Technical Paper

Critical material issues in cast aluminium

cylinder heads
Yucel Birol* and Ali Aslan Ebrinç** flame-deck wall because of the prevailing cyclic
tensile stresses, while the thin-walled valve
*Materials Institute, Marmara Research Center, bridge areas exposed to high temperatures may
Tubitak, Kocaeli, Turkey suffer cracking due to low-cycle fatigue (LCF). LCF
**Gebze Engineering, Ford Otosan, Kocaeli, Turkey resistance requires high temperature strength
which in turn relies on alloy composition while
This paper was presented at the World Foundry the microstructure, porosity and surface quality
are the key parameters for HCF life(2-11).
Congress in Chennai, India in February 2008. It
received a ‘commended’ award from the World Any feature that promotes stress concentration
and contributes to crack initiation or crack
Foundrymen Organization (WFO). It is reprinted with
propagation is a major threat both to strength
permission of the WFO and the Institute of Indian and ductility. Manufacturing practice takes every
Foundrymen. measure to reduce the quantity and size of such
features in order to improve the overall quality of
Abstract
the casting. This paper reports the results of an
Regulations calling for emission control and fuel savings and thus
extensive investigation of cylinder head castings
lightweighting of vehicles have increased the use of aluminium in
taken from serial production and submitted to
the manufacture of vehicles. A major application of aluminium in
performance tests. The material issues critical for
the automotive industry is cylinder heads. In this application, the
aluminium cylinder head castings are outlined.
main advantage of aluminium, besides its lower density, is the
excellent thermal conductivity, which allows the combustion heat to
Experimental
be extracted more rapidly. As a result of the permanent increase of
Cylinder heads were cast in Sr-modified EN1706
combustion pressures and temperatures, material issues play a key
AlSi10Mg alloy (AC43100) which provides
role in the performance of aluminium cylinder heads. These material
higher ductility than most secondary alloys
issues and their effect on performance are discussed in this paper.
thanks to its lower limits for impurity elements.
A low pressure casting process was employed
Introduction and made sophisticated and complex water-
Aluminium alloys have almost completely replaced grey iron in jacket designs possible with the application
cylinder head castings during the past 20 years thanks to their of sand cores. The solution heat treatment
lower density and excellent thermal conductivity, which allows the was performed in a drop-bottom furnace for
combustion heat to be extracted more rapidly(1). Exposed to high- five hours at 530°C. The heads were plunged
cycle fatigue (HCF) due to millions of combustion cycles and to low- into either water or polymer-water maintained
cycle fatigue (LCF) resulting from thermal expansion and contraction at a minimum temperature of 75°C and 50°C,
during start-up and engine stop up to 10 times a day, cylinder heads respectively. Artificial ageing was performed at
are certainly a demanding application for aluminium alloys. Critical 170°C for four hours. A total of forty generic
Figures fatigue (HCF) areas are on the water jacket side of the
high-cycle cylinder heads were taken from serial production,
were machined and assembled before they were
b submitted to the following performance tests.
a
Gasket thermal test cycle, designed to verify
the cylinder head gasket, is one of the best
test procedures for the engine cylinder heads.
Test duration is 635 hours. The purpose of this
dynamometer test is to investigate durability
and reliability characteristics of static and
dynamic sealing systems under extreme thermal
conditions and the cylinder head. Test cycle is
200!m 40!m
based on the cooling and heating of coolant and
c d oil circulating in the engine externally. Coolant
temperature varies between -20°C and 98±2°C
as inlet and exit, respectively. Oil sump outlet
temperature should be as high as possible (target
145°C).
Customer correlated general durability is the
test to correlate the customer driving behaviour
on the dyno engine. Test duration is 1202 hours
20!m 100!m and covers various aspects of engine running
conditions. Test cycle includes certain number
Fig. 1 Optical (a,b) and scanning electron (c,d) micrographs of gas (a,c) and of full power and maximum engine rpm steps.
Figure 1. (b,d)
shrinkage Optical (a,b)
porosity and scanning
in cylinder electron (c,d) micrographs of gas
head castings
(a,c) and shrinkage (b,d) porosity in cylinder head castings.

