Structural properties of Silicon-Germanium and Germanium-Silicon Core-ShellNanowires

Conn O’Rourke, Shereif Y. Mujahed, and Chathurangi KumarasingheLondon Centre for Nanotechnology, University College London, 17-19 Gordon St, London, WC1H 0AH and

International Centre for Materials Nanoarchitectonics (MANA),National Institute for Materials Science (NIMS), 1-1 Namiki, Tsukuba, Ibaraki 305-0044, Japan

Tsuyoshi MiyazakiInternational Centre for Materials Nanoarchitectonics (MANA),

National Institute for Materials Science (NIMS), 1-1 Namiki, Tsukuba, Ibaraki 305-0044, Japan

David R. Bowler∗

London Centre for Nanotechnology, University College London, 17-19 Gordon St, London, WC1H 0AHInternational Centre for Materials Nanoarchitectonics (MANA),

National Institute for Materials Science (NIMS),1-1 Namiki, Tsukuba, Ibaraki 305-0044, Japan and

Department of Physics & Astronomy, University College London, Gower St, London, WC1E 6BT(Dated: September 7, 2018)

Core-shell nanowires made of Si and Ge can be grown experimentally with excellent control fordifferent sizes of both core and shell. We have studied the structural properties of Si/Ge andGe/Si core-shell nanowires aligned along the [110] direction, with diameters up to 10.2 nm andvarying core to shell ratios, using linear scaling Density Functional Theory (DFT). We show thatVegard’s law, which is often used to predict the axial lattice constant, can lead to an error of up to1%, underlining the need for a detailed ab initio atomistic treatment of the nanowire structure. Weanalyse the character of the intrinsic strain distribution and show that, regardless of the compositionor bond direction, the Si core or shell always expands. In contrast, the strain patterns in the Ge shellor core are highly sensitive to the location, composition and bond direction. The highest strains arefound at heterojunction interfaces and the surfaces of the nanowires. This detailed understanding ofthe atomistic structure and strain paves the way for studies of the electronic properties of core-shellnanowires and investigations of doping and structure defects.

I. INTRODUCTION

Scaling down the size of the current generation ofelectronic devices has led to an increased interest insemiconductor nanostructures, such as nanowires andnanotubes1–5. Quantum size effects and high surface tovolume ratios in these structures as a result of one ormore reduced dimensions can lead to highly tunable andunique electronic, optical and transport properties, whichhave the potential to be exploited in next generation elec-tronic devices6–8. Recently, Si/Ge core-shell nanowireshave been studied extensively, both experimentally andtheoretically since they are promising candidates for suchapplications with the valence band offset between Ge andSi offering a unique opportunity to control spacial carrierconfinement and carrier transport6,7,9. In this work, weuse linear scaling DFT to study the structural and strainproperties of Si/Ge and Ge/Si core-shell nanowires, asa function of nanowire composition and diameter, from∼5 nm to ∼10 nm.

Pure Si and Ge nanowires are typically grown by chem-ical vapour deposition within the VLS method10, and thisapproach affords a significant amount of control duringthe deposition process; the smallest nanowires grown bythis method vary between 1nm and 7nm in diameter11,12.As well as pure nanowires, it is possible to create core-

shell heterostructures13 using simple CVD after the VLSgrowth of a nanowire; these can be created with a Sicore and Ge shell (and vice-versa), simply by changingthe growth conditions and reactants during the depo-sition process7,13. Shells as thin as 2–3nm have beengrown14,15, though on cores which are a little larger thanthose studied here (15–20nm).

VLS grown Si nanowires are known to form essentiallyexclusively along the 〈110〉 direction11,12 for diametersup to 10 nm, exposing the (100) and (111) surfaces11

(beyond this diameter, they transition to a 〈111〉 direc-tion, with shapes that tend towards round; ). It has alsobeen shown that growth of shells on these small diame-ter nanowires leads to smooth, dislocation-free shells16.We have therefore modelled the core-shell nanowire sys-tem for diameters up to this cross-over point, using themodel depicted in Fig. 1, for both Si-core and Ge-corenanowires. In all cases, the surfaces have been passivatedusing hydrogen atoms.

Strain has a significant impact on the structural stabil-ity as well as electronic properties of nanostructures17–19.The bonds in Si/Ge core-shell nanowires are intrinsi-cally strained, primarily due to the lattice mismatch ofaround 4% between Si and Ge, though there are alsostrains due to surface reconstructions. The nature ofthese anisotropic strain patterns will be determined by

arX

iv:1

805.

