You are on page 1of 12

Available online at www.sciencedirect.

com

Acta Materialia 59 (2011) 1259–1270


www.elsevier.com/locate/actamat

Residual stress and thermo-mechanical properties of cold spray


metal coatings
V. Luzin a,⇑, K. Spencer b,c, M.-X. Zhang b,c
a
The Bragg Institute, Australian Nuclear Science and Technology Organisation, Locked Bag 2001, Kirrawee DC, NSW 2232, Australia
b
Division of Materials, School of Mechanical and Mining Engineering, The University of Queensland, St. Lucia, QLD 4072, Australia
c
ARC Centre of Excellence for Design in Light Metals, Australia

Received 19 May 2010; received in revised form 25 October 2010; accepted 27 October 2010
Available online 22 November 2010

Abstract

The residual stress profiles in Cu and Al coatings cold sprayed using kinetic metallization have been studied using neutron diffraction.
To interpret results and to describe them quantitatively, the measured profiles were fit to Tsui and Clyne’s progressive coating deposition
model, which demonstrated that the residual stresses are largely due to kinetic and not thermal effects. The residual stress state of the
coatings was found to depend mainly on the deformation behaviour and properties of the coating material, and the kinetic parameters
of the cold spray process. Young’s modulus and impact strain were measured and used along with published material data for Cu and Al
to approximate the residual stresses, using a model developed for shot peening. The properties of the Cu coatings such as Young’s mod-
ulus and porosity were found to be closer to their bulk values than in the case of the Al coatings, and this was related to the amount of
particle deformation on impact.
Crown Copyright Ó 2010 Published by Elsevier Ltd. on behalf of Acta Materialia Inc. All rights reserved.

Keywords: Cold spray; Residual stress; Neutron diffraction

1. Introduction it is suitable for depositing thermally sensitive materials,


and little or no oxidation is thought to occur during the
Cold spray [1] is a rapidly developing technology for spray process. The low temperature also makes this process
depositing materials in the solid state. In this process the suitable for coatings on light metal substrates such as mag-
feedstock material, which should have a metallic powder nesium alloys. In all such applications coating integrity is
as the principal component, is injected into a gas stream important, which may be loosely defined as the quality of
and accelerated to speeds from 500 to 1000 m s1 and is bonding between particles within the coating, and between
impacted onto a metallic substrate. If the powder reaches the coating and the substrate. Coating integrity may be
a critical velocity defined by the material properties and influenced by the residual stresses present in the coating,
process conditions, then metallurgical bonding is obtained and in the case of thermal spray coatings residual stresses
and the properties of the deposit approach those of the have been shown to lead to peeling and delamination of
equivalent bulk material. the coating [2]. Among other factors, understanding, pre-
The primary application of cold spray coatings is for the diction and control of internal stress accumulation can
surface enhancement of metals to improve properties such contribute to improved coating performance. To date there
as wear and corrosion resistance, electrical/thermal con- is a limited amount of data detailing the residual stresses in
ductivity, etc. Due to the low temperature of the process cold spray coatings [3,4]. Also, there have been no attempts
to interpret the measured values based on the kinetics of
⇑ Corresponding author. Tel.: +61 2 9717 7262; fax: +61 2 9717 3606. the spray process and material properties other than
E-mail address: vladimir.luzin@ansto.gov.au (V. Luzin). straightforward finite element calculations.

1359-6454/$36.00 Crown Copyright Ó 2010 Published by Elsevier Ltd. on behalf of Acta Materialia Inc. All rights reserved.
doi:10.1016/j.actamat.2010.10.058
1260 V. Luzin et al. / Acta Materialia 59 (2011) 1259–1270

A distinct feature of the cold spray process is the com- aration is required. Although laboratory X-rays do not
pressive residual stress that arises during the deposition have the penetration power of neutrons and are not suited
process. A state of compressive residual stress on the sur- for stress profiling, they are indispensable for stress mea-
face is generally thought to be beneficial for fatigue resis- surements on the surface of coatings and films [17] or the
tance, and several surface treatment processes are stress measurements in individual splats [18], or they can
designed to produce this state of stress such as shot peen- be adapted for the through-depth stress profiling in lm-
ing, laser shock peening, low plasticity burnishing, ultra- thin films [19].
sonic impact treatment, etc. In all of these methods the In the present work we present the results of residual
compressive stress is produced through plastic deformation stress measurements obtained by using all discussed advan-
of the surface region. In a similar manner, the cold spray tages of the neutron diffraction with the focus on stress
process induces compressive stress by high speed impact profiling. Cold spray coatings produced using different
of the sprayed particles on the surface, causing a peening coating/substrate material combinations have been stud-
effect. In Ref. [4] the role of the particle kinetic energy in ied. Through-thickness stress profiles are fit to the progres-
the accumulation of residual stress was studied, by compar- sive deposition model of Tsui and Clyne [20–22], which has
ing the internal stress in coatings produced by “high ther- been used to assess the relative importance of thermal vs.
mal impact” spray processes such as plasma spray to kinetic effects in the accumulation of residual stresses dur-
those produced by “high kinetic impact” spray processes ing the cold spray process. Observations of the coating
such as cold spray or the high-velocity oxy-fuel process. microstructure are used to interpret the measured stresses,
As far as experimental studies are concerned, several which may be qualitatively predicted using an analogy to
experimental techniques have been used to study residual shot peening. This is done using a physically based model
stress in coatings and films produced by very different tech- incorporating readily available material parameters.
niques and methods. Historically, the curvature measure-
ment technique has been employed first for studying 2. Experimental procedures and material characterization
stress in the electrolytically deposited thin metallic films
[5] and it was a major stress measurement technique in 2.1. Materials
coatings for the following decades [6]. In its original formu-
lation, the Stoney formula, the method works well for thin Copper and aluminium were selected as coating and
films; for thicker coatings some significant corrections are substrate materials. Different combinations of the substrate
required [7]. The modern versions of this technique are still and coating materials enabled an analysis of the different
in use and known as the continuous curvature measure- factors influencing the residual stress state. Four samples
ment [8] or in situ curvature measurement [9–12] and allow in total were produced forming a 2  2 matrix, two coating
even through-thickness stress profile reconstruction but materials vs. two substrate materials, Al/Al, Al/Cu, Cu/Cu
with some experimental and numerical difficulties. The and Cu/Al.
main drawbacks of the method are apparent – it is impos- The substrate materials used were electrical grade pure
sible to retrace stresses in the samples with no recorded copper sheet and 5xxx Al–Mg alloy sheet, chosen because
in situ thermal/displacement history, and it works only their grain size of 20–30 lm enables reasonable counting
for samples of certain geometry and design. A group of times for the neutron diffraction stress measurements.
material removal techniques, which includes the layer The thickness of the Cu and Al substrates was 3.1 mm
removal method [2], slitting method, hole drilling [13] and 2.6 mm respectively. Before spraying they were cut into
and some other variations, can overcome these problems square coupons 30  30 mm, sized to obtain a state of bal-
but at the price of sample destruction. Although these anced biaxial plane stress in the central part. Prior to spray-
methods have been applied successfully, extraction of the ing the substrate surface was ground using 1200 grit SiC
residual stress might require considerable numerical com- paper and rinsed with ethanol.
putations, especially for the samples of complex geometry, The feedstock powders used were commercially pure
and accuracy might suffer. The neutron diffraction tech- copper and aluminium powders. The particle size for each
nique is non-destructive and the most direct way of stress material was measured using a laser diffraction particle
determination in coatings thicker than 0.5 mm. Because size analyser, with the volume-weighted average given in
of their penetration power, neutrons can be used for stress Table 1. Both powders were produced using an atomiza-
measurements in relatively thick deposits (cm) and/or with tion process, and are almost perfectly spherical in shape.
fine resolution (better than 0.5 mm). Through-thickness
stress profiles have been experimentally determined with 2.2. Spraying procedure
success for different types of coatings nature (metals,
ceramics and composites) produced by different spraying The cold spray coatings were produced using a kinetic
techniques [14,15]. Unlike the two mechanical methods metallization (KM) system, which is a commercially avail-
described above, the neutron diffraction technique is phase able cold spray variant (Inovati, Santa Barbara, CA,
sensitive and stresses can be assessed for each phase inde- USA). Whereas a conventional cold spray system uses a
pendently [16]. In addition, no special sample/surface prep- convergent–divergent nozzle to accelerate the process gas
V. Luzin et al. / Acta Materialia 59 (2011) 1259–1270 1261

