arXiv:1009.0336v2 [cond-mat.mes-hall] 30 Aug 2011 nld nstoi opst emo mass Fermion composite anisotropic include contacts point tum achieved wires quantum been overgrown cleaved-edge has in system multi-valley one-dimensional ineet,clbaigvle tanssetblt and susceptibility strain mass valley effective valley calibrating degener- effects, valley tion uniaxi- induce with to splitting realized QWs acy been (001)AlAs has strained degeneracy ally valley the of cuac nwd 10AA wells (110)AlAs wide in occupancy shown wells been has (001)AlAs occupation for single-valley to double- ferromagnetism from identify- Hall QWs in quantum AlAs useful like in proven effects also exchange have ing (411) as such facets as t ev nstoi lcrnms losfrlarge for allows effects mass electron interaction be- anisotropic 1) heavy (Fig. interest its particular cause of three- is bulk degeneracy valley its fold with semi- (AlAs) arsenide blende aluminum zinc multi-valley conductor various indirect-bandgap the the Of semiconductors, contribute quantum scales. all energy strain, since comparable at fields of piezoelectric treatment and confinement, comprehensive a quires ino nstoi 10-ae QWs (110)-facet anisotropic of QWs tion (001)-facet in oe unu el (QWs) wells modulation- quantum high-mobility, doped for allows substrates GaAs to ec oiiiso h re of order the of mobilities reach acltn alydgnrc naqatmwl re- well quantum a in degeneracy valley Calculating alydgnrc nbailysrie lmnmarsenide aluminum strained biaxially in degeneracy Valley 1 ASnmes 73.21.Fg,73.50.Bk,73.61.Ey,71.70.Fk,71. numbers: PACS momen component. in-plane momentum purely out-of-plane distinguish an p to require scattering one allows valley biaxi which multi-valley a space other to scattering, adapted tha intervalley be so anisotropic shear-t can planes a treatment non-Miller-indexed for the solve address and and to high define strain generalized the we shear is and for additional QWs An calculated double (411)-oriented is index QWs. from ratio (111)-oriented crossover Poisson and for (110)-, biaxial width strai The well and system. confinement critical wh electron the facet, of determine (411) (111)-orien unconventional Calculations the and is (110)-, literature. as (001)-, examined, of are cases wells specific scatt The and t fields, bases. mass piezoelectric effective index, ratios, whereby strain basis be spin anisotropic transport interval transform the well vanishing to quantum to the defined the and to is analogy matrix due In transformation index coordinate pseudospin wells. a grow as quantum wells justified AlAs quantum multi-valley for strained results in degeneracy and 1,10,11 lcrclEgneigadCmue cec,Northwester Science, Computer and Engineering Electrical 6,7 hspprdsrbsacmlt nltclfraimfrca for formalism analytical complete a describes paper This 2 .INTRODUCTION I. vdnefraQ it crossover width QW a for Evidence . atrShtk nttt ehiceUniversit¨at M¨unch Technische Institut, Schottky Walter 1 uhsuiscnqatf interac- quantify can studies Such . .Prabhu-Gaunkar S. n t erpretltiematch lattice perfect near its and , 17 8,9 10,12–14 4 oe neato ffcsi QWs in effects interaction Novel . shseiec o igevalley single for evidence has as , n ntehg oiiydirec- mobility high the in and unu ofieett a to confinement Quantum . 2,3 lsAGA W can QWs AlAs/AlGaAs . µ 5 0,0 cm 100,000 = yai control Dynamic . 5 Unconventional . 1 .Birner S. , 15,16 18 n quan- and n valley and 2 .Dasgupta S. , 2 /Vs in ftovlesa once at valleys two superposi- of linear tions populate electrons whereby skyrmions, ftedffrn lcrnvles n eti rwhsym- growth certain and valleys, degeneracy electron and different energy the subband of electron strain so the and substrate, confinement estimate quantum to GaAs both the consider to must one respect are with QWs AlAs strained slightly system. multi-valley a in occupancy valley pseudospin. functions SU(2) a freedom as of perfect degree valley a the where of applications, splitting exchange symmetry from result that uidwt lcrn sidctdb lisia equi-ene ellipsoidal by indicated as contours. ba- electrons conventional-cubic-cell with the cupied defining structures AlAs QW sis bulk axes, AlAs for of basis depiction zone transport space the Real (Left) defining 1: FIG. 70.Gm rn etr r l endi hi respective their in defined all are vectors ering aiu omlsshv endvlpdt understand to developed been have formalisms Various ,y z y, x, iiieui eli endi momentum in defined is cell unit rimitive 2 .Knaak C. , ictsbtae a lob treated, be also can substrates miscut t quantum (AlAs) arsenide aluminum ted nte(0) 10,ad(1)facets (411) and (110), (001), the in n nvriy vntn L628USA 60208 IL Evanston, University, n c so atclriprac nAlAs in importance particular of is ich we h ovninlcbccl basis conventional-cubic-cell the tween e xhneitga.Astandardized A integral. exchange ley oetetredgnrt aly htaeoc- are that valleys degenerate three the Note . u cteigeet rmtoethat those from events scattering tum classify help To systems. strained ally opnn rssi h ihrMiller higher the in arises component n acig -54 Germany D-85748 Garching, en, osnl-alydgnrc neach in degeneracy single-valley to - smer oe ilridx(001)-, index Miller lower -symmetry ln n retto ihexplicit with orientation any along n nos alyvcos tanmatrices, strain vectors, valley ensors, 20 -ixa tanrto h notation The ratio. strain o-biaxial cltn lcrnsbadenergy subband electron lculating a rv sfli uueqatmdevice quantum future in useful prove may h alydge ffedmis freedom of degree valley the 2 n .Grayson M. and , unu wells quantum 19 uhitrcineffects interaction Such . ,b c b, a, 1 Rgt Brillouin (Right) . rgy 2 metries also introduce in-plane or out-of-plane piezoelec- change energy calculations demonstrate why the valley tric fields. Stern and Howard modeled two-dimensional index can function as a pseudospin index. In Section III, (2D) confinement of electron valleys in Si for arbitrary we develop the notation and formalism for determining crystal plane orientations neglecting strain effects21. Van valley degeneracy in multi-valley strained semiconductor de Walle theoretically studied the absolute energy level QWs. The three subband energy components are defined for an unstrained semiconductor heterojunction, con- – kinetic energy, confinement energy and strain energy – sidering strain effects only for bulk systems22. Smith as are the two useful coordinate bases – the conventional- et al. calculated the strain tensor theoretically for the cubic-cell basis and the transport basis – as well as the case of (001)- and (111)-oriented QW superlattices23. Al- coordinate transformation matrix which transforms be- though Caridi and Stark derived the complete strain ten- tween them. Single-electron analytical solutions are pro- sor for arbitrarily oriented substrates with cubic symme- vided to allow easy identification of characteristic en- try, they ignored a critical shear component which arises ergy scales for multi-valley systems. The explicit crite- in the high Miller index directions under relaxation24. ria for valley degeneracy in arbitrarily oriented biaxially De Caro et al. calculated the shear strain with com- strained QWs are defined in Section IV. Section V de- mensurability constraint equations accurately only for scribes how shear strain can induce piezoelectric fields in the low index (100)-, (110)- and (111)-orientations25. the QW. In Section VI, we introduce the valley-scattering Yang et al. corrected the constraint equations and cal- unit cell – a 3D primitive unit cell with the same volume culated piezoelectric fields arising from the shear strain as the Brillouin zone, but which shares the full symme- components for the general case of a pseudomorphic try of the reciprocal lattice vectors that lie in the Miller film26. Adachi calculated the strain tensor theoretically index plane. Momentum-space illustrations provide in- for the case of bulk and superlattice structures but did tuition for visualizing intervalley scattering. Section VII not address the case of single QW structures27. Ham- applies the developed formalism to the case of AlAs QWs, merschmidt et al. focus on the case of isotropic planar whereby the projected in-plane transport masses, strain strain arising in QWs but do not carry out a combined tensor, degeneracies, piezoelectric fields, and valley split- treatment of strain energy along with quantum confine- ting energies for the specific cases of (001)-, (110)-, (111)- ment energy and how these affect valley degeneracies28. as well as the unconventional (411)-oriented QWs illus- Theoretically, Rasolt proposed an understanding of the trate how valley degeneracies can be engineered. We re- various continuous symmetries and symmetry-breaking view the crossover width calculation for double-to-single in a multi-valley system in terms of an SU(N) symme- valley occupation in the (001)-, (110)-, and (411)-facet try where N is the valley degeneracy number20, but did QWs. For the high-symmetry low Miller index orien- not explicitly derive the vanishing intervalley exchange tations (001), (110) and (111) we calculate the biaxial integral which underlies this theory. Material specific Poisson ratio, and for the low-symmetry high Miller- calculations of quantum confinement and strain in high- index (411) facet there arises a shear component in the mobility Si and Ge structures are well studied for high- strain tensor and consequently a shear-to-biaxial strain symmetry facets29–40 and recently, such calculations have ratio. We calculate intervalley scattering ratios for valley- been been made in other quantum confined systems as degenerate AlAs QWs on various facets in Section VIII well41,42. However, a combined treatment of quantum and refer to the valley scattering unit cell for intuition. confinement, strain, and piezoelectric fields to determine We complete our analysis by comparing the standard 2D valley degeneracy in QWs is lacking, especially for low Bravais lattice valley representation to our valley cell rep- symmetry facets. resentation. To generalize the formalism for calculating To eventually model transport in such valley degener- valley subband energies in arbitrarily oriented facets, we ate systems, one must also consider the extra scattering end in Section IX by adapting this notation for miscut 43 channel not present in single valley systems, namely in- samples . tervalley