Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

1.

Sample preparation

Nickel powders with grain sizes of 3 nm, 20 nm, 40 nm, 70 nm, 100 nm and 200 nm

were purchased commercially from various companies including nanoComposix,

Nanomaterial store, and US nano. Grain sizes of 8 nm and 12 nm were made by heating

3 nm-sized nickel in a reaction kettle at 80 0C in ethanol medium for 24 h and at 150 0C

in isopropanol medium for 12 h, respectively. The size distribution of nickel grains

(Extended Data Fig. 2) was confirmed with TEM, Scanning Electron Microscopy

(SEM) (Extended Data Fig. 3) and XRD measurements. High-resolution TEM

examinations confirm that the raw samples contain very few defects. We also examined

potential chemical contamination of our samples. Some content of oxygen was detected

in sub-10 nm nickel. However, our mechanical characterization is based on the

characteristic XRD from the crystalline sample, and the effect of surface impurities on

the results is relatively small as long as the XRD peaks do not overlap. When needed,

low-pressure annealing treatment of the sample in hydrogen or vacuum environments

was used to remove oxygen contamination, resulting in a clean XRD signal of pure

nickel (Fig. 1a).

2. High-pressure radial XRD measurements

X-ray transparent boron-kapton gaskets with a thickness of 80 μm and a sample

chamber of 60 μm diameter was mounted on top of 300 μm culet DAC with the

support of clay. Ni powders were pre-compacted before being loaded into the sample

chamber. A small piece of Pt foil was placed on top of the Ni sample to serve as a

1
pressure calibrant using the Birch-Murnaghan equation of state of Pt31. No pressure

medium was used in order to maximize the differential stress between axial and radial

directions. Monochromatic synchrotron x-rays with energies of 30 keV (for 3 nm Ni

sample) or 25 keV (for other Ni samples) were used to conduct the XRD experiments

at beamline 12.2.2 of the Advanced Light Source, Lawrence Berkeley National Lab.

The experimental setup is shown in Extended Data Fig. 1. A Mar345 detector was used

to collect the diffraction images. A CeO2 or LaB6 standard were used to calibrate the

instrumental parameters. Raw sample characterizations and some make-up

measurements were made at 15U1 of SSRF.

3. Deviatoric strain theory and data analysis method

The stress state in a diamond anvil cell can be resolved into a hydrostatic part and

a nonhydrostatic part associated with the maximum stress component in the sample

along compression direction (σ11), and two minimum stress components in the plane of

the gasket with a cylindrical symmetry about the compression axis (σ22=σ33) (Extended

Data Fig. 1). The difference between these maximum and minimum stress components

is called the differential stress16, t, which induces deviatoric strain in the sample. The

differential stress provides a lower-bound estimate of a material’s shear strength τs and

yield strength σy, based on the von Mises yield criterion

t = 11 −  33  2 s =  y (4)

The produced deviatoric strain Qhkl can be given by the measured d-spacings from

different diffraction directions16:

2
Qhkl = (d00 − d900 ) / (d00 + 2d900 ) (5)

in which d 00 and d900 are the d-spacings measured from  = 00 and  = 900 ( the

angle between the diffraction vector and the load axis), respectively. The differential

stress can be calculated from the deviatoric strain by using the following equation16:

3Qhkl
thkl = (6)
 ( 2GR ( hkl ) )  + (1 −  ) ( 2GV )
−1 −1

 

where GR ( hkl ) and GV are the shear modulus of the aggregates under the Reuss

(iso-stress) condition and Voigt (iso-strain) condition, respectively;  is the factor to

determine the relative weight of Reuss and Voigt conditions. As an approximation, the

macroscopic stress of aggregates can be treated by a Hill average which is the

arithmetic mean of the Reuss (isostress) and Voigt (isostrain) bounds32 (  =0.5).

For a cubic system, GR ( hkl ) and GV can be described by the following

equations16, 33:

5 ( S11 − S12 ) S44


( 2GV )
−1
= (7)
2 3 ( S11 − S12 ) + S44 

 S 
( 2GR ) = S11 − S12 − 3  S11 − S12 − 44   ( hkl )
−1
(8)
 2 

h 2 k 2 + k 2l 2 + h 2l 2
 ( hkl ) = (9)
( h2 + k 2 + l 2 )
2

where Sij are the single crystal elastic compliances and can be inverted from the

measured elastic stiffness constants Cij of materials, where Cij = Cij (0) + Cijʹ• P. In

our calculations, Cij values of nickel at zero pressure, Cij(0), and its pressure

derivative, Cijʹ, are cited from Ref. 34.