Table 1
Process parameters used for cold spray experiments.
Powder Average particle size (lm) Driving pressure (kPa) Nozzle temperature (°C) Powder feed Estimated particle
material rate (g min1) exit velocity (m s1)
Al 15 620 140 15 585
Cu 6 620 200 15 645

to supersonic speeds between Mach 1 and Mach 3 [23], KM between horizontal nozzle movement lines of the spray
uses a convergent barrel nozzle to accelerate the process nozzle. We assume that this is due to dynamic effects of
gas to Mach 1 [24]. In order to increase the impact veloc- the spraying process. At first, small surface irregularities
ity, helium was used as the process gas, as the speed of form because of non-uniformities in flux density across
sound in helium is almost three times that in nitrogen. Fur- the stream of particles. As deposition progresses, these ini-
thermore, we have used a smaller particle size than is typ- tial undulations are amplified: hills become bigger since
ical in supersonic cold spray systems. Therefore, in this they are exposed to more particle flux, but the valleys do
work the particle exit velocities are similar to those attained not fill in because they are shadowed by hills, and they
in a conventional supersonic cold spray system. The pro- become relatively deeper. This defect is observed only in
cess conditions and resultant particle velocities are detailed very thick coatings; in the present case they were made
in the next section. Useful comparisons of convergent vs. thick specifically to enable easier neutron stress measure-
convergent–divergent nozzles in cold spray are given in ments. Usually in practical applications the coating thick-
Refs. [25,26]. ness is less then 0.5 mm, and this pronounced pattern is
not developed.
2.3. Spraying conditions
2.4. Neutron powder diffraction measurements
All coatings were sprayed using a temperature optimized
for maximum deposition efficiency while avoiding fouling Neutron powder diffraction was used to measure the
of the nozzle. The process parameters used are as follows. oxide volume fraction of the sprayed coatings, to compare
The nozzle standoff distance from the substrate was 12 mm it against the feedstock powder, and to confirm the chem-
and the nozzle traverse speed was 50 mm s1. The nozzle ical purity of the feedstock powders. Neutron diffraction
temperature was measured in the mixing chamber just was chosen over X-ray diffraction because the latter
upstream of the nozzle throat (Table 1). This temperature method is surface quality sensitive and can bias the result
is used to estimate the exit velocity of the particles from toward a higher oxide content. Also, the oxide content of
the nozzle using a one-dimensional isentropic model based the coating surface may be different from that of the coat-
on that of Dykhuizen and Smith [27]. In a simple conver- ing bulk, and neutron diffraction enables a better average
gent barrel nozzle the gas speed through the barrel is less measurement. Neutron diffraction data were collected
than Mach 1 due to friction effects. In the KM system using a wavelength of 1.622 Å from the high resolution
the barrel of the nozzle has a slight divergence to compen- powder diffractometer ECHIDNA at the ANSTO OPAL
sate for friction [28]; therefore, the velocity along the barrel
is Mach 1, which simplifies calculation of particle speed.
The gas properties are calculated assuming ideal gas behav-
iour, and the average particle size is used. The particle drag
coefficient is calculated using the correlation of Henderson
[29], based on the change in particle Reynolds number as it
accelerates through the nozzle. The process parameters are
summarized in Table 1. It should be noted that the nozzle
temperature is in fact the stagnation temperature of the
mixed gas/particle stream just prior to being accelerated
through the nozzle throat.
The resultant thickness of the deposited coatings was
3 mm to enable through-thickness stress measurements
using a reasonable diffraction volume. It is typical that
thick cold spray coatings have an uneven surface, as shown
in Fig. 1. A pattern appears with streak-like features in the
direction of nozzle movement. Quantitatively, roughness
has been estimated through maximum profile height, Fig. 1. General appearance of the surface of the Al/Cu cold sprayed
Rt = 1.0 mm, and average Ra = 0.2 mm. The spatial period sample. Inset: the top coating surface profile obtained by digitizing of the
of the ripples is 1 mm and this conforms to the 0.5 mm step sample cross-section.
1262 V. Luzin et al. / Acta Materialia 59 (2011) 1259–1270

