Journal of the Mechanics and Physics of Solids 112 (2018) 552–562

Contents lists available at ScienceDirect

Journal of the Mechanics and Physics of Solids

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

A simple model for indentation creep

Ryan S. Ginder a, William D. Nix b, George M. Pharr c,∗
a
Department of Mechanical, Aerospace, and Biomedical Engineering, University of Tennessee, Knoxville, TN, USA
b
Department of Materials Science and Engineering, Stanford University, Stanford, CA, USA
c
Department of Materials Science and Engineering, Texas A&M University, College Station, TX, USA

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

Article history: A simple model for indentation creep is developed that allows one to directly convert
Received 11 November 2017 creep parameters measured in indentation tests to those observed in uniaxial tests through
Revised 31 December 2017
simple closed-form relationships. The model is based on the expansion of a spherical cav-
Accepted 4 January 2018
ity in a power law creeping material modified to account for indentation loading in a man-
Available online 5 January 2018
ner similar to that developed by Johnson for elastic-plastic indentation (Johnson, 1970).
Keywords: Although only approximate in nature, the simple mathematical form of the new model
Power law creep makes it useful for general estimation purposes or in the development of other deforma-
Nanoindentation tion models in which a simple closed-form expression for the indentation creep rate is
Expanding cavity model desirable. Comparison to a more rigorous analysis which uses finite element simulation
Contact mechanics for numerical evaluation shows that the new model predicts uniaxial creep rates within
a factor of 2.5, and usually much better than this, for materials creeping with stress ex-
ponents in the range 1 ≤ n ≤ 7. The predictive capabilities of the model are evaluated by
comparing it to the more rigorous analysis and several sets of experimental data in which
both the indentation and uniaxial creep behavior have been measured independently.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction

Over the past decade, great progress has been made in the development of nanoindentation testing systems that can be
used at high temperatures, with some now reporting operation at up to 800 °C (Wheeler et al., 2015; Duan and Hodge, 2009;
Phani and Oliver, 2016). This new experimental capability paves the way for measuring creep properties at very small scales,
thereby facilitating the creep characterization of thin films, thin surface layers, and even small particles or individual phases
in complex multiphase materials. Despite this progress, how one converts data obtained in nanoindentation creep tests
to the parameters normally used to characterize uniaxial creep, i.e., those used by designers, is not at all straightforward
because of the complex, non-uniform stress states produced during indentation contact (Su et al., 2013; Martinez and Shen,
2016; Takagi and Fujiwara, 2014; Dean et al., 2014; Goodall and Clyne, 2006).
Understanding how creeping materials deform over their lifetime can be essential to the long-term operation of engi-
neering devices and structures. The creep behavior of a material is most often characterized in experiments conducted in
simple uniaxial tension or compression. The data obtained in such tests relate the steady state strain rate, ˙ , to the applied
uniaxial stress, σ , through the power-law relation

˙ = ασ n , (1)

∗
Corresponding author.
E-mail address: [email protected] (G.M. Pharr).

https://doi.org/10.1016/j.jmps.2018.01.001
0022-5096/© 2018 Elsevier Ltd. All rights reserved.
 R.S. Ginder et al. / Journal of the Mechanics and Physics of Solids 112 (2018) 552–562 553

Fig. 1. Indentation contact geometries used in this work illustrating the various measures of indentation depth, h, and contact radius, a, used for analysis:
(a) cylindrical flat punch; (b) conical indenter.

