Grain Morphology of As-Cast Wrought Aluminium Alloys*

Mark Easton1;2, Cameron Davidson1;3 and David StJohn1;4

1CAST Co-operative Research Centre, Monash University, Vic 3800, Australia2Department of Materials Engineering, Wellington Road, Monash University, Vic 3800 Australia3CSIRO Process Science and Engineering, 1 Technology Court, Pullenvale, Qld 4069, Australia4School of Mechanical and Mining Engineering, Materials Division, The University of Queensland, Qld 4072, Australia

Two of the most important microstructural features of alloys are the grain size and the secondary dendrite arm spacing (SDAS) and thesetwo factors are shown to combine together to describe the grain morphology. Both grain refinement and the SDAS depend upon alloycomposition through constitutional undercooling, but in different ways. It is shown that there is a ‘characteristic’ SDAS for each alloy at aparticular cooling rate, and reducing the grain size causes the grain morphology to change from dendritic to cellular/rosette-like to globular orspherical. Increasing the cooling rate refines both the SDAS and the grain size, but reduces the SDAS more rapidly leading to finer, moredendritic grain structures. Particular ratios of grain size to SDAS are used to define each morphology and it is shown how these can assist withobtaining a required grain size and morphology through the use of solidification conditions and alloy chemistry.[doi:10.2320/matertrans.L-MZ201118]

(Received October 5, 2010; Accepted December 14, 2010; Published May 1, 2011)

Keywords: grain size, dendrite arm spacing, cooling rate, constitutional undercooling

1. Introduction

Two of the key microstructural parameters in the as-castmicrostructure of an alloy are the grain size and thesecondary dendrite arm spacing (SDAS). Fundamentally,these parameters can be used to describe the morphology of agrain. Where the grain size is large and the dendrite armspacing is small, the grains are dendritic; when the grain sizeis only a little greater than the dendrite arm spacing then arosette or cellular grain morphology is observed, and ifthe grains are spherical or globular then no dendrites areobserved. The grain size, SDAS and grain morphology affectvarious properties of the alloy including their castability,1,2)

especially hot tearing resistance,3) their mechanical proper-ties,4) and are especially important in semi-solid process-ing.5–7)

The equiaxed grain morphologies have some similarity tothe planar, cellular and dendritic growth that have beencharacterized in columnar solidification,8) where the classi-cal approach attributes the growth morphology to thepropensity of a perturbation to grow into a constitutionallyundercooled region. Equiaxed grains tend to grow dendriti-cally because their growth is unconstrained by the temper-ature gradient. The morphology of equiaxed grains has oftenbeen related to the breakdown of spherical growth.5,9)

However, it is also known that dendrites coarsen duringsoldification10) to the point that the grains become spher-ical.11) There is also evidence in peritectic systems such asAl-Ti and Mg-Zr that whilst initial growth is dendritic asshown by the morphology of the Ti-12) or Zr-enrichedregions,13) the final grain morphology can be spherical orglobular.

Hence the grain morphology can be described as acompetition between the nucleation processes related to

grain refinement and the ripening processes, if it is assumedthat the grain will grow dendritically. This is shownschematically in Fig. 1. Whilst there are some situationswhere grains will grow spherically,5,9) e.g. skip stage 2 andgo directly to stage 3 or 4 in the figure, it is likely that thesame processes are acting. In the end for the solidification ofalloys, the SDAS will be determined by when in thecoarsening process the eutectic reactions that delineate thedendrites occur.14) Hence, it appears that analytical and/orempirical models for grain size and SDAS can be used todescribe the grain morphology and be complimentary tocomputational approaches undertaken by other research-ers.15)

There has been considerable work on determining thefactors that affect grain size. These include the alloycomposition,16–18) the number density and potency ofnucleant particles19) (which is influenced by particle size20)),and the cooling conditions.21,22) Some of these factors areknown to also affect the SDAS, e.g. the cooling rate and thealloy composition.14,23,24)

The factor that is influenced by both alloy composition andcooling rate is constitutional undercooling (CU) and thisinfluences both the grain size and the SDAS. The parameterthat is used to link both factors is the growth restrictionfactor, Q0,

Fig. 1 A schematic of the growth of grains during solidification. Initially

nucleation occurs, the grains begin to grow dendritically, until they

impinge and then the dendrites coarsen potentially to rosette-shaped and

then to globular or spherical grains.