research reactor [30]. The phase volume fraction was deter- 2.6. Thermo-mechanical properties
mined using the GSAS Rietveld refinement software [31]
with the EXPGUI front-end [32] for the Al and Cu feed- The planar geometry of the coatings and their small
stock powders and corresponding coatings. thickness make it difficult to obtain specimens large enough
for direct mechanical tests. However, two methods were
2.5. Neutron diffraction stress measurements found to be appropriate to measure the Young’s modulus
of the coatings:
The bulk of the neutron diffraction residual stresses
measurements have been carried out at the NIST Center (i) Microindentation tests were performed on the pol-
for Neutron Research using the residual stress diffractom- ished cross-section of the specimens using a UMIS
eter BT8 [33]. A gauge volume 0.5  0.5  18 mm was cho- 200 universal material tester together with IBIS soft-
sen as a result of the fine balance between different ware to analyse the indentation data. The Bercovich
competing factors: (i) it should be small enough to provide test was employed in load control mode using a max-
necessary through-thickness resolution and avoid edge imum load of 250 mN in the case of Al and 100 mN
effects; (ii) at the same time it should be large enough to for Cu. The Young’s modulus was extracted by ana-
produce a count rate sufficiently high such that experimen- lysing indentation curves (unloading–displacement).
tally determined strains can be measured with statistical Indents were made in several different locations to
uncertainty better than 5  105, within a reasonable mea- obtain an average value and a standard deviation.
surement time. Since Cu is a stronger scatterer than Al the (ii) The Young’s modulus was also measured in rectan-
measurement time was varied correspondingly: 10 min per gular specimens (27  5  2 mm for Cu and
measurement point for Cu and 90 min for Al. 32  5  2 mm for Al) using the impulse excitation
The measurements were done in several different technique (IET) according to the ASTM standard
through-thickness locations to cover the entire sample thick- E1876. The accuracy of this method is generally high
ness, forming a line profile with 0.3 mm spacing between and limited in the present case by the parallelism of
points. In order to optimize localization of the gauge vol- the specimen faces and the uncertainty of the smallest
ume, the take-off angle 2hM of the Si (400) monochromator dimension. It was estimated that the accuracy was
was varied to maintain a 90° geometry (2hB  90°). The better than 0.5%.
instrument settings are summarized in Table 2.
For each measurement point, d-spacings (diffraction The coefficient of thermal expansion (CTE) was mea-
peak positions) were measured in the two principle direc- sured using a dilatometer. Samples were measured in the
tions, normal to the surface and in-plane. From the mea- temperature rage of 22–300 °C in steps of 0.5 °C with a
sured d-spacings in-plane stresses were calculated using heating ramp rate of 5 °C min1. The CTE behaviour
the assumption of a balanced biaxial plane stress state fol- above 180 °C was the most stable, so it was used to judge
lowing the procedure described in Ref. [3]. The diffraction the closeness of the CTE of the cold spray coatings to
elastic constants used for stress calculation were computed the equivalent bulk value.
using the self-consistent method of Kröner [34], and are The rectangular specimens used to measure Young’s
reported in Table 2. modulus in (ii) were also used to calculate coating porosity
To separate stresses originating from the cold spray pro- using Archimedes’ method. Bulk specimens of Cu and Al
cess from pre-existing stresses (e.g. residual stress from cold were used to assess the accuracy of the method, and the
rolling of the substrate) neutron stress measurements were results are within 1% of published values.
done on the uncoated substrate. They were treated as sep-
arate samples and measured using the same procedure. 3. Numerical modelling of the residual stress
One sample, a Cu coating on an Al substrate, was mea-
sured on a different neutron diffractometer – the KOWARI The basis of our analysis of residual stress accumulation
strain scanner at the ANSTO OPAL research reactor. The is according to Tsui and Clyne’s progressive coating depo-
measurement protocol was kept as close as possible to what sition model [20]. A given spraying process such as cold or
is described above with only one major difference: for mea- thermal spray is complex; however, it has been demon-
surements in the copper coating the Cu (222) reflection was strated that the model is applicable to the prediction of
used. Use of the appropriate diffraction elastic constants, the stress distribution in coatings produced by different
S1 = 1.94 and ½S2 = 8.23 TPa1, however, should ensure techniques under various spray conditions [4]. Thus, we
there is no practical difference. can consider that the progressive deposition model is ade-
Table 2
Instrument setting and material constants for the measured reflections.
d (Å) 2hM (°) k (Å) 2hB S1 (TPa1) ½S2 (TPa1) t (min)
Cu (311) 1.10 70.0 1.55 89.7° 3.38 12.58 10
Al (311) 1.22 79.8 1.72 89.5° 5.16 19.57 90
V. Luzin et al. / Acta Materialia 59 (2011) 1259–1270 1263

quate to physical process and it adequately describes the mentally measured stress profiles; this serves to validate the
elastic stress state in coatings unless stress development microscopic modelling approach. The fitting process
during the deposition process reaches yield conditions. In enables the relative significance of rd and De to be deter-
this case, plastic effects intervene and should be addressed mined for a given coating, and also to evaluate the signifi-
in a more general elasto-plastic model, see e.g. Ref. [35]. cance of these terms for different materials. This can be
Although the model was originally developed to model used to assess the importance of the thermal vs. kinetic
the residual stress accumulation in thermal spray coatings, components of the coating process, and the sensitivity of
it has been found to work equally well for cold spray coat- a given system to accumulating residual stress.
ings in the present work. Similar to the model applications
used before in the case of the thermal spray for separation 4. Results
of quenching stress from thermal stress [12,15], in our study
we use the same model to separate the peening stress from 4.1. Microstructure
the thermal stress.
In the model, two components of the spray process are Secondary electron micrographs of a Cu coating on a
accounted for separately using two different fitting Cu substrate and an Al coating (removed from a Cu sub-
parameters: strate for etching) are shown in Fig. 2. Macroscopically,
both coatings are uniform and free of large voids or cracks.
(i) The coating deposition process is considered as the At higher magnification some details of the microstructure
formation of a new, nth layer on the top of the previ- are clear: while there is some deformation of the particles
ously formed system comprising the substrate plus all from their original spherical shape, the overall structure
previously deposited coating layers. This new layer is is still granular, as compared to the typical splat-like micro-
formed with a characteristic deposition stress rd. For structure of thermal spray coatings. There was virtually no
thermal spray coatings rd is treated as a tensile observable porosity in the Cu coatings, and a small amount
“quench” stress originating from shrinkage of a solid- of porosity in the Al coatings.
ifying splat on the surface. For cold spray coatings The oxide content of the feedstock powders and coating
this stress is compressive, characteristic of a peening materials is given in Table 3, as measured by neutron pow-
process. der diffraction. The oxide content of the Al powder and
(ii) The coating process occurs at or above room tempera- coating was below the resolution of the measurement,
ture. If after coating deposition the substrate + coat- 0.5 vol.%.
ing system is cooled to room temperature, then a Table 4 gives the porosity of the Cu and Al coatings esti-
thermal misfit term De is needed to account for thermal mated using Archimedes’ method. The oxide content of the
stresses arising from a difference in thermal expansion coatings in Table 3 was used to correct the measurement,
coefficients of the substrate and coating materials, and measurements of the substrate materials were done
De = DaDT. The significance of this term depends on
the cooling range and difference in thermal expansion
Table 3
coefficients of the coating and substrate materials. Results of powder diffraction phase analysis for different coating
materials.
Determination of the two physical parameters of the Oxide species Oxide content (vol.%)
model can be made by studying the coating formation pro-
Cu powder vs. coating Cu2O 5 ± 0.7 vs. 4 ± 0.7
cess on the microscopic scale. In the present case it was Al powder vs. coating Al2O3 <0.5 vs.<0.5
found more practical to obtain rd and De from the experi-