where the material parameters of interest are the stress exponent, n, and the uniaxial creep coefficient, α . The creep coef-
ficient is usually a strong function of temperature of the form exp(−Qc /kT), where Qc is the activation energy for creep, k
is Boltzmann’s constant, and T is the absolute temperature. Through this, the temperature dependence of the steady state
creep rate can also be explored by nanoindentation testing (Wheeler et al., 2015; Duan and Hodge, 2009; Phani and Oliver,
2016; Lucas and Oliver, 1999; Maier et al., 2013).
Under steady state flow conditions, numerous experimental studies have demonstrated that nanoindentation creep data
exhibit a similar power law form
˙ i = β pnm , (2)
where ˙ i is the indentation strain rate, pm is the mean pressure applied by the indenter (i.e., the hardness), and β and n are
the material specific indentation creep constants (Phani and Oliver, 2016; Su et al., 2013; Martinez and Shen, 2016; Takagi
and Fujiwara, 2014; Dean et al., 2014; Goodall and Clyne, 2006; Lucas and Oliver, 1999). Eq. (2) illustrates two important
issues in analyzing and understanding nanoindentation creep data. First, to interpret indentation data, a physical definition
of what the indentation strain rate is must be established, and second, since the creep coefficient β measured in nanoin-
dentation experiments is different from that in uniaxial testing due to differences in the testing geometries, one must have
a means of determining α from measurements of β .
Early experimental work on nanoindentation creep focused mostly on the problem of measuring the stress exponent, n
(Lucas and Oliver, 1999; Mayo and Nix, 1988; Raman and Berriche, 1992; Poisl et al., 1995; Chu and Li, 1977). This work
included flat punch “impression creep” experiments conducted by Chu and Li in succinonitrile in which they successfully
measured n by treating ˙ i as the rate of indenter penetration h˙ divided by the radius of the cylindrical punch, ac (Chu and
Li, 1977). Other early experiments by Mayo and Nix in Pb, Sn, and their alloys using the Berkovich triangular pyramidal in-
denter frequently used in nanoindentation testing showed that h˙ divided by the instantaneous indenter depth, h, is a useful
measure of ˙ i for this indenter geometry (Mayo and Nix, 1988). Although these early successes provided some empirical
evidence for what form ˙ i should take and how the stress exponent can be measured, they did not address the problem of
the conversion of β to α .
The first complete theoretical solution to these issues came from Bower et al. in their seminal 1993 paper (Bower et al.,
1993), hereafter referred to as the Bower analysis or the Bower model. The derivation is based on the generalized nonlinear
creep constitutive law for a multiaxial stress state
γ˙ /τ = 3α /2(σ ∗ )n−1 , (3)
where γ˙ is the shear strain rate, τ is the shear stress, and σ ∗ is the Mises flow stress defined in its usual way. The analysis
assumes that there is no primary or tertiary creep and that the material exhibits no history dependence, i.e., the strain rate
at any point in time depends only on the instantaneous state of stress through Eq. (3) and not on the strain path or history.
Using nonlinear elastic analysis procedures developed by Hill (1992), Bower et al. showed that the generalized indentation
creep rate can be written as
˙ i = h˙ /ac = α ( pm /F )n , (4)
where F is called the “dimensionless contact pressure” or “reduced contact pressure” and ac is the true contact radius. Re-
ferring to the geometries in Fig. 1, this relation applies to any axisymmetric indenter geometry, e.g., cylindrical flat punches,
cones, spheres, etc., where the radius of contact at the contact periphery, including any effects from pile-up or sink-in, is
ac . The amount of pile-up or sink-in is very important in this development because it directly affects the true mean inden-
tation pressure, pm . In the Bower analysis, this is characterized by a “pile-up/sink-in parameter” c = hc /h, where hc is the
depth along which contact is made by the sample and the indenter (the contact depth) and h is the total depth of indenter
penetration relative to the initial surface (see Fig. 1). For the geometrically self-similar conical indenter, c = hc /h = ac /anom ,
554 R.S. Ginder et al. / Journal of the Mechanics and Physics of Solids 112 (2018) 552–562

Fig. 2. Modification of expanding cavity models to describe indentation contact: (a) Johnson’s model for indentation in an elastic-perfectly plastic material
(Johnson, 1970), and (b) adaptation of the model to a power-law creeping solid as developed in this work.

where anom is the nominal contact radius defined as the contact radius that would be observed if there were no pile-up
or sink-in. Note that while Bower’s analysis suggests that h˙ /ac is the natural choice for ˙ i , for conical indentation h˙ /hc is
also appropriate given the linear geometric dependence of ac on hc . More importantly, however, the analysis provides a di-
rect way to relate the indentation creep coefficient, β , to the uniaxial creep coefficient, α . Specifically, comparing Eq. (4) to
Eq. (2) reveals that
β = α /F n . (5)
In order to use Eqs. (4) and (5) in the experimental evaluation of the uniaxial creep parameters from indentation creep
data, one must have numerical values for the pile-up sink-in parameter, c, and the reduced contact pressure, F. In their
development, Bower et al. show that F is a unique function of the stress exponent, n, and c is a function of both the stress
exponent and the indenter geometry. Since closed-form expressions for these relations do not exist, finite element methods
were employed to evaluate them, with tables of results presented in the paper (Bower et al., 1993). Subsequent work has
shown that for conical indenters, c and F also depend weakly on the indenter half-included angle, θ (Su et al., 2013). It
should be noted that the reduced contact pressure plays a role similar to the Tabor constraint factor of approximately
3 used to relate the hardness, H, to the uniaxial flow strength, σ , that is σ = H/3. In fact, the reduced contact pressure
asymptotically approaches a value close to 3 at large n’s, but is much smaller than this for stress exponents less than 5.
Although the analysis of Bower et al. is undoubtedly the most complete to date, the fact that it involves parameters
that can be derived only from finite element analysis detracts slightly from its use in practice, in addition to which some
physical insight into the indentation process is lost through the dependence of the model on numerical evaluation. Here,
we develop a simple new model and analysis that can be used to make the conversion from β to α based on a simple
closed-form algebraic expression. Although only approximate in nature, comparison to Bower’s analysis shows that the new
model may be quite useful for materials creeping with stress exponents in the range 1 ≤ n ≤ 7, which includes many of the
more important engineering materials. The predictive capabilities of the new model and Bower’s analysis are discussed by
comparing them to several experimental data sets for which both the indentation and uniaxial creep behavior have been
independently characterized. The new model may be particularly useful when a simple closed-form relation is needed to
estimate uniaxial parameters from indentation data or in the development of other deformation models where a simple
closed-form approximation of the indentation creep rate is required.