*The Paper Contains Partial Overlap with the ICAA12 Proceedings by

USB under the Permission of the Editorial Committee.

Materials Transactions, Vol. 52, No. 5 (2011) pp. 842 to 847Special Issue on Aluminium Alloys 2010#2011 The Japan Institute of Light Metals

Q0 ¼X

jmlc0; jðkj � 1Þ ð1Þ

where ml is the liquidus gradient, c0 is the alloy compositionand k is the partition coefficient for each element j in thealloy. A simple equation has been found to relate the grainsize, d, and the alloy constitution through this parameter,19,25)

i.e.

d ¼1ffiffiffiffiffiffiffiffif Nv

3p þ

D ��Tn

v � Q0

ð2Þ

where �Tn is the nucleation undercooling, Nv is the numberof those particles, and f is the proportion that are active, D isthe diffusion coefficient and v is the growth velocity. Thecooling rate affects the growth velocity through a square rootrelationship.17,21)

The relationship between the SDAS and Q is not quite assimple. Using a diffusion ripening model, the SDAS, �2 canbe related to the solidification time, tf , through

�2 ¼ 5:5ðMtf Þ0:33 ð3Þ

M ¼��

Xnj¼1

ðQ0; j � Qf ; jÞ=Dj

� ln

Xnj¼1

�Qf ; j=Dj

Xnj¼1

�Q0; j=Dj

266664

377775ð3aÞ

for multi-component alloys.26) In this case there are twoparameters related to Q: Q0 at the beginning of solidification,which is the same as used in eq. (2) and Qf at the end ofsolidification which is determined using the composition ofthe final segregated liquid. This formulation works reason-ably well for the prediction of SDAS in many alloys, but islimited by the assumptions in the model and factors otherthan diffusion controlled-ripening that influence the SDAS incommercial alloys.27)

Previous publications have focused on the effect of alloyand solidification variables on the grain size19,21,22) and theSDAS27) separately. The aim of the paper is to emphasisehow this work can be used in a description of grainmorphology.

2. Experimental

The details of the experimental procedure are describedelsewhere.19,21,27) In summary, four wrought alloys from the

1000 to 6000 series, which have a variety of alloyingelements and Q values, were solidified at four cooling ratesranging from 0.3 to 15�C/s. A constant addition of 0.005mass% TiB2 by an Al-3 mass%Ti-1 mass%B master alloywas made to each alloy so that the nucleant population wasessentially equivalent for all of the alloys investigated. At acooling rate of 1�C/s an additional three alloys were studiedat two additional TiB2 contents. Castings 40 mm high and30 mm in diameter were made in pre-heated graphite moulds.Typical compositions of the alloys and the Q0

28) and Qf27)

values are given in Table 1.Metallographic samples were sectioned approximately

10 mm from the base of the ingots. These were mechanicallypolished and anodized using a 0.5% HBF4 solution forapproximately 2 min at 20 V. These samples were viewedoptically with polarized light and the grain sizes weremeasured using a linear intercept technique (ASTM 112-96).At least two fields were calculated for each measurementwith approximately 100 intercepts in each field depending onthe grain size. The dendrite arm spacing was measured byfinding a number of adjacent dendrite arms from one primarydendrite arm and measuring the distance between thedendrite centres. Wherever possible at least two fields wereanalysed for each measurement with approximately 50secondary dendrite arms from multiple primary dendritearms being counted in each field.

3. Experimental Observations

The addition of Ti solute leads to a substantial decrease inthe grain size for low solute alloys,18,29) such as alloy 1050(Fig. 2(a)), where the value of Q0 is approximately 2. This isbecause Ti has a relative Q value 40 to 100 times that of mostalloying elements.30) Likewise in high solute alloys such asalloy 2014, the value of Q0 is approximately 22 and theaddition of Ti has little effect on the grain size of alloys(Fig. 2(b)). However, the Ti content has little effect on theSDAS in either case. The other alloys studied had Q0 valuesbetween these two alloys and the effect that Ti soluteadditions had on the grain size decreased as the Q0 valueincreased. The addition of Ti had no observable effect on theSDAS in any of the alloys.27)