0569

4v2

[co

nd-m

at.m

trl-

sci]

6 S

ep 2

018

2

FIG. 1. Si core, Ge shell nanowire model illustrating exposedsurfaces and growth direction, along with a side view of thepassivated 2× 1 surface reconstruction of the (100) surface ofthe SiGe-NW.

several factors including shape and composition of theNWs, characteristics of the heterostructure interface, theamount of lattice mismatch and the elastic parameters ofthe constituent materials20–22. Detailed analysis of thelocal strain distribution will enable us to identify highlycompressed and extended areas in the nanowires, wheredislocations and other defects are most likely form to re-lieve the strain. This will also inform studies of wheredopants are stable in such nanowires, since it is knownthat certain dopants such as As tend to be more stablein areas under tensile strain, while others such as P areknown to be stable under compressive strain23.

Strained areas, particulary near the surfaces, can havea significant effect on chemical reactivity of nanowiresbecause strain can lift degeneracy and cause bandsplitting24,25, changing the electronic structure. Thiscan alter reaction rates or allow reactions that wouldotherwise not proceed. Precise knowledge of the strainpatterns will help to identify local changes in electronicstructure and hence determine how strain will affect re-action rates near the surfaces. Further, it will help topredict spatial variation of electron transport and opticalproperties in nanowires. Even though the intrinsic strainhas been demonstrated to critically affect both structuraland electronic properties, it has not been given much at-tention in the literature, especially when it comes to rela-tively large nanowires with diameters over 5 nm (previousstudies have maximum sizes of 4.7 nm20).

The axial lattice parameter, which is sensitive to thechanging of the NW composition, is a key input to anycore-shell nanowire atomistic or electronic structure com-putation, and has a significant effect on the structure andhence electronic properties of the nanowire. Determiningthe optimum axial lattice parameter using first principlemethods can be time consuming, so an estimate based onVegard’s law26,27 is often used as a first approximation,though is not always accurate28. Vegard’s law determinesthe lattice parameter of a solid solution based on lat-tice parameters of the pure components and their relativeconcentrations in the solution29. The law is purely empir-ical, and was first derived in the context of solid ionic so-lutions. Deviations from the linear behaviour assumed inVegard’s law are often observed for other materials29,30.

In situations where the lattice parameter of a solid so-lution is known experimentally (e.g. through diffractiondata), Vegard’s law is often used effectively in reverse, toestimate the relative compositions31. It has been shownthat the lattice constant of Si1−xGex alloy nanowires canbe well predicted by means of Vegards law32, in line withbulk and thin film alloys. Here, we use linear scalingDFT to explore the applicability of Vegard’s law esti-mates to Si/Ge core-shell nanowires, even though theseheterostructures do not fall within the area where the lawwas first derived. This analysis is needed for core-shellnanowires, particularly due to highly inhomogeneous andanistropic strain distribution generated as a result of non-statistical distribution of Si and Ge atoms; however, it islacking in the literature, particularly for nanowires withdiameters over 5 nm28,33.

Ab initio approaches based on DFT are commonlyused in studiying these structural factors of nanostuc-tures. Due to computational demands, the overwhelm-ing majority of the work based on the DFT frameworkhas examined NWs with diameters of around 5 nm orless19,33–39. Given that the dimensions of experimentallystudied core-shell nanowires are typically of the order oftens of nanometers, it is clear that significantly largersystems must be examined theoretically. Recent imple-mentations of DFT methods which scale linearly withsystem size40–44 have made accurate modeling of suchlarger scale systems possible with reasonable computa-tional cost, thereby affording a much better insight intothe properties of core-shell nanowires of physically real-istic size.

In this work, we study the relationship between struc-tural properties and overall composition of Si/Ge andGe/Si core-shell nanowires with diameters in the range4.9—10.2 nm using the linear scaling DFT code Con-quest. We explore the accuracy of Vegard’s law in de-termining the axial lattice constant, and investigate theintrinsic strain patterns of the core-shell nanowires, par-ticularly how these change with core to shell ratio, core-shell composition and diameter.

II. APPROACH

All the calculations of nanowire structure used thelinear scaling DFT code, Conquest42, using the PBEGGA functional45. Conquest is a linear-scaling, orO(N) DFT code with the capability to perform first-principles DFT calculations on systems of up to millionsof atoms40,46. Since the details of implementation ofConquest have been discussed elsewhere40–42, we sum-marize only the main principles needed to explain thecurrent approach.

It is well known that the DFT ground state can beobtained by minimising the total energy with respect tothe Kohn-Sham (KS) density matrix ρ(r, r′), which is

3

formally defined as,

ρ(r, r′) =∑n

fnΨn(r)Ψn(r′), (1)

where Ψn and fn are the nth KS orbital and its occupa-tion number, respectively.