196 FTJ July/August 2008

a b B
 Figure 1. Optical (a,b) and scanning electron (c,d) micrographs of gas
Technical Paper
(a,c) and shrinkage (b,d) porosity in cylinder head castings.

The engine faces a major challenge at max a b B

power/torque for about 180 hours of the total
test duration. Cyclic gas load is the the biggest
contribution to the HCF. Engine coolant and oil B
are cooled internally by engine oil cooler and
radiator.    1 mm

High speed test lasts 825 hours and the engine c A

runs mostly at high engine speeds together with
the maximum power/load to the cylinder head. A
Cycling loading on the cylinder head is effective
again for HCF under high engine rpms.
1 mm
Maximum torque test is a 500 hour long test
Fig. 2 (a)2.Distinction of porosity in cylinder head castings,
head(b) site B is nearly
(b) Sitefree
which forces piston cylinders under high cylinder Figure (a) Distribution of porosity
a any porosity while (c) site A is crowded b
in cylinder castings: B is
of with pores
peak fire pressure. Together with pistons, nearly free of any porosity while (c) site A is crowded with pores.
cylinder head is naturally tested with cyclic high
peak fire pressure weakining the cylinder head a b
for HCF’s.
1
Forty cylinder heads from 10 different
production campaigns over a period of 12
months were submitted to the performance tests
described above. Non-destructive techniques
were employed to find out about the presence 200!m 0,5 mm
and location of fatigue craks. These heads were
subsequently sectioned. Small samples were Figure 3. (a) Pore which is involved in crack initiation near the water
removed from predetermined critical regions, jacket side of the flame deck; (b) crack surface with a dense distribution
200!m 0,5 mmof

adjacent to the water jacket side of the flame pores.

Fig.3 (a) Pore which is involved in crack initiation near the water jacker side of the
Figure 3. (a) Pore which is involved in crack initiation near the water
deck shown by finite element analysis to suffer flame deck, (b) crack surface with a dense distribution of pores
jacket side of the flame deck; (b) crack surface with a dense distribution of
the highest cyclic stresses. Those with fatigue pores.
cracks were further investigated with the a b
common failure analysis techniques in order to
identify the cause of cracking and were tested
also for hardness, chemical composition and a b
microstructural features. Standard metallographic
practices were employed for the preparation of
samples.

Results and Discussion

The chemical analysis, the tensile test results, 40!m 40!m
and the hardness values are given in Tables 1 Fig. 4 AlFeSi
Figure intermetallic
4. AlFeSi particles (a)
intermetallic which initiate
particles a fatigue
(a) which crack,
initiate (b) which
a fatigue crack;
and 2, respectively. The quality index (Q.I.= UTS block interdendritic
(b) which melt flow and
block interdendritic thus
melt leadand
flow to porosity
thus lead to porosity.
40!m 40!m
+ 150 log (%elongation) estimated from tensile
test results were found to vary between 312 Figure 4. AlFeSi intermetallic particles (a) which initiate a fatigue crack;
MPa and 487 MPa. Several of the head castings (b) which block interdendritic melt flow and thus lead to porosity.
Si Fe Cu Mn Mg Zn Ti Sr
were found to contain fatigue cracks after the a b
min 9.40 0.092 0.018 0.010 0.241 0.004 0.131 0.009
performance tests. The features which were max 10.33 0.182 0.077 0.023 0.409 0.026 0.183 0.032
found to have an impact on the performance of a b
cylinder heads are discussed in the following
Table 1. Chemical composition range of the cylinder heads invesitigated in the
section. present work

Porosity
Both gas and shrinkage porosity were UTS, MPa YS, MPa Elongation, % Hardness, HB
encountered in aluminium cylinder head min 240 169 20!m 4 93.4 20!m
castings (fig. 1). The distribution was rather max 351 273 13 106.9
heterogeneous with many more pores Figure 5. (a) Silicon particles adequaely modified and (b) silicon plates
ave 329 ± 9 238 ± 1220!m 9 ± 2 100 ± 9
located near the water jacket core where the which have initiated fatigue cracking. 20!m

solidification rate is relatively slower (fig. 2). Table 2. 5.