Fig. 2. Secondary electron micrographs of the cross-section of: (a) Cu coating on a Cu substrate and (b) Al coating on an Al substrate.
1264 V. Luzin et al. / Acta Materialia 59 (2011) 1259–1270

Table 4 30
Porosity measurements for the cold spray coatings.
q (g/cm3) qref (g/cm3) Vol.% 25 Al

CTE [x10 -6 1/K]


Cu substrate 8.951 8.954 (Cu 100%) 99.96 ± 0.3
Cu coating 8.670 8.806 (Cu 95% + Cu2O 5%) 98.5 ± 0.5 20
Al coating 2.528 2.700 (Al 100%) 93.6 ± 0.6
Cu
15

to verify the accuracy of the technique. Higher porosity in


10
the Al coatings was expected because of the use of coarser
200 250 300
powder (15 lm) than for the Cu coatings (6 lm). Although temperature,°C
the use of a mixture of powders or finer powder can reduce
the porosity of Al coatings to the level of 1–2%, the finer Fig. 3. Comparison of the temperature dependence of CTE of the sprayed
material (solid line) and bulk value (dashed line).
powders agglomerate and do not feed as consistently.

4.2. Thermo-mechanical properties ison of expansion behaviour is given in Fig. 3. There is no


significant difference between the coating and substrate
The Young’s modulus measurements of the cold spray behaviour. The thermal expansion coefficient values
coatings are given in Table 5. They are closer to those of obtained are summarized in Table 5, and are equivalent
a bulk material than is found in the case thermal spray to those of a bulk material.
coatings. For example, a plasma sprayed Cu coating has
a typical Young’s modulus of 30–40 GPa, while it is 4.3. Residual stress
100 GPa in the cold spray coatings examined here. In
thermal spray coatings the deviation of mechanical proper- The in-plane residual stress profiles determined by neu-
ties from those of a bulk material is mainly caused by tron diffraction are shown in Fig. 4a–d. Both Cu coatings
imperfections in interparticle bonding, oxidation of the show a significant compressive residual stress at the surface
splat surface and, to a lesser extent, coating porosity. While 50–80 MPa, while the surface residual stress of the Al
the indentation test and impulse excitation test give similar coatings is less than 10 MPa. The overall stress profile of
modulus values for the Cu coating, the two techniques give the two Cu coatings is similar, regardless of the substrate
significantly different modulus values for the Al coating. material, and the same can be said of the Al coatings.
Any calculations in the present work requiring Young’s Some internal stress was measured in the freestanding Al
modulus were done using the values obtained from the substrate before the cold spray coating was applied, and this
impulse excitation measurements, as this is generally con- was subtracted to isolate stresses induced by the cold spray
sidered the most accurate measurement of elastic constants. process. There was no significant residual stress in the free-
The difficulty of Young’s modulus measurements in coat- standing Cu substrate. The error bars in Fig. 4 comprise
ings by indentation technique is known but with certain both neutron counting statistical error and positioning
precautions can be successfully employed [36–38]. Issues error, and each of these gives approximately equal contribu-
such as the presence of a splat-like microstructure, the pres- tions to the total uncertainty. The achieved uncertainties of
ence of porosity as well as surface preparation can impact 5–10 MPa are typical of this type of experiment. In the
results. The better agreement between results obtained by points close to the free surface or the coating/substrate inter-
indentation and IET in the Cu coatings is in accord with face the measured stress values can be biased due to the edge
the fact that the Cu coating is denser than the Al coatings, effect in neutron stress measurements, and it is difficult to
so for the Cu coatings porosity had a smaller influence on eliminate all of the error from this effect in the measure-
the measurements. Thus, the modulus value measured by ments. With these issues in mind, the stress balance condi-
the indentation technique can be more distorted in the case tion is fulfilled for the experimental data: the volume
of the Al coatings and larger measurement errors are integral of the residual stress adds to zero, which gives added
observed, as shown in Table 5. confidence regarding the accuracy of the stress profiles.
The thermal expansion coefficient was measured for The stress profiles were used as an input to fit experi-
both the coatings and the substrate material and a compar- mental data with the model of Tsui and Clyne as described

Table 5
Experimental values of the thermal and mechanical properties for coating materials.
Young’s modulus (GPa) CTE (106 K1)
Indentation test Impulse excitation technique Bulk material value Measured value Bulk material value
Cu coating 98 ± 10 104.0 ± 0.5 124a 17 17
Al coating 74 ± 20 49.4 ± 0.2 71 22 22
a
This value was also measured by IET on the specimen taken from the Cu substrate.
V. Luzin et al. / Acta Materialia 59 (2011) 1259–1270 1265

(a) 100 (b) 100


80 80
60 60
40 40

stress,MPa
stress,MPa
20 20
0 0
-20 -20
-40 -40
-60 -60
-80 -80
Cu substrate Cu coating Al substrate Cu coating
-100 -100
-3 -2 -1 0 1 2 -2 -1 0 1
through-thickness position,mm through-thickness position,mm

(c) 60 (d) 60

40 40
stress,MPa

stress,MPa
20 20

0 0

-20 -20

-40 -40
Cu substrate Al coating Al substrate Al coating
-60 -60
-3 -2 -1 0 1 2 -2 -1 0 1 2
through-thickness position,mm through-thickness position,mm

Fig. 4. Measurement (symbols) and model fit (solid lines) of the through thickness in-plane stress distributions for (a) Cu/Cu sample, (b) Cu/Al sample, (c)
Al/Cu sample and (d) Al/Al sample.

