Materials Science and Engineering A259 (1999) 53 – 64

Effect of welding parameters on the solidification microstructure

of autogenous TIG welds in an Al–Cu–Mg–Mn alloy
A.F. Norman *, V. Drazhner, P.B. Prangnell
Manchester Materials Science Centre, Uni6ersity of Manchester/UMIST, Gros6enor Street, Manchester, M1 7HS, UK

Received 4 November 1997; received in revised form 7 August 1998

Abstract

The weld metal microstructures of autogenous TIG welds have been investigated for a range of welding conditions using an
Al –Cu–Mg–Mn alloy. It was found that a combination of high welding speeds and low power densities provide the thermal
conditions required for the nucleation and growth of equiaxed grains in the weld pool, providing heterogeneous nucleation sites
are available. The most likely origin of the nucleants is from a combination of dendrite fragments and TiB2 particles that survive
in the weld pool. The finest microstructure was observed in the centre of the weld and is attributed to the higher cooling rates
which operate along the weld centreline. Composition profiles across the dendrite side arms were measured in the TEM and were
found to follow a Scheil type segregation behaviour where there is negligible back diffusion in the solid. The measured core
concentration of the dendrite side arms was found to rise with increasing welding speed and was attributed to the formation of
significant undercoolings ahead of the primary dendrite tip, which enriched the liquid surrounding the dendrite side arms. © 1999
Elsevier Science S.A. All rights reserved.

Keywords: TIG welding; Al–Cu–Mg–Mn alloys; Grain structures; Microsegregation

1. Introduction siderable advantage in the manufacturing process

would result, both in terms of weight savings and a
Heat treatable AA2000 series aluminium alloys are reduction in the fabrication time. At present the weld-
among the most widely used materials for structural ability of the most widely used AA2000 series alu-
applications where high strength to density ratios are minium alloys is known to be poor [1,3]. This is in
important. In many applications, the dominant joining contrast to alloys used in the former Soviet Union,
process is through the use of mechanical fasteners where a number of weldable aluminium compositions
(rivets). Typically a large numbers of fasteners are used, have been developed, and are used in both aerospace
which makes assembling primary structures an ex- and automotive applications [4–6]. Furthermore there
tremely tedious and time consuming process. Addi- is also some evidence to suggest that welded joints
tional problems arise because a wide flange is required perform better in fatigue than riveted joints [2].
and the section thickness of each component is usually The fusion welding of high strength AA2000 series
increased around the joint producing a further weight
aluminium alloys was given some attention in the early
penalty. Riveted joints are also relatively inefficient, in
1960s [7,8]. Despite the fact that at the time, welding
terms of strength, and act as sites for rapid fatigue
technology was more primitive, several problems asso-
crack initiation [1].
ciated with the welding of AA2000 series aluminium
If welding could be utilised without significantly in-
creasing the section thickness of the joint, then a con- alloys were identified. These included the poor strength
and ductility of the weld metal, solidification cracking,
and grain boundary melting in the heat affected zone
* Corresponding author. Tel.: + 44 161 2003588; fax: + 44 161 (HAZ) [7]. Recently the macroscopic grain structures of
2003586; e-mail: [email protected] Al alloy TIG welds have been investigated by a number

0921-5093/99/$ - see front matter © 1999 Elsevier Science S.A. All rights reserved.
PII S0921-5093(98)00873-9
54 A.F. Norman et al. / Materials Science and Engineering A259 (1999) 53–64

Table 1
Welding conditions for the samples under investigation

Sample No. Welding speed (mm s−1) Welding current (A) Powera (W) Power densityb (J mm−3)

1 7 100 630 14.8

2 13 130 819 12.3
3 19 160 1008 10.4
4 25 190 1197 8.3

a
Calculated from Q =hIV, where Q is the total heat input, V is the welding voltage and, h is taken to be 0.70 [10].
b
The volume of weld per second is estimated from width×thickness×welding speed.

