Experimental Mechanics

https://doi.org/10.1007/s11340-022-00937-4

RESEARCH PAPER

Creep Characterization of Amorphous ­SiO2 in the Transmission

Electron Microscope Using Digital Image Correlation and Finite
Element Analysis
Y. Zhang1 · S. Dillon2 · J. Lambros1

Received: 15 July 2022 / Accepted: 29 December 2022

© Society for Experimental Mechanics 2023

Abstract
Background Amorphous silica (a-SiO2) exhibits creep behavior under electron beam irradiation in the transmission electron
microscope (TEM) even at room temperature. This effect is invariably present during in situ TEM microscale mechanical
testing of a-SiO2, thus necessitating creep characterization of this material in the TEM environment.
Objective In this paper, we extract creep properties of a-SiO2 during electron beam irradiation induced creep (IIC) by com-
bining experimental measurements with a 2D finite element model (FEM) based on an assumed creep behavior modeled
by power creep law.
Methods Micron sized a-SiO2 beam samples deposited with gold nanoparticles are machined by focused ion beam milling
and loaded in the TEM via indentation. The applied load-displacement profile at the loading point is recorded by the indenter,
while full-field deformation is measured from the TEM images by correlating deformed and undeformed nanoparticle speckle
patterns using digital image correlation (DIC).
Results The elastic modulus and creep properties are obtained by solving an inverse problem in the FEM analysis based
on the experimentally measured load-displacement data, and are validated by full-field displacement comparisons between
FEM results and DIC measurements.
Conclusion FEM and DIC results show good agreement, indicating applicability of the power creep model and the accuracy
of extracted creep properties. A linear dependance between creep strain rate and applied stress is derived. Possible error
sources from both the experiment and simulation are discussed.

Keywords Irradiation induced creep · Digital image correlation · Inverse analysis · Finite element model · Transmission
electron microscopy

Introduction obscured by the complexity of the bulk response. In some

cases, small-scale testing provides a unique route to obtain-
Small-scale mechanical testing has been developed and ing mechanical property data of interest. For example, irra-
applied to mirror the broad range of testing modes and diation induced creep (IIC) affects the lifetime and safety
configurations of bulk material testing, including tension, of nuclear reactors, and knowledge of IIC response up to
compression, bending, etc. [1–3]. Such small-scale experi- high levels of damage is extremely valuable. The associated
ments are often valuable because they can provide insights neutron irradiation experiments, however, cannot be suffi-
into local microstructural processes whose physics might be ciently accelerated to obtain such information on laboratory
timescales. Charged particle irradiations are, instead, used
to simulate accelerated IIC. MeV energy heavy ion irradia-
* J. Lambros tions simulate the primary recoil damage spectra of neutrons
[email protected] sufficiently well, but only penetrate to depths of ≈ 1 μm [4].
1
Department of Aerospace Engineering, University of Illinois This has motivated the use of small-scale testing for acceler-
at Urbana-Champaign, Urbana, IL 61801, USA ated studies of IIC [2, 5]. In situ small-scale mechanical tests
2
Department of Materials Science and Engineering, are particularly valuable because the deformation modes
University of California Irvine, Irvine, CA 92585, USA involved may be observed directly.

Vol.:(0123456789)
 Experimental Mechanics

Creep deformation under the effects of irradiation was understand electron IIC in amorphous materials, since test-
initially discovered in fissionable materials under neutronic ing of glassy materials via in situ electron microscopy has
irradiation in the 1940s [6]. Since then, considerable work grown in popularity and the influence of the electron beam
has been done to model so-called irradiation induced creep is not well understood [19]. A goal of this work is to better
(IIC) [7–12]. Irradiation induced creep occurs in both crystal- understand how electron IIC affects in situ mechanical test-
line and amorphous materials. The latter exhibit large creep ing, whether the deformation response is uniform, and how
compliance whose mechanism is not fully understood and the small-scale testing configuration can affect analysis and
for which two competing mechanisms have been proposed. interpretation of the results.
The first mechanism is the thermal spike model proposed by Digital image correlation (DIC) is a non-contact optical
Trinkaus et al. [11] where creep happens due to the stress method that has seen widespread use in mechanics since
relaxation around thermal spikes resulting from locally its introduction [21, 22]. It is used for measuring full-field
increased temperature. This mechanism is often applied to displacement and strain by correlating deformed and unde-
both crystalline and amorphous materials where significant formed images of a speckle pattern deposited on the sample.
thermal spikes can occur driven by high temperature [13, 14]. For most experimental studies that measure continuous dis-
The second mechanism is the point defect model proposed by placement/strain fields (e.g., uniaxial tension, beam bending),
Mayr et al. [12] where, when the recoil energy exceeds a cer- the classical subset-based local DIC algorithm [21] shows
tain value, Frenkel defect pairs are generated to release local good performance and high effectiveness and will be used
stresses and thus produce creep deformation. This mecha- here. Since the inception of DIC, the method has been widely
nism is well suited for IIC in amorphous materials since they applied in studying the bulk mechanical behavior of materials
can undergo IIC even at low temperature where the thermal at the macroscale [23, 24]. With improvements in microscopy
spikes would generally be limited. The point defect model technology, it has also been extended to mesoscale studies
has also been numerically employed in the finite element performed using optical microscopy [25, 26], and to micro-
framework to predict anelasticity behavior of nanolattices scale studies performed by atomic force microscopy (AFM)
[15]. In the mathematical formulation of the creep model, the [27, 28] and scanning electron microscopy (SEM) [29–31].
creep strain rate is usually considered to be proportional to a Depending on the experimental environment present at dif-
coefficient, called irradiation creep compliance, multiplied ferent scales, different patterning techniques have been devel-
by a power law of the stress, a power law of neutron flux, oped. For example, spray painting a speckle pattern [32] is
and an exponential thermal activation term [16]. However, often employed at the macroscale, while airbrushing [33]
IIC can be observed at temperatures that are far below the can produce finer patterns for optical microscopy studies.
thermal creep regime solely under mechanical loading, even For DIC applications in electron microscopes, several pat-
at room temperature. In this case, the activation energy is terning techniques have been developed including scratching
sufficiently low, indicating that IIC is a nearly athermal phe- (which originated from AFM testing) [34], TEM grid pattern-
nomenon (i.e., the thermal activation term is close to 1). The ing [35], electron beam lithography [36] and nanoparticle
stress exponent and neutron flux exponent are often consid- deposition [37]. These techniques have been widely applied
ered to be close to 1 for most of materials [16, 17]. for use in DIC in SEM studies. Recently, we have developed
Electron irradiation does not produce thermal spikes, a nanodot patterning approach by thermal de-wetting of a
yet IIC was detected and measured under 200 KeV elec- deposited gold thin film, which has proven effective and sta-
tron beam irradiation in the TEM [18, 19]. Jawaharram, ble for DIC measurements in the TEM [38].
recently, reported a positive correlation between irradia- Small-scale mechanical test can be prone to large errors
tion creep compliance and electron beam current density due to non-ideal boundary conditions, loading geometry,
from in situ TEM bending tests of a-SiO2, FeBSi and CuTi and deformation response. Commonly employed micropil-
[20]. That work also suggests a point defect mechanism lar and nanopillar compression experiments, for example,
may be active. The definition of a point defect is unclear in often do not recover appropriate elastic moduli due to non-
amorphous materials, and might instead be generalized to ideal sample and loading geometries [2]. Steady-state creep
a concept wherein irradiation affects local free volume that experiments are typically performed to large values of strain,
might relax via shear transformation zones. Unfortunately, which will exacerbate any non-ideal loading conditions and
a good theory does not exist to describe the threshold dam- associated errors. The major goal of this work is to assess
age energy necessary to induce IIC in amorphous materi- the soundness of small-scale creep experiments in determin-
als. While the existing data strongly indicates IIC is active ing creep properties. Amorphous S ­ iO2 is used as a model for
in these systems, idealized experiments have not been per- this study because it exhibits electron IIC, which allows its
formed to accurately measure creep compliance and stress creep compliance to be tuned via the current density of the
dependence. The mechanism for ICC is relevant to amor- electron beam during in situ imaging in the TEM. The amor-
phous materials in nuclear systems. It is also important to phous structure of ­SiO2 is well suited to characterization by
Experimental Mechanics

