Journal of Non-Crystalline Solids 358 (2012) 2753–2758

Contents lists available at SciVerse ScienceDirect

Journal of Non-Crystalline Solids

journal homepage: www.elsevier.com/ locate/ jnoncrysol

Estimation of critical cooling rates for formation of amorphous alloys

from critical sizes
Sheng Guo a,⁎, Yong Liu b
a
Department of MBE, City University of Hong Kong, Kowloon, Hong Kong, PR China
b
State Key Laboratory for Powder Metallurgy, Central South University, Changsha, Hunan, 410083, PR China

a r t i c l e i n f o a b s t r a c t

Article history: The critical cooling rate for marginal glass formers, like Al-based alloys, is difficult to measure experimentally.
Received 29 February 2012 In this work, we acquired the critical cooling rate for formation of amorphous alloys by estimating the cooling
Received in revised form 19 June 2012 rates of the specimens with critical dimensions. Analytical solutions were given to estimate the cooling rates
Available online 24 July 2012
for the gas-atomized powders and melt-spun ribbons, and as an example the critical cooling rate to form an
amorphous Al82Ni10Y8 alloy was estimated to be ~1.0 × 10 6 K s−1. The effect of melt temperature on the
Keywords:
Metallic glasses;
cooling rate was quantitatively evaluated and its effect on the glass forming ability was also discussed.
Aluminum alloys; © 2012 Elsevier B.V. All rights reserved.
Cooling rate;
Heat transfer

1. Introduction authors there lacks such an obvious link between the cooling rate
and sample size in the literature, or at least not in a convenient
Amorphous alloys, or metallic glasses have been developed rapid- way. In this work, we prepared amorphous Al82Ni10Y8 alloys in both
ly and studied extensively during the past 30 years [1–3], due to their ribbon and powder forms, and estimated their cooling rates by
many unique properties over the crystalline counterparts. A large va- using simplified heat transfer models. The target alloy was chosen be-
riety of amorphous alloy systems has been developed so far, including cause it is a simple ternary alloy system and the alloying components
Zr-, Fe-, Pd-, Pt-, La-, Cu-, Mg-, Co-, Ni- and Ti-based alloys [4]. How- well satisfy the three empirical rules proposed by Inoue to possess
ever, the preparation of bulk (typically above 1 mm in size) Al-based high GFA [3]. The melt treatment has been known to affect the GFA
metallic glasses has been proved very difficult [5–9] and a break- [11–15], but how the melt temperature affects the cooling rate is
through in the achievable size is yet to be reached. Considering the not quantitatively studied previously, and this issue is to be addressed
wide applicability of Al alloys and their light weight, the capability here as well.
to make bulk Al-based metallic glasses is very attractive. The difficulty
of making bulk Al-based metallic glasses is due to the fact that the
critical cooling rate, Rc, to obtain amorphous Al alloys is fairly high, 2. Material and methods
and such a demanding cooling rate limits the achievable sample size
that can be fully amorphous in structure. To advance the develop- The ribbon and powder specimens used in this work were pre-
ment of bulk amorphous alloy with marginal glass forming ability pared by melt spinning and gas atomization methods, respectively.
(GFA), our understanding to the GFA and particularly to the improve- Starting materials were pure Al (99.995 mass%), Ni (99.9 mass%)
ment of the GFA need to be advanced greatly. In the meantime, the and Y (99.99 mass%). The master alloy was prepared by mixing the
characterization of the GFA of Al alloys is also an issue. The critical pure metals in a molar composition of Al82Ni10Y8 using arc melting
cooling rate, which is the unambiguous criterion defining the GFA of (for melt spinning) or induction melting (for gas atomization) in
various metallic glasses, can be experimentally measured for easy the high-purity argon atmosphere. Repeated melting was carried
glass formers but not for marginal glass formers like Al alloys [10]. out at least five times to ensure the chemical homogeneity of the
However, the critical size (Dc) of the amorphous Al alloys can be alloy. For the gas atomization process, a close-coupled nozzle was
experimentally determined. If we can calculate the cooling rates (R) used to promote the efficient atomization of the melt [16]. The master
from these critical sizes, we can then get the critical cooling rates alloy was first remelted at ~ 1473 K in an alumina crucible in a vacu-
for Al alloys indirectly. Unfortunately, to the knowledge of the um of 10 −2 Pa. The melt was teemed through a guide tube, and then
atomized by a jet of argon at ~ 3 MPa. The atomized powder was col-
⁎ Corresponding author. Tel.: +852 34428416; fax: +852 34420172. lected and sieved in a closed system filled with nitrogen. The ribbons
E-mail address: [email protected] (S. Guo). were prepared by remelting the master alloy in a quart tube in argon

