Incident at Ajinomoto

OmniChem
Wetteren

Aug 23rd 2010- 12:40 h

Reaction between Starting material
and Me3SiN3 in tBuOH

J. De Keijser
Ajinomoto OmniChem
OmniChem Division
Safety manager
Dept - Date / 1
 Incident – 23 Aug 2010- 12:40 h

Reaction between Starting Material and Me3SiN3

(TMSA) in t-Butanol (in R824 – cf infra for a reactor
drawing)

Reaction under reflux-conditions 12 h

Explosion occured after 3 h of reaction

Batch 250

Process dates from early 90’s

Reactor
 Glass lined
 Reactor setup: glass

Condensors: glass/SiC, cooling medium at 45°C
(avoid HN3 condensation).
Ajinomoto OmniChem
OmniChem Division
Dept - Date / 2
 Trimethylsilylazide (TMSA)

Volatile organic azide (bp:95°C)

Reacts with water and forms hydrazoic acid
(HN3) (bp: 37°C)

Hydrazoic acid is unstable as a liquid and
may decompose by shock or vibrations

TMSA is considered as an alternative to HN3

Azides may form explosive compounds with
some metals (Cu, Pb ...)

Azides are toxic

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 3
 R824

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 4
 Consequences

No person injured !

Fire after explosion for approx 20 minutes:
cooling medium of condensors took fire

Pressure relief panels opened

Fire and explosion destroyed reactors setup
(glass)

Fire proof separations avoided spreading of
the fire and damage was limited to only that
part of the building. Other units within the
building were not affected

Butterfly valve closed – reactor contents were
unaffected
Ajinomoto OmniChem
OmniChem Division
Dept - Date / 5
 Location of the affected unit

SY1&2
IPC lab SY3&4

SY5&6
Main WH

DRY428

SY7
DRY478

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 6
 Compartimentation inside the
building

Ground floor 1st floor 2nd floor

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 7
Ajinomoto OmniChem
OmniChem Division
Dept - Date / 8
 R824 after the fire

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 9
 Where did the explosion start?
Timings of instrumentation failure
Tn t=10s Tn t= 0 s

Tr Tr
t=40s t=10 s

t= 20s t=20s
P P
T T0
t=
t= 40s t= 0 s

Tn Tn
t=40s t=40s

Tr Tr
t=40s t=40 s

t=30s t=50s
P P
T T
t= >80s t= 40s

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 10
 Potential Causes

Uncontrolled reaction (runaway)

Vapor explosion
 Internal
 External

Formation of excess of hydrazoic acid
 Explosion in vapor
 Decomposition of liquid

Formation of instable or shock sensitive
compounds

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 11
 Uncontrolled reaction?

High temperature increase of reaction mixture

at time of the event
 Temperature increase of 200°C/s and a pressure
increase of 10 bar in 10s is not consistent with
runaway (cfr Chilworth)
 T & P increase was due to failure of
instrumentation at the moment of the explosion or
just after

ARC’s on reaction mixture and on new R&D
samples did not give any concern

Adiabatic Dewar Calorimetry confirmed this
Ajinomoto OmniChem
OmniChem Division
Dept - Date / 12
 Vapor explosion

Inside reactor
 Reaction at reflux for 3h: oxygen concentration low
 No ignition sources inside reactor
 Integrity of glass lining checked during shutdown
 PSV on reactor still sealed
 Product profile of reaction mixture similar to normal reaction
 Reactor volume still the same
 Investigation of glass pieces by De Dietrich revealed that
there was an internal pressure on certain parts of the
glassware of 55 bar

Outside reactor
 No detection of components in ambient air (via FTIR)
 No pressure differences seen in reactor just before explosion
• Rupture disk OK
• Pressure transmitter: no differences
• Pressure test of reactor before startup: OK
Ajinomoto OmniChem
OmniChem Division
Dept - Date / 13
 Formation of hydrazoic acid?

In addition to what until then was known,new

R&D work using NIR showed that the normal
reaction goes via HN3

Temperature
after gas cooler
32-36°C

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 14

What about batch 1 to 249? Impurities?
 Water?
• KF before start of reaction was in spec
 Cooling medium?
• No Shellsol nor Marlowtherm found in reaction mixture
 Nitrogen
• All normal
 Metals
• Cu, Ta, Ni,
• Could not be detected
 User testing of the precise starting materials, utilized in the
batch # 250, have shown an acceleration of the reaction
 Analyses of starting material showed higher amounts of K in
the 2 input batches used for #250
• 660 and 340 ppm, compared to less than 50 ppm (mostly not
detectable) in other batches

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 15
Metal analysis on starting material input lots
mg/kg
metal B #1 B#2
Al --- 17,2 Were analysed for, but not detectable:
Ba --- 0,2 Ag Fe Sn
Ca --- 32,9 Al K Sr
Cr --- --- As Mg Ti
Fe --- --- Ba Mn te
K 11,4 348,6 Be Mo Tl
Mg 0,2 0,6 Ca Na V
Mn --- --- Cd Ni W
Na 3,7 4,2 Co Pb Zn
Ni 0,2 --- Cr Sb Zr
Sb 0,3 --- Cu Se

