded on smelt and precipitator dust samples.
the pipe surface near the tip, and a cooling system
that regulates the compressed air flow through the Table 1. As-fired Black Liquor and Flue Gas Anal-
inner probe to keep the sampling tip at a desired yses at the Sampling Locations
temperature. The outer pipe acts as a shield to
prevent fume from condensing on the surface of
the sampling tip. A small tube is inserted through
the inner pipe to collect flue gas samples from the
boiler using a vacuum pump.

Gas Inner Shield Pipe Thermocouple

Sampling pipe
Tube Air Out

Pump
Air In Carryover
Sampling
Tip

Figure 6. Air-cooled sampling probe used for col-

lecting carryover particles from operating boilers.

The probe was inserted 1.5 to 2 m into each boiler The temperature and O2 and SO2 concentrations in
through an opening ahead of the superheater at the the flue gas at the sampling locations. The flue gas
bull nose elevation. The carryover sampling tip temperature in Boiler A was the highest, followed
of the inner pipe was kept inside the outer shield by Boiler B and Boiler C. Excess O2 and SO2 con-
pipe when the probe was first inserted in the boil- centrations were similar in all three boilers.
er. As the sampling tip temperature reached about
500ºC, it was pushed out of the shield pipe to be Enrichment Factors of Cl and K
exposed to the flue gas. This procedure was nec- Figure 7 shows the enrichment factors for Cl and
essary in order to prevent condensed fume from K in carryover deposits collected from boilers A,
depositing on the sampling tip while it was still B, and C, along with those for smelt (shown at
“cold” at the start of the test. The probe was ex- time = 0). In all cases, the enrichment factors were
posed in the boiler for different time intervals to less than 1, indicating that both carryover deposits
examine how carryover composition may change and smelt in each boiler contained less Cl and K
with time. At the end of each test, the sampling tip than the as-fired black liquor of the boiler. The av-
was pulled back into the outer shield pipe before erage EFCl value was about 0.8 for smelt, and sig-
the entire probe was retracted from the boiler. As nificantly lower (about 0.5) for carryover depos-
the probe had been sufficiently cooled down, the its. In each boiler, no significant difference in EFCl
sampling tip, along with the carryover sample on value was observed between carryover deposits
its surface, was removed from the inner pipe tube, from different sampling times, except for the 12-
and wrapped and stored in a shielded plastic bag. minute sample from Boiler B, which showed a
A new or clean sampling tip was used for the next significantly higher EFCl value than other samples.
test. This outlier (noted with an asterisk (*) in the Fig-
ure was due probably to the contamination of the
At about the same time as carryover sampling, sample by fume deposits which usually contain
samples of as-fired black liquor, smelt, and pre- much more Cl and K than carryover deposits.
cipitator dust of each boiler were also collected
and analyzed. Similar results were obtained for potassium, al-
though the average EFK value was higher than the
Table 1 summarizes Na, K, and Cl contents in the average EFCl value (Figure 8). It was close to 1
as-fired black liquors of the boilers. Boilers B and for smelt, but varied between 0.8 and 0.9 for car-
C had a similar black liquor composition, reflect- ryover deposits.

160
The higher EFK value compared to EFCl value is nition, as-fired black liquor samples have an en-
due to the fact that potassium compounds are not richment factor of 1. Carryover deposits and smelt
only in the form of KCl, but also K2SO4, K2S and samples have enrichment factors of less than 1,
K2CO3, whereas chloride compounds can only while precipitator ash samples, which are sig-
be in the form of NaCl and KCl, which are much nificantly enriched in Cl and K, have enrichment
more volatile than other alkali compounds. factors of much greater than 1. The correlation
between EFK and EFCl of these samples for boil-
1.0 C HL O R IDE
ers A and B appears to be linear. These results are
Smelt * consistent with those reported by Janka et al [6].
0.8

0.6 2.0
B oile r A
0.4 1.5
Boiler A
0.2 Boiler B
Boiler C 1.0

EFK
Black liquor
0.0 Sm elt
0 5 10 15 0.5 Precip . ash
Carryover
Sampling Time (min)
0.0
Figure 7. Enrichment factors for chloride in car- 0.0 0.5 1.0 1.5 2.0 2.5 3.0
ryover deposits and smelt collected from boiler A, EF C l
B, and C (oxidized smelt shown at time = 0
2.0
Smelt B oile r B
1.0 P O T A S S IUM
1.5
0.9
* 1.0
EFK