scattering. It is therefore useful to introduce a k-space unit cell which permits visualization of momen- tum scattering events in a geometry that is natural to II. VALLEY EXCHANGE INTEGRAL AND the quantum confinement direction. However, the stan- VALLEY INDEX AS A PEUDOSPIN dard depiction of a 2D Brillouin zone of a quantum con- 21 fined valley-degenerate system , projects all valleys to a Just as electrons with different spins are distinguish- single two-dimensional plane. Such a depiction loses in- able and have no exchange term in their interaction en- formation about the out-of-plane momentum scattering ergy, electrons in different valleys can be shown to be component that was projected out, which is necessary to effectively distinguishable from each other due to the van- determine the full momentum scattering matrix element. ishingly small intervalley exchange energy. The valley in- Thus it is useful to develop a graphical representation dex can thus be treated as a pesudospin index. Rasolt20 for the unit cell in k-space that can clearly elucidate how discusses symmetry-breaking in a multivalley system but valleys are coupled with both in-plane and out-of-plane does not explicitly derive this exchange integral. Previ- momentum-scattering events. ous work on Si quantum dots to study Kondo effect44–46 The paper is organized as follows. In Section II, ex- calculates the exchange integral only for the special case 3 of identical spatial wavefunctions on Si quantum dots. larger than a few lattice constants. On the other hand if For completeness, in this section we explicitly derive the τ = τ, Qττ = 0 and the exponential term is unity and inter- and intravalley exchange integrals for arbitrary the integral remains finite. Therefore, the dependence wavefunctions to justify the valley index as a pseudospin, of the exchange integral Eq. (4) on valley index can be with exchange interaction proportional to δττ where τ approximated with a second delta function to notate the and τ are valley indices. vanishing exchange integral between different valleys An electron wavepacket within a single valley is de- τ,τ,σ,σ δσσδττ 2 scribed with a weighted integral over Bloch functions. E = dr φ(r )φ∗(r ) uqτ (r ) ex V 2 1 1 1 | 1 | The normalized wavepacket ψτ,σ in valley τ is chosen to Z 2 be in a spin polarized state σ described by the spinor χσ e 2 dr2 φ∗(r2)φ(r2) − uqτ (r2) . (6) r1 r2 | | 1 3 i(qτ +k) r Z | − | ψτ,σ(r)= d k Ak e · uqτ +k(r)χσ . (1) √V We note that Eq. (6) cannot be simplified because the Z Coulomb potential can vary rapidly over small distances uqτ +k(r) is the component of the Bloch function peri- of order a lattice constant. odic in the Bravais lattice, Ak is the complex amplitude We conclude by virtue of this vanishing intervalley ex- τ for a particular k, q + k is the total crystal momentum change interaction, that wavefunctions in different valleys τ where q denotes the center of the valley and k denotes can be treated like distinguishable particles, and the val- the additional small momentum deviation away from this ley index is a valid pseudospin index. As was pointed out valley center, and V is the volume of the system. Assum- by Rasolt20 the valley pseudospin constitutes an SU(N) ing k is much smaller than the Umklapp vector for the group where N is the number of electron valleys. lattice, and taking the envelope function approximation uqτ +k ∼= uqτ , we obtain III. VALLEY SUBBAND ENERGY 1 iqτ r 3 ik r ψτ,σ(r) = uqτ (r)e · χσ d k Ake · √V Z Crystal symmetry dictates that multiple energy- 1 iqτ r = uqτ (r)e · χσφ(r) (2) degenerate valleys will occur whenever a local conduc- √V tion band minimum exists away from the origin in mo- mentum space (the Γ-point). When such a multi-valley where φ(r) is the slowly varying envelope function. The system is then quantum confined in a layer with planar Coulomb exchange energy integral between the two elec- Miller indices M = (hkl), the energy Eτ (k) in the lowest tron wavefunctions ψ (r) and ψ (r) in valleys τ and τ,σ τ,σ subband in valley τ is τ with coordinates r1 and r2 is given by τ τ τ τ 2 E (k)= E0 (k)+ T (k) + ∆E (7) τ,τ,σ,σ e Eex = dr1dr2 − ψτ,σ(r1)ψτ,σ∗ (r2) r1 r2 where k is the 2D in-plane momentum relative to the σσ Z Z | − | τ X τ-valley minimum, E0 (k) is the ground confinement en- ∗ τ τ ψτ,σ(r2)ψτ,σ(r1) . (3) ergy, T (k) is the in-plane kinetic energy, and ∆E is the strain induced energy shift caused by lattice mismatch of Substituting Eq. (2) in Eq. (3), we can show that two the QW with respect to the substrate. Note that the particles with different valley indices behave like distin- Miller index M is not explicitly superscripted because guishable particles by virtue of their vanishing exchange it is common to all valleys. To calculate these terms in integral, Eq. (7), we need to find the in-plane (parallel to the QW) and out-of-plane (confinement direction) components of τ,τ,σ,σ δσσ E = dr dr φ(r )φ∗(r )φ(r )φ∗(r ) ex V 2 1 2 1 2 2 1 the inverse mass tensor of the corresponding electron val- Z Z ley as well as the various strain tensor components in the 2 ττ e iQ (r2 r1) − e · − uqτ (r )u∗τ (r )u τ (r )u∗ τ (r ) QW. r r 1 q 2 q 2 q 1 | 1 − 2| We start by introducing two useful bases, the (4) conventional-cubic-cell (CCC) basis x = (x,y,z) and the transport basis a = (a,b,c), along with the coordinate with intervalley scattering wavevector transformation matrix RM which transforms between them Qττ = qτ qτ , (5) − a = RM x . (8) where the inner product of the spinors is given by χσ† χσ = δσσ. This delta function denotes that the exchange in- The CCC basis has the x, y, and z axes aligned along the tegral vanishes when the two spin indices are different axes of the cubic cell of the reciprocal lattice of the crys- σ = σ. Analogously, if τ = τ, Qττ is of order an Umk- tal, Fig. 1 (right). The transport basis is chosen with the lapp6 vector, and the rapidly6 oscillating complex expo- a-b plane parallel to the QW, and the c-direction perpen- nential makes the integral vanish for envelope functions dicular to this plane, Fig. 1 (left). Thus if the Miller index 4 of the growth plane is (hkl), the perpendicular unit vec- The confinement potential V (c, k) can be expressed as tor in the transport basis is c = (h,k,l) 1 . To √h2+k2+l2 0 if c < W/2; uniquely define the a and b directions in the transport V (c, k)= | | , (12) basis, we take a to be the in-plane unit vector with the V (k) if c > W/2 b ( 0 | | lowest Miller index, and b the unit vector which main- where the height of the potential barrier V is given by tains a right handedb coordinate system a b = c. With 0 this definition we obtain ab unique set of axes.× These also ~2 τ τ τ τ define the components of the coordinate transformation V0(k)= Ec,B Ec,W + k (w′ 2 2,B w′ 2 2,W) k (13) b b b − 2 · × − × · M matrix R whereby a, b, and c are the top, middle, τ where Ec,W is the conduction band energy of the τ-valley and bottom rows of the coordinate transformation ma- τ M in the QW, Ec,B is the energy in the barrier, and W is trix R . In what follows,b b vectorsb and tensors which are expressed in the x-basis are unprimed, and in the a-basis the well width. The last term accounts for differences in-plane k they are primed. in the effective mass as a -dependent barrier height. This last term is often neglected in calculations The mass tensor is naturally expressed in the unprimed under the assumption that k is small and the tensor mass CCC x-basis, so by transforming it to the primed trans- τ difference w w τ is also small, making Eτ and ψ(c) port a-basis we can easily extract the in-plane and out- ′ B ′ W 0 independent of− in-plane momentum. of-plane confinement masses for a given electron valley. The matrix inverse of the mass tensor for the τ-valley τ τ 1 in the x-basis is denoted by w = (m )− . We as- B. Kinetic energy sume parabolic bands so that the mass is independent of the wavevector k. Applying the transformation (w τ ) ′ We can now calculate the second term in Eq. (7), the = RM (wτ )(RM ) 1 we obtain the inverse mass tensor of − kinetic energy T (k), from the in-plane inverse mass ten- the τ-valley in the primed transport a-basis τ sor w′ 2 2. The in-plane inverse mass tensor is deter- mined by× solving the ground state from Eq. (10) and then w τ w τ w τ ′ aa ′ ab ′ ac determining a weighted average of the well and barrier τ τ τ τ  w′ ba w′ bb w′ bc  inverse mass tensors given by (w′ )= . (9) W/2   τ  w τ w τ w τ  w τ =2w ψ(c) 2dc +  ′ ca ′ cb ′ cc  ′ 2 2 ′2 2,W   × × 0 | |   Z τ ∞ 2 w τ 2w′2 2,B ψ(c) dc . (14) The upper left 2 2 sub-matrix ′ 2 2 gives the inverse × × τ × W/2 | | in-plane mass tensor and w′ cc gives the inverse of the Z confinement mass perpendicular to the QW. In most cases it is safe to assume the wide-well limit, whereby the effective mass is equal to that in the well material only. The kinetic energy is A. Quantum confinement energy ~2 τ τ T (k)= k (w′ 2 2) k . (15) 2 · × · We can now define the first energetic term in Eq. (7), τ The projected in-plane effective masses can be solved the quantum confinement energy E0 (k), using the co- from the determinant of this same sub-matrix ordinate transformation matrix and out-of-plane compo- τ nent of the inverse mass tensor in the primed a-basis. det(w′ 2 2 λI)=0 . (16) This is calculated from the ground state energy solution × − of the 1D Schr¨odinger equation for a particle confined The eigenvalue solutions λ1 and λ2 directly give the re- along c-direction with in-plane momentum k, ciprocal masses along the major and minor axes of the projected mass tensor. The cyclotron mass is the same ~2 as the 2D density-of-states mass, and is given by d τ dψk(c) τ − w′ cc(c) + V (c, k)ψk(c)= E0 (k)ψk(c) τ τ 1/2 1/2 2 dc dc m2D = det(w′ 2 2)− = (λ1λ2)− . (17)   (10) × τ where w′ (c) is the diagonal component of the recipro- The spin-degenerate two-dimensional energy density-of- cc τ τ ~2 cal mass tensor in the a-basis. Due to the different re- states per unit area in the τ-valley is n2D = m2D/π , so τ τ that the total energy density-of-states is given by a sum ciprocal masses in the barrier (w′ B) and the well (w′ W) layers, wavefunction derivatives at the boundary c0 must over all valleys 47 satisfy mτ n (E)= 2D Θ(E Eτ ∆Eτ ) . (18) 2D π~2 − 0 − τ τ dψk(c) τ dψk(c) X w′ B,cc = w′ W,cc . (11) dc c=c− dc c=c+ where Θ(x) is the Heavyside step function. 0 0