of researchers [9 – 12]. It is widely accepted that (TEM), EDX analysis, and X-ray diffraction were used
equiaxed dendritic grains tend to form in the centre of to measure the effect of the welding speed on the
the weld, at high welding speeds, whereas columnar segregation behaviour of the major alloying elements,
structures (otherwise known as axial or stray grains) and the phases which formed on solidification. The
usually form at low welding speeds. results are compared to existing models to see if it is
Only a limited number of researchers have focused possible to predict the main features of the weld mi-
on the microsegregation behaviour in Al alloy TIG crostructure in multi-component alloys, as a function of
welds [13,14]. However, understanding the redistribu- the welding parameters, without resorting to thermody-
tion of solute during solidification is particularly impor- namic calculations which require extensive computing
tant in heat treatable alloys as the weld metal strength power.
will depend on the solute supersaturation, which deter-
mines the subsequent ageing response and yield
strength. In order to develop models capable of predict- 2. Experimental
ing the weld metal properties, it is equally important to
be able to predict the grain structure and dendrite arm The commercial aluminium alloy AA2024 (Al–4.4
spacing and volume fraction of eutectic, which will wt% Cu–1.5 wt% Mg–0.8 wt% Mn, and Fe and Si
dominate the toughness and ductility. A number of impurities) was TIG welded autogenously, using the
workers have demonstrated that a relatively simple range of conditions given in Table 1. The sheet thick-
approach, based on estimating the cooling rate can be ness was 1.6 mm and a simple butt geometry was used
used to reliably predict dendrite arm spacings in model for each experiment. The experiments were designed
aluminium alloys [15], but this has not yet been at- such that, as the welding speed was increased, the
tempted with more complex alloys of near commercial welding current was also increased to just maintain full
composition. Furthermore, there is little experimental penetration of the weld. The data in Table 1 shows
data available in the literature comparing measure- that, with increasing welding speed, the overall power
ments of solute profiles across dendrites to models of density required to maintain full penetration decreases.
solute redistribution. Conflicting results exist between After welding, the samples were sectioned and mounted
measured solute profiles; some workers suggesting a either along the direction of the weld in plan view or
Scheil behaviour [16], where there is effectively no sectioned through the weld (transverse view). Each
diffusion in the solid phase, and other profiles based on sample was examined using conventional optical mi-
the model of Brody and Fleming [17], where there is croscopy. Samples for SEM examination were prepared
limited back diffusion in the solid. by polishing to 1 mm finish and electropolishing in a
In this paper, results are presented on the commercial solution of 10 vol% nitric acid in methanol at 20 V, and
aluminium alloy AA2024 which has been autogenously − 30°C.
TIG welded using a range of processing conditions. Thin foils for TEM observation were prepared from
Although in practice it is unlikely that structures would the welds by spark eroding 3 mm discs from different
be welded without the use of filler materials, the addi- positions in the weld sample. The discs were ground to
tion of a filler material makes the prediction of the a thickness of ~100 mm and jet-polished using a solu-
microstructure in the weld more difficult. The welds tion of 30 vol% nitric acid in methanol at −30°C and
were characterised using optical metallography to re- 12 V. The foils were examined in a Philips CM 200
veal the different grain structures, produced with differ- analytical TEM operated at 200 kV. Quantitative EDX
ent welding conditions. The sub-grain structures were analysis was performed by traversing across several
studied using a combination of techniques. Scanning dendrite side arms for each welding speed. The core
electron microscopy (SEM) was used to measure the concentration of the side arms was determined as a
dendrite secondary arm spacing at different positions in function of welding speed (approximately 15 arms were
each weld, whereas transmission electron microscopy measured for each speed). Because of the strong likeli-
 A.F. Norman et al. / Materials Science and Engineering A259 (1999) 53–64 55

hood of dendrites not being sectioned through their structure along the central region of the weld, through-
centre in a random section, only the widest dendrites out the entire weld length. It has been reported by
were analysed and the lowest core concentration mea- Ganaha et al. [9] that stray grain structures usually
sured was used as the most representative value. All form at intermediate welding speeds and heat inputs,
EDX data was ZAF corrected and compositions were and are characterised by the continual appearance of
measured across dendrite side arms in positions where new anisotropic grains aligned along the central region
the eutectic phase had been polished out of the foil, of the weld, in the welding direction. As the nucleation
thus eliminating the possibility of unrepresentative mea- rate at the interface is very low, many of the grains
surements near the dendrite boundary. grow a considerable distance in the direction of the heat
source, and therefore appear elongated in morphology.
At all higher welding speeds, the central region of the
3. Macroscopic grain structures weld consisted of an equiaxed-dendritic structure with
an average grain size of ~250 mm (Fig. 1(d)). The
3.1. Formation of fusion zone microstructures remaining weld microstructure still consisted of epitax-
ial columnar grains growing in from the weld edge. The
In autogenous TIG welding, the macroscopic grain equiaxed region covered almost a third of the width of
structure is controlled by a combination of the thermal the weld produced at 19 mm s − 1, which increased to
conditions that prevail at the solid – liquid interface and almost half the weld width at 25 mm s − 1. These welds
the crystal growth rate which is directly related to the were also examined in the transverse direction and
welding speed [14]. The thermal conditions are deter- show that for the two highest welding speeds, a region
mined by the heat input and the weld speed for a given of equiaxed grains formed throughout the thickness of
sheet thickness. Furthermore, the conditions vary con- the weld (Fig. 1(e) and (f)).
siderably depending on the position at the solid–liquid
interface along the trailing edge of the weld pool. For 3.2. Columnar to equiaxed transition (CET)
thin sheet Al alloys, the high thermal diffusivity of
aluminium favours the formation of an elliptical weld In this work, a transition series from axial, to stray,
pool shape, even at relatively high heat inputs and to equiaxed grains, was observed as the welding speed
welding speeds [18]. was increased (shown in Fig. 2(d)). The same type of
The different types of grain structures observed, as a transition has previously been reported for a range of
function of the welding conditions, are shown in Fig. 1 other TIG welded Al alloys; namely AA6061 [10],
and are summarised in Fig. 2. For the lowest welding AA1100, AA5052, and AA7004 [9], and as can be seen
speed considered (7 mm s − 1) a few extremely long in Fig. 2(d), the columnar to equiaxed transition occurs
columnar-dendritic grains were observed (often termed at roughly the same welding speed/heat input combina-
‘axial’ [11]), which grew along the centre of the weld in tions, despite the difference in alloy compositions. Hunt
the direction of the motion of the heat source (Fig. [19] has analysed the conditions necessary for the
2(a)). The overall width of this weld was ~4 mm and the growth of equiaxed grains ahead of a columnar inter-
columnar-dendritic region was around 1 mm wide. The face during directional solidification. Hunt’s analysis,
axial grain structure shown in Fig. 1(a) has been re- which assumes that a 50% equiaxed volume fraction is
ported for other TIG welded Al alloys, such as AA3003 sufficient to block columnar growth, results in the
(Al–Fe–Mn) and AA5454 (Al – Mg) [9] which were following expression for predicting the columnar-to-