In Conquest, ρ(r, r′) is represented in terms of local-ized orbitals φ centered on the atoms, known as ‘supportfunctions’:

ρ(r, r′) =∑iα,jβ

φiα(r)Kiα,jβφjβ(r′). (2)

where i(j) indicates an atom and α(β) runs over the sup-port functions on the ith(jth) atom. While the supportfunctions themselves can be represented in terms of basisfunctions, we use a one-to-one mapping between supportfunctions and pseudo-atomic orbitals (PAOs)47. For allour calculations we have employed a single-ζ plus polari-sation orbital basis of PAOs. This basis has been chosento give a balance between accuracy and speed, and is thelargest basis that can be easily used with linear scaling.The PAO cutoffs are chosen to give an optimised bulklattice paramter within 1% of the experimental valuesfor both Si and Ge respectively.

The coefficients Kiα,jβ are the density matrix elementsin the basis of support functions. In Conquest, the den-sity matrix (K) can either be calculated by the conven-tional direct diagonalization with O(N3) scaling or byusing the density matrix minimization method proposedby Li, Nunes and Vanderbilt (LNV)48 with O(N) scaling.In the LNV method, which is used for our calculations,K is expressed in terms of an auxiliary density matrix Lby the matrix relation:

K = 3LSL− 2LSLSL (3)

where Siα,jβ = 〈φiα | φjβ〉 is the overlap matrix of sup-port functions. To achieve linear scaling with the LNVmethod, a spatial cut-off RL must be imposed on theL-matrix so that its value is zero when the distance be-tween the centres of the support-functions exceeds RL.Imposing this spatial cutoff is justified by the the near-sightedness of electronic matter49 and needed to ensureO(N) scaling. The RL value is a compromise betweenthe accuracy and the computational cost, with largerRL leading to better accuracy at a higher computationalcost. For all our calculations, a spatial cut-off RL = 16a0was selected, at which O(N) forces in the system wereconverged to within the force tolerance of the exact di-agonalisation results for Si and Ge surfaces.50

Numerical integration of PAOs in space is required toform the local part of the Hamiltonian matrix in Con-quest. This integration takes place on an integrationgrid and its spacing will contribute to determining theoverall accuracy of the calculations. The accuracy can beimproved by using a finer grid, i.e. increasing the integra-tion grid cutoff, however this leads to a rapid increase in

the computational costs. For the inputs including pseu-dopotentials and basis functions used in our calculations,an integration grid cut-off of 100 Ha has been identifiedas sufficient for the DFT energy to converge.

The nanowires present an interesting challenge forstructural optimisation: the nanowire is constrainedalong the [110] axial direction, but is free along the radialdirection. We perform a two-stage relaxation to find theoptimal axial lattice parameter: a value for the lattice pa-rameter is chosen, and we perform a structural relaxationof the nanowire. This process is repeated for different val-ues of the axial lattice parameter, alat, to find the lowestenergy value. The simulation cell includes ∼13A vacuumin lateral x- and y-directions to avoid any interactionsbetween the images of neighboring nanowires from pe-riodicity. Structural optimisations have been performedusing the FIRE algorithm51, until force components oneach atom were less than 0.0005 Ha/a0.

Our intention is to examine how the relative core-shellthicknesses will affect the structural properties of thesenanowires, and as such we have performed calculationsvarying independently the shell and core size for both Si-core Ge-shell (SiGe-NWs) and Ge-core Si-shell nanowires(GeSi-NWs). The shape of the core is set by the free ener-gies of the surfaces11,12, and we have chosen to maintainthe shape in the shell for simplicity. Cross-sections of themodels examined for the SiGe-NW variants can be seenin Fig. 2, with the same motifs used for the GeSi-NWs.The smallest of our models is approximately 4.9 nm in di-ameter, containing 612 atoms, and the largest is approx-imately 10.2 nm in diameter and contains 2404 atoms.

3 3 3 5 3 7

6 3 6 5 6 7

FIG. 2. SiGe-NW models used throughout, labelled C Swhere the index C represents the number of layers in the coreand S the surface. Shell thickness increases left to right, andcore thickness top to bottom. (The same motifs have beenused for the GeSi-NW models).

4

Model

3 3 3 5 3 7 6 3 6 5 6 7

SiGe-NWaminlat 5.593 5.619 5.624 5.574 5.532 5.565

alinlat 5.590 5.628 5.631 5.549 5.584 5.604

error(%) 0.054 -0.160 -0.125 0.449 -0.940 -0.701

GeSi-NWaminlat 5.531 5.475 5.460 5.574 5.554 5.492

alinlat 5.500 5.471 5.457 5.539 5.505 5.485

error(%) 0.561 0.073 0.055 0.628 0.882 0.128

TABLE I. Calculated axial lattice parameters using DFT, aminlat , for SiGe model nanowires depicted in Fig. 2, along with the

interpolated axial lattice parameter, alinlat based upon Eq. 4. Error is the percentatge of deviation of alin

lat from aminlat . All

measurements are given in A.