0022-3093/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.jnoncrysol.2012.06.023
2754 S. Guo, Y. Liu / Journal of Non-Crystalline Solids 358 (2012) 2753–2758

and then ejecting the molten liquid through a nozzle onto a rotating
Cu wheel with a rotation speed up to 2700 rotation per minute (rpm).
The surface morphology of the powders with different size was
observed using scanning electron microscopy (SEM). The amorphicity

Intensity, I/cps
of ribbons was identified by X-ray diffraction (XRD) using Cu-Kα
radiation.
1500 rpm, 50 μm

3. Results

The surface morphology of the gas atomized powder is shown in 2700 rpm, 25 μm
Fig. 1. It can be seen that the powders are almost perfectly spherical
in shape (although a few ellipsoidal powders were also seen, not
shown here). Those particles with diameter smaller than ~ 10 μm 10° 20° 30° 40° 50° 60° 70° 80° 90°
(Fig. 1a) are featureless in the microstructure, possibly having an Angle, 2θ
amorphous structure; while in those particles with larger diameter
hence slower cooling rate (~ 30 μm in Fig. 1b) dendrites appear, Fig. 2. XRD patterns of the melt-spun Al82Ni10Y8 alloy ribbons with different thickness.
suggesting they are already crystalline.
The XRD patterns for melt spun ribbons with thickness of 25 μm 4. Calculation
and 50 μm are shown in Fig. 2. The ribbons prepared using a faster
Cu wheel rotation speed are thinner in thickness. Both ribbons are In this section, we will model the heat transfer during the gas at-
identified to be XRD amorphous. Compared to the critical powder omization and melt spinning processes, and estimate the cooling
size, the determination of the critical ribbon thickness to achieve rates for these two processes based on the models, from intrinsic
the fully amorphous state is more time consuming, as the rotation properties of the materials and also properties of the cooling media
speed has to be varied continuously. We will show later that the crit- (argon gas or the Cu wheel). The heat transfer for the gas atomization
ical ribbon thickness is estimated to be ~ 85 μm for the Al82Ni10Y8 process is relatively more straightforward and we deal with it first.
alloy. For the melt spinning process, the heat transfer is more complicated
and we made many simplifications on it. We want to emphasize
here that these treatments actually over-simplify the real heat trans-
fer in the melt spinning process and the mathematical solution is
hence not robust. However, the estimated critical cooling rates for
the melt spinning process, out of apparently simplified treatments,
(a) agree well with that for the gas atomization process. It is hence our
belief that the scheme we put forward here is still of practical value,
to reasonably estimate the cooling rates for the melt spun ribbons.

4.1. Cooling rate calculation for the gas-atomized powders

In the gas atomization process, the high-speed gas flow impinges

on the liquid flow ejected from the nozzle and the liquid shatters
into small melt drops. The melt drops release heat through their sur-
faces during the flight in the chamber, while the gas in the surround-
ing environment cools the melt drops via the convective heat transfer.
A few assumptions are made here for the heat transfer process:
1) The melt drops reach the thermal equilibrium state instantly. As
2 μm such, their internal temperature could be assumed homogeneous;
2) The surrounding environment is assumed to be an infinite space.
The heat released from the melt drops diffuses to the surrounding
(b) environment rapidly, and the environmental temperature is as-
sumed to be constant.
3) Since there exists a huge difference between the melt drops and
the environment, the heat transfer is assumed to be completely in-
terface controlled and the cooling follows the Newtonian heat
transfer mode.
The thermal equilibrium condition is thus the heat release from
the melt drops equaling to the heat conducted to the environment
via the melt drop surfaces. Mathematically, it can be written as [16]:

dT d hAðT d −T e Þ
¼− ð1Þ
dt ρVC p
5 μm
where V is the volume (m 3); ρ is the density (kg m −3); Cp is the spe-
Fig. 1. Surface morphology of the gas-atomized Al82Ni10Y8 alloy powders with different cific heat capacity (J kg −1 K −1); A is the surface area of the melt drop
particle size. (a) b10 μm and (b) ~30 μm. (m 2); h is the coefficient of heat transfer of the gas (W m −2 K −1); Td
 S. Guo, Y. Liu / Journal of Non-Crystalline Solids 358 (2012) 2753–2758 2755