2. The model

The model is based on many of the ideas developed by Johnson in 1970 in his classic analysis of how the mean inden-
tation pressure, pm , depends on the yield stress, σ y , and elastic modulus, E, during the indentation of an elastic-perfectly
plastic solid (Johnson, 1970). Johnson’s analysis borrows heavily on Hill’s description of the expansion of spherical cavity of
radius, R, in an elastic-plastic medium uniformly pressurized at a pressure, p̄ (Hill, 1950). As shown schematically in Fig. 2a,
Johnson suggested that the indentation problem can be related to the cavity expansion problem by cutting the cavity in
two and assuming that there is a hemispherical hydrostatic core of material adjacent to the indenter that radially transmits
the pressure of the indenter to the surrounding material, thus producing radially directed elastic-plastic flow. Analytically,
the two problems are coupled by assuming that the radius of the hydrostatic core is equal to the radius of contact of the
indenter in the surface, i.e., R = ac , and that during an increment of indenter displacement, dh, the increase in volume of the
core, dVcore , is equal to the volume displaced by the indenter, dVindent , or
dVcore = dVindent . (6)
Letting the hardness, H, be equal to the pressure in the hydrostatic core, p̄, the model leads to
  
H 2 1 E
= 1 + ln cot θ (7)
σy 3 3 σy
 R.S. Ginder et al. / Journal of the Mechanics and Physics of Solids 112 (2018) 552–562 555

for indentation with a conical indenter of half-included angle, θ . Johnson showed that Eq. (7) provides a good estimate of
the hardness based on a variety of experimental hardness measurements provided that E cotθ /σy is smaller than about 100.
At larger values, the model breaks down because the larger plastic zones allow for flow of material to the free surface to
form pile-up which disrupts the radial flow pattern required for the expanding cavity model to work. Because of this, the
free surface plays an important role in limiting the model.

2.1. Application to conical indentation in an isotropic elastic material

To illustrate the utility of Johnson’s approach in modeling other indentation problems, we first apply it to the frictionless
elastic contact between a rigid conical indenter of half-included angle, θ , driven into an isotropic elastic medium character-
ized by Young’s modulus, E, and Poisson’s ratio, ν . The well-known solution to this problem is

π E 2 E
P =   cot θ a2c =   tanθ h2 , (8)
2 1 − ν2 π 1 − ν2

where P is the indentation load and h is the displacement of the indenter relative to the initial flat surface (see, for instance,
Sneddon, 1965). Noting that the mean indentation pressure is pm = P/(π a2c ), Eq. (8) can be re-written as

1 E
pm = cot θ . (9)
2 (1 − ν ) (1 + ν )

Johnson’s methodology can be used to approximate this solution starting from Lamé’s solution for the stresses in an
infinite, isotropic elastic medium containing a pressurized spherical cavity of radius, R (see, for instance, Hill, 1950). Letting
the cavity pressure be p̄, the stresses in the surrounding material in spherical coordinates (r, θ , φ ) are
 3
R
σrr = − p̄ 3 (10a)
r

and
 
p̄ R3
σθ θ = σφφ = . (10b)
2 r3

From Hooke’s law, it follows that the tangential strain, θ θ , is

p̄ (1 + ν ) R3
θ θ = . (11)
2 E r3
Because of the pure radial symmetry of the problem, the only non-zero displacement is ur , which is related to the tangential
strain through θ θ = ur /r. From this and Eq. (11), it follows that the radial displacement at the cavity surface r = R is

p̄ (1 + ν )
ur ( r = R ) = R. (12)
2 E
To relate this back to the indentation problem, we assume a contact geometry like that in Fig. 2a, and in a manner
similar to Johnson, use Eq. (6) to relate the differential increase in volume of the indentation during an increment of dis-
placement dh to the differential increase in the volume of the hydrostatic core. Since the latter is directly related to the
radial displacement at the surface of the core, ur (r = R), the differential volume equivalency in combination with at R = ac
gives

π a2c dac /tanθ = 2π R2 dur (R = ac ) (13)

or
d ur ( R = ac ) cot θ
= . (14)
dR 2
Combining Eqs. (12) and (14), and assuming that the pressure in the hydrostatic core is equal to the mean indentation
pressure, pm , gives

E
pm = cot θ . (15)
(1 + υ )
Comparison with Eq. (9) reveals that the expanding cavity modeling approach and the exact solution differ by only a factor
of 2(1−ν ). In most materials, this is not very large, and when the material is incompressible (ν = 0.5), the two solutions are
identical.
556 R.S. Ginder et al. / Journal of the Mechanics and Physics of Solids 112 (2018) 552–562

2.2. Application to indentation in a power-law creeping solid

The same sort of analysis can be applied to model the indentation of a power law creeping solid. The geometry we
assume is outlined in Fig. 2b, where the plastic zone in Johnson’s model is replaced with a zone of power law creeping solid
embedded in surrounding elastic material. The boundary between the creeping and elastic material is not well-defined, but
could be approximated by determining the radius at which the stresses drop so much that elastic deformation dominates
that produced by creep. Since the position of this boundary depends in complicated ways both on material parameters and
time, we ignore it in the current development, as did Bower et al., and assume that all the material outside the hydrostatic
core deforms entirely by power law creep. As discussed elsewhere (Su et al., 2013), ignoring the elastic deformation may
have important influences on transient behavior, but should not significantly affect the indentation creep rates once the
steady state is established.
The mathematical analysis begins with the relation for the rate of expansion, R˙ , of a pressurized spherical cavity of
radius, R, embedded in an infinite solid that creeps according to the basic uniaxial constitutive law of Eq. (1). Finnie and
Heller have thoroughly analyzed this problem (Finnie and Heller, 1959) and showed that
 n
R˙ α 3 p̄
= , (16)
R 2 2n

where p̄ is the uniform pressure in the cavity. To apply this to the indentation problem, we rewrite Eq. (13) for the differ-
ential volume equivalency in its time dependent form involving the rate of growth of the cavity R˙ = dR/dt and the velocity
of the indenter h˙ = dh/dt, yielding