III. RESULTS: CALCULATED AXIAL LATTICEPARAMETERS

One approach for calculating the ideal axial lattice pa-rameter is to treat the nanowire as a solid solution, and touse Vegard’s law26,27, taking a linear interpolation basedon the relative numbers of Si and Ge atoms as follows:

alinlat =NSiaSilat +NGeaGelat

NSi +NGe, (4)

where NSi and NGe are the total number of Si andGe atoms present in the nanowire respectively. The bulkSi and Ge lattice constamts, aSilat and aGelat are taken as

5.432A and 5.658A52 respectively, for all calculations.Another approach is to plot DFT calculated total sys-

tem energy vs overall lattice parameter(alat) by varyingthe lattice parameter explicitly, and fitting to find the op-timum lattice parameter corresponding to the minimumenergy structure. This lattice parameter correspondingto the minimum energy structure is the ideal axial latticeparameter(alat) used for fixing the axial atomic distancesand axial simulation cell length in structural relaxationsperformed in the next sections.

The effect of varying the 3 3 SiGe-NW lattice parame-ter can be seen in Fig. 3, along with the lattice parameteras calculated from Eq. 4. From our result for the 3 3 SiGenanowire we can see that, in this case, a linear interpo-lation gives a good approximation, with the minimum ofthe curve fitted to the data residing extremely close tothe interpolated point.

Results from carrying this procedure out for the rest ofthe SiGe and GeSi nanowires are given in Table I, alongwith the lattice parameter calculated via Eq. 4. We cansee that, in general, a linear interpolation using Eq. 4provides a reasonable prediction of this ideal axial latticeparameter, although far from perfect. In a heteroepitax-ial system such as this, with an intrinsic strain of ∼4%, afurther strain from an approximate lattice constant willhave a considerable effect.

For the GeSi-NWs we can see that the interpolatedaxial lattice parameter consistently decreases towards theSi bulk lattice parameter with increasing shell size, as

FIG. 3. Energy variation with axial lattice parameter forthe 3 3 SiGe-NW. The red line is a cubic spline fitted to thedata, and the black dotted line represents the axial latticeparameter calculated analytically by eq. (4)

the number of Si atoms in the shell increases relative tothe number of Ge atoms in the core. For the optimisedcalculations, the same trend is seen, though the valuesare all larger than the interpolated values.

The interpolated axial lattice parameter of the SiGe-NWs increases with shell size as the system tends to-wards the Ge bulk lattice parameter, though this be-haviour is not seen in the optimised parameters. Witha small silicon core, the lattice parameter follows theinterpolated value reasonably; with the larger siliconcore, the behaviour is quite different, and the latticeconstant behaves non-monotonically, with large differ-ences to the interpolated values. Previous calculations ofcylindrical core-shell NWs have also shown diameter de-pendent behaviour: very small (∼1.5nm diameter) NWsshow non-linear behaviour33, though the behaviour wasmonotonic; slightly large NWs (up to 4nm diameter)20,39

show behaviour similar to our large hexagonal NWs—

5

non-monotonic for cylindical SiGe-NWs, with GeSi-NWsbehaving monotonically.

This non-monotonic departure from the simple lin-ear interpolation of Vegard’s law highlights an impor-tant point: while it is desirable to understand the struc-tural properties of these nanowires in the simplest pos-sible terms, the many different interfaces, coupled withthe elastic anisotropy of both Si and Ge, provide an ex-tremely complex system which will require careful, firstprinciples simulation to explore fully. The differences inlattice parameter found here (up to 1%), and the depar-ture from simple, expected behaviour, may give signifi-cant deviations in atomic and electronic structure whichcan only be fully explored using a technique such as lin-ear scaling DFT, which can reach realistic simulation cellsizes. This ability will help to understand systems in thefield of next generation electronics, allowing systems ofphysically realistic size to be examined and trends ex-posed.

IV. RESULTS: INTRINSIC STRAIN

Having calculated the axial lattice parameter corre-sponding to the minimum energy for each of our nanowiremodels, we now proceed to analyse the relaxed struc-tures for this axial lattice in each case. We can see fromFig. 1 that we have four (111) surfaces in our hexagonalnanowire, and two (100) surfaces. The (111) surface isvery simple, without reconstruction, and is unlikely tohave any effect beyond the influence of the surface asa boundary, while the (100) is more complex with Si-Sidimerisation along the [110] direction, typically leadingto strained bonds and changes in bond angle.