When grain refiners are added, both TiB2 particles andsolute Ti are usually added to the alloy. It has been

Table 1 The composition of the alloys made and an estimation of the Q0 and Qf values of the alloys for an average composition of the

alloys cast. Both Q0 and Qf were determined using the additive methodology (eq. (1)). Whilst this approach has shown to be suitable for

these alloys for Q0,28Þ it is less well understood for Qf . The compositional values for Qf were determined using data from the Gulliver-

Scheil model in Thermocalc and were chosen at a major eutectic reaction towards the end of solidification where intermetallics that

outlined the dendrites were observed. More details have been provided previously.27Þ

AlloySi

(mass%)

Cu

(mass%)

Fe

(mass%)

Mg

(mass%)

Cr

(mass%)

Mn

(mass%)

Q0

(K)

Qf

(K)

1050 0.2 0.27 2.0 29.7

2014 1.09 4.78 0.51 0.45 0.45 22.8 89.8

3003 0.55 0.12 0.58 0.01 1.37 5.4 50.7

5083 0.35 0.29 4.24 0.17 0.86 16.5 53.8

6060 0.54 0.12 0.31 4.5 56.8

6061 0.83 0.21 0.05 0.73 7.8 84.6

6082 1.09 0.04 0.6 0.46 8.4 93.5

Grain Morphology of As-Cast Wrought Aluminium Alloys 843

established that unlike the addition of Ti solute, the additionof TiB2 particles reduces the grain size in all alloys byapproximately the same amount,19) in line with eq. (2).However, again the SDAS is unchanged.

The effects of the combination of TiB2 additions and Tisolute additions on the grain size and morphology are shownin Fig. 3. Especially in alloy 1050 but also alloy 2014, theinitial addition of grain refiner substantially decreases thegrain size whilst the grain morphology remains dendritic. Afurther reduction in grain size leads to the development of acellular or rosette morphology. It is impossible to measureSDAS on grains with such morphology, and consequentlythis point is missing from Fig. 2(a). It appears that when thegrain size to SDAS ratio is below 3 and especially when it isclose to 2, the measurement of SDAS becomes impossible.

Another important factor that affects both the grain sizeand the SDAS is the cooling rate. This is particularlyobserved in Fig. 4, where it is apparent that when the coolingrate increases the grain size decreases, but there is also anobvious change in grain morphology from spherical to

(a)

(b)

Fig. 2 Variation of grain size and SDAS with the amount of Ti for (a) alloy

1050 and (b) 2014 at a cooling rate of 1�C/s and a TiB2 addition of

0.005 mass% as Al-3Ti-1B. There is no point for SDAS at 0.05 mass% Ti

for 1050 in (a) as the grains were no longer dendritic. Data sourced from

Easton et al.19,27)

(d)(a)

(e)(b)

(f)(c)

500µm

500µm

500µm

Fig. 3 Optical micrographs viewed under polarised light. Alloy 1050 is

shown in (a) to (c), and alloy 2014 is (d) to (f). Alloy 1050: (a) without

grain refiner addition; (b) with 0.01 mass% TiB2 using a Al-3Ti-1B master

alloy; (c) with 0.02 mass% TiB2 and the addition of 0.05 mass% Ti solute.

Alloy 2014: (d) without grain refiner addition (e) with 0.01 mass% TiB2

using a Al-3Ti-1B master alloy and (f) with the addition of 0.05Ti solute.

The scale bar on all images is 500 mm.

(a)

(c)

(b)

(d)

500µm500µm

500µm 500µm

Fig. 4 Optical micrographs viewed under polarised light of Alloy 1050.

(a) Grain refined with 0.005 mass% TiB2 addition at a cooling rate of

0.3�C/s; (b) with a further addition of 0.05 mass% Ti solute addition also

at 0.3�C/s; (c) 0.005 mass% TiB2 at a cooling rate of 15�C/s and (d) with a

further addition of 0.05 mass% Ti solute addition also at 15�C/s. The scale

bar on the images is 500mm.