5

C. Strain energy constants22,50. Using Hooke’s law in the x-basis, stress (σ) and strain (ǫ) tensors are related by We can now define the final term of Eq. (7), the strain τ σ = c ǫ i,j,k,l x,y,z . (24) energy ∆E , in terms of the strain tensor and the co- ij ijkl kl ∈{ } ordinate transformation matrix. Following the notation Here c is the fourth-rank elastic stiffness tensor in the of Van de Walle22 and Herring and Vogt48, the absolute x-basis. For crystals with cubic symmetry c can be sim- energy shift ∆Eτ of the τ-valley for a homogeneous de- plified using the Voigt notation51 into a matrix form formation described by the strain tensor ǫ in the x-basis is given by σi = Cij ǫj i, j =1, 2, ..., 6 . (25) ∆Eτ = (Ξτ δ + Ξτ qτ qτ )ǫij (19) d ij u i j The indices 1 through 6 denote xx, yy, zz, yz/zy, zx/xz where qτ is a unit vector in the direction of the τ-valley, and xy/yx, respectively. For cubic materials most of τ τ b b and Ξd and Ξu represent the deformation potentials due the elements of the matrix Cij vanish and C11 = C22 = to a bulkb dilation and a uniaxial deformation, respec- C33, C12 = C13 = C23 and C44 = C55 = C66 simplifying tively. The average shift in the energy of the subband the stress tensor σ to extrema is given by σ1 C11 C12 C12 0 0 0 ǫ1 τ 1 τ ij ij ∆E = (Ξ + Ξ )δ ǫ = a δ ǫ , (20) σ2 C12 C11 C12 0 0 0 ǫ2 av d 3 u ij c ij       σ3 C12 C12 C11 0 0 0 ǫ3 where a is the hydrostatic deformation potential for the   =     , c  σ4   0 0 0 C44 0 0   ǫ4  conduction band and δ ǫij = Tr(ǫ) is the trace of the       ij  σ   0 0 0 0 C 0   ǫ  strain tensor ǫ + ǫ + ǫ . Often the relative strain  5   44   5  xx yy zz       energy shift ∆τ from the mean value is all that is needed  σ6   00 00 0 C44   ǫ6  to determine subband occupancy:      (26) where ǫ1 = ǫxx, ǫ2 = ǫyy, ǫ3 = ǫzz, ǫ4 = 2ǫyz, ǫ5 = 2ǫzx τ τ τ τ τ 1 ij and ǫ6 = 2ǫxy. Table I lists the values of C11, C12, and ∆ = ∆E ∆Eav = Ξu(qi qj δij)ǫ . (21) − − 3 C44 for AlAs and GaAs. The strain tensor ǫ and the ǫ elastic constants cklmn are used to obtain the free energy To determine the strain tensorb b it is easiest to explic- 28 itly determine its components in the a-basis, which we of isothermal elastic deformations of a medium , ǫ will denote with ′, and then apply a rotational transfor- 1 mation to express it in the x-basis for use in Eqs.(19)- F (ǫ)= cklmnǫklǫmn . (27) 2 (21). Heteroepitaxial QWs are biaxially strained relative klmnX to an unstrained substrate with a different lattice con- stant. Thus the strain tensor ǫ is diagonal in the upper For a cubic crystal using the Voigt notation for the elastic ′ constants, the elastic energy reduces to 2 2 block ǫ′aa = ǫ′bb = ǫ′ , where ǫ′ is the in-plane × k k strain component given by22,49 ǫ C11 2 2 2 2 2 2 F ( )= (ǫxx + ǫyy + ǫzz)+2C44(ǫxy + ǫxz + ǫyz) asubstrate alayer 2 ǫ′ = − . (22) k alayer + C12(ǫxxǫyy + ǫxxǫzz + ǫyyǫzz) .(28) M Note that lattice matching to the unstrained substrate The coefficients Di can be determined for a given fixes ǫab′ = ǫba′ = 0. The remaining components of crystal facet orientation by transforming the strain ten- the strain tensor are linearly proportional to the parallel sor of ǫ′ into the x-basis with the rotation matrix strain ǫ′ as follows: M 1 M k ǫ = (R )− (ǫ′)R , (29) 1 0 DM 1 and inserting these components into the elastic energy of M M  0 1 D2  Eq. (28) and minimizing with respect to each D ǫ′ = ǫ′ (23) i k   M 1 M  DM DM DM  F (ǫ)= F [(R )− (ǫ′)R ] (30)  1 2 0   −    M where D0 = ǫcc′ /ǫaa′ = ǫ′ /ǫ′ is the biaxial Poisson dF (ǫ) − M − k ⊥ 22 = 0 (31) ratio (notated D by Van de Walle ), and we define dDM M M i D1 = ǫca′ /ǫaa′ and D2 = ǫbc′ /ǫaa′ as the shear-to-biaxial strain ratios. to deduce a set of three equations which can be simul- M M M The coefficients Di can be derived for a crystal with taneously solved to give the Di values. With the Di cubic symmetry and arbitrary Miller index M by mini- values solved for this material and this facet, the strain mizing the free energy of the layer in terms of its elastic tensor in the x-basis can now be used in Eqs.(19)-(21) 6

001 110 111 411 411 X material D0 D0 D0 D0 D2 C11 C12 C44 ac Ξu ml mt ex,4 2 [GPa] [eV] [eV] [me] [me] [C/m ] AlAs 0.854 0.616 0.550 0.775 0.176 125.0 53.4 54.2 2.54 6.11 1.1 0.20 -0.23 GaAs 0.934 0.580 0.489 0.820 0.250 122.1 56.6 60.0 -0.16 8.61 1.3 0.23 -0.16

