A MICROSCOPIC MECHANISM FOR STEADY

STATE INHOMOGENEOUS FLOW IN

METALLIC GLASSES
FRiSS SPAEPENt
Harv-ard University, Division of Engineering and Applied Physics.
Cambridge, MA 01138, U.S..%

(Recrkrd 7 Srptemher 1976)

Abstract-An empirical deformation map for metallic glasses is introduced and the two modes of
deformation, homogeneous and inhomogeneous Row are reviewed. The microscopic mechanism for
steady state inhomogeneous How is based on a dynamic equilibrium between stress-driven creation
and diffusional annihilation of structural disorder. The formalism is developed using the free volume
as the order parameter. The boundary line between the homogeneous and inhomogeneous flow regiotx
on the deformation map is calculated. The stress-strain relation in inhomogeneous flow approaches
ideally plastic behavior.

R&urn&-On introduit une carte empirique de la deformation des verres mttalliques et Ton passe
en revue les deux modes de deformation. I’ecoulement homogene et I’ecoulement htteroeens. Le
m&canisme microscopique de I’&coulement httirogine en regime permanent repose sur un ;quilibre
dynamique entre la creation produite par la contrainte et I’annihilation par diffusion du desordrt
structural. On utilise un formalisme dans lequel le volume libre est considtrt comme parametre d’ordrs.
On calcule dans la carte de la deformation la ligne de separation des regions d’ecoulement homogtce
et htterogene. Les relations contrainte-deformation en ecoulement htdrog2ne sont proches du comporte-
ment idealement plastique.

Zusammenfassung-Eine empirische Verformungskarte wird fir metallische GlPser eingefuhrt ; zwei

Arten der Verformung, homogenes und inhomogenes Fliel3en. werden besprochen. Der mikroskopische
Mechanismus des stationlren inhomogenen FlieDens baut auf einem dynamischen Gleichgewicht
zwischen spannungsgetriebener Erzeugung dsr struktureilen Unordnung und ihrer Annihilation durch
Diffusion auf. Der Formalismus wird mit dcr freien Energie als Ordnungsparameter entwickelt. Die
Grenzlinie zwischen Gebieten homogenen und inhomogenen FlieDens in der Verformungskarte wird
berechnet. Die Spannungs-Dehnungsbeziehung bei inhomogenem FlieDen kommt ideal plastischem
Verhalten nahe.

1. INTRODUCTION: AN EklPIRICAL and T (usually normalized by the shear modulus p

DEFOR.tLATION lMAP FOR and the melting point TM, respectively), and to outline
iMETALLIC GLASSES regions in which each of the various mechanisms is
dominant.
The deformation map. developed by Ashby and
Plotting experimental flow data $T. T) in the same
Frost for crystalline systems [l-3] is a convenient
coordinate system results in an empirical deformation
means of surveying the various modes of plastic
map. A comparison of the calculated and empirical
deformation of a material. To construct a map. each
maps provides a test of the flow equations. But even
mode of deformation is described by a steady state
if no calculated map is available yet. an empirical
constitutive flow law:
map is useful for surveying the data and outlining
I; = f(s, ?;structure) the various flow mechanisms. Therefore, the available
flow data for Pd-based metallic glasses are presented
where ? is the shear strain rate, T the shear stress,
on such an empirical map in Fig. 1. Ideally. only data
T the temperature; ‘structure’ represents all the rele-
obtained from samples with identical composition
vant structural parameters of the material. The steady
should be compared, but the relative paucity of the
state condition implies that the structural parameters
data, together with the large flexibility in preparing
are uniquely determined by the external parameters
amorphous metal alloy compositions make it necess-
stress and temperature, and hence remain constant
ary to group the data together according to their
during the course of the flow. An appropriate super-
major constituent. With one exception, all the alloys
position of the flow mechanisms makes it possible
represented in Fig. 1 contain 74-82 at.?,, Pd and
to draw constant 5 contours on a map with axes r
16.5-20 at.% Si. Their structural characteristics (den-
t 1.B.M. Postdoctoral Fellow, 1975-76. sity, radial _distribution function, modulus.. . .) and
407
408 SPAEPEN: STEADY STATE FLOW Iy lMETXLLIC GLASSES

(MN mm21
.
1000
1

Pd BASED METALLIC GLASSES

10

/
I * wr4 h 5% CnW I WPEN, ISTS (12)
, / ,I
100 200 300 400 500 600 ‘$6I )O
TPK)