π a2c h˙ = 2π R2 R˙ . (17)
Re-arranging, combining with Eq. (16), and assuming that the pressure in the cavity p̄ is equal to the mean indentation
pressure, pm , leads to the desired simple closed-form relation for the indentation strain rate

h˙
 3 p n
˙ i = =α
m
. (18)
ac 2n
This may be written in the form of Eq. (2) provided that
 3 n
β=α . (19)
2n
Comparison to Bower’s results in Eqs. (4) and (5) shows that the two sets of relations are very similar in form and identically
equal if F = 2n/3.
In order to directly compare the predictions of the two models, one must also account for the influences of pile-up and
sink-in since in the Bower model, the pile-up/sink-in parameter c = hc /h is also a function of the stress exponent, n, in a
manner that must be taken into account. In a later section, we will compare predictions of the models to real experimental
results obtained with two indenters: (1) a cylindrical flat punch, and (2) a Berkovich triangular pyramid that we will model
as a conical indenter with a half-included angle of 70°. As a result, we focus here on these two specific indenters but note
that the same principles could be applied to other indenters as well, e.g., spheres. To make the comparison, we will assess
how the uniaxial creep coefficient, α , can be deduced from measurements of the indentation creep coefficient, β , noting
that careful consideration must be given to what can actually be measured in the indentation experiment. For example,
in a nanoindentation creep experiment, one typically measures load, displacement, and time. From the displacement mea-
surement, h, and the geometry of the indenter, one can deduce the nominal contact radius, anom , and the nominal contact
area, Anom , i.e., those that would occur if there were no pile-up or sink-in (see Fig. 1). However, one cannot deduce the true
contact radius, ac , which would be greater than anom if there is pile-up and less than anom if there is sink-in.

2.3. Comparison of models – cylindrical flat punch indentation

For cylindrical flat punch indentation, the contact geometry is relatively simple because there is no pile-up at the con-
tact edge, so the radius of the contact remains fixed and equal to the radius of the punch, ac (see Fig. 1a). Consequently,
comparison of the two models, which we will refer to as the Bower model and expanding cavity model (ECM), is relatively
simple because it involves only the dependence of the reduced contact pressure, F, on n. For the flat punch indenter, the
indentation strain rate is conveniently defined as ˙ i = h˙ /ac , where ac is simply the fixed radius of the punch. Note that the
indentation strain rate defined in this way is consistent with the definitions in Eqs. (4) and (18). Under these circumstances,
the β to α conversion is achieved using

α = β F n (Bower ) (20a)
and
 2n n
α=β (ECM ), (20b)
3
 R.S. Ginder et al. / Journal of the Mechanics and Physics of Solids 112 (2018) 552–562 557

Fig. 3. Comparison of the Bower and expanding cavity models based on their predictions of the uniaxial creep coefficient, α . The models give identical
results when αBower /αECM = 1.

and the ratio of these two measures of α is given by

αBower  3F n
= . (21)
αECM 2n
By curve fitting, we have found that the values of F determined by Bower et al. in their finite element analyses (Bower et al.,
1993) are well fit by the fourth order polynomial:
1  1 2  1 3  1 4
F (n ) = 3.1454 + 0.39371 − 7.9638 + 7.9228 − 2.6501 . (22)
n n n n
With values of F so determined, the filled circle data points in Fig. 3 show how the ratio αBower /αECM given in Eq. (21) varies
with the stress exponent, n, for the flat punch indenter. Over the range n = 1 to 5, which encompasses many creeping
materials, the predictions of the two models are well within a factor 3 and generally much better than this. For larger n, the
expanding cavity model grossly overestimates α , presumably because of a breakdown in the presumed radial flow geometry.

2.4. Comparison of models – conical indentation

For conical indentation, pile-up and sink-in play an important role and must be carefully included in the analysis. To
address this, we note that in a typical nanoindentation creep experiment, only the nominal depth of penetration, h, can
be measured (see Fig. 1b), which means that data analyses based on the true contact depth, hc , the true contact radius,
ac , or true contact area, Ac , are not particularly useful. This also holds for the true mean contact pressure, pm , since it is
derived from the true contact area. As a result, it is useful to rewrite the relations describing conical indentation creep in
terms of the nominal indentation parameters which can be directly measured or deduced from the nanoindentation load-
displacement-time data.
This is conveniently achieved in Bower’s analysis by redefining the indentation strain rate in Eq. (4) as ˙ i = h˙ /h, relating
the true mean contact pressure to the nominal contact pressure pnom = P/Anom = P/(π a2nom ), and making use of the relations
anom = h tanθ and c = hc /h = ac /anom . For the conical indenter, this leads to
 
h˙ α tanθ pnom n
˙ i = = , (23)
h c2n−1 F

and the β to α conversion is given by

 
α = β F n c2n−1 cot θ (Bower for cone). (24)
By the same reasoning, the expanding cavity model applied to conical indentation yields

h˙
 3 n
˙ i = = α tanθ pnnom (25)
h 2n
and
 2 n n
α=β cot θ (ECM for cone ). (26)
3
558 R.S. Ginder et al. / Journal of the Mechanics and Physics of Solids 112 (2018) 552–562