 −n
and Te are melt drop temperature (K) and environment temperature spacing (SDAS), λ2, in the larger sized powder (λ2 ¼ A′ dT dt
, A′ and
(K), respectively; and t is the time (s). n are constants [20]), where the slower cooling rate leads to crystalli-
For simplification, a spherical shape with a diameter of d (m) is as- zation (see Fig. 1 (b)). The two methods gave highly consistent re-
sumed for the melt drop. This assumption is also a close reflection of sults (2.1 × 10 4 K s −1 estimated from SDAS, and 2.4 × 10 4 K s −1
the real powder shape, as shown in Fig. 1. Eq. (1) then becomes: calculated from Eq. (7), for a 65 μm-sized powder particle) [21], and
this certifies the validity of using Eq. (7) to estimate the cooling rate
dT d 6hðT d −T e Þ for the gas-atomized powder.
¼− : ð2Þ
dt ρC p d For powders with diameters of 10 μm and 30 μm, their cooling
rates are 9.8 × 10 5 and 1.1 × 10 5 K s −1, respectively according to
The critical parameter for calculating the cooling rate is the heat Eq. (7). The critical cooling rate to obtain amorphous Al82Ni10Y8
transfer coefficient, h. Between the inert gas and the molten drops, alloy powder is thus ~ 10 6 K s −1.
h has the form of [17,18]

kg
1=2 1=3

4.2. Cooling rate calculation for the melt-spun ribbons
h¼ 2 þ 0:6Re P r ð3Þ
d
There exist many discussions on the heat transfer for the melt
where kg is the thermal conductivity of the gas (W m −1 K −1), Re is
spinning processes in the literature (for example, see [22–31]). How-
the Reynolds number and Pr is the Prandtl number. Re can be further
ever, a straightforward expression to give the cooling rate for the
expressed as [19]
melt-spun ribbons as a function of ribbon thickness is rarely seen.
  The main reason is that the heat transfer in the melt spinning process
 
vd −vg ρg d is very complicated and many factors need to be taken into account
Re ¼ ð4Þ
μg [22–27]. To treat the heat transfer issue rigorously is not the target
of this paper, and instead we aim at simplifying the mathematical de-
where vd and vg are the velocities for the melt drop and the gas (m s −1), scription of the melt spinning process and generating an expression
respectively, ρg is the density of the gas (kg m−3) and μg is the kinetic that can reasonably estimate the cooling rate of the melt-spun rib-
viscosity of the gas (kg m−1 s−1). bons, even though some of our assumptions are not physically robust.
From Eqs. (2), (3) and (4) we have As the critical cooling rate to form an amorphous alloy for the same
0 0  11=2 1 composition shall be in principle an independent process, if the criti-
  
dT d  6ðT d −T e Þ B2kg kg ρg vd −vg  cal cooling rate estimated from the melt spinning process is compara-
  @ 2 þ 0:6 @ A Pr C
1=3
 dt  ¼ ρC p d d μgd
A: ð5Þ ble to that estimated from the gas atomization process as we
discussed above, we then regard the scheme we use here for the
melt spinning process as reasonable, at least for the sole purpose of
The symbol ‖ indicates the absolute value and this makes the cooling estimating the cooling rate for the melt-spun ribbons.
rate a positive value. After the metal liquid is shattered into drops with Here is our simplified treatment for the heat transfer in the melt
different sizes by the flowing gas, the smaller drops get higher acceler- spinning process. The molten liquid solidifies rapidly once it is ejected
ation and they catch up with the gas speed easily and solidify rapidly; from the quartz nozzle and forms thin ribbons on the rotating Cu
on the other hand, those larger drops get lower acceleration and they wheel. Compared to the thickness of the ribbons, their width and
fail to catch up with the high-speed gas and hence break away from length can be assumed to be infinitively large and so are the width
the gas flow before solidification. We can then assume the relative and diameter of the Cu wheel. The Cu wheel can be seen as a
speed between the melt drop and the gas, (vd − vg), to be zero, and semi-infinite solid. If we ignore the radiation and convection of heat
Eq. (5) now simplifies to the form of during the solidification process, the heat transfer between the liquid
  alloy and the Cu wheel can be assumed to be a one dimension
dT d  12kg ðT d −T e Þ non-steady heat transfer process. The cooling rate of the ribbons
 