Black liquor
0.8
Sm elt
0.5 Precip . ash
EFK

0.7 Boiler A Carryover

Boiler B
0.6 Boiler C 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
0.5 EF C l
0 5 10 15
Sampling Time (min)
2.0
Figure 8. Enrichment factors for potassium in car- B oile r C
Expected
ryover deposits and smelt collected from boiler A, 1.5
B, and C (oxidized smelt shown at time = 0).
1.0
EFK

Black liquor
Extrapolating the carryover sample data to zero Sm elt
sampling time in Figures 7 and 8 respectively 0.5 Precip . ash
gives EFCl and EFK values of about 0.5 and 0.8 Carryover
for “freshly formed” carryover. These values are 0.0
significantly lower than those for smelt. This is 0.0 0.5 1.0 1.5 2.0 2.5 3.0
plausible since carryover particles were exposed EF C l
to a more oxidizing atmosphere and a higher tem-
perature than smelt in the lower furnace. High Figure 9. Correlation between the enrichment
temperatures cause more NaCl and KCl to vapor- factor of Cl and K for as-fired liquor, smelt, car-
ize and thus more Cl and K to be lost from the ryover and dust from boilers A, B, and C.
particles. The results are consistent with the data
obtained using EFR (Figures 3 and 5).
For Boiler C, the data point for the precipitator
ash did not fall on to the same line as the data
Figure 9 shows the average EFK values plotted
points for carryover, smelt and black liquor (bro-
against the average EFCl values for all types of
ken line). The expected linear relationship implies
samples taken from boilers A, B, and C. By defi-
the measured EFCl and EFK values for Boiler C

161
 precipitator ash must have been wrong. Either the Figure 10 shows the total sulphate, expressed as
EFCl value should have been higher or the EFK mole% (S+SO4)/(Na2+K2) and carbonate contents
value measured K content been lower, or both. of carryover samples from Boilers A, B, and C,
along with the composition of oxidized smelt
CO3, SO4 and S Contents (shown in these figures at time = 0), which was
taken to be the composition of smelt with all sul-
In order to enable a comparison between carryo- phide converted to sulphate. The carryover depos-
ver composition and smelt composition, mole% its contained more total sulphate and less carbon-
CO3/(Na2+K2) was used for carbonate, mole% ate than oxidized smelt. For carryover samples
SO4/(Na2+K2) for sulphate and mole% S/(Na2+K2) collected over short exposure times, the total sul-
for sulphide. phate content increased slightly while the carbon-
ate content decreased with sampling time. These
results suggest that although some sulphation of
the carbonate in carryover may have occurred
Smelt on the probe, the majority of the sulphation must
80 B oile r A
have occurred during black liquor combustion,
CO 3 /(Na 2 +K 2 ) before deposition.
60
Note also from Figure 10 that while the total sul-
mole %

40 phate and carbonate contents of oxidized smelt in

Boilers B and C were similar (since these boilers
(S+SO 4 )/(Na 2 +K 2 )
20 were burning the same black liquor at the same
pulp mill), the total sulphate content of carryover
0 in Boiler C was significantly higher than in Boiler
0 5 10 15 B. This, together with the higher carbonate content
Sampling Time (min.) of Boiler C precipitator dust, suggests that Boiler
C may have been operated at a higher temperature
80 Smelt B oile r B in the lower furnace than Boiler B. Higher tem-
peratures minimize sulphur release and thus help
CO 3 /(Na 2 +K 2 )
keep the sulphur in carryover.
60

Theoretical Consideration
mole %

40
(S+SO 4 )/(Na 2 +K 2 ) The apparent linear relationship obtained in this
study between Cl and K enrichment factors for the
20
as-fired black liquor, carryover, precipitator ash
and smelt (Figure 9) and that obtained by Janka et
0 al [6] for different boilers at different mills is in-
0 5 10 15
triguing. It suggests that there is a good reason for
Sampling Time (min.)
such a linear correlation to exist. It is, therefore,
worthwhile to examine if such a correlation can
80 B oile r C be theoretically verified.
Sm elt
(S+SO 4 )/(Na 2 +K 2 )
60 Consider a system involving three streams, in
which either stream 1 mixes with stream 2 to
mole %

40 form stream 3, or stream 3 splits into stream 2 and

CO 3 /(Na 2 +K 2 )
stream 1, as shown in Figure 11.
20

0
0 5 10 15
Sampling Time (min.)