also produced at low welding speeds and heat inputs. equiaxed (CET) transition:
On either side of the central axial grains, columnar-den-
dritic grains were found, which solidified on the base GL B 0.061N 01/3 1−
(DTN)3
DTC
n (1)
metal (epitaxial nucleation) and grew towards the cen- (DTC)3
tral axial region (Fig. 1(b)). The epitaxially grown where GL is the maximum thermal gradient, N0 is the
dendrites curve towards the heat source, so that the density of heterogeneous nucleants, DTN is the critical
maximum thermal gradients present at the solid–liquid undercooling for heterogeneous nucleation, and DTC is
interface, are maintained as growth proceeds. For the the growth undercooling at the columnar front. Al-
sample with the lowest speed (7 mm s − 1), near the end though these quantities cannot be easily established,
of the weld, the central region became slightly wider Eq. (1) suggests that the formation of equiaxed grains
and the density of grains increased, producing a more in the centre of the weld can be expected providing
‘stray-like’ grain structure (Fig. 1(c)). This effect was that; (i) there is a supply of nucleation sites from which
probably due to the sheet metal increasing in tempera- new grains may develop, and (ii) thermal conditions are
ture as the weld proceeded which reduced the thermal present which favour the nucleation and growth of new
gradients at the solid – liquid interface. The weld with grains. It should be noted that the transition from
the next highest speed (13 mm s − 1) had a stray grain columnar to equiaxed grains also requires a sufficiently
56 A.F. Norman et al. / Materials Science and Engineering A259 (1999) 53–64

Fig. 1. Examples of the different grain structures found in 2024 autogenous TIG welds for the different welding speeds. (a) Axial, 7 mm s − 1
(plan); (b) epitaxial, 7 mm s − 1 (plan); (c) stray, 13 mm s − 1 (plan); (d) equiaxed, 25 mm s − 1 (plan); (e) equiaxed, 25 mm s − 1 (transverse); and
(f) equiaxed, 25 mm s − 1 (transverse).

high nucleation rate relative to the growth rate. The at second phase particles present in the melt, or impuri-
two conditions noted above are explained in more ties in the weld pool [14]. It is well known that the grain
detail below. structure of cast ingots is controlled through the delib-
erate addition of inoculants [20]. During welding it is
3.2.1. Supply of nucleants possible for some of the inoculants to survive in the
In the welding of Al alloys, the formation of new weld pool and act as nucleants for new grains. The
grains can originate from two sources; the growth of alloy examined in this work also contains a small
pre-existing nuclei created by dendrite fragmentation, addition of TiB2 particles, which are used to control the
or by grain detachment, and heterogeneous nucleation grain structure during DC casting, and therefore it is
 A.F. Norman et al. / Materials Science and Engineering A259 (1999) 53–64 57

possible for some of these particles to act as nucleants sition occurs is reduced from 15 to around 6 mm s − 1
in the weld pool. Recently there has been some direct [24].
evidence of Ti rich particles acting as nucleants for
grains in TIG welds of Al alloys AA6061 [10] and 3.2.2. Thermal considerations
AA7004 [9]. It is also interesting to note that for the Al For the welds studied in this work, an increase in the
alloy AA5083, which contains a significant amount of Ti welding speed was accompanied by an increase in the
(0.024 wt%), a region of equiaxed grains was observed welding current so that full weld penetration was main-
at both low and high welding speeds [9]. Also, when the tained. However, the power density (see Table 1) de-
alloy AA2024 is welded with the filler AA2319 (which creased as the welding speed was increased. This
combination of conditions can cause a decreasing, or
contains grain refining additions of Ti and Zr), the
nearly constant thermal gradient in the liquid, whilst
welding speed at which the columnar to equiaxed tran-
promoting an increase in the crystal growth velocity. If
the thermal gradient decreases, with increasing weld
speed, this could result in significant undercoolings
ahead of the solid–liquid interface, due to a combination
of solutal and dendrite tip curvature effects (see also
Section 5) [21].
In many of the previous studies on the grain structure
of Al TIG welds, the temperature gradient along the
weld centreline has not been measured experimentally,
but has been calculated using the following equation,
which is derived from the Rosenthal thin plate solution