M X TABLE I: Strain ratios Di for different facet orientations, elastic constants Cij , deformation potentials ac, Ξu , X-valley mass 21,22,52–54 tensor components ml,mt, and piezoelectric coefficients ex,4 for AlAs and GaAs . to deduce the strain energy contribution to each valley meaning that all valleys with the same polar angle from ∆Eτ . the c-axis are degenerate. For the more general case of M As discussed in Section VII, for high-symmetry crys- arbitrary Di , the full Eq. (36) has to be solved to deter- tal orientations like (001), (110), and (111), there is no mine valley degeneracy. However, given the shear vector shear component in the strain tensor when expressed in α in the plane of the QW as defined in Eq. (32), a spe- M M the transport a-basis so that D1 = D2 = 0, whereas cial case can be defined, and two valleys τ and τ are M for (411) growth the D2 coefficient is nonzero, and the degenerate if they simultaneously satisfy Eq. (37) and direction of shear is defined by the vector α given by τ α τ α α M M q′ = q′ , (38) = D1 a + D2 b . (32) | · | | · | For AlAs and GaAs, we tabulate the nonzero DM in or equivalently, if the valleys are mirror symmetric about b i b b b b Table I along with the elasticb stiffness components in the c-α plane. Voigt notation. For completeness, the Appendix provides The second condition for robust valley degeneracy is a compact list of equations for determining the strain ten- that the out-of-plane inverse masses are equal, sor for arbitrary substrate orientation. From the above τ τ analysis, we can now calculate all the terms in Eq. (7) w′cc = w′cc . (39) for any valley and any QW orientation. guaranteeing equal confinement energies. As long as the mass ellipsoid associated with each valley is oriented with IV. ROBUST VALLEY DEGENERACY AND its longitudinal mass parallel to the unit vector q τ , as CROSSOVER VALLEY DEGENERACY is usually the case, then the Eq. (39) is automatically satisfied under the same condition as Eq. (37). b Two valleys τ and τ will be degenerate if Eτ (k = 0)= Eτ (k = 0) . (33) V. PIEZOELECTRIC EFFECTS ON THE When this condition holds for τ and τ regardless of the QUANTUM WELL POTENTIAL QW width, we call the valley degeneracy robust, and if it holds only for a specific well width W0, we call it crossover In crystals which are inversion asymmetric like zinc degeneracy. The crossover degeneracy condition can arise blende crystals, strain may also produce piezoelectric only when the valley electrons with a higher confinement fields, modifying the confinement potential V (c, k), and τ mass are strain-shifted to a higher energy than the valley resulting in shifts of the ground energies E0 for the var- electrons with a lower confinement mass. ious valleys as solutions to the Schr¨oedinger equation, Robust valley degeneracy can only occur if the strain Eq. (10). The out-of-plane piezoelectric field which re- energies and out-of-plane confinement masses are inde- sults will not break robust valley degeneracy, but will pendently equal. The strain energy must satisfy shift the crossover degeneracy condition to a different well width, W . ∆Eτ = ∆Eτ , (34) 0 In our treatment, we will calculate the piezoelectric for all well widths, which simplifies via Eq. (19) to the field inside the QW and assume that any out-of-plane x-basis condition component will be canceled with an external gate bias, restoring the condition of the flat square well. The piezo- q τ ǫ q τ = q τ ǫ q τ (35) · · · · electric field inside the QW and/or the barriers is propor- 27 or equivalently in the a-basis, tional to the shear strain in the crystal basis b b b b τ τ τ τ q ǫ′ q = q ǫ′ q . (36) ′ · · ′ ′ · · ′ 1 = − e ǫ . (40) When DM = DM = 0, this condition for degeneracy Ei ε ε i,j j 1 b 2 b b b s 0 simply reduces to τ τ Here i lists the unprimed piezoelectric field components q c = q c , (37) in theE x-basis indexed by i = x,y,z, ε and ε are | ′ · | | ′ · | 0 s b b b b 7 the free-space and relative semiconductor dielectric con- For the present treatment, we reiterate our assumption stants, and ei,j represents the piezoelectric tensor with j that any nonzero out-of-plane component of piezoelectric indices in Voigt notation. We assume that i is zero in the field c′ will be assumed to be canceled by an external gate E M 1 ′ M E unstrained substrate. Recall that ǫ = (R )− (ǫ )R voltage. is the strain tensor derived from Eq.(23) but expressed in the unprimed x-basis. Even though the facet orienta- tions (110) and (111) have no shear strain in the growth VI. VALLEY SCATTERING UNIT CELL a-basis, they do have shear strain in the crystal x-basis. The (001) facet orientation, on the other hand does not In addition to calculating valley subband energies and have any shear strain. The strain in Eq.(40) is repre- degeneracies, it is useful to visualize the valley orienta- sented in Voigt notation as for Eq. (26) with j =1, 2, ...6. tions in three dimensions in order to map relevant inter- For zinc blende crystals, the only nonzero piezoelectric 27 valley scattering vectors. In this section, we lay down a coefficients are ex,4 = ey,5 = ez,6 and we obtain general derivation of a 3D primitive unit cell which we call the valley scattering unit cell (VSC), that aids in ǫ1 identifying in-plane and out-of-plane intervalley scatter- ǫ2 ing vectors. The proposed VSC is not a 3D Brillouin x 0 0 0 ex,4 0 0   E 1 ǫ3 zone, but a simple unit cell defined to have the 2D sym-  y  = −  000 0 ex,4 0    . E εsε0  ǫ4  metry of the Miller index plane of the QW, while pre- z 000 0 0 ex,4   serving the total volume of the original 3D Brillouin zone.  E     ǫ5        This cell is to be also distinguished from the standard de-  ǫ   6  piction of a 2D Brillouin zone which neglects information   (41) about the out-of-plane momentum which is necessary for Table I lists the piezoelectric coefficients for GaAs and calculating momentum scattering matrix elements. AlAs, and the coefficient for Al Ga As can be linearly x 1 x To define such a unit cell, we first identify the 2D sub- interpolated according to the x-Al content− 27. The general set of reciprocal Bravais lattice points which lie in the shear strain components in a QW are reproduced from Miller index plane of the QW. The original 3D Bravais Cadini and Stark24,55 and simultaneously derived using lattice is composed of parallel planes of this 2D sublat- Eq.(29), Eq.(23), Eq.(30) and Eq.(31). It is useful to tice which are displaced and separated by an interplanar express the electric field in the primed growth a-basis wavevector 2π/d. We now take the primitive unit cell of this 2D sublattice and extend it in the perpendicular di- ′a ǫyz E 2ex,4 M rection by an amount π/d, resulting in a primitive unit  ′b  = − R  ǫzx  . (42) ± E εsε0 cell with the same volume as the original 3D Brillouin ′c ǫxy zone. We call the resulting unit cell the valley scattering  E        unit cell. We define coordinates for each valley τ so that The c-component of the piezoelectric field ′ will affect Ec it lies within the VSC. the electron subband energy by modifying the confine- QW structures induce characteristically anisotropic ment potential V (c, k) of Eq.(12) both in the strained scattering potentials, defined in the a-basis as Vs(a, b, c), well and in the strained barrier which can result in anisotropic intervalley scattering. V (c, k)= (43) Scatterers such as a single monolayer step in the side- wall of a QW or a miscut substrate with periodic side- q ′ c if c < W/2; wall steps are sharp on the order of a single lattice period Ec,W | | and could in principle induce large momentum intervalley  V0(k)+ q c′,W W/2+ q c′,B (c W/2) if c>W/2; E E − scattering in the k-space plane parallel to the QW.  V0(k) q ′ W/2+ q ′ (c + W/2) if c< W/2, − Ec,W Ec,B − Recalling that the intervalley scattering vector is Qττ ,  we calculate the intervalley scattering matrix element, where c,′ W and c,′ B are the out-of-plane c-components of the electricE fieldE in the well and barrier, respectively. 27 v = ψ (r) V (a, b, c) ψ (r) (44) As pointed out by Adachi for (111) GaAs and AlAs, ττ h τ | s | τ i this piezoelectric field points from the (111)B surface (As- iQττ (a+b+c) = φ(r)uqτ (r) Vs(a, b, c)e · φ(r)uqτ (r) terminated) to the (111)A surface (Ga-terminated) when h | | i ττ ττ 3 iQ⊥ c iQk (a+b) under positive c-axis strain. = d rφ∗(r)uq∗ τ (r)Vs(a, b, c)e · e · The piezoelectric field in the QW and barrier mate- Z rial alters both the quantum confinement energy and the φ(r)uqτ (r) (45) wavefunction, and in the limit of weak piezoelectric fields τ the energy shift can be estimated to be Epz = c′W/2. For To emphasize the importance of the out-of-plane inter- example, in strained (111) AlAs on a GaAs substrate,E the valley scattering vector, we write Qττ in the exponential 6 ττ piezoelectric field is c′ =8.794 10 V/m, so for a QW as the sum of an out-of-plane component Q and the width of W = 10 nmE the piezoelectric× energy shift would in-plane vector Qττ . If we use the standard⊥ planar 2D τ k be around Epz = 43.97 meV. Brillouin zone view, we lose the representational impor- 8

FIG. 2: Depiction of the valley degeneracies for various QW orientations. Each row is labeled with its Miller index. Left column: 3D representation of the X-valleys showing both the CCC kx,ky,kz and the transport basis ka,kb,kc. In all diagrams, the doubly-degenerate Xx,y valleys are red, and the singly-degenerate Xz valley is blue. For the (111)-oriented AlAs QW, the three degenerate Xx,y,z valleys are purple. We note that the red and purple valleys exhibit robust valley degeneracy condition whilst the intersection point of the red and blue traces identifies the crossover degeneracy condition. The purple valleys satisfy the robust valley degeneracy condition as well as the crossover degeneracy condition trivially for all well widths. Center column: 2D projection and 3D representation of the valley scattering unit cell described in the text. Right column: Valley degenerate ground energy as a function of QW width for a QW barrier of Al0.45Ga0.55As at T = 4 K. Dashed lines represent the calculated strain energy shift of the respective electron valley relative to unstrained AlAs. The results of the heterostructure simulation software nextnano are depicted with scatter plots, and are verified with the analytical calculations shown as continuous lines. The (111) and (411) calculations assume that the piezoelectric field has been canceled by a top gate to result in a flat QW. tance of the out-of-plane scattering component. There- fore, we choose to depict the VSC in all three dimensions. 9

Specific examples will be shown in the following section (0.566110 nm) is the lattice constant of the AlAs at 0 K 52 which demonstrate how all possible intervalley scattering (300 K) . The perpendicular strain component ǫ′ is M ⊥ events can be easily identified in such a cell. calculated using Eq. (23) where D0 is the biaxial Pois- son ratio in the respective growth direction. The strain energy is determined from the absolute deformation po- X VII. VALLEY DEGENERACY AND THE tential for uniaxial strain at the X-point, Ξu = 6.11 eV VALLEY SCATTERING UNIT CELL FOR ALAS in AlAs. QWS In the following subsections, we compare and contrast the calculated results for the (001)-, (110)-, (111)-, and We apply the above analysis to the AlAs multi-valley (411)-oriented QWs, as summarized in Fig. 2. In the left system, determining the valley vectors, mass tensors, and panel of the figure, the transport basis vector c is taken strain tensors for the various growth directions, and then along the growth direction. The differently colored el- we calculate the valley subband energies as a function lipsoids distinguish the triple (purple), double (red),b and of well width W for each orientation. The three de- single (blue) degenerate valleys, each of these satisfy the generate conduction band electron valleys are composed robust degeneracy condition given by Eqs. (37) or (38). of six half valleys located at the X-points of the Bril- For the (001), (110) and (411) orientations, the strain en- louin zone edge, with valley vectors q100 = (2π/a) x, ergy breaks the three fold valley degeneracy into twofold 010 001 q = (2π/a) y, and q = (2π/a) z for the Xx,Xy, and valleys and a single valley. The growth direction governs Xz points, respectively. The superscripts express τ asb which of these valleys will have the lowest energy for momentum-space directions after the notation of Van de a given range of well widths. For the (111) orientation, 22 b b Walle . The AlAs electron valleys have anisotropic elec- symmetry dictates threefold degenerate Xx,y,z valleys for tron mass, with heavy longitudinal mass ml = 1.1 me and all well widths. 1,12,13 light transverse mass mt = 0.20 me, respectively . The right panel of Fig. 2 shows the results of the an- x mτ τ In the conventional crystal -basis, the mass tensor alytical simulation of E0 which gives the valley subband τ is diagonal with mii = ml for the mass component par- energies as a function of the well width. The horizon- τ τ allel to q and mjj = mt for the two transverse mass tal dashed lines indicate the pure strain splitting energy components repectively. ∆Eτ , equivalent to the energy splitting in the asymptotic Quantum confinement is created by sandwiching wide-well limit. The analytical results are confirmed by the AlAs QWs between aluminum gallium arsenide the semiconductor heterostructure simulation software (AlxGa1 xAs) layers which have a high aluminium con- nextnano60,61. For the crossover QW widths W in- − 0 tent x > 0.4. Properties of the barrier alloy are deter- dicated for the (001), (110) and (411) orientations, all mined by interpolating between AlAs and GaAs, either valleys satisfy the degeneracy condition of Eq. (33) simul- linearly or with a bowing term where applicable. We taneously. The intervalley scattering momentum vectors assume the reciprocal mass tensor in the AlxGa1 xAs Qττ for all the facets are listed in Table II. barrier layer to follow a linear interpolation for use− in Eq. (10) and (11)56,