 dt  ¼ ρC d2
: ð6Þ
can be then estimated by solving the mature one dimension
p
non-steady equations.
The relevant parameters in Eq. (6) are listed in Table 1 and after The melt-Cu wheel contact is schematically shown in Fig. 3. The
substitution of these known parameters, Eq. (6) finally becomes: ribbon thickness is along the x direction and we set the origin at the
 
dT d  1  10−4
 
 dt ≈ d2
: ð7Þ

The cooling rate of the gas-atomized powder estimated by Eq. (7)

is compared with that estimated from the secondary dendrite arm

Table 1
Parameters used for calculating the cooling rate of the gas-atomized powder [33,34].

Thermal Density, Specific heat Temperature, Temperature,

conductivity, ρb/kg m− capacity, Cpb/J kg− Td/K Te/K
kga/W m−1 K−1 3 1 −1
K

1.79 × 10−2 2385 1080 1473 298

a
Thermal conductivity at 300 K.
b
For simplicity, density and specific heat capacity for pure Al liquid were used for the
Al82Ni10Y8 alloy. It is noted here that although the alloying effect and the temperature
dependence of these parameters will certainly affect the estimated cooling rate, the
variation shall be within one order of magnitude. This applies to the cited parameters Fig. 3. Schematic of the simplified contact between the ribbon and the Cu wheel during
in Table 2 as well. the melt spinning process.
2756 S. Guo, Y. Liu / Journal of Non-Crystalline Solids 358 (2012) 2753–2758

interface between the ribbon and the Cu wheel. The one dimension At the interface where x = 0, Fourier's law entails that:
heat transfer along the x direction can be mathematically described  
by [16]: ∂T 1 ∂T 2
k1 ¼ k2 : ð14Þ
∂x x¼0 ∂x x¼0
2 2
∂T ∂ T k ∂ T
¼α 2 ¼ ð8Þ
∂t ∂x C p ρ ∂x2 From Eq. (14) we obtain the interface temperature, Ti, as a func-
tion of T10 and T20:
where T is the temperature (K); t is the time (s); x is the distance to pffiffiffiffiffiffi pffiffiffiffiffiffi
k1 α 2 T 10 þ k2 α 1 T 20
the interface (m); α is the thermal diffusivity (W m 2 J −1); k is the Ti ¼ pffiffiffiffiffiffi pffiffiffiffiffiffi : ð15Þ
thermal conductivity (W m −1 K −1); Cp is the specific heat capacity k1 α 2 þ k2 α 1
(J kg −1 K −1); and ρ is the density (kg m −3).
The general solution to Eq. (8) is in the form of [16] To calculate the cooling rate of the ribbon, ∂T∂t1 , from Eq. (11), we
 still need to know the cooling time, t. Geller et al. [28] used the con-
x cept of temperature gradient to estimate t. It is assumed that ribbon
T ðx; t Þ ¼ A þ B erf pffiffiffiffiffiffi : ð9Þ
2 αt cooling rate is constant (within a very short cooling time), and the
x temperature variation of the Cu wheel as a function of time follows:
erf() in Eq. (9) is the error function and erf ðxÞ ¼ p2ffiffiπ ∫e−y dy. A and
2

0 T 2 ð0; t Þ ¼ T 20 þ βt ð16Þ
B are constants in the infinite integral. It is noted here that the solu-
tion in the form of Eq. (9) is a solution for a semi-infinite solid [16], where β is the temperature gradient in terms of time.
so basically it is suitable to be used for the Cu wheel but not for the From Eqs. (13) and (16), we can derive:
ribbons. However, out of consideration for easy mathematical treat-
pffiffi
ment, we use Eq. (9) as well to describe the heat transfer for the rib- ∂T 2 t
¼ −β pffiffiffiffiffiffiffiffiffi : ð17Þ
bons. We will see later that by doing this, the estimated critical ∂x πα 2
cooling rate from the melt spinning is comparable to that estimated
from the gas atomization process. We therefore think it is reasonable The boundary conditions at the interface are:
to use the scheme we put forward here to estimate the cooling rate of
the melt-spun ribbons. T 1 ð0; t Þ ¼ T 2 ð0; t Þ ¼ T ð0; t Þ ð18Þ
We then treat the heat transfer function for the ribbon and the Cu
wheel separately, both by solving Eq. (9) using the boundary condi- ∂T ∂T
C p1 ρ1 δ1 ¼ −k2 : ð19Þ
tions. For clarity, we use the subscripts 1 and 2 to denote the ribbon ∂t ∂x
and the Cu wheel, respectively.
For the ribbon (x b 0): In Eq. (19), δ1 is the thickness of the ribbon and ∂T
∂t
is β defined in
Eq. (16). From Eqs. (17) and (19), the cooling time, t, can be given