DIC since the sample intensity only depends on mass thick-

ness, and a speckle pattern of higher density material may be
affixed to the surface. By performing in situ loading experi-
ments of ­SiO2 microbeam beam samples machined by focused
ion beam (FIB) milling, experimental load-displacement
data is recorded. Creep properties are extracted by inverse
approach in a 2D finite element model (FEM). TEM images
with undeformed and deformed patterns are also analyzed by
DIC to measure the full-field deformation during the loading Fig. 2  Secondary electron images of the a-SiO2 beam sample with two
tests, which will serve as a validation of the FEM results. Here built-in ends acquired using the ion beam. (a) plan view of the beam show-
we will discuss the procedure of deriving creep properties, the ing length L and width w, and (b) side view showing beam thickness t
potential errors introduced by the experimental set up and the
limitation of the FEM model, and how these factors affect the
calculated/measured mechanical response. DIC under the TEM settings given in “In Situ Loading Exper-
iment in the TEM”.
Experimental Methods A microbeam with built-in ends on both sides was milled
using a Scios 2 FIB with 30 KeV accelerating voltage.
Sample Preparation and Loading Figure 2(a) shows one such microbeam sample in plane view
(i.e., top view where the beam length and width are visible).
The material used in this work is a 1 μm thick amorphous The beam dimensions were length L ≈ 20 μm, width w ≈ 2 μm
­SiO2 substrate which is thermally grown on a 500 μm thick and thickness t ≈ 400 nm. Because of the presence of pre-
Si wafer (UniversityWafer, Inc.). A small bulk piece of the existing residual stresses, we observed buckling to occur dur-
wafer with in-plane dimensions approximately 3 mm by ing the FIB process if directly reducing the beam thickness
2 mm was cleaved using a diamond scribe, and was then pol- to 400 nm [38]. To prevent this buckling, we released the
ished on alumina lapping films on the Si side into a wedge residual stress by irradiating the beam sample using a TEM
shape with the thinnest edge being a few microns thick. To electron beam part way through the thickness reduction pro-
generate patterns for use in DIC, a uniform 5 nm thick gold cess. The thickness of the sample was first thinned down to
layer was deposited on the surface of the ­SiO2 side by an ≈ 1.3 μm through coarse FIB milling (7 nA) so that buckling
e-beam evaporator. Randomly distributed gold nanoparticles would not initiate but the sample would be thin enough for
were then generated by solid-state de-wetting [39] during TEM electron beam transmission. Residual stress in the beam
thermal annealing at 630 ℃ for an hour. Based on the study sample was then released by 200 KeV electron irradiation
in [39], the nanoparticle size is mainly affected by deposited [38] for 20 min at a current density of ≈ 14.5 A ­m− 2. Finally,
film thickness, and the pattern used in this work was found the beam thickness was further thinned down to ≈ 400 nm
to be suitable for DIC measurements in the TEM after a by removing redundant material using fine FIB milling with
series of parametric studies on gold layer thickness. Figure 1 50 pA beam current with a final side view of a milled beam
shows SEM images of finalized patterns after depositing shown in Fig. 2(b). More details of sample and speckle pat-
3 nm, 5 and 7 nm thick films, where 5 nm was chosen for tern preparation techniques can be found in [38].

Fig. 1  SEM images of gold nanoparticles generated by thermal annealing of a deposited gold thin film at 630 ℃ for different deposited film
thicknesses (f). The corresponding mean particle diameter (d) and standard deviation (s) of diameters are also indicated
 Experimental Mechanics

In Situ Loading Experiment in the TEM

In situ mechanical testing was conducted in a JEOL 2100

Cryo TEM. The beam sample was loaded using a Bruker PI
95 indenter system with a 1 μm diameter flat diamond punch.
The point of contact was at the mid-span of the beam, which
was found by measuring half of the beam length at 1000x
magnification in the TEM. The indenter tip was then aligned
to this position using mechanical/piezoelectric manipulation.
Then at a magnification of 5000x, a 4 μm × 2 μm region
near the loading point was selected as the field of view. The
Fig. 3  In situ TEM images of the undeformed and deformed a-SiO2
indenter tip was aligned in the out-of-plane (thickness) direc- beam with deposited gold nanoparticles near the loading area. The
tion via piezo manipulation by determining the two height time stamp shown is with respect to time of 0 s denoting the onset of
positions where a load reading was recorded, indicating two contact between indenter and beam
edges at the front and back of the beam sample. The location
corresponding to the midplane between these two positions
was used for the application of the load in the experiment. sequence of images, namely six images every 30 s, was taken
Load-displacement-time data of the diamond punch were without moving and/or loading the sample. To evaluate possi-
recorded from the indenter during the entire experiment with ble intensity changes in the TEM during this baseline imaging
an acquisition rate of 200 data points per second. sequence we calculate the mean intensity within every subset
In situ TEM images of the beam sample were manually in the beam sample. The (possible) change between subset
acquired at selected time instances during the loading (see intensities in the reference image (chosen as the first image
“Loading History”) using a Gatan UltraScan 1000 CCD taken) and subsequent images was quantified by the Pearson
camera (2048 × 2048 ­pixel2) with 0.4 s exposure time. With correlation coefficient (PCC) defined by equation (1).
a ramp-up loading rate of 10 µN/s, typical of what is used ∑n
(x − x̄ )(yi − ȳ )
i=1 i
this study, an indentation displacement increment of approx- PCCxy = � � (1)
∑n ∑n
imately 9 nm (indicating the largest deformation based on (x
i=1 i
− x
̄ ) 2
̄ 2
(y − y)
i=1 i
the beam geometry) takes place during the 0.4 s exposure
time. For the creep portion of the experiment, this motion where n is the sample size, xi and yi are the individual sample
during the exposure time can be much smaller, < 0.4 nm. points in two datasets, x̄ and ȳ are the sample means. PCC is a
Therefore, a 0.4 s exposure time is sufficient to minimize measure of linear correlation of two datasets ranging from − 1
motion blur while providing the necessary image contrast. to 1. Ideally, if all six images have identical grey values the
Figure 3 shows two selected TEM images of an undeformed correlation would yield a PCC equal to 1. However, the PCC
image (left) before contact between the indenter tip and the is usually smaller than 1 due to noise introduced in images
beam sample, and a deformed sample image (right) near (mainly by the image drift rate ≈ 0.079 nm/s reported in [38])
maximum load, where curved edges of the bent beam are and moreover the PCC will depend on the subset size.
observed under loading. Note that the gold nanoparticles did By changing the subset size from 3 to 199 pixels (the
not detach from the sample during the experiment, indicat- largest subset size in VIC-2D), the typical variation of the
ing stability of the deposited patterns. With the TEM imag- PCC in a subsequent image is shown in Fig. 4(a). PCC vari-
ing settings mentioned above, gold nanoparticles and sharp ation is stabilized to a value near 1 after a subset size of
beam edges are clearly identifiable in Fig. 3. about 90 pixels. The approximate “3 to 5 speckles” criterion
suggests a subset size of around 90 pixels by 90 pixels to
110 pixels by 110 pixels for our pattern and magnification.
DIC Displacement Measurements However, according to the PCC in Fig. 4(a) and 75 pix-
els by 75 pixels subset size may also be usable (and in fact
On the images such as those shown in Fig. 3, we applied DIC did work when implemented in VIC-2D, though with more
to measure full-field displacements of the deforming sample. uncorrelated points). Ultimately the subset size chosen in
A two-dimensional DIC analysis was conducted using the this work was 95 pixels by 95 pixels (200 nm by 200 nm) as
commercially available VIC-2D software (from Correlated a compromise of data resolution and successful rate of cor-
Solutions) [40]. Although it is suggested to choose a minimum relation. A correlation interval of 20 pixels (40 nm) was used
DIC subset size that contains 3–5 speckle features per side in order to increase the data acquisition spatial density (≈ 50
[40], we nonetheless conducted an evaluation of possible sub- data points/µm). Figure 4(b) shows two consecutive subsets
set size selection by performing “baseline” imaging in which a indicating the selected subset size and correlation interval.
Experimental Mechanics