Consequently, the ratio αBower /αECM can be written as

αBower  3F n 2n−1
= c , (26a)
αECM 2n
which differs from the expression for the flat punch indenter (Eq. (21)) by the factor c2n−1 .
In Fig. 3, the conical indenter results are plotted along with the flat punch results for comparison, where the values used
for F and c were taken from the curve fits of finite element results that account for the mild dependence of F and c on the
indenter angle θ (Su et al., 2013). For the 70° cone, these relations are:
1  1 2  1 3  1 4
F (n, θ = 70◦ ) = 2.588 + 2.745 − 12.428 + 11.915 − 3.996 . (27)
n n n n
and
1  1 2  1 3
c (n, θ = 70◦ ) = 1.310 − 0.948 + 0.540 − 0.252 . (28)
n n n
Fig. 3 shows that the values of α for the conical indenter predicted by the expanding cavity model are within a factor of 2.5
of those in the Bower analysis for n values up to nearly 7. Thus, to the degree that estimating the uniaxial creep coefficient
from indentation measurements to within a factor of 2.5 is valuable, there is merit to the simple relations provided by the
expanding cavity analysis and its simple closed form relationships. It is also notable that at n values of ∼1.1 and ∼5.5, the
expanding cavity and the Bower analyses provide identical results.

3. Comparison to experiment

Although there are not many experimental data sets that can be used to test the models, there are a few worth con-
sidering. Here, we examine: (1) succinonitrile tested at 37 °C with a cylindrical flat punch indenter (impression creep); (2)
amorphous selenium tested at 35 °C with a Berkovich triangular pyramid; (3) the solid acid CsHSO4 tested at 145 °C with
a Berkovich indenter; and (4) high purity indium tested a room temperature, also with a Berkovich indenter. For each of
these, predictions of the Bower and expanding cavity models are compared directly to the experimental results, with the
Berkovich data analyzed assuming it is well-described by a 70° conical indenter. The equivalence of these indenters is based
on a geometry that produces equal cross-sectional areas at a given distance back from the tip. For a Berkovich indenter, the
half-included angle of the equivalent conical indenter is 70.3°. Although some of the important experimental procedures are
included here, the reader is referred to the original publications for details.

3.1. Succinonitrile

Succinonitrile is a waxy hydrocarbon that melts at approximately 57 °C and creeps readily at temperatures slightly
greater than ambient. Chu and Li performed indentation creep tests on reagent grade polycrystals of this material at 37 °C
using a series of flat cylindrical punch indenters of different radii (impression creep) (Chu and Li, 1977). The indentation
pressure was kept constant by applying a fixed load to the flat punch, with the creep rates measured once the steady state
had set in based on observations of constant punch velocity. Since the grain size in the samples varied in the range 0.1 to
5 mm and the flat punch radii from 0.2 to 0.8 mm, there is a distinct possibility that single grains or a small number of
grains were sampled in many of the tests. The uniaxial creep behavior was characterized independently using conventional
uniaxial compression tests at the same temperature.
The basic results are shown in Fig. 4, where the indentation data are shown in red and the uniaxial compression data in
black. The indentation data are plotted as the indentation strain rate ˙ i = h˙ /ac , where ac is the radius of the punch, vs. the
nominal mean pressure, pnom , which for the cylindrical flat punch indenter is the same as the true mean pressure, pm . The
uniaxial compression data are plotted on the same axes in the normal way, i.e., as the uniaxial strain rate, ˙ , vs. the uniaxial
stress, σ . The dashed lines through the data points are least squares linear regressions of the log-log data. The slope of the
regression for the indentation data indicates a stress exponent n = 3.8, with the uniaxial data exhibiting a similar slope, as
expected.
To assess the predictive capabilities of the models, the β to α conversion was numerically evaluated for both. According
to Eq. (20a), the conversion for the Bower model requires values for n and F. Using n = 3.8 from the indentation data, it
follows from Eq. (22) that F = 2.83, and this results in a β to α conversion factor of 52.0. This means that the indentation
data needs to be shifted vertically by a factor of 52.0 to predict the uniaxial behavior. For convenience, this shift has been
shown in the plot as the equivalent horizontal shift, with the discrete blue data points showing how the indentation data
must be horizontally shifted to predict the uniaxial behavior according to the Bower model. The shift is identified in the
plot as the blue arrow labeled “Bower shift”. The equivalent conversion for the expanding cavity model (ECM) follows from
Eq. (20b), for which the β to α conversion requires only the stress exponent (n = 3.8). In this case, evaluating Eq. (20b) leads
to a slightly smaller β to α conversion factor of 34.2, for which the equivalent horizontal shift is shown in the plot as the
green arrow labeled “ECM shift”. Comparison of the shifted indentation data to the uniaxial data shows that both models
do a reasonably good job, with the Bower model being slightly better. The similarity of the two predictions follows from
 R.S. Ginder et al. / Journal of the Mechanics and Physics of Solids 112 (2018) 552–562 559

Fig. 4. Uniaxial and indentation creep behavior of succinonitrile at 37 °C based on data from Chu and Li (1977). The indentation creep experiments were
performed with a cylindrical flat punch indenter. The blue arrow and data points show how the Bower model predicts the uniaxial behavior from the
measured indentation data. The green arrow and data points show the predictions of the new expanding cavity model (ECM). (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 5. Uniaxial and indentation creep behavior of amorphous selenium at 35 °C based on data from Su et al. (2013). The indentation creep experiments
were performed at constant load with a Berkovich triangular pyramid indenter. The blue arrow and data points show how the Bower model predicts the
uniaxial behavior from the measured indentation data. The green arrow and data points show the predictions of the new expanding cavity model (ECM).
(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

the flat punch predictions in Fig. 3, which show that for a stress exponent a little greater than 4, the Bower model and the
expanding cavity model yield identical conversions.