Fig. 1. Empirical deformation map for Pd-based metallic glasses. The number at each data point
is log k (set-I). Both stress (a) and strain rate (4) are uniaxtal. The glass transition temperature (T,)
and the crystallization temperature (T;) are for Pda,Si,, [13].

transformation parameters (glass transition tempera- 2. THE FLOW hIECHASISMS

ture Tg,crystallization temperature T,, heat of crystal-
lization) vary within acceptable limits, given the typi- Inspection of the samples and fracture surfaces after
cal accuracy of mechanical testing. Since all the data testing shows that there are two basic modes of defor-
were obtained in either uniaxial tension or compres- mation: homogeneous flow. where each volume ele-
sion, the stress (a) and strain rate 4) are plotted as ment of the specimen contributes to the strain, and
such and have not been converted to the equivalent inhomogeneous flow, where the strain is localized in
shear quantities. Where possible, some approximate a few very thin shear bands.
constant strain rate contours have been drawn in. The
flow stresses obtained from creep tests [4,7], com-
pression tests [9-ll], and ultimate tensile strength
measurements (the three highest temperature points
in Masumoto and Maddin’s tensile tests [S]) are
probably close to the steady state values. In the other
tensile tests [5-S, 121, however, fracture intervenes
before steady state is reached. In these cases, the frac-
ture stress has been plotted, since it comes closest
to the steady state value. Above ?; (the liquidus tem-
perature of the crystalline material with the same
composition-not shown in Fig. 1) flow is measured
by viscometric methods. No measurements of this
type have been reported for Pd-based glasses, but ex-
perience from other liquid metal alloys [14] shows
that the viscosity in this temperature regime is usually
around 10L3 Nm-‘sec. HOMOGENEOUS FLGW