Fig. 5  (a) Input ramp-step loading profile of the indenter. TEM

Fig. 4  (a) Variation of Pearson coefficient, defined by equation (1),
images were taken at timepoints marked by green crosses. Zero load
with the subset size generated by correlating two images from the
was aligned with the 0 s time stamp, indicating the onset of contact
baseline experiments, i.e., successive images with no loading. (b) Two
between the indenter tip and the a-SiO2 beam sample. (b) Displace-
successive DIC subsets shown in yellow and green with the selected
ment data of the indenter tip recorded by the indenter. Elastic and
DIC subset size, and the data acquisition density indicated by the dis-
creep properties were extracted from the data in first two steps by fit-
tance between the two subsets
ting the experimental measurements to FEM results at the three fit-
ting points shown. The following three steps were used as a validation
of the fitting results
The zero-normalized squared differences (ZNSSD) criterion
was applied during the correlation to avoid errors caused
by global intensity variation in the image which typically Loading History
occur in TEM imaging. As seen in Fig. 4(b), for subsequent
comparison of DIC with FEM results, the TEM images were To perform a creep experiment, a ramp-step loading is com-
rotated by aligning the top edge of the beam sample horizon- monly applied [42]. Here, a loading profile with up to five
tally in the undeformed image. ramp-step parts was designed since we are aiming at both
determining and validating creep properties from a single
experiment. The approach is to use the first few ramp-steps
Creep Model Calibration to extract properties and then validate the extent of applica-
bility of these properties on the remaining steps. Although
Creep Model these experiments are difficult to conduct and cannot thus
be done in large numbers, nonetheless, we did perform a
Based on the experimental data collected, we investigated series of experiments. However, it was found that, unlike
the feasibility of modelling the electron beam-induced IIC elastic response, transferability of creep response among
behavior of amorphous ­SiO2 at the microscale in the TEM experiments was an issue since IIC creep response depended
using the continuum model of Norton [41]. This relation, intimately on the amount of time that the electron beam was
which is one of the simplest and most widely used creep hitting the sample, as well as the intensity and size of the
models, defines the relation between steady-state creep beam. Although attempts were made to do so, it is diffi-
strain rate and stress as: cult to control these parameters precisely between different
experiments. Thus, it was decided to have the multiple ramp-
𝜖̇ cr = C1 𝜎 C2 e−C3 ∕T , (2) step profile, an example of which is shown in Fig. 5, so that
we can conduct property extraction and perform validation
where 𝜖̇ cr is the uniaxial equivalent strain rate, 𝜎 is the uni- on a single data set. Having said this, we will still show com-
axial equivalent deviatoric stress, T is the temperature and parisons between experiments, although each experiment
C1 , C2 , C3 are undetermined creep constants. We neglect the was essentially designed to be independent.
variation of temperature T during the experiment, in which Figure 5(a) shows an example of the indenter load input
case we can further simplify the model by setting C3 = 0: for a five ramp-step test–all creep experiments were con-
𝜖̇ cr = C1 𝜎 C2 . (3) ducted under load control. The total resulting displacement
at any time instant, shown as measured in Fig. 5(b), is com-
Thus, in such an isothermal case the creep response will posed of elastic deformation, determined by the elastic modu-
be governed by two the parameters C1 , C2, and by the instan- lus, and creep deformation, determined by creep constants in
taneous elastic modulus E. To extract the model parameters, equation (3) (to the extent that this model is valid). In order to
this creep model will be employed in a 2D FEM simulation, isolate the elastic effect and be able to extract elastic modulus
and C1 , C2 , and E will be obtained by an inverse parameter individually, the electron beam was blanked during the first
optimization discussed in "FEM parameter optimization". ramp loading stage (i.e., during 0–20s in Fig. 5(a)). During
 Experimental Mechanics

this 20 s time interval, the electron beam was irradiating the

sample only when an image was being taken (indicated by
the crosses in Fig. 5(a)) and then only during the 0.4 s expo-
sure time. Assuming that in the experiment the largest creep
displacement rate found takes place, the calculated creep dis-
placement during the exposure time in the first ramp loading
can introduce a maximum error of 0.04%. Thus, the response
during this first ramp up is expected to be almost solely elas-
tic. The beam sample was then continuously irradiated by Fig. 6  Mesh for 2D FEM model of a rectangular beam using pla-
the electron beam from 20 s onwards until the end of the nar elements (PLANE183), with uniformly distributed load over the
experiment, thus inducing IIC after this point. measured contact region and fixed ends as boundary conditions. The
Elastic modulus was obtained by fitting the displacement total load applied is that recorded experimentally Point A, shown in
the inset, is the midpoint of loading area and is used for calibration
response of the first ramp-up loading period. Creep proper- purposes between the experiment and FEM results
ties C1 , C2 were then derived according to the creep response
in the first two load plateaus and verified using the next three
steps. In theory, when using Norton’s law with two constants loading profile according to the load history shown in
(i.e., under isothermal conditions), the first two steps should Fig. 5(a), the displacement response can vary depending
be sufficient for numerically determining these two creep on the undetermined parameters of the elastic modulus E
constants. By taking the logarithm of equation (3) we obtain: and creep constants C1 , C2 in equation (3). To extract these
parameters, inverse problems were solved by the param-
ln 𝜖̇ cr = ln C1 + C2 ln 𝜎. (4) eter optimization method based on a response surface in
This equation indicates that ln 𝜖̇ cr and ln 𝜎 are linearly gov- ANSYS Workbench [44]. An input parameter set and an
erned by ln C1 and C2 and thus theoretically, C1 , C2 can be output parameter set were defined by the user: here the
uniquely determined by two different creep states (i.e., a two input set includes unknown constants E, C1 , C2 and the out-
ramp-step loading). However, considering the fluctuations put set contains measured load point displacement at cer-
of load-displacement data collected by the indenter (noise, tain times in the loading history. The location used for the
tolerances in the fitting procedure etc.), even for a two-ramp optimization was the load point displacement measured by
case the results of creep constants C1 and C2 are expected to the indenter during the experiment shown in Fig. 5(b) as a
be determined within a range of values (i.e., contain some function of time, and we co-registered this measurement
level of non-uniqueness). The more fitting points that are to the corresponding FEM result in the middle of the beam
included, the narrower this range will become. bottom edge, i.e., the location of maximum deflection.
As mentioned previously, a purely elastic response was
separated by design during the first ramp-up loading stage,
FEM Parameter Optimization allowing individual extraction of the elastic modulus prior
to obtaining creep properties. We derived elastic modulus
The microbeam sample shown in Fig. 2 is modeled using by a single fitting point at the end of the first ramp-up
ANSYS as a 2D rectangular elastic/creep material with loading (labeled “Fitting point 1” in Fig. 5(b)). Subse-
boundary conditions of fixed displacement on both ends, as quently, the two creep constants C1 and C2 were obtained
shown in Fig. 6. The geometry of the beam sample, length by fitting the measured and calculated load point deflec-
L = 20.2 μm and width w = 2.1 μm, was measured from the tion at the end of first two step loadings (labelled “Fitting
undeformed TEM image, and the thickness t = 400 nm was point 2&3” in Fig. 5(b)). A response surface (RS) based
measured from the ion beam image acquired in the FIB. A parameter optimization has advantages in solving a multi-
uniformly distributed load is applied at the midspan of the parameter problem, but is still effective in the degenerated
beam length, with the area of load application being equal single-parameter case. A complete optimization task was
to the width of the indenter tip (measured in the TEM). performed according to the following procedure:
The model is meshed by planar elements (PLANE183) with
mesh size = 0.1 μm, and with 0.05 μm refinement close to (a) Generate Design Points (DPs) by assigning differ-
the loading area as shown in the close-up in Fig. 6. A mesh ent values in the design space spanned by the input
sensitivity study was conducted to ensure convergence of parameters, and solve the FEM for each design point
results and Fig. 6 shows the 2D-FEM configuration with the to extract corresponding output parameters. To avoid
final meshing used which consisted of 4284 total elements. clustering points and to achieve a more uniform dis-
Using Poisson’s ratio v = 0.19, which is typically tribution in the parameter space, optimal space-filling
applied for thermally grown ­S iO 2 [43], and inputting a design [44] was applied for DP generation.
Experimental Mechanics