previously, using the fitting parameters rd and De, which lated peening stress from successive deposited layers that
are characteristics of the deposition stress and thermal mis- is largely responsible for the residual stress. The evidently
match respectively. All other parameters such as coating higher values of rd in the case of the Cu coatings suggest
thickness and elastic modulus were taken from measured the amount of plasticity on impact is more significant than
values. The fitting parameters are given in Table 6 and in the Al coatings, leading in part to higher residual stress
the corresponding stress profiles are shown in Fig. 4a–d accumulation.
as solid lines. The fitted stress profiles are almost entirely The thermal mismatch term De has a minimal effect on
determined by rd, with De having relatively little effect: the fitted residual stress profiles. This is due to the small
the small shift resulting from setting De to zero is less than macroscopic variation in temperature during the spray pro-
the experimental error (5 MPa) in the neutron diffraction cess of 100 °C, which is discussed in Appendix A (ii.2)
stress measurements. and this makes the thermal mismatch term De = DaDT
almost negligible. In the case of high-velocity oxy-fuel coat-
5. Discussion ings both impact and thermal effects play a role in the accu-
mulation of residual stresses. In contrast, the residual stress
5.1. Interpretation of the result in terms of empirical model accumulation in the cold spray coatings examined here is
parameters primarily a result of plastic deformation resulting from
particle impact, and thermal effects are relatively minor.
The fitted stress profiles in Fig. 4 are almost entirely Macroscopically, the residual deposition stress in the
shaped by rd. Its negative sign (compression) is indicative coatings rd is the net result of the stress resulting from
of “peening” on impact and the evidence of a certain degree shock loading, then unloading with, in general, reverse
of plastic strain of the sprayed material. It is the accumu- yielding and then possible stress relief due to recovery
and relaxation processes. The microscopic interpretation
Table 6 of rd (in terms of micromechanics, on the scale of a parti-
Fitting parameters of the model and the quality of fit of the experimental cle) is less straightforward and requires the knowledge of
data. many details of the deposition process such as non-uniform
Coating/substrate rd ± error (MPa) De ± error (l strain) v2 local deformation and recrystallization, that can be studied
Cu/Cu 41 ± 12 150 ± 200 2.0 by some experimental methods such as EBSD [39] and
Cu/Al 85 ± 20 140 ± 200 2.4 TEM [40], multiple modes of stress relaxation, etc. (see
Al/Cu 9 ± 2 140 ± 200 0.8 the Appendix A for further discussion). Such complexity
Al/Al 9 ± 2 3 ± 200 2.2
in the cold spray process makes it attractive to use finite
1266 V. Luzin et al. / Acta Materialia 59 (2011) 1259–1270

element methods for numerical simulation [41,42], rather graphically by fitting the deformed particles to an oblate
then analytical models. The task of establishing an accurate ellipsoid, which is the closest regular shape with a reason-
link between macroscopically available experimental data able fit. The deformation from a sphere to an ellipsoid of
and phenomena at the microscopic level may ultimately a given aspect ratio gives the average impact strain.
require the use of statistically meaningful averaging proce- Although there is generally strain localization at the parti-
dures difficult for practical applications. In the next section cle boundaries, for the purpose of estimating the average
an intermediate approach is described. In an attempt to residual stress it is the macroscopic particle strain that is
establish a link between the micro- and macro-scales we of interest, since strain localization occurs over a relatively
discuss the importance of some factors and the relative small volume of the particle. The final deformed state of
insignificance of others based on available experimental the particles in a cold spray coating is the result not just
data and easily attainable estimates. of the initial particle impact, but also of successive particle
impacts as the coating is built up. This is a similar situation
5.2. Interpretation of the result in terms of physical process to shot peening, where the substrate material is subject to
parameters multiple impacts. Using the calculated particle velocity
and impact strain, the impact duration can be estimated
While the fit of Tsui and Clyne’s empirical model to the by assuming the impacting particle decelerates linearly, as
residual stress data in Fig. 4 demonstrates the relative is done with the analysis of a Taylor impact test [47]. This
importance of kinetic aspects of the cold spray processes also gives the average strain rate. The average impact pres-
on residual stress formation and provides quantitative sure is calculated based on momentum transfer over the
measure of the accumulative result, it would be useful to calculated impact time, as done in Ref. [48].
understand why the residual stress is so much larger in Of interest are the short impact duration, 10–
the Cu coatings, namely, why rd is larger in Cu than in 30  109 s, and the high average strain rate of 107 s1
Al. Given the complexity of the cold spray process a finite (local strain rates can be even higher). The other interesting
element simulation is best able to capture details the point is that the impact strain is significantly larger in the
mechanics of residual stress accumulation. Several publica- Cu coating than in the Al coating. Papyrin et al. have
tions detail the use of finite element simulation to calculate observed that the impact strain in cold spray coatings of
the temperature and stress inside individual impacting par- different materials can be fit empirically to the ratio of
ticles [42,43], or a small number of successive impacting the kinetic impact energy to the dynamic flow stress [49],
particles [44,45], in order to gain an understanding of the and this makes sense when considering dissipation of the
cold spray bonding process. Surprisingly there has been impact energy through plastic deformation. This ratio is
no attempt to extract residual stress data from any of these higher in the case of Cu, so a larger impact strain is
simulations, even though they incorporate the details of expected.
impact pressure, stress accumulation using a rate sensitive The maximum impact pressure can be roughly estimated
material model, and elastic unloading, e.g. Ref. [46]. In based on consideration of momentum transfer, by assum-
principle it should be possible to calculate at least the aver- ing impact of a moving plate against a fixed, stationary
age residual stress in the coating and substrate respectively. plate of the same material and size [47]. For the impact
While a finite element simulation should be able to pre- speeds in Table 7 this gives maximum impact pressures of
dict residual stresses with reasonable accuracy, it is often dif- 12.7 GPa and 4.5 GPa for Cu and Al respectively, well
ficult to extract simple predictive relationships from the beyond the Hugoniot elastic limit (HEL) for both materi-
results. In the present work a physically based approach will als, of 0.6 GPa. As a result, the impact will propagate
be used instead, incorporating data that are either measured through the material in the form of a plastic shock wave.
or easily estimated. An analogy of the cold spray process to a It is important to note that this calculation assumes a pla-
Taylor impact test [47] will be used to estimate some process nar geometry. Since the impacting particle is roughly
parameters, combined with a model of residual stress accu- spherical during the impact process, attenuation of the
mulation in the shot peening process, since this process has plastic shock wave will occur as it traverses the particle
many similarities to cold spray. [49]. As a result the pressure will rapidly drop and it is unli-
The relevant process parameters are given in Table 7. kely that the entire particle will deform plastically. How-
The particle impact velocity was calculated as described ever, once a given particle (now incorporated into the
in Section 2.3. The impact strain was measured metallo- coating) is impacted by another particle, complete plastic