Significant increase in K, Ca and Al in both lots

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 16
 Evaluation of influence K, Ca and Al in
reaction at 80°C after 3 hours

Imp. 1 Final Intermedia Starting

product te material
CaCO3 (2200 ppm Ca) 4,1 78,0 2,3 10,7
KHCO3 (9700 ppm K) 0,9 90,1 3,4 1,9
Al powder (2000 ppm Al) 10,0 78,3 2,3 4,5
Al2O3 (2600 ppm Al) 6,5 77,2 2,4 8,9
Blank 5,1 74,4 2,4 14,9

Taken into account the high amounts of “catalysts” added,

only KHCO3 increases reaction speed significantly

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 17
 Potassium in Starting Material
Starting sulphated
material lot ppm K ash
60530060 20 <0,05

Increased amounts of K 60530061 50 <0,05

60530067 nd <0,05
found in batches 605360084
60530068 nd <0,05
and 605360085: input of
60530069 nd 0,05
486 ppm 60530070 nd <0,05

Confirmation by analyses on 60530073 nd <0,05
reaction mixture: 57 ppm 60530074 nd 0,07
calculated on 582 ppm input 60530075 nd <0,05
60530076 279 0,07
60530077 10 <0,05
Confirmation that lots used in 60530082 <100 <0,05
batch 250 are out of trend 60530083 nd <0,05
for potassium content 60530084 659 0,13
60530085 340 0,06
Ajinomoto OmniChem
OmniChem Division 60530086 <100 <0,05
Dept - Date / 18
 tBuOH + TMSA: Catalysis by KHCO3

Influence of 780 ppm K added as KHCO3 on reaction
between tBuOH and TMSA at 80-82°C

120

100
residual TMSA (%)

80

blanc
60
KHCO3

40

20

0
0 1 2 3 4
time (h)
Ajinomoto OmniChem
OmniChem Division
Dept - Date / 19
 Possible potassium impurities

Potassium source: bicarbonate from extractions in previous reaction stage

Consequence HN3 build up in closed reactor – Yellow line = input B58

160

140

120

100

80

60

40

20

0
0 1 2 3 4

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 20
 NIR Measurements
B58 containing increased K: Accumulation after Explosion behind
fast increase HN3 to 10% condensor condensor

12

10

8
HN3. (%v/v)

0
0:00:00 1:12:00 2:24:00 3:36:00 4:48:00 6:00:00 7:12:00
time (hh:mm:ss)

Blue curve =Normal batch: slow

Ajinomoto OmniChem
OmniChem Division
increase HN3 (closed lab reactor)
Dept - Date / 21
 Shock sensitive compounds

Non-ferro metals in reactor or glassware

Formation of unstable compounds in the
process

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 22
 Root cause

We have proven that a weak base (under the form of potassium

or sodium bicarbonate, acetate and butoxide) increases the
reaction between t. butanol and TMS azide (stage one)
significantly. After 1 hour of reaction at 80°C onl y 13% of TMS
azide is left in the reaction mixture

As a consequence of this previously unknown catalytic effect,

the formation of hydrazoic acid is occurring in a shorter period of
time and concentration of hydrazoic acid increases, leading to
free hydrazoic acid in the gas phase. The diffusion of the gas in
the dead volume beyond the condenser (condensation) has led
to the explosion, initiating the fire.

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 23
 Countermeasures (1)

Future research on Process improvements:
 Modification of the actual process avoiding large quantities of
hydrazoic acid formed (catalytic effects, temperature, dosage
control)
 Alternative process avoiding the use of trimethylsilylazide or use it
at other process-temperature
 Monitoring of reaction mixture and headspace of the adapted
process to assure safe operation

Screening of all other azide reactions: carried out on running

processes / required for future processes

Headspace monitoring by NIR

 All azide reactions
 Emergency levels defined and implemented

Tracing of lines above condensation temperature of HN3

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 24
 Countermeasures (2)

Reducing the fire load in the production cells

Revision of the internal emergency plan

Increased process understanding

 NIR in liquid – done
 Consider investment in FT-IR or Raman

Task force “specs & process safety”

 How monitor O.O.T. in raw materials and intermediates
 Idea “safety by monitoring”  Plan ready and in execution
• Triggers for dangerous reactions defined
• Purchasing ICP-MS ongoing
• Safety by Design => design space formed by metal imps, organic imps,
water, pH value of input materials

Ajinomoto OmniChem
OmniChem Division
Dept - Date / 25
 Actual status – example HN3 gas phase trending in Pilot

1,80 25,0

Quench using NaNO2/lactic acid

1,60
20,0

1,40

15,0

1,20

HN3 (Quant 2 - destruction)

10,0 NaNO2 dosage
1,00
HN3 (v/v%)

Temp. (°C)
lactic acid dosage
water dosage
TMSA dosage
0,80
5,0 temp reactor
temp bottom

0,60

0,0

0,40
Azide reaction
-5,0
0,20

0,00 -10,0
0,00 5,00 10,00 15,00 20,00 25,00 30,00 35,00
Time (h)
Questions ?