Based on this empirical information. it is now poss-

ible to show an approximate picture of what a calcu-
lated map shouid look like. Such a schematic map,
with stress and temperature axes extended further
than in Fig. 1, is shown in Fig. 2. The temperature
region between z and 7; is unaccessible for mechani- J
cal measurement because crystallization intervenes, r. r, rt r
but it is possible to make a reasonable extrapolation Fig. 2. Schematic deformation map of a metallic glass. The
of the strain rate contours through this region, various modes of deformation are indicated.
 SP.\EPEN: STE;\DY STXTE FLOW IN MET.ALLIC GLASSES
As indicated on Fig. 2. this mode of deformation
occurs at low stresses and high temperatures. In this
regime. the How is close to Newtonian viscous (i.e.
$ x Y). as is clear from the spacing of the strain rate
contours on the map. In uniaxial tension, a specimen
thins dov.n uniformly during deformation. Fracture
occurs after extensive plastic flow. a-hen some section
of the specimen has narrowed down to zero thickness.
For convenience. the region of homogeneous dcfor-
mation can be subdivided. admittedly somewhat arbi-
trarily. in three parts. depending on the viscosity of
the material.
(i) For 7 > T;. the material has a viscosity
q : 10-3Nm-‘sec. and can be characterized as
-fl&‘. Flow measurements must be performed by
viscometric methods. Since the viscosity changes only
slonlq u-ith temperature [i-l]. the strain rate con-
tours in this regime are almost horizontal.
(ii) For T : G. the material has. by definition of
T’. a viscosity between IO” and lOLo Nm-’ sec.. and
could be called ‘dscous’. In this regime. the viscosity,
which is measured in a creep test. falls very steeply
with increasing temperature [4, 151. Consequently, the
strain rate contours are almost vertical around &.
(iii) For P< q. the material has a viscosity
rl > 10” Nm-’ sec., and can be called ‘solid’. From
the limited number of creep tests available in this
regime [7. 161. it seems that the viscosity does not
change rapidly with temperature, which is retlected
on the map by the strain rate contours becoming
more horizontal. However, conclusions about flow in
this region must be regarded with caution because
of the limited number of experiments available. and’
the largely unstudied influence of the thermal and
mechanical history of the material on the properties.
Some careful investigations of these effects are necess-
ary.
This mode of deformation occurs at high stress
levels (See Fig. 2). In this regime, the stress is very
strain rate insensitive (or, in other words: the stress
exponent II = ? log ?/? log r is very large), so that the
flow is almost ideally plastic. The flow stress for a
given strain rate. when normalized by the temperature
dependent shear modulus (i.e. r,‘,~), is almost constant
with temperature. except at low temperatures and
near r,. On the deformation map, this results in very
closely spaced. almost horizontal strain rate contours.
In inhomogeneous flow the strain is localized in
a few very thin shear bands. In ribbons pulled in ten-
sion these shear bands are planes, usually perpendicu-
lar to the thin side of the ribbon and at a 45’ angle
with the tensile axis [j, 6, 171. Slip on these planes
can be very extensive. This weakens the specimen
locally by decreasing the cross section, until finally
fracture occurs nior7g the plar7rs of‘ Chow shrar
bnrds [6. 121. The fracture surface exhibits a typical
309
vein-like pattern [6, 11. 181. which indicates that the
mechanism of fracture is the Taylor instability
[l&,0].
The basic physical process underlying this inhomo-
geneous Row phenomenon is a local softening of the
material. This becomes clear from a number of obser-
vations
(i) Localization of the flow in shear bands requires
that there is some structural change of the material
inside the bands. such that they deform much faster
there than in the rest of the specimen [20].
(ii) Fracture in a tensile specimen occurs, as indi-
cated above. along the plane of a shear band. which
makes a 45’ angle with the tensile axis. Ordinarily,
one would expect fracture to occur along the plane
of maximum normal stress, i.e. normal to the tensile
axis. This means that if fracture occurs along a shear
plane with a smaller resolved normal stress, the
material along this plane must have been weakened
by some shear-related structural change.
(iii) The vein pattern on the fracture surface is very
similar to that obtained by the pulling apart of two
solid surfaces containing a thin viscous layer between
them [Y, 21, 711. This observation is consistent with
(ii): the shear band consists of a layer of material
with a viscosity lower than that of the rest of the
specimen, which weakens the plane against fracture
by the Taylor instability mechanism.
(iv) Differential etching of the shear bands after
deformation [13.24] indicates that there are chemical
changes associated with flow. This is to be expected
if the shear-induced structural changes do not com-
pletely relax out after removal of the load.
3. THE GENERAL FLOW EQUATION
The microscopic mechanism which governs both
homogeneous and inhomogeneous Row is illustrated
by Fig. 3. It is assumed that macroscopic tlow occurs
as a result of a number of individual atomic jumps.
In order for an atom to jump, it must have a nearest
neighbor environment as shown in Fig. 3, i.e. next
to it there must be a hole large enough to accommo-
date its (appropriate hard-sphere equivalent) atomic
volume c*. It is reasonable to assume that the atomic
positions before and after the jump are positions of
relative stability, i.e. local free energy minima. In
order to make the atom jump, some activation energy
of motion AGm must be supplied. If no external force
is present, this is obtained from thermal fluctuations;
the number of jumps across the activation barrier is
the same in both directions; this is the basic micro-
scopic mechanism for diffusion. When an external
force. e.g. a shear stress, is applied, the atomic jumps
are biased in the direction of the force; the number
of forward jumps across the activation barrier is
larger than the number of backward jumps; this
results in a net forward flux of atoms and forms the
basic mechanism for flow. Quantitatively, this can be
410 SPAEPEN: STEADY STATE
I I
FLOW
’ AGmI
Dirtonce
Fig. 3. Illustration of an individual atomic jump, the basic
step for macroscopic diffusion and flow.
expressed as follows. When all the atoms in a speci-
men, deformed in shear, make one jump each of
length i. (2 1 atomic diameter) in the direction of the
shear, this would result in a macroscopic shear strain
7 z 1. If only a fraction of the atoms jump, the shear
strain rate is: 3 = fraction of atoms that make for-
ward jumps/second. Since only a fraction of the total
number of atoms in the sample are potential jump
sites, i.e. have a large enough hole next to them as
in Fig. 3, this can be written as
$ = (fraction of potential jump sites)
net number of forward jumps on each
X ).
of those sites per set
(
The fraction of potential jump sites is calculated in
the free volume theory of Turnbull and Cohen
[25-2q. The free volume of an atom is, intuitively,
that part of its nearest neighbor cage in which the
atom can move around without an energy change.
In an amorphous system, the free volume is distri-
buted statistically among all atoms. In Cohen and
Turnbull’s theory [ZS] the probability p(r)du of find-
ing an atom with a free volume between c and u + do
is calculated to be:
where 7 is a geometrical factor between 1 and l/2,
and uI is the average free volume of an atom.
In order for an atom to be on a potential jump
site, its free volume must be larger than c*, the effec-
tive hard-sphere size of the atom. Therefore, the total
FLOW IX METALLIC GLASSES
probability that an atom is on a potential jump site
is:
I,:$exp(-f)dr=exp(-5). (2)
It is also necessary to include a factor AJ the fraction
of the sample volume in which potential jump sites
can be found. For example. in homogeneous flow,
where the total volume contributes to Row, AIf= 1:
in inhomogeneous flow, where everything happens in
a few very thin bands, Af< 1. Therefore,
fraction of potential jump sites
= AJ-exp(- yc*/c/). (3)
The net number of forward jumps per second on a
potential jump site is obtained from simple rate
theory (see Fig. 3). The shear stress T exerts a force
TO on an atom. where a is the projected area of the
atom onto the plane of shear. When this atom makes
a jump of length i., the work done is sai.. Since 1.
is roughly equal to an atomic diameter, the work
done is rR where R is the atomic volume. The free
energy of the atom after the jump is therefore de-
creased by AG = tR. The net number of forward
jumps per set can now be calculated as the difference
between a forward flux over an activation barrier
AG” - AGi2 and a backward flux over an activation
barrier AG” + AG/2. Assuming an equal distribution
of atoms over the two equilibrium positions gives:
net number of forward jumps per second
per potential jump site
)I
AG” - rRJ2
= v exp - - exp - AGmk;*‘2 ,(4)
[ ( !iT > (
where u is the frequency of atomic vibration (- Debye
(1)
frequency). Substituting (3) and (4) into (1) gives:
j = Afexp(- $)?rsinh(&)exp(- s).
This is the general flow equation. In the case of
homogeneous deformation (Af= I) at low stress
levels (rQ Q ZkT), this leads to Newtonian viscous
behavior (2 sinh(&/2kT) 5 rR/kT), and an expres-
sion for the viscosity can be derived:
q=LkT-2xp[$)2xpf3 (6)
-j VR
The diffusion coefficient in a three dimensional ran-
dom walk process is defined as D = irri.’ where r
and i. are the jump frequency and distance, respect-
ively. An analysis, similar to that of the flow problem
gives :
r = (probability that an atom is on a potential
jump site) x (number of jumps an atom on this site
 SP.AEPEX: STE;\DY ST.ATE FLOW IN SIET.ALLIC GL;\SSES 411