τ τ τ A. (001) AlAs w′ = (x)w′ + (1 x)w′ (46) AlxGa(1−x) As AlAs − GaAs τ 52 For the (001)-oriented QW, the crystal axes are iden- For AlxGa1 xAs, EB is given by the relation − tical to the growth axes, Fig. 2 (top row), thus the co- τ τ τ ordinate transformation matrix is the identity matrix EAlxGa1−xAs = xEAlAs +(1 x)EGaAs bx(1 x) . (47) − − − R001 = I. The inverse mass tensor follows trivially. The 52 57 τ For AlxGa1 xAs , the bowing term b is 0.055 eV for various mass parameters listed in Table II are then used τ − τ to calculate the respective energy terms in Eq. (7). The EGaAs defined to be zero and EAlAs =0.259 eV where τ can be Xx, Xy or Xz respectively. biaxial Poisson ratio derived from Section III is given by22 DM = 2 c12 , and the diagonal strain matrix has For all strain calculations, we assume AlAs to 0 c11 001 001 be strained with respect to the GaAs substrate because in D1 = D2 = 0. The strain tensor in the a-basis is typical structures the thickness of the intervening buffer AlxGa1 xAs layer is below the critical strain relaxation 10 0 thickness− of 0.5 µm58,59. Using Eq. (22) we can calculate  01 0  the in-plane strain from the lattice mismatch between the ǫ′ = ǫ′ . (49) AlAs QW and the GaAs substrate   k  0 0 0.854    aGaAs aAlAs  −  ǫ′ = − , (48)   k aAlAs The results of the Section III analysis for (001)-oriented where aGaAs = 0.564177 nm (0.565325 nm) is the lattice QWs including 2D density-of-states and cyclotron mass constant of the GaAs substrate and aAlAs = 0.565252 nm for each valley are provided in Table II. 10

M (001) τ [100] [010] [001] 0.91 0 0 5 0 0 5 0 0 Q100,010 = ( 1 1 0) 2π       − a w′τ = wτ 010,001 2π 0 5 0 0 0.91 0 0 5 0 Q =( 0 1 0) a       001,100 2π  0 0 5   0 0 5   0 0 0.91  Q =( 1 0 0) a τ m2D [me] 0.469 0.200 nτ [cm−2meV−1] 3.92 1011 8.36 1010 2D × × M (110) τ [100] [010] [001] 100,010 2π 5 0 0 5 0 0 0.91 0 0 Q = ( 0 1.4142 0 ) a ′τ       − w 0 2.95 2.05 0 2.95 2.05 0 5 0 Q010,001 =(1 0.7071 0.7071) 2π  −      − a  0 2.05 2.95   0 2.05 2.95   0 0 5  Q001,100 = ( 1 0.7071 0.7071) 2π − a τ − m2D [me] 0.260 0.469 nτ [cm−2meV−1] 2.18 1011 1.96 1011 2D × × M (111) τ [100] [010] [001] 100,010 2π 2.95 1.18 1.67 2.95 1.18 1.67 5 0 0 Q = (1.2247 2.1213 0) a ′τ  − −      − w 1.18 4.32 0.96 1.18 4.32 0.96 0 2.27 1.93 Q010,001 = (2.4494 0 0) 2π  − −   −    a  1.67 0.96 3.64   1.67 0.96 3.64   0 1.93 3.64  Q001,100 = (1.2247 0.7071 0) 2π − a τ − − − m2D [me] 0.296 nτ [cm−2meV−1] 3.72 1011 2D × M (411) τ [100] [010] [001] 100,010 2π 5 0 0 2.95 1.93 0.68 2.95 1.93 0.68 Q = ( 0.7071 0.6667 0.2357) a ′τ    − −    − w 0 4.55 1.29 1.93 3.18 0.64 1.93 3.18 0.64 Q010,001 = ( 1.4141 0 0 ) 2π    − −   −  − a  0 1.29 1.36   0.68 0.64 4.77   0.68 0.64 4.77  Q001,100 = ( 0.7071 0.6667 0.2357) 2π − − a τ − − − m2D [me] 0.210 0.420 nτ [cm−2meV−1] 8.77 1010 3.52 1011 2D × × ′ TABLE II: Reciprocal mass tensors w τ for the various valleys in the transport basis for (001)-, (110)-, (111)-, and (411)- oriented AlAs QWs. Intervalley scattering vectors Qττ are listed, as are the 2D density-of-states (cyclotron) masses relative to τ the free electron mass. The 2D density-of-states values n2D are combined for robust-degenerate valleys.

The center panel of Fig. 2 (001) depicts the VSC and its B. (110) AlAs 2D projection. The 2D sublattice of coplanar reciprocal lattice points is square symmetric, with the Xz valleys For the (110)-oriented QW in Fig. 2 (110, left), we ap- lying at the corners of the unit cell and the Xx,y valleys at ply the rules of axis identification described in Section the edges. The resulting VSC illustrates that all valleys III. a lies along the lowest Miller index [001] direction, can be connected by purely in-plane scattering events, making the Xz valley in-plane, and b along [110] com- which is not obvious from the usual depiction of (001) pletes the right handed coordinate system. The coordi- valleys. This can be verified in Table II where the c- b nate transformation matrix R110 is b component of the scattering vector Qττ is zero for all intervalley scattering vectors. ⊥ 0 01 110 1 1 R − 0 =  √2 √2  . (50) 1 1 0  √2 √2  As seen in Fig. 2 (001, right), strain alone as indicated   The biaxial Poisson ratio derived from Section III is given by the horizontal dashed lines contributes an energy shift 22 M c11+3c12 2c44 100,010 by D = − , and the diagonal strain matrix of ∆E = 11.62 meV for the doubly-degenerate 0 c11+c12+2c44 − 110 110 Xx,y valleys while raising the singly-degenerate strained has D1 = D2 = 0. 001 Xz valley by ∆E = +9.93 meV, a net difference of The center panel of Fig. 2 (110) depicts the VSC and ∆=21.55meV. The ordering of the degeneracy changes its 2D projection. The strain tensor is diagonal in the 001 a-basis and given by at the crossover well width W0 = 5.4 nm. 10 0  01 0  Because (001) biaxial strain induces no shear compo- ǫ′ = ǫ′ . (51) k nent in the crystal basis, there are no piezoelectric fields    0 0 0.616  for this facet.    −    11

The results of Section III analysis for (110)-oriented QWs a-basis is including 2D density-of-states and cyclotron mass for each valley are provided in Table II. 10 0 The 2D sublattice which defines the VSC has a  01 0  centered-rectangular symmetry. The Xz valley lies in ǫ′ = ǫ′ . (55) this plane and the X valleys lie outside of the central   k x,y  0 0 0.55  plane. In the VSC, it is clear that whereas the Xx,y val-    −  leys can scatter amongst each other with purely in-plane   scattering, the Xz valley is isolated and requires an out- The results of the Section III analysis for (111)-oriented of-plane component for intervalley scattering. Table II QWs including 2D density-of-states and cyclotron mass reflects this result since only the Q100,010 scattering vec- for each valley are provided in Table II. The (111) QW tor has a zero c-component. valleys remain threefold degenerate due to equal contri- As seen in Fig. 2 (110, right), the singly-degenerate Xz butions of the strain and confinement energies for all elec- valley has the lowest strain energy E001 = 12.94 meV tron valleys as seen in Fig. 2 (111) left. − whereas the doubly-degenerate strained Xx,y valleys have The center panel of Fig. 2 (111) depicts the VSC and its an energy of E100,010 = 3.55 meV due to the smaller 2D projection. The planar sublattice is hexagonal, with − compressive strain. A smaller strain differential of ∆ = six half-valleys located at the center of the hexagonal 10.77 meV is observed between the valleys in this orien- facets. The resulting VSC shows that the valleys are all tation as compared to the (001) orientation. We observe connected with coplanar scattering vectors, as seen in Ta- 110 ττ the valley degeneracy crossover at W0 = 5.3 nm. ble II from the zero c-component of all Q . Fig. 2 (111, A purely in-plane piezoelectric field arises in (110) right), shows that all the valleys in this triply-degenerate QWs due to the nonzero shear component of the strain system are equally strained to ∆E100,010,001 = 7.0 meV tensor in the x-basis. When Eq.(23) is transformed to and the robust degeneracy condition given by Eq.− (33) is the x-basis, the shear components take the form55 trivially satisfied for all well widths. A purely out-of-plane piezoelectric field arises in (111) C11 2C12 ǫxy = C−+C−+2C ǫ′ ; (52) QWs due to the nonzero shear component of the strain 11 12 44 || ǫyz = ǫzx =0 tensor in the x-basis. When Eq.(23) is transformed to the x-basis, the shear components take the form24,27 leading to an electric field in the QW plane parallel to the [100] a-direction C11 2C12 ǫxy = ǫyz = ǫzx = − − ǫ′ , (56) C11 +2C12 +4C44 || 2ex,4 ′ = − ǫ . (53) Ea ε ε xy leading to an electric field perpendicular to the QW plane s 0 parallel to the c-growth axis: 6 This electric field a′ = 7.942 10 V/m plays no role in quantum confinementE so that× the flat QW assumption 2ex,4√3 c′ = − ǫxy . (57) remains valid, and the field is screened in-plane by the E εsε0 electrons in the QW within a Thomas-Fermi screening length of the sample edges at a. We will assume that this electric field c′ = 8.794 ± 106 V/m will be canceled by a gate voltage,E so that the× flat QW assumption remains valid. C. (111) AlAs D. (411) AlAs For the (111)-oriented QW, a is chosen along the lowest Miller index [110] direction perpendicular to the growth To demonstrate the utility of the formalism described axis c, and none of the valleysb lie within the plane. b in this paper, we now extend our analysis to the less con- is chosen along the [112] direction to complete the right 20 ventional (411) AlAs QW which has nonetheless shown handed coordinate system . The coordinate transfor-b 6,62 b 111 important results in the literature . Applying the same mation matrix R is rules of axis identification, a is chosen along the [011] di- 1 1 rection which is the lowest Miller index along the plane − 0 √2 √2 of the QW. b is chosen along the [122] direction to com- 111 1 1 2 b R = − . (54)  √6 √6 √6  plete the right-handed coordinate system. The coordi- 1 1 1 nate tranformationb matrix R411 is  √ √ √   3 3 3    1 1 0 − The biaxial Poisson ratio derived from Section III is given √2 √2 411 22 M c11+2c12 2c44 1 2 2 by D = 2 − , and the strain matrix is di- R = − . (58) 0 c11+2c12+4c44  3 3 3  agonal with D111 = D111 = 0. The strain tensor in the 4 1 1 1 2  √18 √18 √18    12