We consider the directions of bonds in the nanowirescarefully, as there is considerable anisotropy in the sys-

(100)

(111)(111) (111)

FIG. 4. Bond labelling: Cross sectional schematic of the NWmodels used, with three different bond types labelled, (100)⊥in blue, (111)⊥ in green and (111)‖ in red.

tem, coming both from elastic anisotropy and from thedifferent boundary conditions. Looking at the cross sec-tional schematic in figure 4, we see that in essence thereare three orientations for the bonds in our nanowiremodel. First, bonds that have a significant vector compo-nent perpendicular to the (100) surface along with a sig-nificant component along the axial [110] direction, whichwe have labelled (100)⊥, irrespective of vector direction.Bonds with a very small vector component along the[110] direction along with a significant vector componentperpendicular to the (111) surface in the first quadrant,which we have labelled (111)⊥. The final bond type,which is symmetrically equivalent to (111)⊥ in the sec-ond quadrant, we have chosen to label (111)‖, as it formsthe (111) surfaces in the first quadrant.

A. SiGe-NWs

There are many ways to present the data on bondlengths and local strain; we show two in Fig. 5. The left-hand side of the figure shows a three-dimensional plot ofvariations in bond length, which gives full informationon the structural variation, but is difficult to interpret,and consistent presentation for different size nanowires isalmost impossible. The right-hand side shows the resultof projecting the average bond length, for each of thebond types shown in Fig. 4, onto a grid in a plane per-pendicular to the nanowire axis (we note that the (111)⊥bonds are not shown as they are symmetrically equiva-lent to the (111)‖ bonds). The variation in strain withlocation in the nanowire and with direction is clearly vis-ible, showing the effects of the surface reconstruction aswell as local strains. A careful inspection will show thatsome information is lost: for instance, in the third layerbelow the (100) surfaces, the surface reconstruction in-duces an alternating compressive and tensile strain alongthe nanowire axis which is not seen in the averaged plots.However, this is a small variation, and does not affect theoverall conclusions.

The bond strain maps for the Si core, Ge shellnanowires are shown in Fig. 6 (shell) and Fig. 7 (core).Significant anisotropy within the shell becomes appar-ent, with bonds in the (111)⊥ directions behaving verydifferently to the (100)⊥ bonds. These bonds are ex-tended when the bond direction is perpendicular to theclosest (111) nanowire surface. This extension is mostpronounced close to the Si-Ge interface and monotoni-cally reduces to the bulk bond length at the surface. Onthe other hand we see a slight compression at the Si-Geinterface in this bond type when it runs parallel to theclosest (111) surface, and compression of this bond ineach of the four (111) shell surfaces it forms in each ofthe four quadrants. The (100)⊥ bond behaves in a signif-icantly different manner, with compression on all of the(111) surfaces and throughout most of the shell interior.The interior compression is most pronounced at the SiGeinterface at the point where the two (111) surfaces inter-

6

(111)‖ (100)⊥

FIG. 5. Left: Percentage variation in bond length, relative to bulk bond lengths for the Si-Si bonds in the core (top), andGe-Ge bonds in the shell (bottom) for SiGe-NWs with a 3-layer core and 5-layer Ge shell thickness. Right: Average bond strainmap for the cross-section of the Si core (top) and Ge shell (bottom) of the same 3 5 SiGe-NW. Maps are shown for bonds types(111)‖ and (100)⊥ (see Fig. 4); strains in (111)⊥ are identical to those in (111)‖ except reflected about the (100) axis, exactlyas the arrows shown in Fig. 4).

sect. However there is also a slight extension of this bondtype at the (100) Si-Ge interface, as well as a consider-able extension due to the (2× 1) reconstruction of (100)Ge surfaces due to Si-Si dimerisation. Overall we can seethat the largest strains are at the Si-Ge interfaces andthe Ge surfaces.

Turning to the core, shown in Fig. 7, we see that itis expanded for all shell thicknesses. Both the 3-layerand 6-layer cores expand, and the expansion is largelyisotropic, particularly in comparison to the shell. This isin agreement with previous theoretical and experimentalwork9,39, in which the Si core is found to be under tensilestrain.