Figure Tensile
(a) tests performed
Silicon on specimens
particles adequaelytaken from the
modified andfront
(b)face of theplates
silicon
Porosity has to be controlled to a minimum at cylinderhave
which heads invesitigated
initiated in cracking.
fatigue the present work
critical locations of the head which are exposed
to HCF, i.e. in the valve areas in the flame deck 2
or the walls between flame deck and water in crack propagation. The gas porosity was found to be reduced
jacket. Pores as small as 100 microns near the substantially when the density index
2 during casting was kept below
water jacket side of the flame deck were found two. Gas porosity was a major quality problem, however, when the
to initiate cracks (fig. 3a). Those that did not density index was allowed to exceed four. An exceptionally poor
take part in crack nucleation were often involved performance in one cylinder which has failed before the test was

FTJ July/August 2008 197

 40!m 40!m

Technical Papermelt flow and thus lead to porosity.

Figure 4. AlFeSi intermetallic particles (a) which initiate a fatigue crack;
(b) which block interdendritic

a b Modification of eutectic silicon

Sr, as much as 0.02 wt%, was found to alter the
shape of the eutectic silicon phase from coarse,
plates and needles to small, finely divided
fibers (fig. 5a). In several heads, however,
Si morphology was found to vary from fully
modified, refined fibers to needles and platelets.
While lower than required Sr addition was
responsible for some of these cases, the poor
20!m 20!m Si morphology was almost always encountered
Fig. 5 (a) Silicon particles adequately modified and (b) silicon plates which have near the water jacket side of the flame deck
Figure 5. (a) Silicon particles adequaely modified and (b) silicon plates
initiated fatigue cracking
a which have initiated fatigue cracking. b and was thus judged to result from slower
soldification. These coarse broad brittle Si plates
a b acted as stress raisers and provided planes of
weakness that promoted both crack initiation
2 and propagation (fig. 5b). Solution treatment
employed in the present work was found to
modify the eutectic Si further owing to high
enough solution heat treatment temperatures to
facilitate Si spheroidisation.
100!m 100!m
Solidification rate
100!m 100!m
Figure 6. Effect of solidification rate on microstructural features: local Rapid solidification provides the greatest
Figure 6. of
solidification
Fig. 6 Effect Effect
time: (a)of100s
solidification
solidification andon(b)
rate rate on microstructural
10s.
microstructural features: local features: local
solidification improvement in microstructural features and
time (a) 100s and
solidification (b) (a)
time: 10s100s and (b) 10s. fatigue strength than any other means available
to the foundryman. Rapid solidification does
this by refining the microstructure and thus
by dispersing shrinkage porosity, by reducing
the extent of segregation and thus the size
and volume fraction of intermetallic phases, by
improving the modification of the eutectic Si.
The rate of solidification is measured by the
secondary dendrite arm spacing (SDAS) which
thus serves as a significant quality parameter.
It is unfortunate, however, that the
solidification is the slowest in those regions of
100!m
100!m the cylinder head which are exposed to HCF and
Figure 7. A notch at a rough surface where a crack has initiated. are thus most critical. The SDAS measured in
Fig. 77.
Figure A notch at a rough
A notch surfacesurface
at a rough where a where
crack has initiated
a crack has initiated.
the water jacket side of the flame deck varies
between 31 to 47 microns in the present work.
120
There were no cracks in heads where the SDAS
120
over, was readily linked with extensive porosity dispersed across the was below 35 microns while the frequency of
section
100 of quench
Water the valve bridge area, once it was fractured after soaking in fatigue cracks almost doubled when the SDAS
liquid
100 nitrogen
Water quench (fig. 3b). While both size and the number of shrinkageW1
was higher than 40 microns.
max principal stress [MPa]

