PHYSICAL REVIEW B VOLUME 51, NUMBER 23 15 SUNE 1995-1 k-p theory of energy bands, wave functions, and optical selection rules in strained tetrahedral semiconductors P. Enders, A. Birwoltf, and M. Woerner ‘Max-Born-Institut fir Nichtlineare Optik und Kurzzitspektroskopie, Rudower Chausse 6, D-12489 Berlin, Germany D. Suisky Institut fir Physik, Humboldt-Universtat2u Berlin, Invalidenstrasse 110, D-10115 Berlin, Germany (Received 14 February 1995) ‘This paper examines an eights :nd k-p theory of strained semiconductors yielding energy bands, wave functions, and momentum matrices. Only ifthe symmetry of the strained erystal is accounted for in all terms of the Hamiltonian, a consistent definition and calcul jon ofthe momentum matrix becomes pos- sible. The band structure and wave functions are nonanalytical functions of strain and crystal momen- tum. For strained crystals, the extrapolation from the T point into the Brillouin zone, such as the ‘effective-mass approximation for the optical-matrix elements, can be misleading. For certain cases, the hheavy- and light-hole isoenergetic surfaces form complex figures resembling the indicatrix of birefringent biaxial crystals. The symmetry of the hole wave functions causes dichroism for photon energies close to ‘the gap energy, while the crystal becomes optically isotropic for larger photon energies. Numerical re- sults are presented forthe eight-band k-p model of biaxilly strained bulklike 1.3um-In,Ga,-.A8,P.- on InP being an important material in optoelectronics. I, INTRODUCTION Strained tetrahedral semiconductors are superior to unstrained materials with respect to various properties and novel effects." For this, they are subject to intensive and comprehensive investigation. For theoretically mod- cling the band structure near the band edges, the k-p method" is the favorite tool. Calculating the electronic properties of direct band-gap semiconductors around the F point, at least an eight-band model is required, that comprises conduction band (C), heavy holes (HH), light holes (LH), and spin-orbit interaction splt-off band (SO) plus spin.” A comprehensive derivation of the corre- sponding Hamiltonian matrix elements has been given by Trebin, Réssler, and Ranvaud.’ Using group-theoretical methods, they extended the work by Bir and Pikus* and included an external quantizing magnetic feld, in order to extract the empirical k-p model parameters from quantum resonances in the valence bands of p-type InSb. ‘The group-theoretical approach has the advantage of more systematically accounting for symmetry, what facil itates the determination of independent model parame- ters, and the clasifiation of the system's behavior, eg, ‘optical transitions. Bahder® has derived various terms by explicitly performing the Pikus-Bir** transformation, with detailed treatment of the spin-orbit interaction terms. This approach is more obvious and will be adopt- ed here. All papers we are aware of account forthe strain strit- ly up to first order. Being consistent within the order of approximation, this has, nevertheless, two disadvantages. First, it violates the von Neumann’ principle, that the physical observables ofa system exhibit the same symme- ‘ry as this system.’ This is our main objective against (0163-1829/95/31(231/16695(101/806.00 a ‘such treatment. Second, the momentum operator, Pip =(™ /A)BH,9/8k, becomes inconsistent, when the matrix elements ‘of the Hamiltonian Hy, are treated nonuniformily. Furthermore, Kane’s* rotation of the z axis on the k axis becomes useless. The goal of this paper is to show that these shortcomings are removed immedi- ately after applying the Pikus-Bir transformation to the terms of second order in k also. Often, it is said that a TE electromagnetic field (the electrical feld vector El|xy plane) is (predominantly) ab- sorbed or amplified by the heavy holes, while a TM fleld (Bliz) is absorbed or amplified by the light holes. This holds true at the I point, if the z axis is the axis of quant- ization. However, the corresponding electron-hole tran- sitions take place not at the F point, since here the densi- ty of states vanishes, but