x in the form of
T 1 ¼ A1 þ B1 erf pffiffiffiffiffiffiffiffi : ð10Þ
2 α1 t
2
δ1 2 π C p1 ρ1
The boundary conditions for Eq. (10) are: t¼ : ð20Þ
k2 C p2 ρ2
1) at x = 0 (t > 0): T1 = Ti, where Ti is the temperature at the interface
between the ribbon and the Cu wheel; The relationship between the cooling time, t, and the ribbon thick-
2) at t = 0 (x b 0): T1 = T10, where T10 is the initial temperature of the ness, δ1, as reflected in Eq. (20) has been verified by the experimental
ribbon when the ribbon just touches the Cu wheel. T10 can be as- results [28]. According to Eq. (20), the time it takes to solidify a 25-μm
sumed to the temperature of the molten alloy. thick ribbon is about 8.5 × 10 −6 s. This short time scale can justify the
neglecting of the time dependence of the temperature for the bound-
From these two boundary conditions, A1 = Ti, B1 = Ti − T10.
ary conditions. From Eqs. (11), (15) and (20) we can calculate the
Eq. (10) now becomes:

cooling rate of the ribbon, ∂T∂t1 , at any distance, x. For the easy com-

x parison, we use the cooling rate at the outer surface (opposite to the
T 1 ¼ T i þ ðT i −T 10 Þ erf pffiffiffiffiffiffiffiffi : ð11Þ
2 α1 t surface that is in contact with the Cu wheel) to denote the cooling


rate of the ribbon, i.e., ∂T 1 x¼−δ .


∂t 1
Similarly, for the Cu wheel (x > 0):


x

  −α 1 e
C
−4πCp2
k 2 ρ2
pffiffiffiffiffi pffiffiffiffiffi
k α 2 T 10 þk2 α 1 T 20
−T 10 þ 1 k p ffiffiffiffiffi pffiffiffiffiffi



p1 k1 ρ1
T 2 ¼ A2 þ B2 erf pffiffiffiffiffiffiffiffi :
2 α2 t
ð12Þ  ∂T 1

 ∂t

j x¼−δ1 ¼

C k ρ 3=2

1 α 2 þk2 α 1
: ð21Þ
 2δ1 2 π2 C p1 k1 ρ1
p2 2 2

The boundary conditions for Eq. (12) are:

1) at x = 0 (t > 0): T2 = Ti; Although Eq. (21) looks complicated, it is basically a simple func-
2) at t = 0, (x > 0): T2 = T20, where T20 is the initial temperature of the tion of the ribbon thickness (1/δ 2) once the experimental conditions
Cu wheel when the ribbon just touches it. T20 is assumed to the (T10 and T20) are set, and all the other parameters are materials con-
room temperature. stants and are easily available in the materials handbooks. All the pa-
rameters needed for the calculation are listed in Table 2. From
From the boundary conditions we have A2 = Ti, B2 = T20 − Ti. Eq. (21), ribbons with thickness of 25 μm and 50 μm have cooling
Eq. (12) then becomes: rates of 1.5 × 10 7 and 3.7 × 10 6 K s −1, respectively. This magnitude
 of the critical cooling rate is comparable to that estimated from the
x gas-atomized powder, giving confidence to the scheme we use here
T 2 ¼ T i þ ðT 20 −T i Þ erf pffiffiffiffiffiffiffiffi ð13Þ
2 α2 t to estimate the cooling rate for the melt-spun ribbons. If we take
 S. Guo, Y. Liu / Journal of Non-Crystalline Solids 358 (2012) 2753–2758 2757