It is well known that both Si and Ge are mechanicallyanisotropic, with the Young’s modulus in the [111] di-rection larger than that in the [110], which in turn islarger than that in the [100] direction. It is clear thatthis anisotropy has a significant impact upon the relativebond compression and extension in each direction withinthe Ge shell. Coupling this fact with the surface and in-terface effects result in the differing strain patterns of thebonds depending upon bond direction. It is interestingto note that the core of the NW does not have the sameradial freedom as the shell, which can expand into thevacuum, nor the strains induced by reconstruction; how-ever, as we will see in the next section, for these nanowiresand the present method, the anisotropy is largely asso-ciated with germanium, while silicon is much more uni-formly strained. (We note that silicon does have a larger

Young’s modulus than germanium53, but it is unlikelythat the cause is anything this simple.)

B. GeSi-NWs

Examining the strain maps for the Ge-core Si-shellnanowires, shown in Fig. 8 and Fig. 9, we can see that thebasic behaviour has notable similarities to the SiGe case:the Si is under tensile stress, and is relatively isotropic,while the Ge is anisotropic, showing both tension andcompression. There are significant differences, however,in both shell and core. The Si shell shows smaller exten-sion, particularly for the 3-layer core, which is often lessthan 1% away from the (100) surface reconstruction. Theradial freedom has allowed more variation in the strainpatterns to emerge. The (111)⊥ bonds suffer the mostsignificant extension when parallel to the closest (111)surface, i.e. in the second and fourth quadrants. The(100)⊥ bonds have the largest extension at the interfacebetween the two (111) Si surfaces, and the two (111) Gesurfaces, as well as at the reconstructed (100) surfacewith Si-Si dimers.

The Ge core shows less anisotropy than the Ge shell inthe SiGe NWs, with a strong dependence on the thick-ness of the Si shell: increasing shell thickness leads toan increase in the tension (or equivalently decrease incompression) for the (111)⊥ bonds. This may seemcounter-intuitive, as a thicker shell might be expected

7

NW Model (111)⊥ (100)⊥

3 3

3 5

3 7

6 3

6 5

6 7

FIG. 6. Average bond strain map for the cross-section of theGe shell of the SiGe-NWs. Maps for the (111)⊥ and (100)⊥labelled bonds are shown, with extension illustrated in blueand compression in red.

to lead to it having more influence, but reflects the in-creasing tension along this direction in the shell. Theshell clearly has more influence for the smaller core, seenparticularly in the (111)⊥ bonds. The anisotropy of theGe is still considerable, with the (100)⊥ bonds gener-ally compressed, while the (111)⊥ bonds vary with NWsize and location, showing particular variation at the in-terface between (111) and (100) surfaces. As with theSiGe nanowires, we see that the germanium shows moreanisotropy than the silicon. It is clear that a careful,detailed ab initio calculation is needed to describe thestructure of nanowires properly: while simple extrapola-tion can approximate the axial lattice constant, the de-tailed structure depends sensitively on the details of thenanowire structure.

C. Conclusion

Using linear scaling DFT calculations, we have studiedand compared the axial lattice parameters and intrinsicstrain patterns of Si/Ge and Ge/Si core-shell nanowireswith different core to shell ratios, with diameters in therange 4.9—10.2 nm.

We found that the axial lattice parameter calculated

NW Model (111)⊥ (100)⊥

3 3

3 5

3 7

6 3

6 5

6 7

FIG. 7. Average bond strain map for the cross-section of theSi core of the SiGe-NWs. Maps for the (111)⊥ and (100)⊥labelled bonds are shown, with extension illustrated in blueand compression in red.

analytically using Vegard’s empirical law gives a reason-able starting approximation to the axial lattice param-eter, but that detailed DFT simulations are needed tofind the correct values. In some cases, the error in lat-tice parameter from Vegard’s law was as large as 1%. Itis not surprising that a simple, empirical law based onsolid solutions should only be approximately correct forthese highly structured, anisotropic nanowires. Our DFTcalculations show that increasing the Si content leads toa reduction in the axial lattice parameter, towards thevalue of bulk Si. With one exception, increasing the Gecontent results in an increase in the axial lattice param-eter towards the value of bulk Ge, though in all casesthe behaviour does not follow the proportions of the con-stituents in a simple, linear fashion.

The lattice constant of bulk Ge is greater than thatof bulk Si by 4% and this generates an intrinsic strainin Ge and Si bonds in the nanowires. In all Si/Geand Ge/Si nanowires we have studied, the Si componentshows generally isotropic expansion, while the Ge compo-nent shows complex, bond-direction and nanowire-size-dependent tensile and compressive strain patterns. Giventhat the core does not have the same radial freedom asthe shell, it is interesting that the Si component behavesconsistently, regardless of whether it is placed as the coreor the shell. The differences in the elastic properties ofsilicon and germanium may play a role in this behaviour,reflecting the different bond strengths in the two materi-

8

NW Model (111)⊥ (100)⊥

3 3

3 5

3 7

6 3

6 5

6 7

FIG. 8. Average bond strain map for the cross-section of theSi shell of the GeSi-NWs. Maps are shown for the (111)⊥ and(100)⊥ labelled bonds, with extension illustrated in blue andcompression in red.

als. The varying strain within core and shell will have avariable effect on the electronic structure54,55, requiringcareful ab initio modelling to quantify the details.