From Fig. 2 (411, right), we obtain a crossover width 411 W0 = 5.0 nm, and the valley occupation as a func- tion of well width follows an analogous discussion as for the (110) case. However, we observe that the Xx val- ley is singly-degenerate and the Xy,z valleys are doubly- degenerate. The strain component of the Xx valley en- 100 ergy is ∆E = 6.82 meV, and that of the Xy,z val- leys which satisfy the robust valley degeneracy condi- tion is ∆E010 = 12.29 meV. A strain differential of ∆ = 19.11 meV is− obtained. The two degenerate val- leys have coplanar momentum scattering as seen in the vector Q010,001 which has zero c-component. A piezoelectric field with both in-plane and out-of- plane components arises in (411) QWs due to the nonzero shear component of the strain tensor in the x-basis. When Eq. (23) is transformed to the x-basis, the shear components take the form24 FIG. 3: Illustration of shear strain for the (411)-oriented QWs 411 due to the nonzero strain-to-shear ratio D1 . ∆a,∆b and ∆c represent strain displacements along the respective vectors of the transport a-basis, and α represents the shear vector in ǫxy = ǫzx = (C11+2C12)(17C11 17C12+2C44) the plane of the QW, which for the (411)-oriented QWs is 2 − − ǫ 8C11 (16C12 C44)(C12+2C44)+C11(8C12+145C44) || parallel to the b-axis. − − (61) ǫyz = (C11+2C12)(8C11 8C12 7C44) 2 − − − ǫ , 8C11 (16C12 C44)(C12+2C44)+C11(8C12+145C44) || Using the derivation for the strain ratios from Section III, − − 411 for the (411)-oriented QWs, we obtain D0 =0.775 and leading to an electric field with components both in the 411 D2 =0.176. The explicit relations for the biaxial Pois- QW a-b plane as well as parallel to the c-growth axis: son ratios for AlAs (411)-QWs are 1 ǫ + 1 ǫ 411 − √2 xy √2 zx D0 = 2ex,4 4 1 E ′ = . (62)  + 3 ǫxy 3 ǫyz  6(c +2c )(4c 4c + 19c ) − εsε0 − 11 12 11 − 12 44 1 2 ǫ + 4 ǫ 8c2 (16c c )(c +2c )+ c (8c + 145c ) −  √18 xy √18 yz  11 12 44 12 44 11 12 44   − − 411 D2 = We will assume that the out-of-plane piezoelectric field 6 √ component ′ = 1.783 10 V/m will be cancelled by a 15 2(c11 +2c12)(c11 c12 2c44) Ec × 2 − − − . gate voltage so that the square QW assumption remains 8c11 (16c12 c44)(c12 +2c44)+ c11(8c12 + 145c44) − − valid. The piezoelectric field component a′ is zero, and (59) E6 the in-plane component ′ = 3.761 10 V/m will be Eb × The results of Section III analysis for (411) including screened at the sample edges by the electrons in the QW. 2D density-of-states and cyclotron mass for each valley are provided in Table II. For this high-index facet, the strain tensor is nondiagonal in the a-basis and given by VIII. THE VALLEY SCATTERING UNIT CELL AND ANISOTROPIC INTERVALLEY 10 0 SCATTERING FOR ALAS QWS  01 0.176  ǫ′ = ǫ′ . (60) This section first discusses the differences between the   k VSC and standard depiction of the 2D Brillouin zone,  0 0.176 0.775    and then determines the inter- to intravalley scattering  −    ratio for valleys near crossover degeneracy. In the VSC The biaxial strain in the structure grown along [411] description for the case of (111)AlAs, Fig. 4 shows the gives rise to nondiagonal components in the strain tensor. (111)AlAs VSC, as well as the 2D Bravais lattice of these The effect of shear is depicted in Fig. 3 by the shear cells. For comparison with this hexagonal unit cell, the vector α (Eq. (32)) which lies in the plane of the QW and standard 2D Brillouin zone is plotted as a smaller grey is parallel to the b-direction. The degeneracy condition hexagon within this lattice21. The hexagons A, B and C 411 takes into account the shear component D1 and is given are intended to illustrate the relative positions of three by Eq. (38). identical planar hexagonal lattices, displaced from each Fig. 2 (411, center) depicts the VSC and its 2D pro- other. When these three layers are stacked in sequence jection. The 2D planar sublattice in Fig. 2 (411, center) ABCABC out of the plane we recover the full 3D recip- has centered rectangular symmetry. rocal lattice. Whereas the standard 2D Brillouin zone is 13