involve states with k >0. Nonhydrostatic stress lifts the degeneracy of the HH and LH band edges (Ty states) and can cause an HH-LH band mixing and anticrossing of the P point (k >0; ef, eg., Chuang’). This means a significant change of the symmetry of the HH and LH wave functions, which is important for the efficiency of various scattering process- 8 or for the interaction with electromagnetic waves. In fact, the band energies are not analytical functions of strain and crystal momentum. "°~?? Calculating the wave functions and the momentum matrix elements as func tions of crystal momentum and stress, the symmetry changes of the wave functions will be shown to be the clue to the understanding of the irregular behavior of the ‘momentum matrix and, thus, oscillator strengths in such materials. In particular, the selection rules are strongly ‘modified for optical transitions near the T’ point. ‘Thus, this paper is organized as follows. In See. II, we first write down the 8X8 k-p Hamilton matrix with 16.695 (©1995 The American Physical Society16.696 strain on the base of the Pikus-Bir transformation. All terms used later for numerical computations are explicit- ly given, in order to facilitate comparison with other ap- proaches and the derivation of the momentum matrix. Some remarks on Kane's rotation of the coordinate sys- tem follow. Then, a consistent definition of the momen- tum matrix will be given which continues to obey the sum rule and the exact relation”? (vklpivk) 'm (MOE. /dk under strain, These results are com- pared with other approaches. In Sec. III, their quantita- tive consequences are illustrated through computations for biaxially strained bulklike 1.3,1m-In,Ga,_,As,P)_ ‘on InP. This is an interesting material for optoelectron- s, but so far its band structure has not been greatly in- vestigated by means of an eight-band model. First, the energy bands are considered. For certain parameter values, the HH and LH isoenergy surfaces form figures resembling the indicatrix of birefringent biaxial crys- tals."* The character of the band states is investigated by ‘means of the symmetry of the wave functions. The nu- ‘merical examination of the effective-mass approximations for the oscillator strength will demonstrate that extrapo- lations from the T point to the Brillouin zone k >0 are not generally allowed. Section IV summarizes the re- sults, The Appendix sketches the calculation of the oscil- lator strengths by means of a projection operator method, which circumvents the computation of the eigenfunctions of the k-p Hamiltonian. IL, k-p THEORY WITH STRAIN A. Hamilton matrix A homogeneous strain is a singular perturbation. This singularity can be removed by means of the Pikus-Bir 0 9 0 0 0 0 0 0-4 9 0 0 0 0 0 o 0 -4 0 0 © 0 0 0 0 0 -$0 0 0 0 H=!0 0 0 0 & 0 0 0 | A Jo 0 0 0 0-4 0 © E, is the fundamental gap of the unstrained material and A the spi ‘band edge (', states) lies at the energy P. ENDERS, A. BARWOLFF, M. WOERNER, AND D. SUISKY st transformation, ie. the transformation of the coordinate system of the strained crystal [x,} to that of the un- strained one {x}: x,=(1+e)-x, where € is the strain ten- sor. Correspondingly, the crystal momentum is transformed as k,=(1—e)'k. Most results of the k-p theory can be carried over from the unstrained to the strained case just by replacing the crystal momentum component k,, With (Byq—Eaa Ik (a,a°=3,y,25 summa tion over repeated indices) and adding the deformation potential contribution. In particular, the Hamiltonian of the strained system, H,(k,e), can be obtained from the Hamiltonian without strain, H(k), as Hike! of 1—e)-k]+Dte) , oy ‘where D(e) denotes the deformation potential contribu- tion (ef. Ref. 5). Tt can be taken apart into H,(k,¢)=Hy+H,+D(e)+Hy(k,) +HAlk, (-erk, @ where Hy represents the energies of the basis states, Hy the spin-orbit interaction, D(e) the deformation potential, interaction, and H(i.) and H,(k,) the k-p interaction of first and second order, respectively. Within the basis of atomiclike states (the superscript T denotes the tran- spose) wH(SEATYEZESLALYIZL" (ky=0) GB) of the eight-band model for the cubic, unstrained materi- al, the dominant contributions are represented through the following matrices (cf. Refs. 3 and 6) (4a) ‘orbit interaction energy there. The upper valence- , when including the k-independent spin-orbit interactionst k-p THEORY OF ENERGY BANDS, WAVE FUNCTIONS, AND . . 