3
Each diffraction image was processed into a file containing 72 spectra over a 5º

azimuthal step by using the program Fit2d35. The diffraction pattern was refined with

the Rietveld method using the MAUD software17. Qhkl of each lattice plane was

obtained by fitting azimuthal variations in lattice spacing using the “Radial diffraction

in the DAC” model17. Differential stress and yield strength were then obtained

according to Eqs. (4)-(9). The uncertainties of calculated differential stress (Fig. 1b

and Extended Data Fig. 4) were calculated with Eqs. (6)-(9) based on the uncertainties

of refined Qhkl . The error bar of t is calculated by averaging the error of thkl of peaks

(111), (200) and (220). The approximation of  may cause additional uncertainty in

the calculation of differential stress, however, that part is difficult to estimate. We use

the differential stress-lattice strain curve to compare with the traditional stress-strain

curve during plastic deformation (Fig. 1b and Extended Data Fig. 4). Lattice strain ε is

calculated from the lattice parameter change:  =


( a0 − a ) , where a0 and a are lattice
a0

parameters of nickel at zero pressure and at a given applied stress, respectively.

4. Elasto–ViscoPlastic Self-Consistent (EVPSC) modeling

In order to remove the potential influence of pressure on our calculated yield

strength such that a more consistent comparison of rDAC XRD results with traditional

tests could be made, we used the EVPSC code18 to model our experimental data. The

EVPSC model18, 36 is a self-consistent method to simulate polycrystal plasticity by

treating each grain in a polycrystal as an inclusion in a homogeneous but anisotropic

medium. The properties of the medium are determined by the average of all the

4
inclusions. At each deformation step, the inclusion and medium interact and the

macroscopic elasto-plastic properties are updated iteratively until the average strain

and stress of all the inclusions equal the macroscopic strain and stress. EVPSC

accounts for both the elastic and viscoplastic behaviors of materials including the

macroscopic stress and strain of the material, intragranular stress and strain, and the

strength and activity of deformation modes as well as strain and pressure hardening,

deviatoric strain and texture evolution18, 36.

We adjust the strain, compression rate (strain rate) and critical resolved shear

stress (CRSS) values of possible slip systems and twin modes to reproduce

experimental deviatoric strain and texture evolution as a function of pressure. As

shown in Extended Data Figs. 5a-5b, both the simulated Qhkl, texture type and texture

intensity of nickel with different grain sizes match quite well with our experimental

results. The large deviatoric strain anisotropy shown in Qhkl of nickel is due to plastic

anisotropy37-39 and is well reproduced by our models. For example, modeling results

show that a 64% {100}<011> slip, 11%{111}<1-10> slip and 25% {111}<11-2>

twinning are active in 3 nm nickel at a pressure of about 32 GPa (17% plastic strain )

whereas a 19% {100}<011> slip, 63%{111}<1-10> slip and 18% {111}<11-2>

twinning are active in 20 nm nickel at a pressure of about 32 GPa (14% plastic strain).

This shows that the activities of different deformation modes accounts for the large

difference of strength anisotropy in small grain sizes in Extended Data Fig. 4.

Once high-pressure simulations match the experimental data, we re-run the

simulations at constant ambient pressure i.e. the unit cell volume of nickel to is kept

5
constant (equivalent to traditional compressive or tensile test at ambient pressure)

during compression. The extrapolated zero pressure differential stress of nickel vs

strain and yield strength of nickel during ambient pressure deformation can thus be

obtained (Extended Data Fig. 5c). Note that this extrapolation does not consider GB

sliding and only eliminates the effects of pressure. Both the deviatoric stress

calculation from Qhkl (Fig. 1b) and stress strain curves from the EVPSC simulations

(Extended Data Fig. 5c) give a similar size strengthening trend.