3.2. Amorphous selenium

Like succinonitrile, amorphous selenium is a material that creeps readily at temperatures slightly above ambient, in this
case because its glass transition temperature is ∼31 °C. This allowed Su et al. to obtain an extensive set of nanoindentation
creep data on this material by modifying the enclosure of a nanoindentation testing system so that everything inside was
uniformly heated to 35.0 °C by forced convection (Su et al., 2013). The indentation creep results we consider here were
performed with a Berkovich indenter at a constant fixed load, producing the data shown in Fig. 5. During the course of a
constant load creep test, the nominal mean pressure decreases as the contact area increases, thus producing a continuum of
560 R.S. Ginder et al. / Journal of the Mechanics and Physics of Solids 112 (2018) 552–562

Fig. 6. Uniaxial and indentation creep behavior of the solid acid CsHSO4 at 145 °C based on data from Ginder and Pharr (2017, 2018). The indentation
creep experiments were performed with a Berkovich triangular pyramid indenter. The blue arrow and data points show how the Bower model predicts the
uniaxial behavior from the measured indentation data. The green arrow and data points show the predictions of the new expanding cavity model (ECM).
(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

data which are shown on the plot as a nearly straight line rather than individual data points. The line exhibits some width
and is multi-colored since it includes data obtained in four separate tests at four different loads. The stress exponent of
n = 1.12 indicates that the creep behavior of this material at 35 °C is close to linear viscous.
The uniaxial creep behavior, also shown in Fig. 5, was determined by constant displacement rate compression testing at
rates in the range 0.66 to 200 μm/s on cylindrical samples nominally 1 cm in diameter and 2 cm in length. In these tests, a
steady state flow stress was approached at a strain near 10%, and this was used to establish the relation between the steady
state flow stress and strain rate.
To evaluate the Bower model, Eq. (24) was employed to determine the β to α conversion assuming that the Berkovich
indenter behaves like a 70° cone. The values used for the computation were: n = 1.12; F = 1.07 (from Eq. (27)); c = 0.715
(from Eq. (28)); and θ = 70°, producing a conversion factor of 0.260. Similarly, Eq. (26) was used to evaluate the conversion
for the expanding cavity model, giving a value of 0.262. Clearly, the conversions are nearly identical, which is a consequence
of n = 1.12 being very close to what is needed for the two models to be identical in their predictions (see Fig. 3). However,
more importantly, both models produce shifts of the indentation data that predict the uniaxial behavior almost exactly, as
shown by the Bower and ECM shifts in Fig. 5. Note that unlike the shifts for succinonitrile in Fig. 4, the shifts in Fig. 5 are to
the right rather than to the left. This behavior is a consequence of the much lower stress exponent, and in the case of the
Bower model, also by the effects of sink-in. This emphasizes the fact that conversion of indentation creep data to uniaxial
data involves much more than just adjusting the nominal contact pressure by a material independent constant.

3.3. Solid acid CsHSO4

A relatively new data set that can be used to assess the predictive capabilities of the models was recently published
for the solid acid CsHSO4 (Ginder and Pharr, 2017; Ginder and Pharr, 2018). This material creeps at relatively rapid rates
at temperatures greater than 141 °C, where it undergoes a superprotonic phase transformation that renders it useful as a
solid-state electrolyte for intermediate temperature fuel cells. Its steady state flow behavior was characterized at several
temperatures in uniaxial compression using cylindrical specimens 1.9 cm in diameter and 5.0 cm in length with a grain size
of about 10 μm (Ginder and Pharr, 2017). Like the compression tests performed on amorphous selenium, these tests were
conducted at constant displacement rate, with the steady state flow stress deduced at strains of about 10% where it became
relatively constant.
Nanoindentation creep tests were performed with a Berkovich indenter using a new high temperature nanoindentation
testing system (Ginder and Pharr, 2018). The tests were conducted at constant P˙ /P (constant loading rate divided by load),
which produces a constant indentation strain rate when the hardness is independent of depth. For this reason, such tests are
often referred to as “constant strain rate tests” (CSR), with the onset of steady state creep evidenced by the onset of a con-
stant nominal mean pressure (or hardness) as the penetration depth increases (Lucas and Oliver, 1999). The nanoindentation
contact impressions were large enough that numerous grains were sampled (Ginder and Pharr, 2018).
Data obtained at 145 °C are presented in Fig. 6, where it is seen that the nanoindentation data correspond to a stress
exponent n = 3.5. Following the same procedures used for amorphous selenium in the previous section but using n = 3.5,
F = 2.61 (Eq. (27)), c = 1.08 (Eq. (28)), and θ = 70°, the magnitudes of the Bower and ECM shifts have been calculated and
are shown in the figure. In this case, the Bower shift (blue shifted data points) not only works well but is clearly better,
with the ECM shift (green shifted data points) falling short of what is needed to match the uniaxial data. The reason for the
 R.S. Ginder et al. / Journal of the Mechanics and Physics of Solids 112 (2018) 552–562 561