FIG. 4: 3D representation of the valley scattering cell for (111)AlAs as well as 2D lattice of such cells. In AlAs, the valleys are located at the q = (2π/a)( xb, yb, bz) The standard depiction of 2D Brillouin zone (small shaded hexagon)21 projects together three identical but laterally and vertically displaced planar hexagonal lattices, translated according to the representative hexagonal cells A, B and C. The area of standard 2D Brillouin zone is one third that of the valley scattering cell since its height is three times that of the valley scattering cell. Arrows illustrate the planar intervalley scattering vectors Qττ for (111)AlAs. defined from the overlay of all three layers, we define the the roughness correlation length along the interface plane VSC from the 2D Brillouin zone of a single layer only. Λ, and is expressed as the Gaussian autocorrelation The area of the hexagonal VSC for (111) is thus three function64 times larger than that of the traditional hexagonal unit 2 cell. The coplanar intervalley momentum scattering vec- 2 r r′ < ∆(r)∆(r′) >= ∆ exp | − 2 | , (63) tors are shown with arrows in the VSC. − Λ ! The VSC is helpful in depicting intervalley scatter- ing events, so we will discuss the two main sources of where < ... > means an ensemble average. The Fourier single-particle scattering for AlAs QWs, namely inter- transform of the roughness autocorrelation at the inter- face roughness and alloy disorder scattering in the bar- valley scattering vector Qττ is rier. We find below that near valley degeneracy, interface ττ 2 roughness scattering does not result in any significant in- 2 2 2 (ΛQ ) < ∆Qττ >= π(∆ Λ )exp . (64) tervalley scattering, and in the presence of alloy scatter- | | "− 4 # ing in the barrier walls, the scattering rate for crossover 65,66 intervalley scattering is significantly repressed relative to Following Quang et al. the scattering potential URS robust intervalley scattering, and this suppression factor for the QW with a square potential and a well width W is calculated. is given by We assume low enough temperatures such that acous- tic and optical phonons can be neglected. Elastic in- ~2 τ tervalley scattering can only occur if the valley splitting ττ 2 w′ W,cc 2 < U >= A < ∆ ττ > , (65) energy is less than the Fermi energy at low temperatures, | RS | 2W 3 | Q | τ τ ! E E < EF, or the thermal energy at high tempera- | − τ| τ tures E E < kBT . Two independent processes can where w τ is the inverse confinement mass in the well | − | ′ W,cc in principle result in elastic intervalley scattering: inter- perpendicular to the interface and A is a unitless nor- face roughness scattering and alloy disorder scattering malization constant. Under optimal epitaxial conditions due to wavefunction penetration into the barrier alloy. of slip-step growth, the AlAs roughness disorder poten- Interface roughness leads to local changes in the well tial will have correlation lengths of tens of lattice periods width ∆(r) at an in-plane position r = (a,b), and the Λ= Na. Since the intervalley scattering vector Qττ is of resulting fluctuation in the quantization energy causes order a Brillouin zone boundary π/a and the intravalley a scattering potential63. The roughness is characterized scattering vector is of order zero, the ratio of intervalley by the rms average displacement of the interface ∆ and scattering to intravalley scattering is of order exp( N 2), − 14 which even for a conservative estimate of N = 5 lattice is expected to be small since the Fermi wavevector kF periods of autocorrelation already yields a suppression of is much smaller than the intervalley scattering vector68 more than 10 orders of magnitude for intervalley scat- Qττ . tering. Thus intervalley surface-roughness scattering is significantly suppressed with respect to intravalley scat- tering. IX. VALLEY DEGENERACY IN MISCUT Alloy scattering is the other possible mechanism for SAMPLES elastic intervalley scattering. The standard model for alloy disorder scattering65,67 with variable alloy compo- In this final section, we show how the analysis of Sec- sition x(z) is described by the autocorrelation function tion III can be applied to miscut samples to determine the projected masses, ground state energies, and strain 3 2 ττ 2 a0V0 ∞ 2 2 energies. Miscut samples are prevalent in the literature, < UAD >= ψτ (c)x(c)[1 x(c)]ψτ (c)dc , | | 8 W − for example (111) GaAs/AlAs growth has been shown to Z 2 (66) have superior morphology with a miscut angle from 0.5◦ where a0 is the in-plane lattice constant, and V0 is the to 4◦ and in general, intentional miscuts improve growth spatial average of the fluctuating alloy potential over the quality due to slip-step growth69–72. In Si, recently in- alloy unit cell. In AlAs QWs, only tails of the wavefunc- vestigated hydrogen terminated (111)Si miscut surfaces 73 tion have an alloy content in the AlxGa1 xAs barrier. have shown very high mobility and the wafer miscut is One notes that the wave functions of electrons− in valleys expected to break the valley degeneracy43,74. with a light confinement mass penetrate to a greater ex- Miscut samples are characterized by two angles. One tent in the barrier and will have a significantly larger alloy angle φ designates the azimuthal angle in the a-b plane scattering potential as compared to the heavy confine- relative to the a-axis towards which the plane is tilted, of- ment mass. Assuming valley degeneracy for all valleys ten expressed as an in-plane Miller index tilt direction T . and plane wave solution for the electron wavefunction in The other angle θ designates the polar tilt angle relative the QW barrier, the scattering matrix element can be to the surface normal c. Experimentally, the angles can written as be deduced from atomic force microscope images of sur- face monolayer steps, where φ is oriented perpendicular ττ 2 b < UAD >= to the steps, and θ = atan(a/2L) is deduced from know- | 3 | 2 a0V0 ∞ 2 2κτ c 2 2κτ c ing the monolayer thickness a/2 and the average width Aτ e− x(c)[1 x(c)]Aτ e− dc 8 W − of the monolayer steps L. We define a new coordinate 2 M Z transformation matrix R ∗ for the miscut samples, by ,(67) introducing the azimuthal and polar rotations Rφ and Rθ, respectively: where Aτ = ψτ (W/2) and Aτ = ψτ (W/2) are the boundary values| of the| wavefunction| at the barrier,| ob- tained by solving for boundary conditions of the square cosφ sinφ 0 wave potential. − R R We get a quantitative estimate of the ratio of the al- T = φ =  sinφ cosφ 0  (69) loy scattering γ between crossover-degenerate valleys 0 01 ×   labelled τ1 and τ3 compared to the scattering γ be-   k tween robust-degenerate valleys labeled τ1 and τ2 in the AlxGa1 xAs barrier by calculating cosθ 0 sinθ − Rθ = 0 1 0 . (70) < U τ1τ3 2> γ   AD = × = (68) sinθ 0 cosθ | τ1τ2 |2 < UAD > γ  −  | | k   2 2κτ1 c 2 2κτ3 c W∞ We obtain a new coordinate transformation matrix Aτ1 e− x(c)[1 x(c)]Aτ3 e− dc 2 − RM for calculating the transport parameters and the 2 2κτ1 c 2 2κτ2 c θ,φ W∞ R Aτ1 e− x(c)[1 x(c)]Aτ2 e− dc 2 − strain tensor analysis of miscut samples where c is theR direction of integration out-of-plane. For M 1 M Rθ,φ = (RφRθRφ− )R . (71) the case of (001)-, (110)-, and (411)-oriented AlAs QWs, γ× = 0.0012, 0.0691 and 0.0235 respectively. The val- γk Note that a VSC cannot be defined for miscut samples, ley with the larger confinement mass penetrates less into but only for those with a pure Miller index. the barrier region and results in a smaller scattering in- We outline here the procedure for determining valley tegral ratio. The VSC makes it easy to visually identify subband energies under miscut. First, the transforma- M this anisotropic scattering possibility along the various tion matrix Rθ,φ from Eq. (71) determines the inverse orientations. mass tensor of each valley with Eq. (9), and the out- Electron-electron intervalley scattering is outside the of-plane inverse mass tensor component can be used in scope of this article and will be treated elsewhere. It Eq. (10) to determine the ground confinement energies 15 of each valley. The in-plane inverse mass tensor is then channel not present in single valley systems, namely calculated with Eq. (14) and used in Eq. (15) to obtain inter-valley scattering. We identified the main inter- the kinetic energy. The stress tensor can be determined valley scattering mechanisms as the interface roughness M M by applying Rθ,φ in Eq. (29), and the strain ratios Di and alloy disorder scattering due to wavefunction pene- can be determined by minimizing the elastic energy with tration into the barrier and calculated the suppression Eqs. (28)-(30). Finally, the strain energy shift can be factor at the crossover well width W0 for inter-valley deduced from Eq. (19) and added to the ground quan- scattering between different sets of valleys. We also tum confinement energies and kinetic energies to arrive defined a valley scattering primitive unit cell to easily at the final valley energies. Any piezoelectric fields can identify scattering events between the robust-degenerate be determined by transforming the strain tensor to the and crossover-degenerate electron valleys. We drew the unprimed frame ǫ and inserting the appropriate shear VSC for the special case of AlAs grown along four differ- components into Eq. (42). ent orientations, the high-symmetry facets (001), (110), and (111), as well as a low-symmetry facet (411) to demonstrate the utility of our model. Furthermore, we X. CONCLUSION explained the relevance of the VSC description for vi- sual identification of anisotropic inter-valley scattering in AlAs QWs. In conclusion, we review the advantages and limita- tions of the formalism we have developed for valley sub- band energy calculations. We first derived how the valley index is a valid pseudospin index in a multi-valley sys- In the final section we demonstrated the power of our tem. A key element of our model is the definition of two formalism to determine valley degeneracy for arbitrary relevant bases: the conventional crystal x- and the trans- substrate orientation, without being restricted to Miller port a-basis. We use this notation to find the projected indexed planes as is prevalent in the literature. We con- in-plane and out-of-plane effective masses, the electron sider the example of a miscut sample and define the addi- subband energy and degeneracy, piezoelectric fields, and tional notation required to address angle deviation from the scattering vectors. conventional growth axes. We then detail the procedure There are five competing energy scales in multi-valley to calculate all the ground energy parameters as well as QWs namely in-plane kinetic energy, confinement energy, the strain tensor analysis for miscut samples. It is worth strain energy, piezoelectric energy, and Fermi energy, all repeating that the formalism developed here is sufficient of which influence valley occupancy. Strain in the QW to calculate valley degeneracy for any substrate orien- breaks valley degeneracy and results in an energy differ- tation, and need not be aligned with any Miller index τ ential, ∆E of 0 - 20 meV between non-degenerate val- facet. leys. The strain energy is independent of the QW width unlike the quantum confinement energy which increases as the QW width decreases and is inversely proportional τ to the confinement mass; typical confinement energies E0 There are some important limitations of our formal- are of the order of 20 meV for a narrow QW about 5 nm ism. Firstly, we assume layer thicknesses are all within wide and 5 meV for a 20 nm wide QW. We defined a ro- the strain relaxation limit such that there are no strain bust valley degeneracy condition to identify valleys which induced defects below or within the QW. These would are degenerate independent of well width, and a crossover relax the lattice constant relative to the substrate, thus degeneracy condition where valleys are degenerate only one would have to first determine the adjusted lattice for a particular crossover well width W0. We showed that constant at the QW layer and then apply the formalism piezoelectric fields in the QW and barrier materials can developed here to deduce valley degeneracy. Secondly, arise in certain facets such as (111) and (411), and cause we assume nearly empty QWs and do not calculate the τ piezoelectric energy shifts Epz as large as 50 meV for a Schr¨odinger equation self-consistently with the Poisson (111) QW about 10 nm wide, and that such a piezoelec- equation. However our formalism for calculating the var- tric field can be canceled by an external gate voltage. ious valley subband energies can be applied to standard The smallest energy scale in these QWs is the Fermi en- self-consistent solvers to obtain the proper Hartree solu- ergy which is a few meV depending on the degeneracy tion. Lastly, whereas the formalism takes into account 11 2 for electron densities around n2D 3 10 cm− and the change in the confinement potential due to piezoelec- τ ∼ × assuming an effective mass m2D 0.3me. With such a tric fields, it does not address larger structural consider- small Fermi energy, valley occupation∼ is strongly defined ations such as modulation doping layers, surface pinning by the interplay of the various other energy scales in the potentials, electrostatic charges, and piezoelectric fields problem. We once again emphasize that it is due to the in the barrier, all of which would have to be calculated similarity of all the energy scales that none of them can together to determine the confinement well potential and be ignored and must be treated carefully to calculate val- Fermi energy. We note, however, that the piezoelectric ley subband energy. equations provided here can be used to determine the Valley degenerate systems have an extra scattering piezoelectric fields in the strained barriers. 16

XI. ACKNOWLEDGMENTS Center of Northwestern University, and the BMBF Nano- Quit Project 01BM470. SPG and MG would like to This work was funded by the NSF grant DMR- thank Wade DeGottardi, Wang Zhou, Florian Herzog, 0748856, the MRSEC program of the National Science and Jens Koch for helpful discussions. Foundation DMR-0520513 at the Materials Research