[22,23]:
2
t
GL = 2parC 6(Tm − T0)3 (2)
Q
In Eq. (2), a is the thermal conductivity, r is the density,
Q is the total heat input, C is the specific heat, t is the
plate thickness, Tm is the freezing temperature and T0 is
the ambient temperature.
The calculated values of GL for the different welding
speeds are given in Table 2. The values range from 81
K mm − 1, at a speed of 13 mm s − 1 to 73 K mm − 1, at
the highest welding speed of 25 mm s − 1, and are thus
predicted to be nearly constant in the present experi-
ment. According to observations made by Kou and Le
[10], of experimentally determined thermal gradients
from thermocouples, the measured results agreed well
with the calculated values at low welding speeds and heat
inputs, but were increasingly inaccurate at higher speeds
and power densities. This difference was attributed to
the basic assumption that the heat source in the calcula-
tion is treated as a line source and neglects the presence
of the weld pool, or heat flow by convection. Therefore
the calculated values quoted in this work should be
taken as an upper bound because it is likely that, due to
convection in the weld pool, the real values would be
somewhat lower resulting in a significantly decreasing
thermal gradient with increasing speed (as the observed
transitions in grain structure might suggest).
In this work, both the welding current and the weld-
ing speed were increased at the same time. If the effect
of this was to keep the thermal gradient along the weld
Fig. 2. Schematic diagrams illustrating the different types of grain
structures found in Al TIG welds. (a) Axial, (b) Stray, (c) equiaxed,
centreline constant, as predicted by Eq. (2), the ratio
and (d) a plot of heat input versus welding speed for the different GL/R at the weld centreline will still decrease with
welds produced in this work. increasing welding speed because R increases (see Table
58 A.F. Norman et al. / Materials Science and Engineering A259 (1999) 53–64

Table 2
Summary of calculations involving GL and R at the weld centreline

Sample No. Welding speed (mm s−1) G aL (°C mm−1) GL/R GLR (°C s−1) Calculatedb l2 (mm) Measured l2 (mm)

1 7 74.2 10.0 519 6.2 5.8

2 13 81.5 6.0 1059 4.9 4.9
3 19 78.7 4.1 1495 4.4 3.9
4 25 73.4 2.9 1835 4.1 3.5

a
Calculated from Eq. (2).
b
Calculated from Eq. (4).