16.697 oo 0 00000 oo 00001 oi 0 0 00 0 ~ 000 00-170 oo 0 00 0 0 Oo a loo 0 -10 07 0 loo 0 -10 +0 0 lor f 00 000 ‘The Hamilton matrix ofthe deformation potential interaction reads a(eq Hey Hey) 0 ° o 0 D3ie) o ° me) 0 0 alee teyte,) 0 |? (4) ° ° ° Date) with Ieee +m (yy +E) ebay Nex D3te)= pe Leb tmel xe HE) Neb : (er) tee Netay Leg tele y tee) where d,, [,, mg, and m, are deformation potentials. D3(e) exhibits the same symmetry as the Shockley matrix S given below. ‘We neglect the first-order k-p coupling between the p-like basis states and the second-order k-p coupling between the ‘s- and the p-like basis states, since they are believed to be small.’ The form of these coupling terms is given in Ref. 4 Then, Hai.) 0 HkO=| 9 make] * an with 0 Phge iPhye Phe ik, 0 0 0 z H4K= |i 9 oo fp Ranke Staak» 4a’) i me -Pk, 0 0 0 where Ps essentially the real-valued momentum matrix element (S|@,|X). The free-electron and second-order k-p interaction contributions are contained in the matrix Ako 00 0 Sik) 0 0 0 0 azo BR tkhtkh (ie) 0 0 0 Stk) Eh} AM Uj +R) Nkaaie Nheckse Stk, Nhyckee LKR +M(k2 +42) Niyckze . (4e’) Nkickee Nksckye Lk +M(k2, +h) SS ‘S(k,) is the Shockley matrix!’ modified through the remote states, R, ¢.g., strain transformation of the reciprocal space coordinate - xtlpIRtMRtlp,lxt) system. This modification can be derived by noting that = 4 {Atle R LED) the parameters L, M, and N contain the free-electron 2m” m? padtrz Ey-Ex term, #?k?/2m, and momentum matrix elements with the (3)16.698 ‘Then, the transformation of the momentum operator in the strained erystal, p,=(1—e)-p+0(€), is carried over conto the quasimomentum k of the unstrained crystal [ef. Ref. 5, Eg, (29.14)]. Remember that the functions R be- long to the unstrained crystal and, thus, are strain in- dependent. Analogously, the term Ak? is treated. ‘The explicit formulas (4) show that the total Hamil- tonian (2) exhibits the symmetry of the strained erystal only, if all k-p terms are transformed in the same way. ‘Some practical consequences of this point of view are ex- amined in the following subsections. 1B, Kane's rotation of the coordinate system fon the k axis inthe case with straln For kliz,z being the unit vector in the [001] direction, the representation (4) of the Hamiltonian @) is particular- ly simple; for shear-free strain (¢,)—t,,—ey,—0), it even diagonalizes into two 4%4 blocks and the HH band separates. For this, Kane" has introduced a k-dependent rotation of the coordinate system (without strain) such that the 2 axis ofthe new coordinate system, 2 is always parallel to the k vector. Obviously, this idea applies to the Hamiltonian with strain only, ital k's are replaced with (K,—Eapk). OF course, this replacement has to be done in the transformation matrices as wel. ‘This transformation works wel for all parts of the ma- trices (4), which are cubicrisotropic. For the Shockley matrix (de), this would mean Syy—cik;Dag-+eaKcakipy Where ¢,'are k independent. “This| is the case for EN=L~M—N=0. LN=0 is never met, however, in ss Oo POO Ak 0 m|—iP 0 .00| ml 0 kee PAI] 0 0 00/tF| 0 Nk. 0 000 0M, ‘and similar expressions for the other components, p, and Pe ., Comparison with other approaches In a reexamination of the signs of the imaginary ele- ‘ments in H, we have obtained the same ones a8 Jones and O'Reilly,” while other authors have written down oppo- site ones (e.g., Refs. 6 and 9). These signs become impor- tant when the phases of the wave functions are significant ‘Since both the first-order and the second-order k terms fare transformed in the same manner, alll terms of the Hamiltonian (4) exhibit the symmetry of the strained crystal and the von Neumann principle is obeyed. How- ever, in the second-order k terms (de), the factor (1—e) creates terms of the order ¢? not usually included in the transformation of the kinetic and potential energy opera- tors. Probably this is the reason that most authors omit the factor (I—e)(1~e) in the k? terms. But itis evident P. ENDERS, A. BARWOLFF, M. WOERNER, AND D. SUISKY Nk, 2Mk, contrast, LN%0 is responsible for the characteristic warping of the valence bands. Fortunately, the terms which prevent the 4X4 block diagonalization of Hy are small and can be neglected in most cases.'