Table 7
Estimated impact parameters based on linear momentum transfer on impact.
Material Density Particle Impact speed True impact Impact Average strain Average impact Maximum impact shock
(kg m3) size (lm) (m s1) strain duration (s) rate (s1) pressure (MPa) pressure (MPa)
Al 2700 15 585 0.6 23  109 2.7  107 680 4520
Cu 8930 6 645 0.95 11  109 7.5  107 2020 12,730
V. Luzin et al. / Acta Materialia 59 (2011) 1259–1270 1267

deformation is likely to result. The average impact pressure into a product that describes in simple terms the elasto-
should better incorporate these effects, and the magnitude plastic state of the deformed material: a is the ratio of
of the average impact pressure for Al and Cu in Table 7 the strain hardening rate (tangent modulus) to the Young’s
are very close to the interface pressures calculated by Gru- modulus, and b is the ratio of the true plastic strain to the
jicic et al. for Al and Cu using a numerical simulation [50]. true elastic strain.
The deformation in cold spray involves very high aver- As can be seen from Eqs. (1) and (2) the impact stress is
age strain rates on the order of 107–108 s1, and local strain the result of several competing factors: spray kinetic condi-
rates as high as 109 s1 [49]. It is interesting to compare the tions (qV2), elastic material properties (E, m), and plastic
strain rate in cold spray to other plastic deformation pro- material properties (ab and rs).
cesses such as low plasticity burnishing (101 s1), deep cold The input parameters in Eqs. (1) and (2) are readily
rolling (102 s1), shot peening (104 s1), laser shock peening determined. The average impact pressure from Table 7 is
(106 s1), and combustion spray (1010 s1). In spite of a used for pmax in Eq. (2) in order to roughly account for
variation in strain rate of nine orders of magnitude, the plastic shock wave attenuation effects. With these points
surface compressive residual stresses in all cases depend in mind the following can be stated:
almost entirely on the material and not the process, e.g.
Refs. [51,52]. This is because the dynamic flow stress of – Cu is more dense than Al and Cu will have a higher
metals exceeds the static yield stress by a factor 61.5–2.0 impact pressure.
[53], due to limitations imposed by dynamic recovery at – The flow stress of Cu is higher than that of Al over a
temperatures at or above ambient. When considering the wide range of strain rates [57–59].
accumulation of surface residual stresses we consider it rea- – This higher impact pressure leads to a greater amount of
sonable to neglect the relative rate sensitivity of materials. plastic strain in the case of Cu, despite its higher
With these points in mind the model of residual stress dynamic flow stress.
accumulation in shot peening due to Li and co-workers is – Al shows minimal shock hardening compared to Cu [60].
appropriate [54], based on Hertzian contact and assuming
plastic deformation on loading–unloading developed for Table 8 gives the estimated residual stress from Eq. (1).
the case of the shot peening. The present work uses a mod- The estimated surface residual stresses for Al and Cu are
ified form [55,56] of the same formalism which is readily qualitatively correct compared to the measured values,
applied to the case of cold spray coatings. This approach and it is thought that this analysis should hold true for
assumes a material volume has been impacted with 100% other metals.
coverage, and this is analogous to the problem in cold While this is by no means an accurate quantitative pre-
spray of a particle deformed partly by its own impact diction, it provides a useful estimate of the residual stresses
energy, and partly by the impact of successive particles that when considering different coating materials, based on
form the coating. The assumption is that the final amount readily available material property data and some simple
of plastic deformation is more important than the deforma- calculations. Even if a more accurate estimate is possible
tion sequence. within this approach, the practical value of it might not
While it is not possible to make an exact analytical pre- be great until a degree of influence of some other factors
diction of the residual stress in the cold spray coatings, it is on the residual stress (such as dynamic and static recovery,
at least possible to rationalize the results. The analysis can rate sensitivity, Bauschinger effects, the role of particle size,
be reduced to the prediction of the maximum residual etc.) is clarified as discussed in the Appendix A.
stress at the surface [55]:
5.3. Coating quality and performance through mechanical
rmax ¼  ð0:333 þ 0:286abÞð1  abÞ½ð1  2abÞrs
properties
þ k  ab  pmax  ð1Þ
 1=5 The results of the deformation modelling are relevant to
2 5 4 2
pmax ¼ pE qV ð2Þ the porosity measurements in Table 4 and the Young’s
p 4 
modulus measurements in Table 5. It is clear that the phys-
where pmax is the maximum pressure calculated in the ical properties of Cu enable it to form a higher integrity
assumption of Hertzian contact, rs is the yield stress, q is coating than Al. Compared to Al, the higher density of
the density, V is the velocity of the particle upon impact, Cu enables a higher impact pressure, without a correspond-
E* is the equivalent modulus, E ¼ E=2ð1  m2 Þ, and k is ing increase in dynamic flow stress. This enables more plas-
a constant close to 1. Two parameters, a and b, are coupled tic deformation on impact in the Cu coatings, and the

Table 8
Estimated residual surface stresses with corresponding input parameters.
Material Yield strength (MPa) ab Average impact pressure (MPa) Estimated residual stress (MPa) Measured residual stress (MPa)
Al 30 0.01 680 12 9
Cu 60 0.04 2020 45 60
1268 V. Luzin et al. / Acta Materialia 59 (2011) 1259–1270