2) and the undercooling achieved in the liquid, ahead of TEM image is shown of the typical eutectic growth
the advancing solid – liquid interface will be mainly a morphology for the welding speed of 7 mm s − 1.
function of the growth velocity, which for columnar
grain structures, is directly related to the welding speed. 4.2. Dendrite secondary arm spacing measurements
At low welding speeds this suggests that even if hetero-
geneous nucleation sites are present in the weld pool, In the solidification of cast structures, there exists a
the thermal conditions do not favour the formation of well established link between the scale of the dendritic
stable nuclei, and that any grain fragments would re- sub-structure and the conditions under which solidifica-
melt, i.e. an axial grain structure would be expected. As tion takes place. Various studies have shown that the
the welding speed is increased, the growth velocity of dendrite arm spacing, measured after processing, de-
the solidification front will also increase producing a pends directly on a combination of the thermal gradient
higher degree of undercooling, thus producing the ob- (GL) and growth rate (R) [14]. The dependence of the
served transition from axial to stray grain structures. At primary (l1) and secondary (l2) arm spacings, on GL
the highest welding speed, the undercooling will be at and R take the following form:
its greatest and can give rise to a significant nucleation
rate resulting in an equiaxed structure. l1 = a1(G 2LR) − n (3)
l2 = a2(GLR) − n (4)
4. Grain sub-structures where a1 and a2 are coefficients, whose values depend
on the alloy system, and n is an exponent whose value
4.1. Introduction lies between 1/4 and 1/2. The primary spacing cannot
be directly related to the cooling rate (e=GLR) since
The observations made in Section 3 focused on the its dependence on GL and R have different exponents.
type of grain structure that formed with different weld- However, the secondary arm spacing depends on the
ing conditions. Of equal importance is the effect of product of GLR and can therefore be related directly to
welding parameters on the grain sub-structure that the cooling rate.
forms during welding. The results in Section 3 clearly The values of a2 = 50 mm [K s − 1]n and n=1/3 have
demonstrate that for the welding conditions considered been estimated for binary Al–4 wt% Cu alloys within
in this work, the alloy always produces a weld which the range of cooling rates considered in this work
contains a dendritic sub-structure (irrespective of (1–105 K s − 1 [15]). Using these constants, the sec-
whether the grain structure is columnar or equiaxed). ondary arm spacing along the weld centreline has been
Fig. 3(a) and (b) show examples of a dendritic sub- estimated from the data given in Table 2 for GL and R
structure taken at the weld centreline for two different (assuming R is equal to the welding speed, this might
speeds. In both optical micrographs, a dendritic growth not be the case for equiaxed microstructures as the
pattern can clearly be seen (light coloured) with a dark growth direction does not necessarily correspond to the
structure present in the interdendritic regions. At the weld traverse direction). This data is plotted in Fig. 4
aluminium rich corner of the ternary Al – Cu –Mg sys- together with secondary arm spacings measured along
tem, the phase diagram is dominated by the ternary the weld centreline for the different welding speeds. Fig.
eutectic reaction between L“a-Al +u(Al2Cu)+ 4 shows that as the welding speed is increased, both the
S(Al2CuMg) which is reported to occur at 503°C with a measured and calculated secondary arm spacing de-
composition of 26.8 wt% Cu and 6.2 wt% Mg [25]. creases, and that there is good agreement between the
X-ray diffraction analysis confirmed that the dendritic measured and calculated values, over the complete
matrix was a-Al and the interdendritic region was a range of welding conditions, although the calculated
mixture of three phases, namely a-Al, u(Al2Cu), and values give consistently large spacings when compared
S(Al2CuMg), i.e. the ternary eutectic. In Fig. 3(c), a to the measured values. There are two reasons why the
 A.F. Norman et al. / Materials Science and Engineering A259 (1999) 53–64 59

Fig. 3. (a) and (b) Optical micrographs showing the dendritic (light) and interdendritic (dark) regions along the weld centreline for the speeds of
7 and 25 mm s − 1 respectively. (c) TEM micrograph showing the typical growth morphology of the ternary eutectic in the interdendritic regions.

calculated and measured spacings might differ. Firstly, and a region near the fusion boundary. The microstruc-
in Section 3 we suggested that the thermal gradients tures clearly show that near the fusion boundary, the
calculated in Eq. (2) were to be considered as an upper sub-structure is at its coarsest, which suggests that the
estimate. If the actual values of the thermal gradient cooling rate at the fusion boundary is much lower than
were somewhat lower (due to convection in the weld that observed in the centre of the weld. This occurs
pool), then the estimated cooling rates along the weld because although the thermal gradients are higher at
centreline would be lower and the calculated arm spac- the fusion boundary, the growth rate will be at its
ings would be even coarser. Secondly, it is clear that a minimum and the overall effect will be to produce a
better fit between the measured and calculated spacings low cooling rate compared to that observed along the
could be achieved by adjusting the values of the con- weld centreline.
stants in Eq. (2). However, these constants have yet to
be established for the commercial Al alloy 2024.
According to Eq. (4), GL and R have the same 5. Microsegregation behaviour
exponent and are therefore equally important in deter-
mining the scale of the dendritic sub-structure. How- 5.1. Composition profiles
ever, the calculated thermal gradients along the weld
centreline are almost constant for the different welding TEM was used to study the segregation behaviour of
speeds, so it is largely the increase in the welding speed the alloying elements Cu and Mn within the dendrite
which gives rise to the increase in cooling rate and side arms. Microanalyses were performed at different
reduction in the secondary arm spacing. Fig. 5 shows a positions across dendrite side arms for the different
comparison between the scale of the sub-structure welding speeds, and the measured variation in Cu and
along the weld centreline for the highest welding speed Mn solute contents, from the centre of a dendrite side
60 A.F. Norman et al. / Materials Science and Engineering A259 (1999) 53–64

Fig. 4. Plot of measured and calculated dendrite secondary arm spacing (along the weld centreline) as a function of welding speed.