(b) Generate and refine the RS according to DPs to inter- the largest deformation. From Fig. 7 we can see that most
pret the relation between input and output parameters. of the sample area was correlated although some points
Here we used genetic aggregation, which generates failed to correlate mainly because of a locally insufficient
all types of RS parallelly including Full 2nd-order speckle pattern density resulting from occasional speckle
polynomial, kriging, non-parametric regression, neu- nanoparticle loss during the FIB milling process. Addi-
ral network and sparse grid. Evaluated by the genetic tionally, a line of points near the top edge of the sample
algorithm, the genetic aggregation RS (can be a single failed to correlate at large deformation (see DIC results of
RS or a combination of several RS) which is the most t = 377.3 s in Fig. 7). The shape of the contours, neverthe-
appropriate for outputs is used and updated by adding less, agrees with the behavior of a bent beam and effective
new DPs until the outputs from the current RS are close DIC results in most of the sample region were obtained.
enough to the FEM results. We will use the loading point displacements measured
(c) By setting the objectives and tolerances of output from both the indenter tip and DIC to extract properties
parameters according to the fitting points from experi- by solving the inverse problem and then, full-field DIC
mental measurements, multi-objective genetic algorithm results are used for validating the corresponding 2D FEM
is applied to analyze the finalized RS and derive proper displacement results.
input parameter sets. More details about the methods
and algorithms mentioned above can be found in [44]. Elastic Analysis

At the first ramp loading stage (0–20 s), the electron beam was
Results and Discussion blanked and the displacement response of the diamond punch,
which is also considered as the displacement of the middle
DIC Measurements point of the bottom edge (point A in Fig. 6), increased lin-
early with load. Elastic modulus E, which is the only unknown
By correlating deformed TEM images to the undeformed parameter governing the displacement response in this range,
image Fig. 7 shows the DIC results of u and v displace- was obtained by the parameter optimization process. Follow-
ment components at three instances of time: at t = 9.8 s ing the same procedure, results of elastic modulus from 12
approaching the end of the first ramp loading, at t = 213.1 s experiments spanning a range of loading rates (to establish a
that nears the end of the second creep portion loading, and possible rate effect) are shown in Fig. 8. In the range of loading
at t = 377.3 s which is the last TEM image taken and shows rates used (1–17.5 µN/s), the elastic modulus does not show

Fig. 7  DIC results of u and v displacement measurements at the end of the first loading ramp (9.8s), the end of the second step loading (213.1s),
and the end of the fifth step loading (377.3s). All values are in µm
 Experimental Mechanics

any significant trend (neither increasing or decreasing) sug- The accuracy of the fitting is also verified by substituting
gesting no rate independence of elastic modulus in this range. the derived modulus E in equation (4) and the corresponding
Overall, most of the experiments show a modulus of 65–75 load value of F = 97 µN into the Euler-Bernoulli expres-
GPa, which is in reasonable agreement with reported values sion for the amplitude term A1 = 48EI F
, where I = tb3/12.
in previous microscale studies on this material, as shown in The green curve (beam theory, t = 400 nm) in Fig. 9, which
the box area in Fig. 8 [43, 45]. shows this beam theory result, also agrees with both DIC
In order to further evaluate the elastic results, the derived measurements and the DIC-fitted magenta curve.
elastic modulus (E = 68.04 GPa with the loading of 100 µN) It should be noted that the variance of these theoreti-
was used in 1D Euler-Bernoulli beam theory to predict beam cal results can be larger since the uncertainty of geometric
deflection. In agreement with the FEM configuration, we measurements can result in considerable variation in elastic
considered a beam with two fixed ends loaded in the middle. response prediction. The thickness of the beam, for example,
Setting the geometric center of the beam as the origin, the was measured from the ion beam image in the FIB. Diffrac-
theoretical solution of vertical deflection vi located at (xi, yi) is: tion on the two boundaries along the thickness direction can
( ( )) cause a measurement uncertainty of ± 30 nm. To visulaize
L 3 L 2 the sensitivity of the solution to this geometric variablity/
| |
vi = A1 4(|xi | + ) − 3L |xi | + | | + A2 xi + A3 , (5)
2 2 uncertainty, beam theory results for thicknesses of t = 380 nm
and t = 420 nm are also plotted in Fig. 9. Similarly to thick-
where L is the beam length and three terms on the right of
ness variations, the analytical results after changing the beam
equation (5) represent beam deflection, rigid rotation and
length and width by similar amounts were also investigated
rigid translation respectively. A1 , A2 , A3 are coefficients
(though not shown here in the interst of brevity). By compar-
related to applied load and flexural rigidity, rigid rota-
ing the variation of v displacement, we observed that elastic
tion angle and rigid translation displacement, respectively.
response is most sensitive to the beam thickness measure-
These coefficients are determined by fitting DIC results with
ment (Fig. 9), then width, and then length. Coupling of these
equation (5) via nonlinear regression. For the TEM image
three measurement errors can possibly lead to a variability of
acquired at t = 9.8 s, fitting results are A1 = 9.58 × ­10− 5 μm− 2,
more than 10 GPa, which can account for the measurement
A2 = 8.47 × ­10− 4 rad, and A3 = 0.001 μm. The total v dis-
scatter of elastic modulus seen in Fig. 8.
placement is dominated by the beam deflection, while rigid
To assess two-dimensionality in the experimental con-
rotation and translation due to image drifting are small in
figuration, we can compare the 2D-FEM results to the DIC
relation. The magenta curve (beam theory fitted result) in
Fig. 9 represents this beam deflection and shows a good
agreement with the DIC measurements, the individual points
at three different beam heights, implying high fitting quality.