makes per 9521

or:

i- = espj-
LJycsp(-g’1. (7)

This gives for the diffusion coetTicienr:

It is worth noting that combining equations (61 and

(8) giv-es
kT
Dty =
6(RiL)’

which is. save for a numerical factor of order 1. the

D,S:3-ft
Stokes-Einstein equation.
Fig. 4. Illustration of the creation of free IKxme. b)
squeezing an atom of volume c* into a neighbc<ny hole
0i smaller volume C.

4. THE SOFTESISG MECHANISM IN

INHO~IOGENEOUS FLOW (~1) Crrilfion ~J‘f/<r ~.oJu~tte
The energy necessary to squeeze an atom with
As has been discussed in Section 7tb). during inho-
volume r* into a smaller hole of volume I‘ isee Fig.
mogeneous How, the material in the shear bands un-
1) can be approsimated by the elastic distortion
dergoes some structural change that leads to a local
energy required to squeeze a sphere with voiume L.*
lowering of the viscosity.
into a spherical hole with volume L‘ in 3 matrix
Polk and Turnbull [l-l] have proposed that this
of the same material. This has been calcu1xed by
structural change is the net result of two competing
Eshelby [28]:
processes: a shear-induced disordering and a diffusion
controlled reordering process. This concept will be lG’ = $L.* - 0’
extended here. by deriv-ing a formal expression for .
L