arm towards the interdendritic region. For each weld- adopted where the tie-lines are re-calculated at temper-
ing speed, five scans were made and the average values ature increments using a thermodynamic database [26]),
are shown in Fig. 6(a) and (b) respectively. The data in in a system where one alloy component dominates,
Fig. 6 was obtained from regions where the intermetal- such as Cu does in this case, it can be used as a rough
lic eutectic phase had polished out of the sample on approximation. The Scheil equation is based on the
preparing the specimen, so that the results near the basic assumptions that the liquid is of uniform compo-
interdendritic region would not be affected by their sition, local equilibrium is maintained at the solid–liq-
presence. Mg could not be reliably analysed due to the uid interface (k is constant), and that there is negligible
overlap with the Al Ka peak. The spatial resolution of solid-state diffusion.
the measurements was ~0.2 mm which is substantially Whether or not there is a significant contribution of
less than the dendrite arm spacings of between 4 and solid state diffusion to the solute redistribution process
5.5 mm for the welding speeds considered. Fig. 6(a) can be estimated from the parameter, h [17]:
shows that, as expected, the lowest Cu concentration
Dstf
occurs in the core of the dendrite side arm with the Cu h= (6)
content increasing towards the interdendritic boundary. l 22
The profile in Fig. 6(b) also demonstrates that for Mn, where Ds is the solid diffusivity of the solute and tf is
the lowest concentration also occurs in the centre of the the local time of freezing. For the welds produced in
side arm, although it should be noted that the Mn this work, the values of tf are in the range 0.05–0.2 s
content in this alloy is a lot lower than Cu. which produces an a parameter in the range 2.4×
The results clearly suggest that the segregation be- 10 − 3 –6× 10 − 4. As it can be shown that solid state
haviour is more similar to the scenario envisaged by diffusion during freezing only begins to become an
Scheil [16], where there is negligible back diffusion in important factor in the solute redistribution process for
the solid, than that expected by the Brody and Flem- values of h\ 0.1 [17], the freezing conditions in the
ings model [17]. The Scheil equation for solute parti- weld should be similar to those predicted by the Scheil
tioning [16] is given by: equation.
k−1
C*=C
s 0k(1−fs) (5)
5.2. Variation of core concentration with welding speed
where C *s is the solid composition, C0 is the initial alloy
composition, fs is the fraction of solid and k is the As the solute profiles were found to be affected by
partition coefficient. While the Scheil equation cannot the processing conditions, the effect of welding speed
be applied directly to multi-component alloys (for non- on the core composition of the dendrite arm was inves-
binary systems, an iterative approach should be tigated in more detail by analysing the centre of ap-
 A.F. Norman et al. / Materials Science and Engineering A259 (1999) 53–64 61

Fig. 5. Optical micrographs showing the difference in microstructural scale between (a) the weld centreline, and (b) near the fusion boundary for
a welding speed of 25 mm s − 1.

proximately 15 side arms for each welding speed. Be- than that measured in any of the weld samples. Fur-
cause of the difficulties associated with the random thermore, the results shown in Fig. 6(c) demonstrate
sectioning of dendrite side arms during the production that the core concentration varies from ~2C0k for the
of TEM foils, it is possible that many of the measured lowest welding speed to ~3C0k for the highest speed
concentrations correspond to the composition some considered. Similar observations have been made by
distance away from the centre. To reduce this risk, the Brookes and Baskes [13] who studied GTA binary
widest side arms were analysed and the minimum mea- Al–Cu welds containing between 1 and 2 wt% Cu.
sured concentration from 15 arms was used for each They also observed core concentrations greater than
welding speed. This data is plotted in Fig. 6(c) and used equilibrium (around 2C0k–3C0k) which is in good
in the subsequent discussion. Despite the difficulties agreement with the core concentration enrichment mea-
mentioned above, the results in Fig. 6(c) clearly show sured in this work.
that as the welding speed is increased, the dendrite core To try and explain why the core concentration in the
concentration of both Cu and Mn also increases, but side arms increase with increasing welding speed, it is
this effect is more noticeable for Cu. Even higher core necessary to consider how the growth of the dendrite
concentrations have been observed for the laser welding side arms are related to the growth of the dendrite
of AA2024 where much greater welding speeds were primary tip. A schematic solute profile through the
used [27]. liquid in the mushy zone for dendrite growth is shown
To try and explain this variation in the core composi- in Fig. 7. In this figure, the composition in the bulk of
tion with welding speed, the arguments will be confined the liquid is given by C0. It is possible that for the range
to analysing the behaviour of Cu, the major alloying
of welding speeds considered in this work, significant
element in 2024, as this element produced the most
undercoolings can be achieved at the primary dendrite
reliable data. Although this is a multicomponent alloy,
tip which will enrich the liquid at the tip, to a composi-
the behaviour is compared to that of the binary Al–Cu
tion Ct. Microstructural observations have shown that
phase diagram for simplicity, and the effect of the other
a significant volume fraction of eutectic phases are
alloying elements on the microsegregation behaviour of
formed in the mushy zone at the end of solidification,
Cu in Al are neglected. As under the welding conditions
i.e. at the root of the dendrites in the interdendritic
used, the solid state diffusion of Cu in Al appears to be
negligible (see above), it is assumed that the core con- regions (see Fig. 3(c)). Therefore, the solute profile in
centration in the dendrite side arm corresponds to the the liquid will change as it moves through the mushy
dendrite side arm tip concentration during growth. zone, from Ct to CE, the eutectic composition, as shown
Under true equilibrium solidification conditions, the in Fig. 7. In the region where the secondary arms form,
composition of the first solid to form will be given by the local liquid composition will thus be enriched with
C0k, where C0 is the initial alloy concentration (equal to solute to a level somewhere between Ct and CE, i.e. the
4.46 wt% Cu for a binary Al – Cu alloy of the same Cu core concentration of the dendrite side arm will be
composition as AA2024) and k is the equilibrium parti- given by k(Ct + DC), where DC depends on the relative
tion coefficient (for binary Al – Cu alloys, k= 0.17). distance back from the tip where the secondary arms
Using these values, the equilibrium core concentration form (thought to be of the order of two to three
(C0k) should be 0.76 wt% Cu, which is much lower secondary arm spacings [17]).
62 A.F. Norman et al. / Materials Science and Engineering A259 (1999) 53–64