Figure 10. Total sulphate and carbonate content

of the oxidized smelt and carryover deposit from
boilers A, B and C (oxidized smelt shown at time
= 0).

162
 F1 F1(Ca3Ca1 )
Ca2 Ca3 (8)
F2
F3
Dividing (5) by (6), and (7) by (8) yields:
F2
C b2  C b1 C b3  C b1 (9)
Ca2  C a1 C a3  Ca1

or C b2  C b3 C b3  C b1 (10)
Ca2  C a3 Ca3  Ca1
F1
Equations (9) and (10) give:
F3
C b2  C b1 C b2  C b3 (11)
F2 Ca2  C a1 C a2  C a3

Equation (11) indicates that the wt% concentra-

Figure 11. System with three streams tion of “b” in streams 1, 2 and 3 is linearly pro-
portional to that of “a”, as schematically shown
F1, F2, and F3 are respectively the mass flows of in Figure 12.
stream 1, stream 2, and stream 3.

F1 + F 2 = F 3 (1) 2
Concentration of “b”

Cb2
If “a” and “b” are two species in 1, stream 2, and
stream 3, then 3
Cb3
F1Ca1 + F2Ca2 = F3Ca3 (2) Cb1
1
F1Cb1 + F2Cb2 = F3Cb3 (3)

where Ca1, Ca2, and Ca3 are concentrations, in wt%, Ca1 Ca3 Ca2
of “a” in streams 1, 2, and 3, respectively; and Cb1,
Concentration of “a”
Cb2, and Cb3 are concentrations, in wt%, of “b” in
streams 1, 2, and 3, respectively. Figure 12. Linear correlation between wt% con-
centrations of species “b” and “a” in streams 1,
Rearrange (3) gives: 2 and 3

F3C b3  F1C b1 (4)

C b2 Virgin Black Liquor, As-fired Black Liquor and
F2
Precipitator Dust
Consider a case around a recovery boiler (Fig-
F3C b3  F1C b1 F2 C b1
C b2 C b1 ure 13), where FVg BL, FDust, FAF BL, Smelt and FStack
F2 respectively represent the mass flows of virgin
Hence black liquor, precipitator dust, as-fired black liq-
F3 (C b3  C b1) (5) uor, smelt, and stack gas.
C b2 C b1
F2
Similarly
F3 (C a3Ca1) (6)
Ca2 Ca1
F2

F1 (C b3C b1)
C b2 C b3 (7)
F2

163
 Or Cl mole 1.1 u
Cl wt (14)
FStack K mole K wt

FDust (13) and (14) give:

Recovery
Boiler Cl mole(Dust)  Cl mole(Vg BL) Cl mole(Dust)  Cl mole(AF BL) (15)
Mix
FVirgin BL Tank FAs-fired BL FSmelt K mole(Dust)  K mole(Vg BL) K mole(Dust)  K mole(AF BL)

Figure 13. Flow around a Recovery Boiler This means that the linear relationship is also held
when the Cl and K concentrations are expressed
as mole% Cl/(Na+K) and mole% K/(Na+K), as
shown in Figure 15.
From the material balance around the mix tank:
FVg BL + FDust = FAF BL (12)

Dust
Applying (11) to these three streams for Cl and

mole% K/(Na+K)
K yields: As-fired
CCl(Dust)  CCl(Vg BL) CCl(Dust)  CCl(AF BL) (13) BL
C K(Dust)  C K(Vg BL) C K(Dust)  C K(AF BL)

Virgin
Equation (13) means that the weight concentra- BL
tions of Cl in precipitator dust, virgin black liquor
and as-fired black liquor are linearly proportional
to the weight concentrations of K, as shown in mole% Cl/( Na+K)
Figure 15. Linear correlation between mole%
Cl/(Na+K) and mole% K/(Na+K) of virgin black
Potassium Conc., wt%