the dynamic steady state equilibrium between the two

with
processes. The free volume will be chosen as the order
parameter. since it makes the problem mathemati- z If\
cally tractable. It should be kept in mind, however,
s = -/I -.
3 l-1
that to the extent that the free volume is equivalent
to other possible order parameters (compositional. If an atom makes the jump over the activation barrier
entropic,. .). the result should be quite generally in the direction of shear, its free ener_e is decreased
valid. by the driving term sR (see Section 3). but it is also
In the free volume formulation of the flow equation increased by the elastic distortion energy X’. The
(5). the structural parameter which governs the visco- driving free energy for creation of free v-olume is
sity is the average free volume L,.. Therefore. if there therefore :
is to be a lowering of the viscosity in the shear bands.
there must be an increase of the free volume.
It is proposed that. at high stress levels, free volume
can be created by the mechanism shown on Fig. 4. The volume u,,, of the smallest hole that an atom can
At a sufficiently high stress, an atom with hard-sphere still be squeezed into by a given stress is determined
volume c* can be squeezed into a neighboring hole by AG = 0. or:
with a smaller volume L’.This makes the neighbors
of the new position move out. and creates a certain ;-) = p* - 1‘J
amount of free volume. (7)
L’,
Competing with this is a relaxation process. which
tends to annihilate the extra free volume created and If T 4 S. this gives:
restore the system to its initial structural state. The
-
structural rearrangements necessary to annihilate free (‘m 2 c*(l - \ r S)
volume consist of a series of diffusional jumps.
In steady state a dynamic equilibrium is reached The amount of free volume created per sec. A-c,.
between the two competing processes: an equal can now be calculated from simple rate theoor];.
amount of free volume is created by the stress-driven Consider the amount of free volume created per
process. as is annihilated by the diffusional process. second associated with the sque’ezing open of holes
 412 SPAEPEN: STEADY ST.ATE FLOW IN METALLIC GLASSES

with a volume between c and c + dc. Therefore,

d(A-cf) = (number of potential sites)
$exp(-g;texp(-T)
x (net number of forward jumps on each
of those sites per set)
p* - C)’ __s(c* - c)’
x (amount of free volume created per
ZV &,.*
jump).
This gives:
The number of potential sites is equal to the total
number of atoms (1%‘)multiplied by the probability
that they have a neighboring hole with volume
between c and c + dc. This factor is obtained from
the free volume theory (see equation 2).
so:
i: 7.
number of potential sites = S -exp - L dr.
“f ( “I )
(91
The net number of forward jumps is calculated as
the difference between a forward and a backward flux,
analogously to equation (4). but with a different
AG = TR - AG’. So:
net number of forward jumps per potential site per In view of equation (7). which defines c,. this reduces
set to the final result:

AG” - rR.2 + AG’j2 A’c/ = 5 F[cosh($) - ,]_Vv

-
kT )
AG” + ~f-)“7
* ,- _ A@!7,- x exp(- g)exp(- 5). (12)
- exp - )]. (10)
lir

(b) Annihilnfion offree volume

The amount of free volume created per jump is simply
The structural rearrangements, required for annihi-
the difference between the volume of the atom and
lation of extra free volume, consist of a series of difhi-
the volume of the hole [X3]:
sive jumps. Let n, be the number of diffusive jumps
r* - G (111 necessary to annihilate a free volume equal to c*. It
is expected that nD should be a small number,
The total amount of free volume created per second between 1 and 10.
is obtained by integrating all the elementary contribu- This expectation is made plausible by the observa-
tions d(Afvl) over all the hole sizes c, from c,,, (the tion of a two-dimensional amorphous dynamic hard-
minimum that can be squeezed open) to L.* (above sphere model [29]. In two dimensions. an amorphous
which no free volume is created). This, combined with state of high density can only be obtained by using
equations (9-l 1) gives: different size spheres. In contrast with a crystalline
system, where a vacancy is a stable entity after many
1*
diffusive jumps, in an amorphous system an artifi-
A+c‘, = d(A’u,)
I urn cially created ‘vacancy’ is annihilated after a small
number of diffusive jumps.
Another useful analogy that can be made comes
= J”IlVtexp( - $v
from NabarreHerring creep in crystals. (This ana-
logy must obviously be used with caution, since
AG” - Ti-@ + s(V* - c)‘,Zc amorphous systems are topologically quite different
X exp -
[ ( kT > from an assembly of microcrystals.) In N-H creep,
vacancies travel a distance on the order of the grain