Fig. 9  Variation of elastic response of bending deflection v along the

beam length direction, x, at t = 9.8 s. Red dots are DIC deflection meas-
urements, magenta curve is the deflection derived by fitting DIC full-
field measurements to the theoretical equation shown in equation (5),
Fig. 8  Elastic modulus derived from 12 experiments at different load- showing agreement between experimental and analytical results. The
ing ramps between 1–17.5 µN/s, producing an average value of 69.5 three other solid-colored curves are calculated using different beam
GPa. Red dashed lines indicate the range of moduli measurements thicknesses, illustrating the sensitivity of deflection response with respect
from previous studies available in the literature [43, 45] to the beam thickness
Experimental Mechanics

measurements. √ By looking at total displacement magni- at the displacement profile in each ramp loading step, we
tude results ( u2 + v2 ) from different horizontal lines (y = can observe that the difference for this specific experiment
-0.5 μm, 0, 0.5 μm) in Fig. 10(a), we can observe substantial is mainly from elastic compliance. The creep response meas-
noise in the DIC measurements as the signal-to-noise ratio is ured by DIC and the indenter have nearly identical slopes
lower for the earlier loading stages. This improves at higher in each loading. In fact, the elastic offset between DIC and
loads when deformation accumulates. Due to this noise, it indenter measurement occurs at the end of the very first, and
is hard to observe the 2D effect with varying y coordinate. purely elastic, loading ramp, and thereafter propagates along
For a 2D simulation, however, we can see the difference of through the, very similar, creep response. If we perform the
FEM results from three solid curves in Fig. 10(a). Even so, same fitting procedure as above but using now the DIC meas-
it is also noticed that such a 2D effect is gradually weak- urements instead of the indenter load-displacement measure-
ened from x = 0 towards the two fixed ends and is negligible ment, we obtain the result shown in the magenta curve. In
beyond x ≈ ± 1 μm. We can also see the difference between this case we derived the elastic modulus E = 72.3 GPa and
DIC and FEM results is ≈ 0.02 μm. This is mainly because of creep constants C1 = 1.62 × ­10− 8 ­s− 1, C2 = 1.000, the two lat-
the discrepancy between indenter tip and DIC measurements ter being very close to the range of C1 and C2 obtained earlier.
at the loading point, resulting in a small change of compli- Similarly, good fits for both loading point (magenta curve
ance. However, while fitting with DIC data, we can see the in Fig. 12) and line results along y = -0.5 μm, 0, 0.5 μm
good fitting quality since DIC and FEM results are mostly (Fig. 11(a) and (b)) are observed for the first three ramp-step
coincident according to the corresponding line results shown loadings, whereas FEM results are smaller than DIC meas-
in Fig. 11(a). More analysis about this discrepancy will be urements at last two ramp-step loadings (Fig. 11(c)) due to
given in “Creep Analysis”. the out-of-plane twisting (discussed in “Error Analysis”).
Apart from comparing 2D results at certain times, dis-
Creep Analysis placements of correlation points in terms of time history,
i.e., displacement trajectories, were also investigated. Simi-
Using the elastic modulus derived from “Elastic Analy- lar to our previous study [38], displacements of gold nano-
sis”, we then extracted creep constants by solving inverse particles were also measured by individual particle tracking
problem in the C1 - C2 design space. By setting displace- (PT). Based on the coordinates of tracked particles, results
ment at fitting points 2&3 in Fig. 5(b) with a tolerance of of closest DIC correlation points and FEM nodes that are
± 2 nm, several pairs of (C1, C2) were derived ranging over extracted. Figure 13 shows 10 trajectories color coded with
C1: 1.69 × ­10− 8 – 4.63 × ­10− 8 ­s− 1, and corresponding C2: time, as well as the complete comparison history of DIC, PT
1.023–0.860. We then compare the entire displacement vs. and FEM for the v displacement of the red circled trajectory.
time response predicted by the FEM analysis with the meas- Even though FEM shows slightly smaller v than DIC and
ured displacement at the indenter tip. In Fig. 12 the indenter PT starting from frame 13, overall we observe good agree-
tip displacement measurement is shown as the blue curve ment between these three data sets. Compared to proper-
and the two extremes of the range of pairs C1-C2 are plot- ties derived previously from indenter measurements, here
ted as red and black lines. Looking at the response at first we derived a larger elastic modulus that indicates a smaller
two ramp-steps, which are the ones used in the fitting, the compliance, while creep constants were not severely affected.
difference between the two (C1, C2) sets is quite small, and Therefore, for fitting with either indenter or DIC data, a lin-
both follow the indenter displacement measurement well. ear dependence between creep strain rate 𝜖̇ cr and stress 𝜎
Displacement is also well predicted for the third ramp-step is reached according to the value of C2, which agrees with
although data from this step were not used in the fitting the mechanical model used in ion beam irradiation induced
process. However, we can observe a gradually increasing creep analysis [46]. According to creep constants, we derived
discrepancy between model and measurement starting from the creep compliance normalized by the beam current den-
the fourth ramp-step, where the FEM results predict lower sity ≈ 1.137 × ­10− 9 ­MPa− 1 ­C− 1 ­m2, which agrees well with
displacement values than those measured by the indenter. the creep compliance data reported by Jawaharram [20].
This discrepancy results mainly from the nonnegligible out-
of-plane twisting that can occur at higher load levels, and its Error Analysis
effect will be discussed in more detail in “Error Analysis”.
To further evaluate FEM results, we also compared them In “Elastic Analysis”, we investigated how the beam geom-
to the DIC measurements shown as yellow dots in Fig. 12 etry measurements and the discrepancy between indenter
for a location 15 pixels (15.3% of the subset size) away from and DIC displacement results affect the elastic and creep
the loading point for each image. As can be seen, in this responses. The discrepancy between indenter and DIC dis-
specific case, the DIC results are actually consistently lower placements, however, raises a new issue that the reliability
than the indenter displacement measure (blue line). Looking of these two measurements needs further investigation.
 Experimental Mechanics

Fig. 10  Comparison of total displacement magnitude at three different times (a) 9.8 s, (b) 213.1 s and (c) 377.3 s between full-field DIC and
FEM results along the horizontal axis, x. FEM results along three horizontal lines at heights y = -0.5 μm (blue), 0 (red), and 0.5 μm (black),
are governed by elastic and creep properties derived from the inverse approach based on the indenter tip displacement measurement. The corre-
sponding DIC measurements plotted as colored dots show discrepancies to the FEM results

Figure 14 shows measurements from two different sam- identical contact conditions between the diamond tip and
ples with nominally same geometry and loading profile. the sample are hard to ensure among different experi-
We can observe that the measurements from indenter can ments. In such TEM loading experiments, the diamond
be either larger or smaller than the DIC results. Indenter tip is likely to slide either in front of or behind the sample
measurements are different between two samples, while as loading increases, although the starting load and the
the corresponding DIC results are more consistent. level of sliding are unpredictable. Changing contact condi-
Although nominally the same loading profile is applied, tions may digress from the load-displacement calibration
Experimental Mechanics