844 M. Easton, C. Davidson and D. StJohn

equiaxed dendritic. Also note the evidence of a priordendritic structure in Fig. 4(b) due to the presence of Tisegregation when the final grain morphology is globular orspherical. Figure 5(a) compiles data31) for the grain size andthe SDAS with cooling rate for each of the alloys. The grainsize data is for the alloys with only an addition of0.005 mass% Ti as Al-3Ti-1B master alloy, i.e. no Ti soluteadditions were made and it shows that the SDAS decreasestowards zero as the cooling rate increases. The grain sizedecreases towards a finite value. Hence the SDAS decreasesmore rapidly than the grain size with increased cooling rate(Fig. 5(b)).

4. Discussion

Two effects on grain morphology have been demonstrated.The first one is that when the grain size is reduced the SDASdoes not change and consequently the grain size becomesfiner and the grain morphology changes from dendritic torosette like/cellular to globular/spherical. This occursthrough the addition of nucleant particles such as TiB2 inAl alloys and on the addition of solute Ti to alloys with lowsolute levels, i.e. low Q0 values. The morphology changewith grain refinement has been observed previously in grain

refinement processes for the production of semi-solidslurries5,22,32) and for ultrasonic grain refinement of al-loys.33–35)

Interestingly Ti additions have no discernable effect on theSDAS, even in alloys with low Q0 values where they have avery large effect on the grain size. This has been demon-strated to be because Ti affects Q0 significantly in thesealloys but has no effect on Qf ,

27) as Ti segregates verystrongly into the �-Al phase at the beginning of solidificationand consequently has little effect on the CU at the end ofsolidification. Qf has a greater influence on the SDAS thanQ0. This is because it is the restriction of coarsening towardsthe end of solidification that is more important in controllingthe SDAS. Hence, it is the presence of strongly segregatingeutectic alloying elements that are more important. However,the values of Qf should only be used as a guide as they arevery dependant upon the values chosen as they relate toparticular reactions that take place. Furthermore, values of mand k at the latter stages of solidification are likely to deviatesubstantially from the binary values used. Finally, otherissues such as solidification time are also important. In theend, if the Q0 and Qf values are high and the solidificationtime is short then the SDAS will be very fine and the grainmorphology is more likely to be dendritic.

The second effect is that whilst the grain size and SDASare refined by increasing cooling rate the SDAS was refinedmore rapidly (Fig. 4 and Fig. 5). The relationship betweenSDAS and cooling rate is well established and is of the formof eq. (3), where tf is inversely proportional to the averagecooling rate. Hence, the SDAS reduces with cooling rate witha power-law exponent 0.33 (eq. (3)). The measured powerexponent for these alloys ranged from 0.33–0.527) which is atypical range of observed variation.36)

Grain size has been found to reduce for the samplesanalyzed here according to an equation of the form:21)

d ¼1ffiffiffiffiffiffiffiffiffiffiffiffiffiffi

f ð _TTÞNv3p þ

D:�Tn

Q � c _TT1=2ð4Þ

where the cooling rate affects the number of active nucleants,and the growth velocity v (eq. (2)) is related to the squareroot of the cooling rate via a fitting factor, c. This is becausethe thermal gradient and growth velocity are coupled in thesolidification conditions used in the experimental set-upused. Even though the power index for cooling rate is similarfor SDAS and grain size, the grains become more dendriticbecause increasing cooling rate tends towards a positive finitevalue rather than zero as it does for the SDAS (Fig. 5).

This suggests that it is inevitable that obtaining a very finegrain size by thermal means will inevitably lead to a moredendritic grain morphology. However, another way ofreducing the grain size is to use low temperature pouringinto a chill mould, where the grain size is reducedsubstantially, but the cooling rate changes little. In this case,the grain size can reduce substantially without a significantchange in the SDAS and consequently a more globularmorphology is obtained.32,37)

To be able to characterize grain morphology it is initiallyimportant to define what different grain morphologies are. Todo this, it is important to consider defining a characteristicSDAS for each alloy for a given set of solidification

(a)

(b)

Fig. 5 (a) The effect of increased cooling rate on the grain size (full line)

and SDAS (dashed line) for each of the alloys with a 0.005 mass% TiB2

addition as Al-3Ti-1B. Data sourced from Easton et al.31) (b) The ratio of

grain size to SDAS against cooling rate for each of the alloys.