Table 2 1010
Parameters used for calculating the cooling rate of the melt-spun ribbon [33,34].
(a) ΔT=100K
Material Thermal Specific heat Density, Thermal Temperature, *
109 ΔT=200K
conductivity, capacity, Cp/ ρ/kg m−3 diffusivity, T0/K

Cooling rate, R/ K s-1

k/W m−1 K−1 J kg−1 K−1 α/W m2 J−1 ΔT=300K
b b b −5 c ΔT=400K
Al82Ni10Y8 94 1080 2385 3.6 × 10 1133
108 ΔT=500K
melt1a
Cu 398 384 8900 1.2 × 10−4 298 ΔT=540K
#

wheel2a
107 ΔT=600K
a
Subscripts 1 and 2 denote the Al82Ni10Y8 melt and the Cu wheel, respectively.
b
For simplicity, thermal conductivity, density and specific heat and density for pure increasing ΔT
Al liquid were used for the Al82Ni10Y8 alloy. 106
c
An overheat of 200 K was assumed.

105
0 20 40 60 80 100
1.0 × 10 6 K s −1 as the critical cooling rate for the Al82Ni10Y8 alloy, we Ribbon thickness, δ /μm
can further estimate the critical ribbon thickness for the same alloy to 109
be ~ 95 μm. (b) ΔT=100K
Currently, when estimating the cooling rate of the amorphous al-
108 ΔT=200K
*
loys, the method proposed by Lin and Johnson [32] is most commonly

Cooling rate, R/ K s-1

used due to its simple form: ΔT=300K
107 ΔT=400K
−3 ΔT=500K
dT 10
¼ 2 ð22Þ 106 #
dt R ΔT=540K
ΔT=600K
where R (m) is the typical dimension (thickness or diameter) of the 105
sample. However, detail on how Eq. (22) is derived is not available.
increasing ΔT
The cooling rates for 25 μm and 50 μm sized ribbons calculated 104
according to Eq. (22) are 1.6 × 10 6 and 4 × 10 5 K s −1, about one
order of magnitude lower than those given by Eq. (21). Comparing
103
Eq. (7) with Eq. (22), for the same dimension in thickness or diame- 0 20 40 60 80 100
ter, the former yields a cooling rate one order of magnitude lower Powder diameter, d/μm
than the latter, hence two orders of magnitudes lower than
Eq. (21). For easy comparison, the estimated cooling rates for the rib- Fig. 5. Effect of overheating on the cooling rates for (a) the melt-spun ribbons and (b)
bon and powder are plotted in Fig. 4. the gas-atomized powders. The symbol * in the legend denotes the overheating as-
sumed for the ribbon and # denotes the overheating for the powder.

5. Discussion
overheating) of the alloy, the higher the cooling rate. The melt tem-
The effect of melt temperature on the GFA has been experimental- perature effect on the cooling rate for the melt-spun ribbon is not im-
ly investigated [11–15], with the similar conclusion that increasing mediately obvious from Eq. (21). For clarity, the cooling rates for both
the melt temperature would increase the GFA. Does the melt temper- melt-spun ribbons and gas-atomized powders are plotted in Fig. 5, as
ature have an effect on the cooling rate? This is immediately obvious a function of ΔT. It is clearly seen from Fig. 5 that in both cases, the
from Eq. (6) for the case of the gas-atomized powder: the higher Td cooling rate increases with increasing ΔT, and the increase is more
(= T m þ ΔT, where Tm is the melting temperature, ΔT is the noticeable at larger dimensions, i.e., thicker ribbons or larger powder
particles. However, even at the most significant case the effect of ΔT
on the cooling rate is limited: a difference of 500 K in ΔT would result
1010 1010 in a change of the cooling rate within half an order of magnitude.
Ribbon Fig. 5 tells us that the benefit of the higher melt temperature on the
109 Powder 109 GFA can almost not come from the increase in the cooling rate. Actu-
ally it has been reported that only a moderate ΔT can effectively en-
Cooling rate, R/ K s-1
Cooling rate, R/ K s-1