The most highly strained and anisotropically strainedareas were observed near the surfaces and heterojunctioninterface, and structural deformation may be more likelyin these areas. Areas of sharp transition from tensile tocompressive strain can be seen near the core-shell intefaceof certain nanowires (e.g. Si/Ge 6 5 and 6 7 structures,Ge/Si 3 3 and 3 5). In such cases, the diffusion of Geinto the strained Si layer or Si into strained Ge layers

is likely to be enhanced over standard rates, and as thiswill change the composition of the nanowires, could poseperformance issues in core-shell nanowires. The straindue to both surface reconstruction and relaxation at thesurfaces will affect the local chemical reactivity.

In summary, we have shown that linear scaling DFT isboth possible and necessary for systems of this size, andhave produced a detailed study of the strain patterns insilicon-germanium core-shell nanowires.

NW Model (111)⊥ (100)⊥

3 3

3 5

3 7

6 3

6 5

6 7

FIG. 9. Average bond strain map for the cross-section of theGe core of the GeSi-NWs. Maps for the (111)⊥ and (100)⊥labelled bonds are shown, with extension illustrated in blueand compression in red.

ACKNOWLEDGMENTS

We thank Dr. N. Fukata for useful conversations. TMacknowledges support from JSPS Grant-in-Aid for Sci-entific Research: Grant Number 26246021, Japan.

∗ david.bowler@ucl.ac.uk1 W. Lu and C. M. Lieber, Nat. Mater. 6, 841 (2007).2 C. Thelander, P. Agarwal, S. Brongersma, J. Eymery,

L. Feiner, A. Forchel, M. Scheffler, W. Riess, B. Ohlsson,U. Gsele, and L. Samuelson, Materials Today 9, 28 (2006).

3 P. Avouris, Acc. Chem. Res. 35, 1026 (2002).4 C. Yang, Z. Zhong, and C. M. Lieber, Science 310, 1304

(2005).5 J. Goldberger, A. I. Hochbaum, R. Fan, and P. Yang,

Nano Letters 6, 973 (2006).6 W. Lu, J. Xiang, B. P. Timko, Y. Wu, and C. M. Lieber,

Proc. Natl. Acad. Sci. 102, 10046 (2005).7 J. Xiang, W. Lu, Y. Hu, Y. Wu, H. Yan, and C. M. Lieber,

Nature 441, 489 (2006).8 X. Zhao, C. Wei, L. Yang, and M. Chou, Phys. Rev. Lett.92, 236805 (2004).

9 N. Fukata, M. Mitome, T. Sekiguchi, Y. Bando,M. Kirkham, J.-I. Hong, Z. L. Wang, and R. L. Snyder,ACS Nano 6, 8887 (2012).

10 N. P. Dasgupta, J. Sun, C. Liu, S. Brittman, S. C. An-drews, J. Lim, H. Gao, R. Yan, and P. Yang, Adv. Mater.26, 2137 (2014).

9

11 D. D. D. Ma, C. S. Lee, F. C. K. Au, S. Y. Tong, andS. T. Lee, Science 299, 1874 (2003).

12 Y. Wu, Y. Cui, L. Huynh, C. J. Barrelet, D. C. Bell, andC. M. Lieber, Nano Letters 4, 433 (2004).

13 L. J. Lauhon, M. S. Gudiksen, D. Wang, and C. M. Lieber,Nature 420, 57 (2002).

14 B.-M. Nguyen, Y. Taur, S. T. Picraux, and S. A. Dayeh,Nano Letters, Nano Letters 14, 585 (2014).

15 T. Noguchi, K. Morita, M. Simanullang, Z. Xu, K. Usami,Y. Kawano, T. Kodera, and S. Oda, physica status solidi(a) 212, 1578 (2015).

16 I. A. Goldthorpe, A. F. Marshall, and P. C. McIntyre,Nano Letters 9, 3715 (2009).

17 B. Bhushan, D. Luo, S. R. Schricker, W. Sigmund,and S. Zauscher, Handbook of nanomaterials properties(Springer Science & Business Media, 2014).