>I
AG” + Tfl’2 - .s(c* - V)‘,‘h diameter, and are annihilated at the grain boundaries.
-exp -

x(c*
-c)
( kT In an amorphous system something similar can be
imagined, but with a very small equivalent grain size.
dv. In crystalline systems, vacancies can annihilate at the
grain boundaries because there the structural require-
To simplify the integration, the following assumption
ment of crystalline translational symmetry is relaxed.
is made:
In an amorphous system, this requirement does not
T 4 S and c t c*. exist at all, and hence it is expected that ‘vacancies’
 SP;\EPEN: STEADY ST.ATE FLOW IN hIET.ALLIC GLASSES
or ‘extra free volume’ can be annihilated in just a
few ditfusional jumps.
The amount of free volume annihilated per second
is
A-c,. = (free volume annihilated per jump)
x (number of jumps per set).
The amount of free volume annihilated per jump is,
given the above definitions. c*/nn. The number of
jumps is NT, where the jump frequency r is given
by equation (7). Therefore,
A-c1 = K
n: iVv exp i- $)exp(- $J). (13)
(c) Sreody state
In steady state. the amount of free volume created
is equal to the amount of free volume annihilated:
A+G/ = A-c,
Or. from equations (12) and (13):
$ y[cosh(-$) - I] = $. (14)
Before the sample is loaded, the average free volume
c’/ is set by the structure of the quenched metallic
glass. At low stresses, the left hand term in (14) is
smaller than c*/nr,, i.e. the extra free volume created
by the stress can easily be annihilated by the diffusion,
and the average free volume stays at the initial value
uI. As soon as T exceeds the value for which (14)
is obeyed, however, more free volume is being created
than can be annihilated by diffusion; hence r, will
keep on increasing until condition (14) is obeyed
again. Or, in other words: the free colme c,. is set
by the stress T.
Formally, this can be seen by rewriting equation
(14):
,. .*
$= $&[cosh(&) - 1l-r (15)
When this explicit expression for ~t?‘cI is substituted
in the term exp( - ~L.*/u~)of the general flow equation
(5), a flow equation for inhomogeneous flow is
obtained.
5. DISCUSSION
The only unknown parameter in the flow equations
which prevents us from constructing a complete cal-
culated deformation map is AG”. However, if it is
assumed that AG”, which may be a function of T
and T, is the same for both homogeneous and inho-
mogeneous flow, it becomes possible to calculate the
boundary between the regions in which each of the
mechanisms dominates. The boundary between the
two mechanism regions on the deformation map is
a line r(T), where
413
The only factors in the genral tlow equation (5) that
are different from one mechanism to the other are
AJ”and exp( -;x* r,). So condition (16) becomes then
A/,zxp(- F). = AjIerp[- 5);
or
Since the free volume cf is a structural parameter.
and since all the data in the vicinity of the mechanism
boundary are obtained at temperatures far enough
below T4 where the configuration can be considered
frozen, (~c*,c/)~ is a constant in the equation of the
boundary line. Using equation (15), which determines
(c* r,.)(, equation (17) can be rewritten as:
(18)
with
(19) is the equation of the mechanism boundary line
on the deformation map. It contains only one adjust-
able parameter: 1. The other quantities are known
from experiment.
The temperature dependent shear modulus is:
;og(T-300)
.
P(T) = PO
(
1- -
!
where p. is the shear modulus at 300 K. The values
used here are for Pdi,,,Cu,$i,,.,:
p. = 3.29 x 10” Nm-’ (from Chen et al. (301)
and
1 +
- - = 2.6 x 10’ K-r (from Chen [31]).
POdT
The atomic volume is calculated from the den-
sity measurements [32] on Pds5.5Si16.j: R = 1.46 x
lo-l9 m3. The quantity .S = fdl + u/l - II) is calcu-
lated using a Poisson’s ratio L‘= 0.4, which is typical
for metallic glasses [30].
Figure 5 is an enlarged portion of the empirical
deformation map of Fig. 1, showing the boundary
between the regions of homogeneous and inhomo-
geneous flow. The creep data at the bottom of the
diagram (Maddin and Masumoto [7], open squares)
are obviously all in the homogeneous deformation
region. The transition between inhomogeneous and
homogeneous flow, however, shows up in Maumoto
and Maddin’s tensile data [S] (open circles). The three
data points at the highest temperature (53:+583 K)
are ultimate tensile strengths obtained from stress-
strain curves which show very extensive plastic defor-
414 SPAEPEN: STEADY STATE FLOW IN METALLIC GLASSES

lNHOMOtENEOUS FLOW

Fig. 5. Enlarged portion of the empirical deformation map for Pd-based metallic glasses. The equivalent
shear stress (T) and shear strain ($1 have been calculated from the uniaxial values by the von Mises
criterion (r = a/,‘li; 3 = L/m..
3~). The strain rate contours connect data points from the same reference.
The shear modulus is from Refs. 30 and 31. The heavy line is the boundary between regions of
homogeneous and inhomogeneous flow. calculated from equation (IS).