5. Computational simulations for competition between grain boundary


deformation and dislocation activity-induced deformation under compression

The inverse Hall-Petch relation in nanograined metals comes from dominance of

grain boundary deformation mechanisms such as, grain boundary migration and grain

boundary sliding5, 40. Note that grain boundary migration can be characterized in the

grain boundary diffusion model41 and grain boundary sliding behavior could be

explored in the thermally activated shear transformation model42, 43. Inspired by this

perspective, we propose a theoretical model in which the grain boundary migration

and grain boundary sliding are both involved. Thus, the shear strain rates from grain

boundary deformation can be expressed as

 = GBf +  GBt (10)

where  GBf is the shear strain rate related to the grain boundary diffusion behavior and

 GBt is the shear strain rate associated with thermal-activated shear behavior in grain

boundaries44. Due to the fact that grain boundaries have properties akin to those of the

6
amorphous state, diffusion mass transport induces a shear strain rate at grain

boundaries which can be written as

 (2d −  ) Qgbf
 GBf =Dgbf
0
exp( − ) (11)
kBT d3 kBT

0
in which Dgbf is the self-diffusion coefficient within grain boundary,  the atomic

volume, k B the Boltzmann constant, T the temperature, d the grain size,  the

thickness of grain boundary, and  the applied stress, Qgbf the activation energy for

grain boundary self-diffusion where Qgbf = F . Here, F is the Helmholtz free

energy. Thermal activated-grain boundary shear originates from the many

independent atomic shear events, thus the thermal-induced grain boundary shear

strain could be written as

2 v Q
 GBt = v sinh( ) exp(− v ) (12)
d kBT kBT

in which v is the vibration frequency of atoms, v = b3 the activation volume. b is

atomic diameter. Qv = F is the thermal activation energy related to the Gibbs free

energy G = F − v .

When the nanograined metals are subjected to high pressure, the activation energy

for grain boundary self-diffusion Qgbf and the thermal activation energy Qv

become sensitive to high pressure because the Helmholtz free energy depends on

pressure. Therefore, one can get

Qgbf = Qv = F + PV (13)

in which P is the pressure, V is the change of volume of grain boundaries. Note

that the bulk strain at grain boundaries can be defined as  e =V / Vgb , and one can get

7
the effective stress (pressure) as  e = K gb e . Here, Vgb is the volume of grain

boundaries and K gb is the bulk modulus of grain boundaries, which is assumed to be

K / 2 . K is the bulk modulus of material. Then, the pressure related term in Eq. (13)

can be rewritten as

PV = PV  e = P 2Vgb / K gb (14)

Combining with Eqs. (10)-(14), one can derive the shear strain rate of nanograined

metals under high pressure, given as

 (2d −  ) F + P Vgb / K gb
2

 =D 0
gbf 3
exp(− )
kBT d kBT
(15)
2 v F + P Vgb / K gb
2

+ v sinh( ) exp( − )
d kBT kBT

Finally, the critical stress for activating the grain boundary deformation  gb can be

determined based on Eq. (15).

The critical stress for the emission of a perfect dislocation can be described as
Gb
f = . The critical stress to activate a partial dislocation can be given as
d
Gbp 
p = + (1 −  ) , where bp is the magnitude of the Burgers vector of the partial
3d Gbp

dislocation,  the ratio of equilibrium stacking fault width to grain size and  the

stacking fault energy. By comparing the critical stresses (  f and  p ) for activating

lattice dislocations and partial dislocations with the critical stress (  gb ) for activating

grain boundary deformation, the dominant deformation mechanism can be determined

when changing the grain size from coarse grains to nano grains. In simulations,

parameters are adopted as below: 0


Dgbf =3.0 10−20 m2 /s ,  = 1.3 10−29 m3 ,

8
v = 1.0 1013 /s ,  = 1.0 10−3 /s , F = 97.5 kJ/mol, K = 180 GPa, G = 76 GPa,  = 90

mJ/m2, b = 0.256 nm (Refs. 41 and 42).

6. Molecular dynamics (MD) simulation

MD simulations were conducted to study the deformation process of nanograined

Ni employing LAMMPS45. 512 nanograins of Ni with an average grain size of 3 nm

were packed in the simulation box using the Voronoi tessellation method, and the

total number of atoms was 3,173,000. Periodic boundary conditions were applied to

the three directions. A highly optimized embedded-atom-method interatomic potential

was used for Ni46. The radial DAC experiment was simulated by exerting uniaxial

compression along the Z direction while keeping the lengths along the x and y

directions fixed. The atomic structures were analyzed and visualized with the Ovito

software47. Dislocation analysis23 was performed to identify the nature of the

dislocations, rendered with different colors.

7. TEM measurements

TEM measurements were performed on a 300 kV aberration-corrected transmission

electron microscope Titan G2 60–300 equipped with high angle annular dark field

(HAADF) and bright field (BF) detectors. Thin pressurized Ni foils for TEM were

prepared by using a Zeiss Auriga focused ion beam (FIB) system. To reduce possible

artifacts during ion milling of the specimen, a protective Pt layer with a thickness of

about 1 μm was deposited on the candidate region.

You might also like