Fig. 11  Comparison of total displacement magnitude at three different times (a) 9.8 s, (b) 213.1 s and (c) 377.3 s between full-field DIC and
FEM results along the horizontal axis, x. FEM results along three horizontal lines at heights y = -0.5 μm (blue), 0 (red), and 0.5 μm (black),
are governed by elastic and creep properties derived from the inverse approach based on the DIC measurement near the contact area. Improved
agreement is seen in (a) and (b), while (c) is dominated by a 3D twisting effect (discussed in “Error Analysis”)

of the indenter, which is provided by the manufacturer. DIC or indenter measurements are used for fitting. One pos-
As shown in Fig. 14, indenter measurement can be more sible reason is that after the fourth loading step, the von-
sensitive to the contact condition than DIC measurement, Mises stress at the load point calculated by FEM exceeds
and therefore the usefulness of DIC measurements can the yield stress of 2.52 GPa reported in [20]. Another related
extend beyond the validation shown earlier and may serve but important source of error is the 3D out-of-plane twist-
as the parameter of choice for the fitting procedure as well. ing that exists in almost every experiment at higher loads.
As shown in Fig. 12, the FEM results are always smaller In our TEM loading experiments, the thickness of the sam-
than experimental measurements at higher loads, either if ple is usually limited to hundreds of nanometers to ensure
 Experimental Mechanics

Fig. 14  Displacement measurements from indenter tip and DIC of

Fig. 12  FEM total displacement response (pink line) at point A, defined two samples with the same geometry and loading profile. The results
in Fig. 6, obtained by the applying creep constants derived from an suggest that the DIC measurements are more reproducible than the
inverse analyses based on the DIC measurements (orange dots). Agree- displacement measured from the indenter
ment is observed in the first three load steps, with discrepancies in the
highest load step resulting from sample out-of-plane twisting. Note that
the DIC-based fitting results showed lower creep displacement response deformation can be observed when nanoparticles originally
than the indenter-based ones, though the creep rates are similar
located exactly on the lower edge of the beam apparently
“move away” from the sample edge after loading, which
transmission of the electron beam. Even though it is easy implies that the lower edge on the back surface becomes
to make contact between the sample and the 1 μm diamond visible underneath the edge on the front surface because of
punch, aligning their central axis perfectly is almost impos- twisting. Figure 15 shows undeformed and deformed TEM
sible. An alignment offset will introduce an out-of-plane images of another sample with large twisting, where the rela-
torque on the sample that gradually becomes larger with tive positions of the DIC pattern and the sample edge move
increasing load. As shown in Fig. 3, excessive twisting more dramatically. Furthermore, with the increase of twist-
ing deformation, sliding of the diamond punch also creates
an apparent “extra” indentation displacement of the indenter,
i.e. additional elastic strain or an apparently lower modulus.
To numerically assess the additional in-plane displace-
ment from this twisting, parametric FEM simulations of
the first two ramp-step loadings with different alignment
offsets from the midplane (denoted as d) were performed.
We selected a fitted set of properties, namely E = 72.27
GPa, C1 = 1.62 × ­10− 8 ­s− 1, C2 = 1.000, and the correspond-
ing displacement response on the edge of the loading area
(marked with green cross) were extracted as a function of
loading offset d, in Fig. 16(a). With increasing offset dis-
tance, larger elastic displacement and creep displacement
rate would appear as measurements on the sample surface (at
the location marked by the cross). It should be noted that this
additional displacement accumulates with time, even before
twisting becomes readily visible in the TEM, which makes
it a challenge not only to align the loading tip, but also to try
Fig. 13  Comparison among DIC, PT, and FEM displacement results to remove the alignment error in such in situ TEM experi-
at selected locations shown as a function of time. Trajectories of the ments. Another possible reason for the twisting effect is that
selected points are shown with time denoted by the color scheme. the indentation direction might not be perpendicular to the
The a-SiO2 beam edges are illustrated by the horizontal dashed
lines. The inset shows a time history plot (by image frame number) bottom surface of the sample spanned by beam length and
for one specific trajectory thickness, but with an angle θ due to the z direction off axis
Experimental Mechanics

experiments. Therefore, actual errors of derived properties are

likely smaller. All these factors discussed above can result in
larger indenter and DIC measurements near the loading area,
which can account for the discrepancy between DIC and FEM
results especially at higher loads.
In addition to the out-of-plane twisting and loading
misalignment, another possible error source in the inverse
property calculation are the boundary conditions of the
beam sample as defined in the FEM analysis. As mentioned
in “FEM Parameter Optimization”, the beam is numerically
constrained by two fixed ends. In reality, however, the beam
Fig. 15  Undeformed and deformed TEM images of a sample with large
out-of-plane twisting. Front and back edges on the lower surface of the on both sides is bonded/connected to the same material but
beam are visible in the deformed image, indicating large twisting defor- with a much larger thickness (see Fig. 2). Even though the
mation. The circled particle has visibly “moved away” from the lower surrounded material pieces are much larger than the beam
edge of the beam which can result in a larger displacement measurement sample itself, the elastic nature of the ­SiO2 material may still
result in an additional compliance in the actual configura-
tion compared to the fixed end boundary conditions in the
trigonometrically associated with the length of the whole 2D FEM simulations. Therefore, the response at the ends of
indenter when aligning the indenter tip with the sample. A the beam may be better simulated by including the material
similar FEM parametric study on θ was conducted. Although volume surrounding the beam, which will necessitate a 3D
the effects of angular misorientation are smaller than those simulation framework. Figure 18(a) shows the correspond-
of loading offset d, they may still contribute to the overall ing 3D configuration in ANSYS which explicitly models a
discrepancies as seen in Fig. 16(b). larger portion of the material surrounding the FIB-machined
By performing the inverse analyses described earlier in microbeam, thus bypassing the constraint of fully built-in
these parametric studies, which are based on the forward FEM beam ends. The mesh size of the beam is identical to the 2D
solution of a set of known nominal properties, the sensitivity model and a coarse meshing has been performed on the sur-
of extracting elastic modulus and creep constants on the loca- rounding material. Norton’s creep model was applied only
tion and angular loading offset can be established. Figure 17 on the beam since the electron beam irradiation mainly cov-
shows the percentage changes from the nominal properties ers the beam sample region with only a small proportion of
used in the forward FEM problem solution as a function of the surrounding material occasionally being irradiated by the
amount of loading offset (both distance and angle) for the electron beam. Using the same elastic and creep properties
three properties extracted, E, C1 and C2. Introducing loading derived in the 2D model, the brown curve (3D FEM results)
offset and angle can result in a maximum deviation of 10% for in Fig. 18(b) shows a larger displacement response than the
extracted elastic modulus and less than 15% for C2. C1 error magenta curve (2D FEM results) in each ramp-up loading.
can be up to 100% due to its small value and thus sensitivity From the deformed model in FEM, we can observe that two
to the variation of C2. It is also noted that a 200 nm offset is edges of built-in ends no longer remain vertical as in the 2D
an extreme case where load would be applied exactly on the FEM, revealing the artificially increased rigidity of the 2D
edge of the beam thickness, which is less likely to happen in simulation with fixed ends. Comparing slopes at each step
loading between 2D and 3D results in Fig. 18(b), we can also