mation (several percent) before fracture. Therefore, Substituting the values for x. Afh, Ah. and (i’t*i;c,-),,
these points belong in the region of homogeneous in (19) leaves
flow. The parameter a in equation (13) can now be
adjusted such that the resulting boundary line passes a*/n = 0.26.
SD
between these points and the rest of the tensile data;
for the line shown in Fig. 5, x = 360.
This is a physically reasonable result. The equivalent
It remains to be shown now that this is a physically
hard-sphere volume of a late transition metal as Pd
realistic number. Therefore, the quantities in (19), the
probably takes up a large fraction of the atomic
defining relation for z, must be investigated.
volume, say, c*jfJ t 0.8. This leaves n, t 3, which
As discussed in Section 3, for homogeneous flow
is consistent with the physical considerations of Sec-
A_$,= 1. For inhomogeneous flow Aj is determined
tion 4(b), where it was argued that the number of
by the thickness of the softened layer in the shear
diffusional jumps necessary to annihilate a free
band. Since no direct measurements on this are avail-
volume equal to D* should be somewhere between
able, the thickness of this layer can only be estimated,
1 and 10.
e.g. from the height and size of the little ‘tributaries’
It may be worth noting that the mechanism bound-
in the fracture surface vein-pattern [Is, 123. A layer
ary on the deformation map has a positive slope, i.e.
thickness of 5OOA seems to be a reasonable estimate.
with increasing tem~rature it is necessary to go to
If the sample length is 5 cm and there are only a
higher stresses to get inhomogeneous flow. Physically,
few shear bands, Ah = IO? This is, of course, only
this is not unexpected, since the ‘diffusional relaxation
an estimate, but, fortunately, the equation is quite in-
process becomes relatively more important at high
sensitive to the exact value of A&
temperatures.
Chen and Goldstein [4] have measured
At this stage, the stress exponent n = i: log $‘Z log r
for the two flow modes can afso be calculated, if it
1,
,o+ 3180
-=- is assumed that AG” is independent of r. The results
Vf T - 515 of this calculation are shown in Fig. 6. The stress
exponent for homogeneous flow is n e 1, as it should
in Pd;l.&ubSilB.5, for homogeneous flow in the
be for Newtonian viscous flow. For inhomogeneous
region TZ q. As mentioned above, in the region
flow n % 50, which indicates that it approaches
under consideration here u, is considered constant,
ideally plastic behavior.
due to configurational freezing at some temperature
Finally, it should be emphasized that the present
around q. For T= 600, the viscosity of
mechanism involves monatomic systems, and does
Pd,,.&u&,,., is 10” Nm-' set [4]; below this the
not explicitly include any chemical effects. However,
system will be considered configurationally frozen.
chemical ordering is reflected partially in the free
The appropriate quantity to be used in equation (19)
volume formulation, for example, if an efficient pack-
is therefore:
ing of atoms of different size (as Pd and Si) requires
.* * a non-random chemical environment for the two
lU 3180
-= = 37.4. atoms. In this case, creation of free volume is equival-
i c’/>h 600 - 515 ent to creation of chemical disorder. On the other
 SP.AEPEN: STEADY ST;\TE FLOW IN 4fET.ALLIC GLASSES 415