18 M. Jaishi and R. Pati, Nanoscale 9, 13425 (2017).19 L. Yang, R. N. Musin, X.-Q. Wang, and M. Y. Chou,

Phys. Rev. B 77, 195325 (2008).20 X. Peng, F. Tang, and P. Logan, J. Phys. Condens. Matter

23, 115502 (2011).21 M. W. Larsson, J. B. Wagner, M. Wallin, P. Hkansson,

L. E. Frberg, L. Samuelson, and L. R. Wallenberg, Nan-otechnology 18, 015504 (2006).

22 N. Liu, N. Lu, Y.-X. Yao, Y.-R. Li, C.-Z. Wang, and K.-M.Ho, J. Phys. Chem. C 115, 15739 (2011).

23 S. T. Dunham, M. Diebel, C. Ahn, and C. L. Shih, Jour-nal of Vacuum Science & Technology B: Microelectronicsand Nanometer Structures Processing, Measurement, andPhenomena 24, 456 (2006).

24 S. Ciraci and I. P. Batra, Phys. Rev. B 38, 1835 (1988).25 C. G. Van de Walle and R. M. Martin, Phys. Rev. B 34,

5621 (1986).26 L. Vegard, Z. Phys. 5, 17 (1921).27 A. R. Denton and N. W. Ashcroft, Phys. Rev. A 43, 3161

(1991).28 A. Nduwimana, R. Musin, A. Smith, and X.-Q. Wang,

Nano Lett. 8, 3341 (2008).29 K. Jacob, S. Raj, and L. Rannesh, Int. J. Mater. Res. 98,

776 (2007).30 A. R. West, Solid state chemistry and its applications

(John Wiley & Sons, 2007).31 N. Nishiyama, J.-h. Lin, A. Okazaki, M. Iwasaka, and

K. Hirakawa, Jap. J. Appl. Phys. 29, 369 (1990).32 F. Iori, S. Ossicini, and R. Rurali, J. Appl. Phys. 116,

154301 (2014).

33 R. Musin and X.-Q. Wang, Phys. Rev. B 71, 155318(2005).

34 M. Amato, S. Ossicini, and R. Rurali, Nano Letters 11,594 (2011).

35 R. N. Musin and X.-Q. Wang, Phys. Rev. B 74, 165308(2006).

36 M. Amato, M. Palummo, and S. Ossicini, Phys. Rev. B79, 201302 (2009).

37 M. Amato, M. Palummo, and S. Ossicini, Phys. Rev. B80, 235333 (2009).

38 D. B. Migas and V. E. Borisenko, Phys. Rev. B 76, 035440(2007).

39 X. Peng and P. Logan, Appl. Phys. Lett. 96, 143119 (2010).40 D. R. Bowler and T. Miyazaki, Rep. Prog. Phys. 75,

036503 (2012).41 D. R. Bowler and T. Miyazaki, J. Phys. Condens. Matter

22, 074207 (2010).42 D. R. Bowler, T. Miyazaki, and M. J. Gillan, J. Phys.:

Condens. Matter 14, 2781 (2002).43 C.-K. Skylaris, P. D. Haynes, A. A. Mostofi, and M. C.

Payne, J. Phys.: Condens. Matter 20, 064209 (2008).44 J. M. Soler, E. Artacho, J. D. Gale, A. Garcıa, J. Junquera,

P. Ordejon, and D. Sanchez-Portal, J. Phys.: Condens.Matter 14, 2745 (2002).

45 J. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett.77, 3865 (1996).

46 A. Nakata, Y. Futamura, T. Sakurai, D. R. Bowler, andT. Miyazaki, J. Chem. Theory Comput. 13, 4146 (2017).

47 A. S. Torralba, M. Todorovic, V. Brazdova, R. Choudhury,T. Miyazaki, M. J. Gillan, and D. R. Bowler, J. Phys.Condens. Matter 20, 294206 (2008).

48 X.-P. Li, R. W. Nunes, and D. Vanderbilt, Phys. Rev. B47, 10891 (1993).

49 W. Kohn, Int. J. Quantum Chem. 56, 229 (1995).50 T. Miyazaki, D. R. Bowler, R. Choudhury, and M. J.

Gillan, Phys. Rev. B 76, 115327 (2007).51 E. Bitzek, P. Koskinen, F. Gahler, M. Moseler, and

P. Gumbsch, Phys. Rev. Lett. 97, 170201 (2006).52 A. Dargys and J. Kundrotas, Handbook on physical prop-

erties of Ge, Si, GaAs and InP (Science and EncyclopediaPubl., 1994).

53 J. Wortman and R. Evans, J. Appl. Phys. 36, 153 (1965).54 P. W. Leu, A. Svizhenko, and K. Cho, Phys. Rev. B 77,

235305 (2008).55 Y.-M. Niquet, C. Delerue, and C. Krzeminski, Nano Let-

ters, Nano Letters 12, 3545 (2012).