porosity and Young’s modulus measurements for the Cu were treated empirically using Tsui and Clyne’s progressive
coatings are closer to their bulk values than those of the deposition model to demonstrate that the residual stresses
Al coatings. Higher deformation within the coating should accumulate mainly by kinetic and not thermal effects, and
lead to better interparticle bonding arising from conformal these effects were characterized quantitatively. The inter-
metal/metal contact by break-up of the particle’s external pretation of the measured compressive residual stresses is
oxide layer. The relatively poor interparticle bonding in that they are determined almost entirely by the plastic
the Al coatings is likely why the Young’s modulus mea- deformation process (peening) of the particles of the spray
surement based on the indentation test gives results that material. An attempt has been made to explain the higher
are so different from the impulse excitation test, with a residual in-plane stress in the Cu coatings by using a phys-
rather large error. ically based model of residual stress accumulation in a shot
Another observation can be made on the basis of the peened surface. Calculations within this model in the case
comparison of our results with experimental data obtained of Al and Cu demonstrate that the balance between the
on the thermal spray coatings and published in the litera- elasto-plastic properties of the material and the spraying
ture. It has been demonstrated that, in the case of thermal conditions defines the resultant residual stress. It is believed
spray, there is a correlation between the coating’s Young’s this interpretation can be applied generally for a range of
modulus and residual (quenching) stress [10–12]. The same materials providing a close numerical estimate.
was also demonstrated for two phase coatings [16]. This is The higher plastic strain on impact of the Cu particles
not surprising considering the nature of the quenching results in higher residual stress in the Cu coatings and bet-
stress: under cooling, with a temperature drop DT quench , a ter compaction, ensuring the mechanical properties of the
splat experiences the thermal misfit Deth ¼ acoat DT quench , Cu coatings are closer to their equivalent bulk values than
where acoat is the thermal expansion coefficient of the coat- is the case with Al coatings.
ing material. If we consider the quenching temperature
drop to be from the melting point then, for any material, Appendix A
the thermal misfit is Deth  1–2%. Because for thermal
spray processes the thermal misfit is so high, it exceeds
the yield strain (for metals) or the brittle fracture limit (i) Although in the preceding arguments we neglected the
(for ceramics) by an order of magnitude since both of these strain rate dependence of the material flow stress, at
limits have the typical value of De0  0:1–0:2%. Only this such strain rates the effects of strain hardening, and
limited strain can be preserved as the elastic strain while strain-rate-dependent hardening coupled with heat
the rest of the total thermal misfit is accommodated transfer, can become important. The rate sensitivity
through inelastic mechanisms (microcracking, sliding, of Cu and Al at the strain rates found in cold spray
yielding, creeping, etc.), which eventually are reflected in impact have not been measured experimentally, and
the coating microstructure. Thus, the maximum quenching the accuracy of results based on extrapolating material
stress which a coating can withhold is rq ¼ Ecoat De0 , i.e. data one or several orders of magnitude in strain rate is
proportional to the Young’s modulus of the coating mate- questionable. On the other hand, in many simulations
rial. In case of cold spray, however, Eq. (1) suggests a dif- the Johnson–Cook or Zerilli–Armstrong material
ferent functional relationship between stress and Young’s models [61] are used up to strain rates of 109 s1, but
modulus than a simple linear proportion. This is the man- none of the data used for verification the model param-
ifestation of the different stress generation mechanism by eters extends beyond 105–106 s1 (achieved in the laser
complex plastic deformation accumulated throughout the peening process). Available data on strain rate sensi-
full loading–unloading history of individual splats. The tivity show that the flow stress of Cu is roughly double
result of that must be connected not only to elastic (e.g. that of Al, at any strain rate lower then 105 s1 [57–59].
Young’s modulus) but also plastic (e.g. hardening rate) (ii) Some of the initial impact stress is relaxed either dur-
properties of the material. For instance, for different alu- ing or immediately after the impact process through
minium alloys cold sprayed in the same conditions, the cor- recovery and recrystallization. Dynamic recovery
relation between the peening stress and UTS is more and recrystallization are expected to lead to stress
plausible, while the Young’s modulus values for different relaxation, probably more so in Al coatings than in
coatings are expected to cluster in a very narrow band. the Cu coatings. Recovery processes are poorly
Experimental verification of this prediction in the case of understood and a quantitative description requires
cold spray coatings will require a larger series of samples knowledge of the dislocation density and spatial
sprayed at different conditions. arrangement, which is beyond the scope of the pres-
ent work (see, e.g. [62]). Roughly, one can estimate
6. Conclusions the relative operation of recovery processes in the
two coatings as follows:
The residual stresses profiles have been measured in Cu 1. Because the thermal conductivity of Cu is higher
and Al cold spray coatings in neutron diffraction stress than that of Al, the impact conditions are expected
experiments. The experimental residual stress distributions to be less adiabatic in the case of Cu.
V. Luzin et al. / Acta Materialia 59 (2011) 1259–1270 1269