108 108 hance the GFA [15], not that the higher ΔT the better, again
suggesting that the improvement of GFA is not due to the cooling
107 107 rate effect. One plausible explanation for the benefit of the increased
melt temperature on GFA is due to that the tendency for atoms to
106 106 form clusters decreases at higher melt temperature [13], and the
amount of pre-existing nuclei in the liquid reduces, hence inhibiting
105 105 the crystallization or in other words enhancing the GFA. However,
one needs to be aware that at higher melt temperature, the absorp-
104 104 tion of oxygen gets more serious and it would lead to easier nucle-
ation in the liquid and hence deteriorate the GFA [15].
103 103
0 20 40 60 80 100
Ribbon thickness, δ/μm 6. Conclusions
Powder diameter, d/μm

Fig. 4. Comparison of the cooling rates for the melt-spun ribbons and gas-atomized 1) The cooling rates for melt-spun ribbons and gas-atomized pow-
powders. ders were estimated. The cooling rates ( dT
dt
) for both ribbons and
2758 S. Guo, Y. Liu / Journal of Non-Crystalline Solids 358 (2012) 2753–2758

powders are inversely proportional to the square of their dimen- [14] C. Fan, A. Inoue, Influence of the liquid states on the crystallization process of
nanocrystal-forming Zr–Cu–Pd–Al metallic glasses, Appl. Phys. Lett. 75 (1999)
sions, thickness (δ) or diameter (d): dT
dt
∝ δ12 or dT
dt
∝ 12 . 3644–3646.
d
2) As an example, the critical cooling to form amorphous Al82Ni10Y8 [15] J. Mao, H.F. Zhang, H.M. Fu, A.M. Wang, H. Li, Z.Q. Hu, The effects of casting tem-
6 −1 perature on the glass formation of Zr-based metallic glasses, Adv. Eng. Mater. 11
alloy was estimated to be 1.0 × 10 K s .
(2009) 986–991.
3) The effect of the melt temperature on the critical cooling rate was [16] J.P. Holman, Heat Transfer, 8th Edition McGraw-Hill, New York, 1997.
quantitatively evaluated and the higher the melt temperature the [17] J. Szekely, N.J. Themelis, Rate Phenomena in Process Metallurgy, Wiley-Interscience,
faster the cooling rate. However, such an enhancement of the New York, 1971..
[18] Y.Z. Zhou, Y. Wu, E.J. Lavernia, Process modeling in spray deposition: a review, Int.
cooling rate by increasing the melt temperature is not significant, J. Non-Equilib. Pr. 10 (1997) 95–183.
and a difference of 500 K in the melt temperature would lead to [19] J.L. Estrada, J. Duszczyk, Characteristics of rapidly solidified Al–Si–X powders for
the change of the cooling rate within half an order of magnitude. high-performance applications, J. Mater. Sci. 25 (1990) 886–904.
[20] C.A. Muojekwu, I.V. Samarasekera, J.K. Brimacombe, Heat transfer and micro-
It was suggested that the enhancement of the glass forming ability
structure during the early stages of metal solidification, Metall. Mater. Trans. B
by increasing the melt temperature does not come from the in- 26 (1995) 361–382.
crease in the cooling rate. [21] S. Guo, Fabrication of Non-equilibrium Al-based Alloys (in Chinese), Master The-
sis, in, Central South University, 2005.
[22] E.M. Gutierrez, J. Szekely, A mathematical model of the planar flow melt spinning
References process, Metall. Trans. B 17 (1986) 695–703.
[23] H. Yu, A fluid mechanics model of the planar flow melt spinning process under
[1] A. Inoue, T. Zhang, T. Masumoto, Production of amorphous cylinder and sheet of low Reynolds number conditions, Metall. Trans. B 18 (1987) 557–563.
La55Al25Ni20 alloys by a metallic mold casting method, Mater. Trans. 31 (1990) [24] B. Karpe, B. Kosec, M. Bizjak, Modeling of heat transfer in the cooling wheel in the