Fig. 6. EDX analyses across dendrite side arms, for (a) Cu and (b) Mn, and (c) a plot of the measured minimum core concentration for dendrite
side arms as a function of welding speed.
 A.F. Norman et al. / Materials Science and Engineering A259 (1999) 53–64 63

Fig. 8. Plot of primary tip velocity versus calculated primary tip

temperature for a binary Al – 4.46 wt% Cu alloy using the models of
Burden and Hunt [30], and Kurz and Fisher [31].

model predicts that this will occur at a greater rate (Fig.

8). Using the Kurz and Fisher model, the variation in
tip undercooling has been translated to a core composi-
Fig. 7. Schematic diagram illustrating a possible solute profile in the tion for the growth of primary dendrites in a binary
liquid during dendritic growth. Al–4.46 wt% Cu alloy, and the data is shown in Fig. 9
together with the measured values for the secondary
Models have been developed to predict the primary arms. By assuming that the primary tip velocity is the
dendrite tip undercooling which can, in turn, be related same as the welding speed, the model of Kurz and
to the primary tip composition. According to Brody Fisher demonstrates that as the welding speed is in-
[28], the tip undercooling can be estimated from the creased, so the primary tip concentration also increases,
following relationship. i.e. in agreement with the trends of this work. As might
be expected from Fig. 7, the predicted composition
DTtip = DTD +DTr +DTk (7)
values for the primary tip (kCt) are somewhat less than
where DTD is the diffusional undercooling, DTr is the the measured values of k(DC +Ct), for the secondary
contribution of the undercooling due to curvature ef- arm core compositions. However the secondary arm
fects, and DTk is the kinetic undercooling (the kinetic core composition does increase with increasing welding
undercooling is the smallest contribution to undercool- speed in line with the predicted primary tip behaviour.
ing in an alloy, and according to Chalmers [29], can be
neglected in this analysis). This approach has been used
by Burden and Hunt [30] to derive an expression for tip
undercooling by assuming the tip grows at the mini-
mum undercooling:
GDL 3 2 −mV(1 −k)C0u
DT = + (8)
V DL
where V is the tip velocity, DL is the solute diffusivity in
the liquid, m is the liquidus slope, and u is the curvature
undercooling coefficient.
More recently, Kurz and Fisher [31] have pointed out
that the minimum undercooling criteria underestimates
the change in tip temperature, and have instead, used
the criteria that the tip radius is equal to the wavelength
of instability of the interface. In both approaches, the
tip temperature is dominated by the growth velocity,
rather than the thermal gradient, when the growth
Fig. 9. Comparison between the predicted core compositions of the
velocities are higher than 10 − 3 cm s − 1 and therefore in
primary dendrite (Ct) taken from the model of Kurz and Fisher [31]
the range of interest to welding. Both models show that and the measured core concentrations from the dendrite side arms as
the tip undercooling starts to rapidly increase at high a function of welding speed (it is assumed that the primary tip grows
velocities (~1 cm s − 1) although the Kurz and Fisher at the same velocity as the welding speed).
64 A.F. Norman et al. / Materials Science and Engineering A259 (1999) 53–64