Fig. 7. Uniaxial and indentation creep behavior of high purity indium at room temperature. The indentation creep experiments were performed by
Lucas and Oliver with a Berkovich triangular pyramid indenter (Lucas and Oliver, 1999). The uniaxial data were obtained in separate experiments by
Weertman (1960). The blue arrow and data points show how the Bower model predicts the uniaxial behavior from the measured indentation data. The
green arrow and data points show the predictions of the new expanding cavity model (ECM). (For interpretation of the references to color in this figure
legend, the reader is referred to the web version of this article.)

difference is that the stress exponent (n = 3.5) is near the peak of the curve in Fig. 3, where there is maximum error in the
predicted ECM shift.

3.4. High purity indium

As a last experimental test case, we consider the creep of high purity indium. Lucas and Oliver examined this material
at several temperatures up to 75 °C in nanoindentation tests with a Berkovich indenter (Lucas and Oliver, 1999). Here, we
focus only on their results from room temperature (∼28 °C). The sample material, which was 99.999% pure, had a grain size
of 2-3 mm, implying that most of the tests were conducted within single grains. Like the solid acid material evaluated in the
previous section, the nanoindentation creep tests we use for comparison here were conducted using the constant strain rate
technique (CSR) at values of P˙ /P ranging from 0.005/s to 0.2/s. The steady state was clearly achieved based on observations
of a constancy of hardness with depth well before the end of the test. The nominal contact pressure (or hardness) was
computed assuming a perfect area function for the Berkovich indenter and no consideration of pile-up or sink-in. Lucas and
Oliver did not conduct separate experiments to characterize the uniaxial behavior, but their nanoindentation data are often
compared to uniaxial data collected much earlier by Weertman (1960) using standard creep testing techniques.
The two sets of data are shown in Fig. 7, where it is seen that the stress exponent for the indentation data is relatively
high, n = 7.3, and slightly higher for the uniaxial data (n = 7.7). Using n = 7.3, F = 2.76 (Eq. (27)), c = 1.19 (Eq. (28)), and θ = 70°,
the magnitudes of the Bower and ECM shifts are shown in the figure. Curiously, this time the expanding cavity model
provides the better approximation, but neither matches well with the uniaxial data. The reason for this is not apparent, but
it should be noted that the two sets of data were taken by two different groups at two different times on two potentially
different materials in terms of purity, grain size, dislocation structure, etc. Thus, we do not consider these data sets as a
reliable test of either model.

4. Concluding remarks

The results presented in this work suggest that with careful experimental procedure, uniaxial creep parameters can in-
deed be accurately measured from data taken in load and depth sensing indentation experiments. The stress exponent, n,
is determined from the slope of log-log plots of the indentation strain rate vs. the nominal indentation pressure, and the
uniaxial creep coefficient α follows directly from n using the procedures developed here. These methods apply to any type
of test, e.g., constant load, constant loading rate, constant strain rate, stress relaxation, or strain rate change tests, provided
that (1) the experimental procedures allow enough time for transient behavior to dissipate and (2) the steady state strain
rates are history independent (Su et al., 2013).
As discussed elsewhere (Su et al., 2013), creep transients in indentation testing, which depend on how the test is con-
ducted, can originate from several sources. The obvious one is microstructural changes like those observed in uniaxial creep
tests, i.e., normal primary creep. Another results from time dependent changes in the contact geometry as the steady-state
pile-up or sink-in profile develops or when the varying contact stresses during the course of a test fall outside the realm of
power-law creep.
562 R.S. Ginder et al. / Journal of the Mechanics and Physics of Solids 112 (2018) 552–562

We need to emphasize that all of the results presented in this work, as well of those of Bower et al., are based on
conventional continuum principles that do not account for size effects. When indentation creep measurements are made
at very small scales, as could be encountered in nanoindentation testing, significant size effects on creep rates could be
observed. For example, if the size of the indentation contact is on the order of or smaller than the grain size, then grain
boundary sliding and/or some diffusional creep processes that operate in the bulk may not be observed.
The focus of this work has been on the β to α conversion for flat punch and pyramidal indenters assuming that pyrami-
dal indenters can be adequately modeled as cones. Comparison to experiment in the previous section shows that Bower’s
analysis written in the form of Eqs. (20a) and (24) provides a remarkably accurate method for doing this. However, when
a quick approximation is desirable, or when a simple closed-form relation is needed for modeling purposes, the new ECM
model through Eqs. (20b) and (26) may prove useful, provided the stress exponents are no greater than 7.
We also wish to comment on the somewhat common practice (Lucas and Oliver, 1999; Poisl et al., 1995) of converting
indentation creep data to uniaxial data simply by modifying the nominal indentation pressure by the factor of 3 suggested
by Tabor to convert hardness to flow strength (Tabor, 1951), that is σ = H/3. The Tabor factor of 3 has also been used in
recent finite element studies in which the β to α shift is argued to involve two parts - a horizontal shift on the stress
axis by an amount equal to the Tabor factor and a vertical shift on the strain rate axis to account for all of the other
influences (Martinez and Shen, 2016; Takagi and Fujiwara, 2014). Although the reduced contact pressure in Bower’s analysis
tends towards 3 at higher values of n and thus serves to modify the indentation pressure in a manner similar to the Tabor
factor, the situation is actually much more complex than this for two important reasons. First, the reduced contact pressure
is not constant at a value of ∼3 but varies with n in a significant way, especially for smaller n’s (n ≤ 3 for the conical
indenter). Second, for many indenter geometries, the pile-up/sink-in geometry varies significantly with n, and this also plays
a significant role. The importance of these influences is most readily seen in the amorphous selenium data in Fig. 5, where
because of the small stress exponent (n = 1.12), the β to α shift moves the indentation data to the right rather than to the
left. These factors are fully accounted for in Bower’s analysis and in an approximate way by the new ECM model.
Lastly, more experimental data are needed to assess the suggested measurement procedures outlined in this work. With
high temperature nanoindentation testing systems becoming more and more common, such data could soon be forthcoming.