Fig. 16  FEM displacements at point A defined in Fig. 6 actuated by the

first two ramp-step loading with identical pre-assigned properties and dif-
ferent (a) loading offset d away from the midpoint of the thickness rang- Fig. 17  Relative errors of elastic and creep properties derived by per-
ing from − 200 nm to 200 nm, (b) loading angle θ deviating from vertical forming the FEM inverse analysis with different (a) loading offset d
by -4° to 4°. A twist is generated with either a non-zero loading offset or ranging from − 200 nm to 200 nm, (b) loading angle θ ranging from
loading angle, resulting in different contact displacement response − 4° to 4°. The zero relative error results correspond to d = 0 and θ = 0
 Experimental Mechanics

used to model the creep behavior. By solving inverse prob-

lems according to the experimental displacement measure-
ments, we derived an elastic modulus of 65–75 GPa for this
material, which is consistent with prior literature results.
It was also consistent with simple beam theory which pre-
dicted results close to those measured in the elastic range.
Linear dependence between creep strain rate and stress was
obtained with C2 ≈ 1 in Norton’s creep model. These prop-
erties can effectively govern the elastic and creep response
of the sample verified by the good agreement of full-field
deformation and time evolution results between FEM, DIC
and PT. However, the accuracy of these properties can be
affected by both experimental factors (e.g., the additional
deformation from out-of-plane twisting), and numerical
factors in the inverse analysis (e.g., the idealized bound-
ary conditions in 2D). Unlike the macroscale measurement
of mechanical properties where load-displacement results
are most commonly used, displacement measurement from
the loading transducer in the TEM environment can be less
stable due to its sensitivity to the uncertain contact condi-
tion. Thus, in the TEM experimental set up perhaps more so
than even in macroscale studies, independent measurement
of kinematic response through DIC (or PT) can result in
much more reliable property extraction during inverse analy-
sis. In addition, when compared to the load-displacement
data from a single transducer, full-field DIC deformation
measurements can provide more experimental data, mak-
ing it possible for characterizing properties in mechanical
behaviors governed by more complicated models (i.e., more
undetermined parameters).
Acknowledgements This work was carried out in the Materials
Fig. 18  (a) 3D FEM configuration including the beam sample and sur- Research Laboratory Central Research Facilities, University of Illi-
rounding material. The same in plane mesh size was used as in the 2D nois. Funding for this research was provided by the National Science
model (0.1 μm), with a coarser mesh generated for the surrounding mate- Foundation (grant No. CMMI 18-25466). The authors acknowledge Dr.
rial. (b) Displacement response governed by the 3D boundary condition, Honghui Zhou, Dr. Waclaw Swiech and Dr. Changqiang Chen for help
using properties extracted from the 2D model. 3D results show larger dis- with FIB, and TEM loading experiments.
placement response and creep displacement rate than the 2D case
Declarations

observe that 3D FEM gives larger displacement rates than Conflict of interest The authors declare that they have no conflict of interest.
2D FEM, and the difference is gradually increasing with
higher applied loads.
References
1. Deng Q, Cheng Y, Yue Y, Zhang L, Zhang Z, Han X, Ma E (2011)
Conclusion Uniform tensile elongation in framed submicron metallic glass
specimen in the limit of suppressed shear banding. Acta Mater
59:6511–6518. https://​doi.​org/​10.​1016/j.​actam​at.​2011.​05.​035
We performed in situ loading experiments of amorphous 2. Dillon SJ, Bufford DC, Jawaharram GS, Liu X, Lear C, Hattar K,
­SiO2 microbeam samples to quantify elastic and creep prop- Averback RS (2017) Irradiation-induced creep in metallic nanolami-
erties under electron beam irradiation in the TEM environ- nates characterized by in situ TEM pillar nanocompression. J Nuclear
ment. Full-field deformation was measured by analyzing Mater 490:59–65. https://​doi.​org/​10.​1016/j.​jnucm​at.​2017.​04.​008
3. Jawaharram GS, Barr CM, Monterrosa AM, Hattar K, Averback
undeformed and deformed gold patterns in TEM images RS, Dillon SJ (2020) Irradiation induced creep in nanocrystalline
using the DIC method specifically tailored for the TEM. high entropy alloys. Acta Mater 182:68–76. https://​doi.​org/​10.​
Two-dimensional FEM governed by Norton’s creep law was 1016/j.​actam​at.​2019.​10.​022
Experimental Mechanics