425–428. melt-spinning process, J. Achiev. Mater. Manuf. Eng. 46 (2011) 88–94.
[2] W. Klement, R.H. Willens, P. Duwez, Non-crystalline structure in solidified gold– [25] J.K. Carpenter, P.H. Steen, Heat transfer and solidification in planar-flow
silicon alloys, Nature 187 (1960) 869–870. melt-spinning: high wheel speeds, Int. J. Heat Mass Transf. 40 (1997) 1993–2007.
[3] A. Inoue, Stabilization of metallic supercooled liquid and bulk amorphous alloys, [26] P.H. Steen, C. Karcher, Fluid mechanics of spin casting of metals, Annu. Rev. Fluid
Acta Mater. 48 (2000) 279–306. Mech. 29 (1997) 373–397.
[4] A. Inoue, A. Takeuchi, Recent development and application products of bulk glass- [27] A. Sundararaj, B.G. Thomas, Heat transfer during melt spinning of Al–7%Si alloy on
y alloys, Acta Mater. 59 (2011) 2243–2267. a Cu–Be wheel, In: in: D.H. DeYoung (Ed.), Light Metals 2008, Wiley, New
[5] H.M. Fu, J. Mu, A.M. Wang, H. Li, Z.Q. Hu, H.F. Zhang, Synthesis and compressive Orleans, 2008, pp. 793–809.
properties of Al–Ni–Y metallic glass, Philos. Mag. Lett. 89 (2009) 711–716. [28] M. Geller, E. Brook-Levinson, V. Manov, Heat transfer during preparation of amor-
[6] B.J. Yang, J.H. Yao, Y.S. Chao, J.Q. Wang, E. Ma, Developing aluminum-based bulk phous metallic alloy ribbon, In: in: M. Zinigrad (Ed.), Proceedings of the Second
metallic glasses, Philos. Mag. 90 (2010) 3215–3231. International Conference on Mathematical Modeling and Computer Simulation
[7] B.J. Yang, J.H. Yao, J. Zhang, H.W. Yang, J.Q. Wang, E. Ma, Al-rich bulk metallic glasses of Metal Technologies, Ariel, Isreal, 2002, pp. 419–424.
with plasticity and ultrahigh specific strength, Scr. Mater. 61 (2009) 423–426. [29] V.I. Tkatch, A.I. Limanovskii, S.N. Denisenko, S.G. Rassolov, The effect of the
[8] L.C. Zhuo, S.J. Pang, H. Wang, T. Zhang, Ductile bulk aluminum-based alloy with melt-spinning of processing parameters on the rate cooling, Mater. Sci. Eng., A
good glass-forming ability and high strength, Chin. Phys. Lett. 26 (2009) 066402. 323 (2002) 91–96.
[9] J. Mu, H.M. Fu, Z.W. Zhu, A.M. Wang, H. Li, Z.Q. Hu, H.F. Zhang, Synthesis and [30] V.I. Tkatch, S.N. Denisenko, O.N. Beloshov, Direct measurements of the cooling
properties of Al–Ni–La bulk metallic glass, Adv. Eng. Mater. 11 (2009) 530–532. rates in the single roller rapid solidification technique, Acta Mater. 45 (1997)
[10] S. Guo, Z.P. Lu, C.T. Liu, Identify the best glass forming ability criterion, Intermetal- 2821–2826.
lics 18 (2010) 883–888. [31] M.J. Kramer, H. Mecco, K.W. Dennis, E. Vargonova, R.W. McCallum, R.E.
[11] H.C. Sheng, X.R. Zeng, H.X. Qian, D.J. Fu, Effect of melt treatment on the micro- Napolitano, Rapid solidification and metallic glass formation—experimental and
structure and magnetic properties of Nd2Fe14B/α-Fe nanocomposites, theoretical limits, J. Non-Cryst. Solids 353 (2007) 3633–3639.
J. Non-Cryst. Solids 356 (2010) 19–23. [32] X.H. Lin, W.L. Johnson, Formation of Ti–Zr–Cu–Ni bulk metallic glasses, J. Appl.
[12] C. Fan, A. Inoue, Influence of liquid temperature on the crystallization behavior in Phys. 78 (1995) 6514–6519.
Zr–Al–Cu–Pd amorphous alloy, Mater. Trans. 40 (1999) 1178–1180. [33] J.H. Lienhard, J.H. Lienhard, A Heat Transfer Textbook, 3rd Edition Phlogiston
[13] C. Fan, D. Chen, P.K. Liaw, H. Choo, C. Benmore, J. Siewenie, G.L. Chen, J.X. Xie, C.T. Press, Cambridge, Massachusetts, 2003.
Liu, Influence of molten status on nanoquasicrystalline-forming Zr-based metallic [34] F. Cardarelli, Materials Handbook: A Concise Desktop Reference, 2nd Edition
glasses, Appl. Phys. Lett. 93 (2008) 261905. Springer Verlag, London, 2008.