Dust liquor, as-fired black liquor and precipitator dust

As-fired
Now let EFCl and EFK be enrichment factors for
BL
Cl and K with respect to as-fired black liquor and
rearrange (15), we have:
Virgin
BL Cl mole(Dust) Cl
 mole(Vg BL)
Cl mole(As-fired BL) Cl mole(AF BL)
K mole(Dust) K
Chloride Conc., W t%  mole(Vg BL)
K mole(As-fired BL) K mole(AF BL)
Figure 14. Linear correlation between wt% Cl
and wt% K of virgin black liquor, as-fired black
Cl mole(Dust) Cl mole(AF BL)
liquor and precipitator dust 
Cl mole(AF BL) Cl mole(AF BL)
K mole(Dust) K
Figure 14.  mole(AF BL)
K mole(As-fired BL) K mole(AF BL)
If Clmole and Kmole are the molar concentrations
of chloride and potassium, and Clwt, and Kwt are
weight concentrations of chloride and potassium, Or
then:
EFCl(Dust)  EFCl(Vg BL) EFCl(Dust)  EFCl(AF BL) (16)
ª Cl wt º ª Na wt K wt º EFK(Dust)  EFK(Vg BL) EFK(Dust)  EFK(AF BL)
Cl mole «¬ 35.5 »¼ «¬ 23  39.1»¼
K mole ª K wt º ª Na wt K wt º Since EF for as-fired black liquor is equal to 1,
«¬ 39.1»¼ «¬ 23  39.1»¼ (16) becomes

164
EFCl(Dust)  EFCl(Vg BL) EFCl(Dust)  1 (17)
where x, y and z are respectively the mass por-
tions of Na, K, and Cl lost to the flue gas.
EFK(Dust)  EFK(Vg BL) EFK(Dust)  1
Cl L Cl I

This indicates that the linear relationship will also Na L  K L Na l  K I Cl L (Na I  K I )  Cl I (Na L  K L )
hold if the Cl and K concentrations in dust, vir- KL

KI K L (Na I  K I )  K I (Na L  K L )
gin black liquor and as-fired black liquor are ex- Na L  K L Na l  K I
pressed as enrichment factors (Figure 16).
zCl I (Na I K I )  Cl I (xNa I  yK I )
yK I (Na I  K I )  K I (xNa I  yK i )

(z-x)Cl I Na I  (z-y)Cl I K I (18)

2 (y-x)Na I K I
Dust
EFK

As-fired BL Cl L Cl F

Na L  K L Na F  K F Cl L (Na F  K F )Cl F (Na L  K L )
1
Virgin BL
KL

KF K L (Na F  K F)K F (Na L  K L )
Na L  K L Na F  K F

zCl I [(1-x)Na I  (1-y)K I ]  (1-z)Cl I (xNa I  yK I )

1 2
yK I [(1-x)Na I  (1-y)K I )]  (1-y)K I (xNa I  yK I )
EF C l
Figure 16. Linear correlation between EFCl and (z - x) Cl I Na I  (z - y)Cl I K I (19)
EFK of virgin black liquor, as-fired black liquor
(y - x)K I Na I
and precipitator dust
[18] and [19] give
Carryover, Fume and Smelt
For carryover, fume and smelt, the material bal- Cl L Cl I Cl L Cl F
ance around the black liquor mix tank does not  
Na L  K L Na I  K I Na L  K L Na F  K F
apply. The material balance around the recovery
KL KI KL KF
boiler (Figure 13) cannot be readily used either,  
because there are more than three streams in- Na L  K L Na l  K I Na L  K L Na F  K F
volved (i.e. as-fired black liquor, smelt, precipita-
tor dust, and stack gas). A different approach is Thus
needed.
Cl mole (Fume) Cl mole (AF BL)
Consider an as-fired black liquor particle en- K mole (Fume) K mole (AF BL)
trained in the flue gas. As it burns and becomes a
carryover particle, some Na, K and Cl in the parti- Cl mole (Fume) Cl mole (Carryover) (20)
cle are inevitably lost to the flue gas due to vapori-
zation and chemical reactions, and subsequently
K mole (Fume) K mole (Carryover)
condense to form NaCl/KCl fume.
And
If NaI, KI and ClI are respectively the numbers
of moles of Na, K and Cl in the initial carryover EFCl(Fume) 1 EFCl(Fume) EFCl(Carryover) (21)
particle; NaL, KL and ClL are respectively the num- EFK(Fume) 1 EFK(Fume) EFK(Carryover)
bers of moles of Na, K and Cl that have been lost
to the flue gas and later become fume, and NaF, Equations (20) and (21) mean that the relationship
KF, and ClF, are respectively the numbers of moles between the Cl and K concentrations (expressed
of Na, K and Cl that remain in the final carryover either as mole% or enrichment factor) in carryo-
particle, then: ver, fume and as-fired black liquor is linear.
Since precipitator dust is a mixture of mostly fume
and a small amount of carryover, the composition
NaL = xNaI, KL = yKI, and ClL = zClI
of fume can be considered very close to that of
NaF = (1-x)NaI, KF = (1-y)KI, and ClF = (1-z)ClI precipitator dust. Furthermore, if the amounts of