4. Mohammadi A, Physics D, University A, Arak, Iran S, Hamidi digital correlation method. Image Vis Comput 1:133–139.
(2018) Department of Physics, Arak University, Arak, Iran, M. https://​doi.​org/​10.​1016/​0262-​8856(83)​90064-1
Asadi Asadabad, materials and nuclear fuel Research School, 22. Peters WH (1982) Digital Imaging Techniques in experimental stress
Nuclear Science and Technology Research Institute, Iran, cal- analysis. Opt Eng 21:213427. https://​doi.​org/​10.​1117/​12.​79729​25
culation of radiation damage for simulation of neutron radiation 23. Wattrisse B, Chrysochoos A, Muracciole JM, Némoz-Gaillard
damage by ion irradiation and development AMTRACK program. M (2001) Analysis of strain localization during tensile tests by
IJPR 18:251–261. https://​doi.​org/​10.​29252/​ijpr.​18.2.​251 digital image correlation. Exp Mech 41:29–39. https://​doi.​org/​
5. Tai K, Averback RS, Bellon P, Ashkenazy Y (2011) Irradiation- 10.​1007/​BF023​23101
induced creep in nanostructured Cu alloys. Scr Mater 65:163–166. 24. Shah SG, Chandra JM, Kishen (2011) Fracture properties of
https://​doi.​org/​10.​1016/j.​scrip​tamat.​2011.​04.​001 concrete–concrete interfaces using digital image correlation. Exp
6. Roberts AC, Cottrell AH (1956) Creep of alpha uranium during Mech 51:303–313. https://​doi.​org/​10.​1007/​s11340-​010-​9358-y
irradiation with neutrons. Phil Mag 1:711–717. https://d​ oi.o​ rg/1​ 0.​ 25. Carroll J, Abuzaid W, Lambros J, Sehitoglu H (2010) An experi-
1080/​14786​43560​82381​45 mental methodology to relate local strain to microstructural texture.
7. Gittus J, Le May I (1976) Creep, viscoelasticity and creep fracture Rev Sci Instrum 81:083703. https://​doi.​org/​10.​1063/1.​34749​02
in solids. J Eng Mater Technol 98:191–191. https://​doi.​org/​10.​ 26. Vieira RB, Sehitoglu H, Lambros J (2021) Representative volume
1115/1.​34433​65 elements for plasticity and creep measured from high-resolution
8. Holt RA (1980) Microstructure dependence of irradiation creep microscale strain fields. Acta Mater 214:117021. https://​doi.​org/​
and growth of zirconium alloys. J Nuclear Mater 90:193–204. 10.​1016/j.​actam​at.​2021.​117021
https://​doi.​org/​10.​1016/​0022-​3115(80)​90256-1 27. Chasiotis I, Knauss WG (2002) A new microtensile tester for the
9. Wolfer WG (1980) Correlation of radiation creep theory with study of MEMS materials with the aid of atomic force micros-
experimental evidence. J Nuclear Mater 90:175–192. https://​doi.​ copy. Exp Mech 42:51–57. https://​doi.​org/​10.​1007/​BF024​11051
org/​10.​1016/​0022-​3115(80)​90255-X 28. Li X, Xu W, Sutton MA, Mello M (2007) In situ nanoscale In-
10. Mansur LK, Reiley TC (1980) Irradiation creep by dislocation Plane deformation studies of ultrathin polymeric Films during
glide enabled by preferred absorption of point defects — theory Tensile Deformation using Atomic Force Microscopy and Digital
and experiment. J Nuclear Mater 90:60–67. https://​doi.​org/​10.​ Image correlation techniques. IEEE Trans Nanotechnol 6:4–12.
1016/​0022-​3115(80)​90245-7 https://​doi.​org/​10.​1109/​TNANO.​2006.​888527
11. Trinkaus H, Ryazanov AI (1995) Viscoelastic model for the plastic 29. Li N, Sutton MA, Li X, Schreier HW (2008) Full-field Thermal
flow of amorphous solids under energetic ion bombardment. Phys Rev deformation measurements in a scanning Electron microscope by
Lett 74:5072–5075. https://​doi.​org/​10.​1103/​PhysR​evLett.​74.​5072 2D digital image correlation. Exp Mech 48:635–646. https://​doi.​
12. Mayr SG, Ashkenazy Y, Albe K, Averback RS (2003) Mechanisms org/​10.​1007/​s11340-​007-​9107-z
of radiation-induced viscous flow: role of point defects. Phys Rev 30. Kammers AD, Daly S (2011) Small-scale patterning methods for
Lett 90:055505. https://​doi.​org/​10.​1103/​PhysR​evLett.​90.​055505 digital image correlation under scanning electron microscopy.
13. Meftah A, Costantini JM, Khalfaoui N, Boudjadar S, Stoquert Meas Sci Technol 22:125501. https://d​ oi.o​ rg/1​ 0.1​ 088/0​ 957-0​ 233/​
JP, Studer F, Toulemonde M (2005) Experimental determina- 22/​12/​125501
tion of track cross-section in Gd3Ga5O12 and comparison to 31. Bourdin F, Stinville JC, Echlin MP, Callahan PG, Lenthe WC, Torbet
the inelastic thermal spike model applied to several materials, CJ, Texier D, Bridier F, Cormier J, Villechaise P, Pollock TM, Valle
Nuclear Instruments and Methods in Physics Research Section V (2018) Measurements of plastic localization by heaviside-digital
B: Beam interactions with materials and atoms. 237:563–574. image correlation. Acta Mater 157:307–325. https://​doi.​org/​10.​1016/j.​
https://​doi.​org/​10.​1016/j.​nimb.​2005.​02.​025 actam​at.​2018.​07.​013
14. Aumayr F, Facsko S, El-Said AS, Trautmann C, Schleberger M 32. Chu TC, Ranson WF, Sutton MA (1985) Applications of digital-
(2011) Single ion induced surface nanostructures: a compari- image-correlation techniques to experimental mechanics. Exp
son between slow highly charged and swift heavy ions. J Phys Mech 25:232–244. https://​doi.​org/​10.​1007/​BF023​25092
Condens Matter 23:393001. https://​doi.​org/​10.​1088/​0953-​8984/​ 33. Novak MD, Zok FW (2011) High-temperature materials testing with
23/​39/​393001 full-field strain measurement: experimental design and practice. Rev
15. Chen I-T, Poblete FR, Bagal A, Zhu Y, Chang C-H (2022) Ane- Sci Instrum 82:115101. https://​doi.​org/​10.​1063/1.​36578​35
lasticity in thin-shell nanolattices. Proc Nat Acad Sci 119. https://​ 34. Grégoire D, Loh O, Juster A, Espinosa HD (2011) In-situ AFM
doi.​org/​10.​1073/​pnas.​22015​89119 experiments with discontinuous DIC Applied to damage identi-
16. Onimus F, Jourdan T, Xu C, Campbell AA, Griffiths M (2020) fication in Biomaterials. Exp Mech 51:591–607. https://​doi.​org/​
Irradiation creep in materials. Compr Nuclear Mater Elsevier 10.​1007/​s11340-​011-​9463-6
310–366. https://​doi.​org/​10.​1016/​B978-0-​12-​803581-​8.​11645-5 35. Biery N, de Graef M, Pollock TM (2003) A method for measuring
17. Chen J, Jung P, Hoffelner W (2013) Irradiation creep of can- microstructural-scale strains using a scanning electron microscope:
didate materials for advanced nuclear plants. J Nuclear Mater applications to γ-titanium aluminides. Metall and Mat Trans A
441:688–694. https://​doi.​org/​10.​1016/j.​jnucm​at.​2013.​04.​024 34:2301–2313. https://​doi.​org/​10.​1007/​s11661-​003-​0294-7
18. Zheng K, Wang C, Cheng Y-Q, Yue Y, Han X, Zhang Z, Shan 36. Ososkov Y, Wilkinson DS, Jain M, Simpson T (2007) In-situ meas-
Z, Mao SX, Ye M, Yin Y, Ma E (2010) Electron-beam-assisted urement of local strain partitioning in a commercial dual-phase
superplastic shaping of nanoscale amorphous silica. Nat Com- steel. IJMR 98:664–673. https://​doi.​org/​10.​3139/​146.​101526
mun 1:24. https://​doi.​org/​10.​1038/​ncomm​s1021 37. Hoefnagels JPM, van Maris MPFHL, Vermeij T (2019) One-step
19. Ebner C, Rajagopalan J, Lekka C, Rentenberger C (2019) Elec- deposition of nano-to-micron-scalable, high-quality digital image
tron beam induced rejuvenation in a metallic glass film during correlation patterns for high-strain in-situ multi-microscopy test-
in-situ TEM tensile straining. Acta Mater 181:148–159. https://​ ing. Strain 55. https://​doi.​org/​10.​1111/​str.​12330
doi.​org/​10.​1016/j.​actam​at.​2019.​09.​033 38. Zhang Y, Feng L, Dillon S, Lambros J (2021) Full-field deforma-
20. Jawaharram GS (2019) The effects of irradiation on the creep tion measurements in the transmission electron microscope using
and hardening behavior of crystalline and amorphous materials, digital image correlation and particle tracking. Mater Charact
thesis, University of Illinois at Urbana-Champaign 111598. https://​doi.​org/​10.​1016/j.​match​ar.​2021.​111598
21. Sutton MA, Wolters WJ, Peters WH, Ranson WF, McNeill 39. Seguini G, Curi JL, Spiga S, Tallarida G, Wiemer C, Perego M
SR (1983) Determination of displacements using an improved (2014) Solid-state dewetting of ultra-thin au films on ­SiO2 and
 Experimental Mechanics

­HfO2. Nanotechnology 25:495603. https://​doi.​org/​10.​1088/​0957-​ 45. Jaccodine RJ, Schlegel WA (1966) Measurement of strains at
4484/​25/​49/​495603 si-sio2 interface. J Appl Phys 37:2429–2434. https://​doi.​org/​10.​
40. Schreier H, Orteu J-J, Sutton MA (2009) Image correlation for 1063/1.​17088​31
shape, motion and deformation measurements. Springer US, Bos- 46. Ashkenazy Y, Averback RS (2012) Irradiation induced grain
ton, MA. https://​doi.​org/​10.​1007/​978-0-​387-​78747-3 boundary flow–a new creep mechanism at the nanoscale. Nano
41. Norton FH (1929) The creep of steels at high temperatures. Mc- Lett 12:4084–4089. https://​doi.​org/​10.​1021/​nl301​554k
Graw-Hill, New York
42. Sabour MH (2013) Creep. In: Wang QJ, Chung Y-W (eds) Ency- Publisher’s Note Springer Nature remains neutral with regard to
clopedia of Tribology. Springer US, Boston, MA, pp 618–627. jurisdictional claims in published maps and institutional affiliations.
https://​doi.​org/​10.​1007/​978-0-​387-​92897-5_​291.
43. Li H, Vlassak JJ (2009) Determining the elastic modulus and hard- Springer Nature or its licensor (e.g. a society or other partner) holds
ness of an ultra-thin film on a substrate using nanoindentation. J exclusive rights to this article under a publishing agreement with the
Mater Res 24:1114–1126. https://​doi.​org/​10.​1557/​jmr.​2009.​0144 author(s) or other rightsholder(s); author self-archiving of the accepted
44. Ansys® Workbench (2018) Help system, DesignXplorer Guide. manuscript version of this article is solely governed by the terms of
ANSYS, Inc such publishing agreement and applicable law.