Grain Morphology of As-Cast Wrought Aluminium Alloys 845

conditions. This work indicates that this is the case, i.e. thatthe addition of grain refiner as TiB2 particles or solute Tiaffects only the grain size and not the SDAS. Hence, reducingthe grain size only makes it more difficult to measure theSDAS as the primary dendrite arms become shorter. Aspherical grain structure is obtained where the number ofsuccessful nucleation events is large enough that the finalgrain morphology is non-dendritic.5,34)

Given that an alloy has a characteristic SDAS, particulargrain morphologies can be defined. In Fig. 4(a) a cellularmorphology is observed and the d : �2 ratio is approx-imately 3, whilst in Fig. 4(b), the ratio is only 1.5 assumingthe same SDAS as measured from Fig. 4(a) and the grainmorphology is definitely globular or spherical. This sug-gests that a criterion can be used to define grainmorphology, i.e. if the d : �2 ratio is less than 2, the grainmorphology is spherical or globular, when it is between 2and 4 it is rosette-like or cellular and above 4 it is equiaxeddendritic.

Figure 6 shows a way of plotting grain size and SDASdata that provides information on the grain morphology fortwo alloys. Semi-empirical relationships are available fordetermining the grain size19) and the SDAS38) for the range

of alloys in this work and empirical equations for otheralloys can be obtained from elsewhere as well, e.g.36)

Figure 6 shows both the change from dendritic to cellular toglobular grain structures as grain refinement occurs and thetrend towards finer, more dendritic grain morphologies asthe cooling rate increases. Such diagrams may be useful inthe future for tailoring grain morphologies for particularalloys for particular processes. A good example is incontrolling the solidification conditions to obtain sphericalgrain morphologies for rheocasting.6,15,39) Another impor-tant use would be as an input for understanding perme-ability that is an important parameter in hot tearingmodels.40–42)

5. Summary and Conclusions

Both the grain size and the SDAS are affected by alloycomposition and cooling rate through CU parameters. Themajor difference is that the grain size is affected by theconstitutional conditions as solidification begins, whilst theSDAS is more affected by the CU generated by the liquidremaining near the end of solidification. Hence soluteadditions that have a large effect on the grain size, such asTi, do not necessarily have a significant effect on the SDAS.Increasing the cooling rate decreases both the SDAS and thegrain size, although it has a greater effect on the SDAS. Thedifferent effects of grain refiner additions and cooling rate ongrain size and SDAS means that a wide variety of grainmorphologies can be generated. It is suggested that the ratioof grain size to SDAS can be used to define the grainmorphology. At ratios less than two the grain morphology isspherical or globular, between 2 and 4 it is rosette-like orcellular and above 4 it is equiaxed dendritic. Understandingthese relationships may assist with engineering problemssuch as optimization of microstructure during rheocastingand the understanding of hot tearing.

Acknowledgements

The CAST Co-operative research centre was establishedunder, and is supported in part by, the Australian Govern-ment’s Cooperative Research Centres scheme. YeannetteLizama is thanked for her assistance with the metallography.

REFERENCES

1) M. A. Easton and D. H. StJohn: Inter. J. Cast Metals Res. 12 (2000)

393.

2) A. Dahle, S. Karlsen and L. Arnberg: Inter. J. Cast Metals Res. 9 (1996)

103.

3) S. Lin, C. Aliravci and M. O. Pekguleryuz: Metall. Mater. Trans. A 38A

(2007) 1056.

4) C. Caceres and Q. Wang: Inter. J. Cast Metals Res. 9 (1996) 157.

5) R. A. Martinez, A. Karma and M. C. Flemings: Metall. Mater. Trans. A

37A (2006) 2807.

6) S. B. Spencer, R. Mehrabian and M. C. Flemings: Metall. Mater. Trans.

B 7 (1972) 1925.

7) P. J. Uggowitzer and H. Kaufmann: Steel Res. 75 (2004) 525.

8) W. Kurz and D. Fisher: Fundamentals of solidification, 4th edition,

(Trans Tech Publications, Switzerland., 1998) pp. 65–69.

9) M. Qian: Acta Mater. 54 (2006) 2241.

10) M. C. Flemings, T. Z. Kattamis and B. P. Bardes: AFS Trans. 99 (1991)

501.

(a)

(b)

Fig. 6 Plots of the range of grain sizes observed versus the SDAS for the

different cooling rates for alloys (a) 1050 and (b) 6060. dmax is the

maximum grain size and dmin is the minimum grain size. Linear best fit

lines are drawn on the graphs to show the trend of cooling rate on the

ranges of grain sizes and consequently the grain morphology.