Acknowledgments

This work was supported by the by the National Science Foundation under grant number DMR-1743343 and by a U.S.
Department of Defense (DoD) through the National Defense Science & Engineering Graduate Fellowship (NDSEG) Program.

References

Bower, A.F., Fleck, N.A., Needleman, A., Ogbonna, N., 1993. Indentation of a power law creeping solid. Proc. Roy. Soc. – Math. and Phys. Sci. 441, 97.
Chu, S.N.G., Li, J.C.M., 1977. Impression creep – a new creep test. J. Mater. Sci. 12, 2200.
Dean, J., Campbell, J., Aldrich-Smith, G., Clyne, T.W., 2014. A critical assessment of the stable indenter velocity method for obtaining the creep stress
exponent from indentation data. Acta Mater. 80, 56.
Duan, Z., Hodge, A., 2009. High temperature nanoindentation: new developments and on-going challenges. J. Metals 61, 32.
Finnie, I., Heller, W.R., 1959. Creep of Engineering Materials. McGraw-Hill, New York.
Ginder, R.S., Pharr, G.M., 2017. Creep behavior of the solid acid fuel cell material CsHSO4 . Scripta Mater 139, 119.
Ginder, R.S., Pharr, G.M., 2018. Characterization of power-law creep in the solid-acid CsHSO4 via nanoindentation, in preparation.
Goodall, R., Clyne, T.W., 2006. A critical appraisal of the extraction of creep parameters from nanoindentation data obtained at room temperature. Acta
Mater. 54, 5489.
Hill, R., 1950. The Mathematical Theory of Plasticity. Clarendon Press, Oxford.
Hill, R., 1992. Similarity analysis of creep indentation tests. Proc. R. Soc. Lond. A 436, 617.
Johnson, K.L., 1970. Correlation of indentation experiments. J. Mech. Phys. Solids 18, 115.
Lucas, B.N., Oliver, W.C., 1999. Indentation power-law creep of high-purity indium. Metall. and Mater. Trans A – Phys. Metall. Mater. Sci. 30, 601.
Maier, V., Merle, B., Göken, M., Durst, K., 2013. An improved long-term nanoindentation creep testing approach for studying the local deformation processes
in nanocrystalline metals at room and elevated temperatures. J. Mater. Res. 28, 1177.
Martinez, N.J., Shen, Y.-L., 2016. Analysis of indentation-derived power-law creep response. J. Mater. Eng. Perform. 25, 1109.
Mayo, M.J., Nix, W.D., 1988. A micro-indentation study of superplasticity in Pb, Sn, and Sn-38wt%Pb. Acta Metall 36, 2183.
Phani, P.S., Oliver, W.C., 2016. A direct comparison of high temperature nanoindentation creep and uniaxial creep measurements for commercial purity
aluminum. Acta Mater. 111, 31.
Poisl, W.H., Oliver, W.C., Fabes, B.D., 1995. The relationship between indentation and uniaxial creep in amorphous selenium. J. Mater. Res. 10, 2024.
Raman, V., Berriche, R., 1992. An investigation of the creep processes in tin and aluminum using a depth-sensing indentation technique. J. Mater. Res. 7,
627.
Sneddon, I.N., 1965. The relation between load and penetration in the axisymmetric Boussinesq problem for a punch of arbitrary profile. Int. J Engng. Sci.
3, 47.
Su, C.J., Herbert, E.G., Sohn, S., LaManna, J.A., Oliver, W.C., Pharr, G.M., 2013. Measurement of power-law creep parameters by instrumented indentation
methods. J. Mech. Phys. Solids 61, 517.
Tabor, D., 1951. Hardness of Metals. Clarendon Press, Oxford.
Takagi, H., Fujiwara, M., 2014. Set of conversion coefficients for extracting uniaxial creep data from pseudo-steady indentation creep test results. Mater. Sci.
Eng. A 602, 98.
Weertman, J., 1960. Creep of indium, lead, and some of their alloys with various metals. Trans. AIME 218, 207.
Wheeler, J.M., Armstrong, D.E.J., Heinz, W., Schwaiger, R., 2015. High temperature nanoindentation: the state of the art and future challenges. Curr. Opin.
Solid State Mater. Sci. 19, 354.