2. Under the impact conditions the average tempera- [13] Buchmann M, Gadow R, Tabellion J. Mater Sci Eng A 2000;288:154.
ture of the Cu and Al as a fraction of the melting [14] Kesler O, Matejicek J, Sampath S, Suresh S, Gnaeupel-Herold T,
Brand PC, Prask HJ. Mater Sci Eng A 1998;257:215.
temperature are estimated as 0.35 and 0.44 respec- [15] Matejicek J, Sampath S, Brand PC, Prask HJ. Acta Mater
tively but during impact it may rise to as much as 1999;47:607.
0.9 locally, as suggested by numerical simulation [16] Luzin V, Matejicek J, Gnäupel-Herold T. Mater Sci Forum
of single particle impacts [41,42,50]. This sudden 2010;50:652.
temperature rise is short, 107 s, and due to heat [17] Welzel U, Ligot J, Lamparter P, Vermeulen AC, Mittemeijer EJ. J
Appl Crystallogr 2005;38:1.
conduction and dissipation this does not in [18] Matejicek J, Sampath S. Acta Mater 2001;49:1993.
general lead to melting in Cu and Al coatings, as [19] Genzel CH. J Neutr Res 2004;12:233.
confirmed by numerous detailed microscopic [20] Tsui YC, Clyne TW. Thin Solid Films 1997;306:23.
examinations, e.g. Refs. [40,63,64]. The nozzle [21] Tsui YC, Clyne TW. Thin Solid Films 1997;306:34.
temperatures given in Table 1 are a measure of [22] Tsui YC, Clyne TW. Thin Solid Films 1997;306:52.
[23] Karthikeyan J. Adv Mater Process 2006;164:66.
the gas temperature immediately before entering [24] Gabel H. Adv Mater Process 2004;162:47.
the nozzle throat, where some cooling takes place. [25] Li W-Y, Liao H, Wang H-T, Li C-J, Zhang G, Coddet C. Appl Surf
The particle temperature will be lower than the gas Sci 2006;253:708.
temperature, though there is general heating [26] Irissou E, Legoux J-G, Ryabinin A, Jodoin B, Moreau C. J Therm
resulting from plastic dissipation of the impact Spray Technol 2008;17:495.
[27] Dykhuizen RC, Smith MF. J Therm Spray Technol 1998;7:205.
energy. It was possible to hold the samples imme- [28] Tapphorn R, Gabel H. System and process for solid state deposition
diately after spraying, which suggests that the and consolidation of high velocity powder particles using thermal
average temperature rise was not very high. In plastic deformation. USA: Office USPaT; 2005.
the absence of detailed information on the thermal [29] Henderson CB. AIAA J 1976;14:707.
history of the material during the cold spray pro- [30] http://www.ansto.gov.au/research/bragg_institute/facilities/instru-
ments/echidna.
cess, it can simply be stated that Al will likely be [31] Larson AC, Von Dreele RB. Report LAUR 86–748: Los Alamos
affected more by thermally activated recovery pro- National Laboratory; 2004.
cesses than Cu during cold spray. [32] Toby B. J Appl Crystallogr 2001;34:210.
[33] http://www.ncnr.nist.gov/instruments/darts/.
(iii) A variety of experimental data shows that the flow [34] Kröner E. Z Phys A: Hadrons Nucl 1958;151:504.
stress begins to increase rapidly beyond a strain rate [35] Totemeier TC, Wright JK. Surf Coat Technol 2006;200:3955.
of 105 s1 due to a change from thermal activation [36] Alcalá J, Gaudette F, Suresh S, Sampath S. Mater Sci Eng A
to drag-controlled motion of dislocations (e.g. Ref. 2001;316:1.
[57]); however, this point has been debated (e.g. Refs. [37] Suresh S, Giannakopoulos AE, Alcalá J. Acta Mater 1997;45:1307.
[38] Choi WB, Prchlik L, Sampath S, Gouldstone A. J Therm Spray
[47] and [58]). Technol 2009;18:58.
(iv) The Bauschinger effect can be important because of [39] Zou Y, Qin W, Irissou E, Legoux J-G, Yue S, Szpunar JA. Scripta
large plastic deformation in cold spray, and impor- Mater 2009;61:899.
tance of yielding upon unloading process. [40] Borchers C, Gartner F, Stoltenhoff T, Assadi H, Kreye H. J Appl
(v) The role of particle size on the resultant stress is Phys 2003;93:10064.
[41] Grujicic M, Zhao CL, DeRosset WS, Helfritch D. Mater Des
unclear. It can be speculated, however, that this role 2004;25:681.
might be significant. At a minimum it can be said that [42] Assadi H, Gärtner F, Stoltenhoff T, Kreye H. Acta Mater
particle size has a significant impact on the defect 2003;51:4379.
microstructure and, together with particle velocity, [43] Schmidt T, Gärtner F, Assadi H, Kreye H. Acta Mater 2006;54:729.
determines critical conditions for particle adhesion. [44] Yin S, Wang X-F, Li W-Y, Xu B-P. J Therm Spray Technol
2009;18:686.
[45] Schmidt T, Assadi H, Gärtner F, Richter H, Stoltenhoff T, Kreye H,
Klassen T. J Therm Spray Technol 2009;18:794.
References [46] Gyuyeol B, Yuming X, Kumar S, Kicheol K, Changhee L. Acta
Mater 2008;56:4858.
[1] Papyrin A. Adv Mater Process 2001;159:49. [47] Meyers MA. Dynamic behavior of materials. New York: John Wiley
[2] McGrann RTR, Greving DJ, Shadley JR, Rybicki EF, Kruecke TL, & Sons, Inc.; 1994. p. 668.
Bodger BE. Surf Coat Technol 1998;59:108–9. [48] Van Steenkiste TH, Smith JR, Teets RE. Surf Coat Technol
[3] Choi WB, Li L, Luzin V, Neiser R, Gnaeupel-Herold T, Prask HJ, 2002;154:237.
Sampath S, Gouldstone A. Acta Mater 2007;55:857. [49] Papyrin PA, Kosarev PV, Klinkov DS, Alkimov PA, Fomin PV. Cold
[4] Luzin V, Valarezo A, Sampath S. Mater Sci Forum 2008;315:571–2. spray technology. Oxford: Elsevier; 2007.
[5] Stoney GG. Proc Roy Soc Lond Ser A 1909;82:172. [50] Grujicic M, Saylor JR, Beasley DE, DeRosset WS, Helfritch D. Appl
[6] Brenner A, Senderoff S. J Res Nat Bur Stand 1949;42:105. Surf Sci 2003;219:211.
[7] Zhang XC, Xu BS, Wang HD, Wu YX. J Appl Phys 2005;98:5. [51] Gabb TP, Telesman J, Kantzos P, Prevey PS. Adv Mater Process
[8] Gill SC, Clyne TW. Thin Solid Films 1994;250:172. 2002;160:69.
[9] Kuroda S, Fukushima T, Kitahara S. Thin Solid Films 1988;164:157. [52] Rodopoulos CA, Kermanidis AT, Statnikov E, Vityazev V, Korolkov
[10] Matejicek J, Sampath S. Acta Mater 2003;51:863. O. J Mater Eng Perform 2007;16:30.
[11] Matejicek J, Sampath S, Gilmore D, Neiser R. Acta Mater [53] Kanel GI, Razorenov SV, Fortov VE. Shock-wave phenomena and
2003;51:873. the properties of condensed matter. New York: Springer-Verlag;
[12] Kuroda S, Clyne TW. Thin Solid Films 1991;200:49. 2004.
1270 V. Luzin et al. / Acta Materialia 59 (2011) 1259–1270

[54] Li JK, Mei Y, Duo W, Renzhi W. Mater Sci Eng A 1991;147:167. [60] Gray III GT. Influence of shock wave deformation on the structure/
[55] Ogawa K, Asano T. Mater Sci Res Int 2000;6:55. property behavior of materials. In: Asay JR, Shahinpoor M, editors.
[56] Franchim AS, Campos VSd, Travessa DN, Neto CdM. Mater Des High-pressure shock compression of solids. New York: Springer-
2009;30:1556. Verlag; 1993. p. 187.
[57] Campbell JD. Mater Sci Eng 1973;12:3. [61] Amarchinta HK, Grandhi RV, Langer K, Stargel DS. Modell Simul
[58] Follansbee PS, Regazzoni G, Kocks UF. The transition to drag- Mater Sci Eng 2009:015010.
controlled deformation in copper at high strain rates. Mechanical [62] Humphreys FJ, Hatherly M. Recovery after deformation. Recrystal-
properties at high rates of strain. Bristol, UK: IOP; 1984. lization and related annealing phenomena. Oxford: Elsevier; 2004. p.
[59] Follansbee PS. High strain-rate deformation of FCC metals and 169.
alloys. In: Murr LE, Staudhammer KP, Meyers MA, editors. [63] King P, Zahiri S, Jahedi M. Metall Mater Trans A 2009;40:2115.
Metallurgical applications of shock-wave and high-strain-rate phe- [64] Balani K, Agarwal A, Seal S, Karthikeyan J. J Scripta Mater
nomena. New York: Marcel Dekker, Inc.; 1986. p. 451. 2005;53:845.

You might also like