REFERESCES
1. M. F. Ashby. .-Lcra.Uer. 20, gY7 (1973).
2. H. J. Frost and M. F. .Ashby. Proc. oj. thr John E.
Darn S.nnposirtm, (edited by J. C. Xl. Li and A. K.
Mukhejee). p. 70. .\.S.M.. Cleveland. OH. 11975).
3. M F. Ashby and H. J. Frost. Consriruticr Equtrrions
in Pfnsticir.v (edited by A. S. Argon), p. 117. hi.1.T.
VHOMOGENEOu! Press. Cambridge. MA. (1975t.
F’DV -I. H. S. Chcn and hl. Goldstem. J. uppl. Phys. -l3. 16-12
11971).
5. T. Masumoto and R. Maddin. .-lcrrr Jfrr. 19, 725
(19il).
6. H. J. Leamy. H. S. Chen and T. T. Wang. .Uer. Trims.
3, 699 (1972).
7. R. Uaddin and T. Masumoto. Ma. Sci. Eq. 9, 153
(1972).
8. H. S. Chen, Scripra .Uer. 7, 931 (1973).
9. C. A. Pampillo and H. S. Chen, Mar. Ssi. Eng. 13.
IS1 (1974).
IO. H. S. Chen and D. E. Polk. J. notwr,w. Solids 15.
17-1(197-l).
11. C. A. Pampillo. J. Mar. Sci. 10, 1194 (1975).
II. C.-P. P. Chou and F. Spaepen, .-lcra Mrr. 23, 609
(1975).
HOMOGENEOUS
13. H. S. Chen and D. Turnbull. ilcrn .Uer. 17, 1021 (1969).
FLOW
14. D. E. Polk and D. Turnbull. .-Icrcr&let. 20. 493 (1972).
15. H. S. Chen and D. Turnbull. J. Chem. Phxs. 48. 2560
/1 I 1 I
(1968).
_
0
_
I
_
8
_
I)
._
I” ‘oq N/r& 16. J. Logan and Xl. F. Ashby. .&ra Met. 11. 1037 (1971).
17. C. A. Pampillo and D. E. Polk, .&m Mer. 22, 741
Fig. 6. Stress-strain rate dependence of homogeneous and (1974).
inhomogeneous Bow. calculated from the theoretical lg. C. A. Pampillo and A. C. Reimschuessel. J. Jfnr. Sci.
models. 9, 718 (197-l).
19. F. Spaepen. Acrn Met. 23, 615 (1975).
20. A. S. Argon and M Salama Proc. 2nd Inr. Con/: on
hand, there are probably some specific chemical Rapidly Quenched ,Lletals (edited by N. J. Grant and
effects, as the metal-metalloid bonding, which should B. C. Giessen). Section II. Mat. Sci. Enn. 23. 219 (1976).
be included in AG”. The differences in strength found 21. F. Spaepen and D. Turnbull, Scripta .Lfet. 8, 563 (1974).
22. T. Masumoto and R. Maddin. .Uat. Sci. Eng. 19, 1
by changing the metalloid components in a metallic
(1975).
glass are probably of this nature. However, it is also 23. C. A. Pampillo, Scripta Met. 6, 915 (1972).
possible that an increased diffusivity of one type of 24. S. Takayama and R. Maddin. .Acm .Cler. 23. 943 (1975).
metalloid to the other might lead to an increased dif- 25. M. H. Cohen and D. Turnbull, J. Chem. Phxs. 31, 1164
fusional relaxation, and hence a higher strength. For (1959).
26. D. Turnbull and XI. H. Cohen. J. Chem. Ph~s. 34, 120
example, the exceptionally high strength of Fe,B
(1961).
glass [33] might conceivably be caused by the higher 27. D. Turnbull and M. H. Cohen, J. Chem. Phn. 52, 3038
diffusivity of the small B-atom. (1970).
28. J. D. Eshelby, Solid Srate Physics (edited by F. Seitz
Aclinowledgements-It is a pleasure to acknowledge several and D. Turnbull) 3. 115 (1956).
useful suggestions and criticisms from Professors D. Turn- 29. D. Turnbull and’ R.’L. Cormia J. appl. Phxs. 31, 671
bull and M. F. Ashby. I am indebted to Dr. H. J. Frost (1960).
for his help in constructing the deformation maps and for 30. H. S. Chen. J. T. Krause and E. Coleman. J. non-crysr.
several helpful discussions. Solids 18, 157 (1975).
This work has been supported in part by grants from 31. H. S. Chen. Scripta Met. 9, 411 (1975).
the Office of Naval Research under Contract 32. H. S. Chen. J. T. Krause and E. Coleman. Scriprn .Cfet.
NOOO14-67-A-0298-0036 and from the National Science 9, 787 (1975).
Foundation under Contract DMR-72-03020. 33. L. A. Davis. R. Ray. C.-P. Chou and R. C. O‘Handley.
I also want to acknowledge a postdoctoral Fellowship Scripto .Cfet. 10, 541 (1976).
from the I.B.;Cl. Corporation.