165
 dust and Cl lost to stack gas are negligible (as in Acknowledgements
the case when the precipitator efficiency is high This work was conducted as part of the research
and there are little HCl emissions), the amounts of program on “Increasing Energy and Chemical
Na, Cl and K in smelt are equal to those in virgin Recovery Efficiency in the Kraft Process”, jointly
black liquor. supported by the Natural Sciences and Engineer-
ing Research Council of Canada (NSERC) and
Based on the relationships between fume and pre- a consortium of the following companies: Abiti-
cipitator dust, and between smelt and virgin black Bowater, Alstom Power, Andritz, Aracruz Celu-
liquor, and Equations (17) and (22), it can be con- lose, Babcock & Wilcox, Boise Paper Solutions,
cluded that the enrichment factors (or concentra- Carter Holt Harvey, Celulose Nipo-Brasileira,
tions) of Cl and K in fume, dust, carryover, smelt, Clyde-Bergemann, Diamond Power International,
virgin black liquor and as-fired black liquor are Domtar, DMI Peace River Pulp, Georgia Pacific,
linearly related, as summarized in Figure 17. International Paper, Irving Pulp & Paper, Metso
Power, MeadWestvaco, Stora Enso Research,
Tembec and Votorantim Celulose e Papel.

References
2 Dust 1. Tran, H. N., “Kraft Recovery Boilers-Chapter
9: Upper Furnace Deposition and Plugging”, Ed-
EFK

As-fired BL Fume ited by T. N. Adams, TAPPI Press (1997)

2. Backman, R., Hupa, M., Uppstu, E., ”Fouling
1
Virgin BL/Smelt
and Corrosion Mechanisms in the Recovery Su-
Carryover perheater Area”, TAPPI Journal, 70(6): 123-127
(1987)
3. Shenassa, R., Kuhn, D.C.S., Tran, H. N., ”The
1 2 Role of Liquid Content in Carryover Deposition
EF C l in Kraft Recovery Boilers-A Fundamental Study”,
Figure 17. Linear correlation between EFCl and Journal of Pulp and Paper Science Vol. 29, No. 4,
EFK of virgin black liquor, as-fired black liquor, p132-136, April 2003
carryover, fume and precipitator dust. 4. Khalaj-Zadeh, A., Kuhn, D.C.S., Tran, H.N.,
and Wessel, R., “Composition of Carryover Par-
ticles in Recovery Boilers”, Journal of Pulp & Pa-
Summary per Science, Vol 32, No.2, p.90-94 (2006)
A systematic study on carryover composition was 5. Khalaj-Zadeh, A., Tran , H.N., and Kuhn,
performed in the laboratory using an Entrained D.C.S., Verrill, C.L., “Field Studies on Composi-
Flow Reactor (EFR) and on three operating recov- tion of Carryover Particles in Recovery Boilers”,
ery boilers. The results show that carryover parti- Tappi Engineering Conference, Philadelphia, Au-
cles contain much less Cl and K than the original gust 2005.
black liquor. The sulphate (SO4) content increases 6. Janka, K., et al. “The Roles of Black Liquor
with an increase in exposure time and tempera- Composition and Boiler Combustion Parameters
ture, while the carbonate (CO3) content decreas- on the Fouling Tendency in a Recovery Boiler”, J.
es. The majority of sulphate in the carryover is Pulp Paper Sci., 26(6):201-207 (2000).
formed during combustion, prior to deposition.

The results also show a linear correlation between

the amounts of Cl and K in samples of as-fired
black liquor, smelt, precipitator dust and carryo-
ver deposits. This correlation can be theoretically
derived based on Na, K, and Cl balances around
the recovery boiler.

166
 167
View publication stats