1 M. Shayegan, E. P. De Poortere, O. Gunawan, Y. P. Shkol- 25 L. De Caro, and L. Tapfer, Phys Rev B 48, 2298 (1993). nikov, E. Tutuc, and K. Vakili, Phys Status Solidi B 243, 26 K. Yang, T. Anan, and L. J. Schowalter, Appl Phys Lett 3629 (2006). 65, 22 (1994). 2 T. P. Smith, W. I. Wang, F. F. Fang, and L. L. Chang, 27 S. Adachi, in GaAs and related materials: bulk semicon- Phys Rev B 35, 9349 (1987). ducting and superlattice properties, 122-129 (World Scien- 3 T. S. Lay, J. J. Heremans, Y. W. Suen, M. B. Santos, tific Publishing Co. Pte. Ltd., 1994). K. Hirakawa, M. Shayegan, and A. Zrenner, Appl Phys 28 T. Hammerschmidt, P. Kratzer, and M. Scheffler, Phys Lett 62, 3120 (1993). Rev B 75, 235328 (2007). 4 S. Dasgupta, C. Knaak, J. Moser, M. Bichler, S. F. Roth, 29 F. Sch¨affler, Semicond Sci Tech 12, 1515 (1997). A. F. I. Morral, G. Abstreiter, and M. Grayson, Appl Phys 30 R. D. Graft, D. J. Lohrmann, G. P. Parravicini and L. Lett 91, 14 (2007). Resca, Phys Rev B 36, 4782 (1987). 5 S. Dasgupta, S. Birner, C. Knaak, M. Bichler, A. F. I. 31 T. B. Boykin, G. Klimeck, M. Friesen, S. N. Coppersmith, Morral, G. Abstreiter, and M. Grayson, Appl Phys Lett P. vonAllmen, F. Oyafuso and S. Lee, Phys Rev B 70, 93, 13 (2008). 165325 (2004). 6 E. P. De Poortere, E. Tutuc, S. J. Papadakis, and 32 T. B. Boykin, G. Klimeck, M. A. Eriksson, M. Friesen, M. Shayegan, Science 290, 1546 (2000). S. N. Coppersmith, P. von Allmen, F. Oyafuso and S. Lee, 7 K. Vakili, Y. P. Shkolnikov, E. Tutuc, N. C. Bishop, E. P. Appl Phys Lett 84, 1 (2004). De Poortere, and M. Shayegan, Physica E 34, 89 (2006). 33 M. O. Nestoklon, L. E. Golub and E. L. Ivchenko, Phys 8 H. W. vanKesteren, E. C. Cosman, P. Dawson, K. J. Rev B 73, 235334 (2006). Moore, and C. T. Foxon, Phys Rev B 39, 13426 (1989). 34 M. Friesen, M. A. Erikssonand S. N. Coppersmith, Appl 9 A. F. W. van de Stadt, P. M. Koenraad, J. A. A. J. Peren- Phys Lett 89, 202106 (2006). boom, and J. H. Wolter, Surf Sci 362, 521 (1996). 35 M.Virgilio,and G.Grosso, J Appl Phys 100, 9 (2006). 10 O. Gunawan, Y. P. Shkolnikov, K. Vakili, T. Gokmen, E. P. 36 M. Friesen, S. Chutia, C. Tahan,and S. N. Coppersmith, De Poortere, and M. Shayegan, Phys Rev Lett 97, 186404 Phys Rev B 75, 115318 (2007). (2006). 37 S. Goswami, M. Friesen, L. M. McGuire, J. L. Truitt, 11 K. Vakili, Y. P. Shkolnikov, E. Tutuc, E. P. De Poortere, C. Tahan, L. J. Klein, J. O. Chu, P. M. Mooney, D. W. van M. Padmanabhan, and M. Shayegan, Appl Phys Lett 89, der Weide, R. Joynt, S. N. Coppersmith,and M. A. Eriks- 17 (2006). son, Nat Phys 3, 1 (2007). 12 O. Gunawan, Y. P. Shkolnikov, E. P. De Poortere, E. Tu- 38 M. O. Nestoklon, E. L. Ivchenko, J. M. Jancu,and P. tuc, and M. Shayegan, Phys Rev Lett 93, 246603 (2004). Voisin, Phys Rev B 77, 155328 (2008). 13 O. Gunawan, E. P. De Poortere, and M. Shayegan, Phys 39 M. Virgilio,and G. Grosso, Phys Rev B 79, 165310 (2009). Rev B 75, 081304(R) (2007). 40 G. Frucci, L. Di Gaspare, F. Evangelisti, E. Giovine, A. 14 Y. P. Shkolnikov, K. Vakili, E. P. De Poortere, and Notargiacomo, V. Piazza,and F. Beltram, Phys Rev B 81, M. Shayegan, Phys Rev Lett 92, 246804 (2004). 195311 (2010). 15 J. Moser, T. Zibold, D. Schuh, M. Bichler, F. Ertl, G. Ab- 41 J. Kim and M. V. Fischetti, J Appl Phys 108, 1 (2010). streiter, M. Grayson, S. Roddaro, and V. Pellegrini, Appl 42 C. Euaruksakul, F. Chen, B. Tanto, C. S. Ritz, D. M. Phys Lett 87, 5 (2005). Paskiewicz, F. J. Himpsel, D. E. Savage, Z. Liu, Yugui 16 J. Moser, S. Roddaro, D. Schuh, M. Bichler, V. Pellegrini, Yao, Feng Liu, and M. G. Lagally Phys Rev B 80, 115323 and M. Grayson, Phys Rev B 74, 193307 (2006). (2009). 17 O. Gunawan, B. Habib, E. P. De Poortere, and 43 K. Eng, R. N. McFarland, and B. E. Kane, Phys Rev Lett M. Shayegan, Phys Rev B 74, 155436 (2006). 99, 016801 (2007). 18 T. Gokmen, M. Padmanabhan, and M. Shayegan, Nat 44 M. Eto, and Y. Hada, AIP Conf Proc 850 (2006). Phys 6, 621 (2010). 45 Y. Hada, and M. Eto, Jpn J Appl Phys 1 43, 10 (2004). 19 Y. P. Shkolnikov, S. Misra, N. C. Bishop, E. P. 46 Y. Hada, and M. Eto, Phys Rev B 68, 155322 (2003). De Poortere, and M. Shayegan, Phys Rev Lett 95, 066809 47 J. Davies, in The physics of low-dimensional semiconduc- (2005). tors (Cambridge University Press, 2006). 20 M. Rasolt, in Advances in Research and Applications, 48 C. Herring and E. Vogt, Phys Rev 101, 944 (1956). edited by H. Ehrenreich vol. 43 of Solid State Physics, pp. 49 M. P. C. M. Krijn, Semicond Sci Tech 6, 27 (1991). 93 – 228, (Academic Press, 1990). . 50 J. F. Nye, in Physical properties of crystals, their represen- 21 F. Stern and W. E. Howard, Phys Rev 163, 816 (1967). tation by tensors and matrices (Oxford University Press, 22 C. G. Van de Walle, Phys Rev B 39, 1871 (1989). 1957). 23 D. L. Smith, and C. Mailhiot, J Appl Phys 63, 8 (1988). 51 W. Voigt, in Lehrbuch der Kristallphysik (Johnson Reprint 24 E. A. Caridi, and J. B. Stark, Appl Phys Lett 60, 12 Corporation (Reprint of 1928 edition), New York, 1966). (1992). 52 I. Vurgaftman, J. R. Meyer, and L. R. Ram-Mohan, J Appl 17

Phys 89, 5815 (2001). tion presents a generalized treatment to express these 53 Landolt-Boernstein, vol. 17 of Halbleiter (Springer, 1982). strain ratios for an arbitrary substrate orientation with 54 S. H. Wei and A. Zunger, Appl Phys Lett 72, 2011 (1998). cubic symmetry, given its coordinate transformation ma- 55 Caridi and Stark incorrectly derived the equation for the M trix R , the elastic constants cijkl in the CCC x-basis strain tensor components of the (110) orientation. Follow- and the in-plane strain ǫ′ in the transport a-basis. ing the appendix of the present paper, the correct terms || can be derived. We first obtain the fourth order rotated elastic stiff- 56 S. Adachi, Adachi et al. found that in order to compare ness tensor cmnop′ in the a-basis by rotating the fourth- theory and experiment interpolating the reciprocal mass rank elastic stiffness tensor cijkl from the CCC basis followed by inverting it gives an equivalent result to invert- x = (x,y,z) to the transport basis a = (a,b,c) ing the reciprocal mass followed by interpolating it even though these two operations are mathematically different. 57 M M M M We have assumed that the bowing term from the expres- cmnop′ = RmiRnj RokRpl cijkl , (A.1) 52 sion for the bandgap in Ref. is accounted for entirely in ijkl the conduction band. X 58 M. Gutierrez, M. Herrera, D. Gonzalez, G. Aragon, J. J. M Sanchez, I. Izpura, M. Hopkinson, and R. Garcia, Micro- where R is the coordinate transformation matrix de- electr J 33, 553 (2002). fined in Eq. (8). The rotated fourth-rank tensor cmnop′ 59 C. Weisbuch and B. Winter, in Quantum Well Struc- in the CCC x-basis should then be mapped to Cij′ using tures. Fundamentals and Applications (Academic Press, the Voigt notation. The strain tensor components ǫij′ in New York, 1991). the transport a-basis are given by 60 C. Knaak, Diplomarbeit. Physics Department, Technische Universit¨at M¨unchen (2007). 61 asubstrate alayer www.nextnano.de (2007). ǫaa′ = ǫbb′ = ǫ′ = − (A.2) 62 E. D. Poortere, PhD thesis, Department of Electrical En- k alayer gineering, Princeton University (2003). ǫ′ = ǫ′ = 0 (A.3) 63 ab ba H. Sakaki, T. Noda, K. Hirakawa, M. Tanaka, and T. Mat- λµ ηω susue, Appl Phys Lett 51, 23 (1987). M ǫac′ = ǫca′ = − 2 ǫ′ = D1 ǫ′ (A.4) 64 λκ η k k T. Ando, A.B Fowler,and F. Stern, Rev Mod Phys 54, 2 − (1982). ω ηDM 65 1 M D. N. Quang, N. H. Tung, D. T. Hien, and H. A. Huy, ǫbc′ = ǫcb′ = − ǫ′ = D2 ǫ′ (A.5) λ k k Phys Rev B 75, 073305 (2007). M M 66 D. N. Quang, V. N. Tuoc, and T. D. Huan, Phys Rev B α 2C34′ D2 2C35′ D1 ǫcc′ = ǫ′ = − − ǫ′ (A.6) 68, 195316 (2003). ⊥ C33′ k 67 F. Murphy-Armando, and S. Fahy, Phys. Rev. Lett. 97, M = D0 ǫ′ 096606 (2006). − k 68 W. Degottardi, private communication. 69 L. Vina,and W. I. Wang, Appl Phys Lett 48, 1 (1986). where the denominators in Eq. (A.4), Eq. (A.5) and 70 T. Hayakawa, M. Kondo, T. Suyama, K. Takahashi, S. Ya- Eq. (A.7) are always nonzero, the coefficients α, β and γ mamoto, and T. Hijikata, Jpn. J. Appl. Phys. 26, (1987). 71 are first order, and the coefficients λ, κ, η, ω and µ are A. Chin, P. Martin, P. Ho, J. Ballingall, T. H. Yu, and J. second order in the elastic stiffness tensor components Mazurowski, Appl Phys Lett 59, 15 (1990). 72 C K. Tsutsui, H. Mizukami, O. Ishiyama, S. Nakamura, and ij′ S. Furukawa, Jpn. J. Appl. Phys. 29 (1990). 73 K. Eng, R. N. McFarland, and B. E. Kane, Appl Phys Lett α = (C13′ + C23′ ) (A.7) 87, 5 (2005). − 74 β = (C′ + C′ ) (A.8) N. Kharche, S. Kim, T. B. Boykin, and G. Klimeck, Appl − 14 24 Phys Lett 94, 4 (2009). γ = (C′ + C′ ) (A.9) − 15 25 λ = 2C33′ C44′ (A.10) Appendix: Analytical equations of the strain tensor κ = 2C33′ C55′ (A.11) and the strain ratios for arbitrary substrate η = 2 (C33′ C45′ C34′ C35′ ) (A.12) orientations − ω = C′ β C′ α (A.13) 33 − 34 M µ = C33′ γ C35′ α . (A.14) In Section III, we derive the strain ratios Di by min- − imizing the free energy of the strained layer. This sec-