846 M. Easton, C. Davidson and D. StJohn

11) Z. Fan: Int. Mater. Rev. 47 (2002) 49.

12) M. A. Easton and D. H. StJohn: Mater. Sci. Technol. 16 (2000) 993.

13) M. Qian, D. H. StJohn and M. T. Frost: Scr. Mater. 46 (2002) 649.

14) Ø. Nielsen, B. Appolaire, H. Combeau and A. Mo: Metall. Mater.

Trans. A 32A (2001) 2049.

15) M.-F. Zhu, C.-P. Hong, D. M. Stefanescu and Y. A. Chang: Metall.

Mater. Trans. B 38B (2007) 517.

16) I. Maxwell and A. Hellawell: Acta Metall. 23 (1975) 229.

17) M. Johnsson: Z. Metallkde. 85 (1994) 781.

18) M. A. Easton and D. H. StJohn: Acta Mater. 49 (2001) 1867.

19) M. A. Easton and D. H. StJohn: Metall. Mater. Trans. A 36A (2005)

1911.

20) T. E. Quested and A. L. Greer: Acta Mater. 52 (2004) 3859.

21) M. A. Easton and D. H. StJohn: Mater. Sci. Eng. A 486 (2008) 8.

22) D. H. StJohn, M. A. Easton and M. Qian: Solid State Phenomena 141–

143 (2008) 355.

23) K. P. Young and D. H. Kirkwood: Metall. Trans. A 6A (1975) 197.

24) M. Bamberger, B. Z. Weiss and M. M. Stupel: Mater. Sci. Technol. 3

(1987) 49.

25) M. Qian, P. Cao, M. A. Easton, S. D. McDonald and D. H. StJohn: Acta

Mater. 58 (2010) 3262.

26) M. Rappaz and W. J. Boettinger: Acta Mater. 47 (1999) 3205.

27) M. A. Easton, C. J. Davidson and D. H. StJohn: Metall. Mater. Trans. A

41A (2010) 1528.

28) M. A. Easton and D. H. StJohn: Light Metals 2001, ed. by J. L. Anjier,

(The Minerals, Metals and Materials Society, Warrendale, Pennsylva-

nia, 2001) pp. 927–934.

29) M. A. Easton and D. H. StJohn: Metall. Mater. Trans. A 30A (1999)

1625.

30) M. Johnsson and L. Backerud: Z. Metallkde. 87 (1996) 216.

31) M. A. Easton, J. F. Grandfield, D. H. StJohn and B. Rinderer: Mater.

Sci. Forum 519–521 (2006) 1675.

32) M. A. Easton, H. Kaufmann and W. Fragner: Mater. Sci. Engng. A 420

(2006) 135.

33) T. V. Atamaneko, D. G. Eskin, L. Zhang and L. Katgerman: Metall.

Mater. Trans. A 41A (2010) 2056.

34) G. I. Eskin and D. G. Eskin: Proc. 12th Int. Conf. on Aluminium Alloys,

(Japan Inst. Light Metals, 2010) pp. 638–645 (USB).

35) G. I. Eskin: Ultrasonics Sonochem. 8 (2001) 319.

36) L. Backerud, E. Krol and J. Tamminen: Solidification characteristics of

aluminium alloys volume. 1: Wrought alloys, (AFS/Skanaluminium,

1993).

37) H. Wang, C. J. Davidson, J. A. Taylor and D. H. StJohn: Mater. Sci.

Forum 396–402 (2002) 143.

38) M. A. Easton, D. H. StJohn and E. Sweet: Mater. Sci. Forum 630 (2010)

213.

39) H. Kaufmann, H. Wabusseg and P. J. Uggowitzer: Aluminium 76

(2000) 69.

40) M. Rappaz, J.-M. Drezet and M. Gremaud: Metall. Mater. Trans. A

30A (1999) 449.

41) S. Vernede and M. Rappaz: Acta Mater. 55 (2007) 1703.

42) M. A. Easton, H. Wang, J. F. Grandfield, D. H. StJohn and E. Sweet:

Mater. Forum 28 (2004) 224.

Grain Morphology of As-Cast Wrought Aluminium Alloys 847