W4
pores were smaller with respect to those originating from dissolved W1
Two arbitrary parts were cast in AlSi10Mg alloy
max principal stress [MPa]

W6
80 W4
hydrogen, they were judged to be more detrimentalHCF due to their W9
Stress
W18 W6
from two different pouring temperatures into
80
shapes. Shrinkage porosity was almost invariably interdendritic
limit
HCF Stress and
P2 W9 a permanent mould merely to achieve different
60
solidification rates by simply adjusting the
P11 W18
was largerPolymer
in size in heads with a coarser microstructure.
quench limit P17 P2

60 (process variation) P22

P11 superheat of the melt. The overall density of the
P12mach
Polymer quench
Intermetallic particles
40
P17
P14mach
P22
parts cast from 730°C and 690°C were measured
(process variation) L8
to be 2.607g/cm3 and 2.657g/cm3, respectively.
The
40 Fe content of the cylinder heads in the present work was L16
P12mach
P14mach
This difference in density corresponds to a
controlled adequately and kept below 0.18 %wt. Nevertheless, the L8
L24
20
porosity difference of 2% by volume and is
number as well as the size of the β-Al5FeSi platelets were found to L16
20 Air quench L24 substantial. The microstructural features of the
increase
0 with increasing Fe content even in this very narrow range.
two parts are also strikingly different (fig. 6).
These particles1 were most frequent 2
near the water jacket
3
surface. The part cast with a higher superheat and which
Air quench strain gage locations
This
0 is basically due to the relatively lower rate solidification which has thus solidified slowly shows a significant
prevail
Figure near the
1
8. Residual water jacketmeasured
stresses core.2
β-Al FeSi platelets
in 5cylinder
3
not only
head castings. amount of shrinkage porosity while that cast with
strain gage locations
degrade the continuity of the aluminium matrix, act as stress raisers limited superheat and thus solidified very fast
and, thus, aid crack initiation (fig. 4a), but also physically block is perfectly sound. It is also worth noting that
Figure 8. Residual stresses measured in cylinder head castings.
the interdendritic melt flow, impair local feeding and thus led to SDAS values of the slowly and rapidly solidified
microshrinkage porosity (fig. 4b). High 3 rates of solidification are well parts were approximately 60 and 20 microns,
established to favor the more compact cubic α phase which forms in respectively. It is fair to conclude that achieving
chinese script morphology. Likewise, replacing some of the Fe with solidification rates which produces a SDAS
Mn produces a similar effect and promotes 3 the transformation of the value of 20 microns ensures a sound casting
β-Al5FeSi platelets into α script which is less damaging to mechanical with no porosity. It is thus advisable to target
properties. a SDAS value as small as possible (preferably

198 FTJ July/August 2008

 Technical Paper
100!m

Figure 7. A notch at a rough surface where a crack has initiated.

below 30 microns) near the water jacket surface 120

by enhancing the solidification rate in this
otherwise, slowly cooling regions of the cylinder
100 Water quench
head. W1

max principal stress [MPa]

W4

Surface roughness 80
W6
W9
HCF Stress
There were several cases where rough water limit
W18
P2
jacket surfaces due to insufficient sand quality 60 P11

used in the manufacture of cores played a key Polymer quench P17

P22
(process variation)
role in crack initiation (fig. 7). This problem was 40
P12mach
P14mach
fixed by using finer and higher quality sand. L8