6. Conclusions References

A detailed understanding of the effect of the process- [1] J.R. Davis (Ed.), ASM Speciality Handbook: Aluminium and
ing conditions is required to produce a good weld Aluminium Alloys, ASM International, Metals Park, OH, 1993,
p. 376.
microstructure for the autogenous TIG welding of the
[2] R. Onjukka, Weld. J. 75 (1996) 29.
commercial aluminium alloy AA2024. At high welding [3] S. Kou, Weld. Res. Council Bull. 320 (1986) 1.
speeds and low power densities, it is possible to pro- [4] R.V. Ilyushenko, Aluminium 69 (1993) 364.
mote the formation of an equiaxed-dendritic mi- [5] V.V. Grinin, A.V. Petrov, V.V. Ovchinnikov, Weld. Prod. 33
crostructure in the centre of autogenous welds. This is (1986) 19.
due to the development of an undercooled liquid ahead [6] I.N. Fridlyander, in: T.H. Sanders, Jr., E.A. Starke, Jr. (Eds.),
Proceeding of the 5th International Conference on Al–Li Alloys,
of the moving solid – liquid interface, which provides
Williamsburg, Virginia, 1989, MCEP, Birmingham, 1989, p.
the correct thermal conditions for the nucleation and 1359.
growth of new grains. The most likely origin for the [7] D.E. Schillinger, I.G. Betz, F.W. Hussey, H. Markus, Welding J.
nucleants is expected to be from a combination of 42 (1963) 269S.
dendrite arm detachment and TiB2 particles that sur- [8] W.I. Pumphrey, J.V. Lyons, J. Inst. Metals 74 (1948) 439.
vive in the melt. [9] T. Ganaha, B.P. Pearce, H.W. Kerr, Metall. Trans. A 11 (1980)
1351.
The finest microstructure was observed at the centre
[10] S. Kou, Y. Le, Weld. J. 65 (1986) 305S.
of the weld and is attributed to the higher cooling rates [11] Y. Arata, F. matsuda, A. Matsui, Trans. Jpn. Weld. Res. Inst. 3
at the weld centre compared to those at the fusion (1974) 89.
boundary. As the welding speed was increased, the [12] M. Rappaz, C.A. Gandin, J.L. Desbiolles, P. Thevoz, Metall.
cooling rate at the centre of the weld also increased Trans. A 27 (1996) 695.
producing a reduction in the scale of the dendrite [13] J.A. Brookes, M.J. Baskes, in: S.A. David (Ed.), Advances in
Welding Science and Technology, ASM International, Metals
secondary arm spacing.
Park, OH, 1987, p. 93.
Both the dendrite side arm spacing and the microseg- [14] S.A. David, J.M. Vitek, Int. Mater. Rev. 34 (1989) 213.
regation behaviour have been investigated using quanti- [15] H. Jones, Monograph No. 8, Institute of Metallurgists, London,
tative EDX analysis. The microsegregation behaviour 1982.
of the dendrite side arms has been shown to follow a [16] E. Scheil, Z. Metallkd. 34 (1942) 70.
Scheil type behaviour in which there appears to be [17] H.D. Brody, M.C. Flemings, Trans. AIME 236 (1966) 615.
[18] Ø. Grong, in: H.K.D.H. Bhadeshia (Ed.), Metallurgical Mod-
negligible back diffusion in the solid. However, the core
elling of Welding, Institute of Materials, London, 1994.
concentration within a dendrite side arm substantially [19] J.D. Hunt, Mater. Sci. Eng. 65 (1984) 75.
changes with the welding speeds used in TIG welding. [20] D.G. McCartney, Int. Mater. Rev. 34 (1989) 247.
This behaviour can be explained in terms of the forma- [21] R. Trivedi, W. Kurz, Int. Mater. Rev. 39 (1994) 49.
tion of significant undercoolings ahead of the primary [22] D. Rosenthal, Weld. J. 20 (1941) 220s.
dendrite tip which enriches the liquid surrounding the [23] D. Rosenthal, Trans. ASME 68 (1946) 849.
[24] A.F. Norman, K.E. Jakielski, G. Scott, P.B. Prangnell, in: J.
dendrite side arms.
Beech, H. Jones (Eds.), Proceedings of the 4th International
Conference on Solidification Processing, University of Sheffield,
December 1997, p. 670.
[25] S.L. Chen, Y. Zuo, H. Liang, Y.A. Chang, Metall. Trans. A 28
Acknowledgements (1997) 435.
[26] N. Saunders, Mater. Sci. Forum 217-222 (1996) 667.
[27] A.F. Norman, P.B. Prangnell, in: T. Sato, S. Kumai, T.
This work was supported by the EPSRC under the Kobayashi, Y. Murakami (Eds.), Proceedings of the 6th Interna-
IMI Programme, Grant: GR/K66901, and is released tional Conference on Aluminium Alloys (ICAA-6), Japan Insti-
with the kind permission of the following partners: tute of Light Metals, Japan, 1998, p. 1435.
British Aerospace, Short Brothers, British Aluminium, [28] H.D. Brody, in: S.A. David (Ed), Advances in Welding Science
DERA, Rolls Royce, TWI, Cranfield University, Uni- and Technology, ASM International, Metals Park, OH, 1987, p.
83.
versity of Liverpool and the University of Essex. The
[29] B.A. Chalmers, Principles of Solidification, Wiley, New York,
authors gratefully acknowledge the assistance of one of 1964.
the referees in interpreting the data produced from [30] M.H. Burden, J.D. Hunt, J. Crystal Growth 22 (1974) 109.
solute profiles. [31] W. Kurz, D.J. Fisher, Acta Metall. 29 (1981) 11.