The flame deck surface quality is critical, not L16

L24
20
only because notches produced by unacceptable
surface roughness serve as sites for fatigue crack Air quench
initiation but also because flow conditions of 0
1 2 3
the incoming gas are critical for the combustion strain gage locations
process.
Fig. 8 Residual
Figure stressesstresses
8. Residual measuredmeasured
in cylinderin
head castings
cylinder head castings.
Residual stresses
Quenching from solution treatment leads to
the quenching process.
varying and sometimes extremely high residual
stresses. Unless eliminated or reduced, these 3
stresses can exceed, together with the stresses
References
1. EAA The Aluminium Automotive Manual: www.
arising from operation, the fatigue endurance
aluminiumtechnologie.de/aam
limit and finally lead to premature cracking
2. Gundlach R B, Ross B, Hetke A, Valtierra S, Mojica J F, ‘Thermal
in service. The highest residual stresses were
fatigue resistance of hypoeutectic aluminum-silicon casting alloys’,
found by experimental measurements to occur
AFS Transactions, Paper no 94-141, 1994, pp 205-223.
at typical crack locations implying that residual
3. Otte M O, McDonald S D, Taylor J A, St John D H, Schenider W,
stresses played a key role in crack initiation and
‘Controlling porosity-related casting rejects: understanding the
propagation.
role of iron in Al-Si alloys’, AFS Transactions, Paper no 99-112,
Residual stresses measured at critical water
1999, pp 471-478.
jacket core regions in cylinder heads quenched
4. Hetke A, Gundlach R B, ‘Aluminum casting quality in alloy 356
in water, water-polymer and in air are shown in
engine components’, AFS Transactions, Paper no 94-137, 1994, pp
fig. 8. Residual stresses vary between 80-105
367-380.
MPa in cylinder heads quenched in water (head
5. Feikus F J, ‘Optimization of Al-Si cast alloys for cylinder head
#:1, 4, 6, 9 and 18). Those measured in cylinder
applications’, AFS Transactions, Paper no 98-61, 1998, pp 225-
heads quenched in polymer-water (head #: 2,
231.
11, 17, 22, 12 and 14) ranged between 15 and
6. Verren O, Evensen J E, Pedersen T B, ‘Microstructure and
55MPa. Finally, residual stresses measured
mechanical properties of AlSi(Mg) casting alloys’, AFS
in air-quenched cylinder heads were all lower
Transactions, Paper no 84-162, 1984, pp 459-466.
than 15 MPa. It is clear from the foregoing that
7. Grunenberg N, Escherle E, Sturm J C, ‘Simulation for process
air-quenching practice gives the lowest residual
design and optimization of aluminum cylinder heads’, AFS
stresses. With an average residual stress level
Transactions, Paper no 99-45, 1999, pp 153-159.
(40 MPa) safely below the fatigue endurance limit
8. Samuel A M, Samuel F H, ‘Porosity factor in quality aluminum
of the present AlSi10Mg alloy (80 MPa) polymer-
castings’, AFS Transactions, Paper no 92-11, 1992, pp 657-666.
water quenching practice was adopted at the
9. Crepeau P N, ‘Effect of iron in Al-Si casting alloys’, AFS
foundry since air quenching requires a rather
Transactions, Paper no 95-110, 1995, pp 361-366.
high initial investment.
10. Zhang B, Sung P K, Poirier D R, Chen W, ‘Effects of strontium
modification and hydrogen content on fatigue behavior of A356.2
Conclusions Al alloy’, AFS Transactions, Paper no 00-42, 2000, pp 383-389.
The cylinder heads which were cast with a
11. Apelian D, Shivkumar S, Sigworth G, ‘Fundamental aspects of heat
sufficient level of eutectic silicon modification,
treatment of cast Al-Si-Mg alloys’, AFS Transactions, Paper no 89-
with a limited amount of hydrogen and shrinkage
137, 1989, pp 727-742.
porosity, with sufficiently high solidification rates
to attain SDAS values below 35 microns, with Fe
Acknowledgements
levels controlled under 0.12 wt%, with sufficient
It is a pleasure to thank Mr O Çakır and F Alageyik for their help in
surface quality were never seen to fail in the
the experimental part of this work and to E Mentese of Componenta
performance tests. It is also worth noting that the
Doktas for providing the tensile test results.
likelyhood of fatigue cracking was almost none
in the cylinder heads which were quenched from Yucel Birol; e-mail: [email protected]
the solution heat treatment in polymer-water so
Ali A. Ebrinç: e-mail: [email protected]
as to limit the residual stresses originating from

FTJ July/August 2008 199