© More difficulties are caused by certain non-negligible off diagonal matrix elements generated by the transforma tion of D3(e). Here, further work is necessary. C. Consistent definition of the momentum operator Within the kp representation of the Hamiltonian, H,.y, without strain, the momentum matrix reads (Ref. 5, See" 36) (ulply) =F VeCulHe lv) (6a) It should be stressed that this holds only for k- independent basis functions |j.), such as those in uy. (3) ‘The discussion above implies that the strain can be in- cluded simply through (ulplv) =F Vi, (el Haep (kee lv) (6) From H, (4d) and H, (4e) we obtain for the x component ‘within the basis (3) the representation Path) 0 pulkse PA (ke) o with o 0 ye Nitze as ” 0 My, ——— that this omission destroys the symmetry of the Hamilton matrix with strain. We believe, however, that consisten- cy in symmetry is superior to consistency resting on the ‘order of approximation. ‘As an example, we consider the momentum operator. Obviously, the representation (7) exhibits the same sym- metry asthe Hamiltonian (4) and, thus, obeys von Neumann's principle. As shown above, its derivation seems to be natural, while any other’ variant either violates the symmetry principle, or introduces artificial terms or ad hoe assumptions. Many authors keep only the k-independent “P terms” (first matrix in p4,, Eq, (7) and neglect the free-electron contributions to the momentum operator and others from the interaction with the remote states [second matrix in 4, Eq. (7)]. These contributions reflect the intimate connection between second-order perturbation theory for the Hamiltonian and first-order perturbation theory for the basis functions. All these terms vanish for k =0, but may become significant for the Fermi momentum kvalues in highly doped or pumped samples, as will be demonstrated in a numerical example below. "Any consistent definition of the momentum operator in different representations should be compatible with the mn, which is exact within the ener- (aki pla mB ie) (Ga) [luk) denotes the band state with energy E,(k)]. Ac- cording to the foregoing discussion, Eq. (8a) generalizes tunder strain to «uk, |p. (nek, Eyelet) « (8b) Equations (8) parallel the relation v,=2e/8k; ie. the mean velocity of an electron or hole in state |jek,) equals the group velocity of the corresponding, wave packet ‘The omission of the free particle and remote states con- tributions (k? terms) in the momentum matrix leads to the result that Eqs. (8) are no longer fulfilled (but at K=O, of course). The proof is straightforward for a two- band model, For the eight-band model, Bq (b) has been verified numerically.'* Consequently, a consistent treat- ment of the k-p momentum operator requires not only t0 keep the k? terms, but to perform additionally the Pikus- Bir transformation in the Hamiltonian and in the momentum matrix for these terms. III, NUMERICAL RESULTS AND DISCUSSION We have computed numerically the band structure, the wave functions, and the oscillator strengths from the 8X8 Hamiltonian (4) for biaxially strained bulklike ng Gt a9A8p qaPoy being closely lattice matched to InP._ Ic is an interesting material for optoelectronic appli- cations (it Iases at a wavelength of A=1.3 sum), but has not been greatly investigated by means of the eight-band model. ‘The material parameters are calculated from the quadratic interpolation formula” and the values of the binary compounds as listed by Adachi and Krijn,?! re- spectively; see Table I. It should be noted that the aceu- rate measurement of the hole band masses in these 3 k-p THEORY OF ENERGY BANDS, WAVE FUNCTIONS, AND 16 699 quaternary compounds is still an unsolved problem, though for the qualitative results to be stressed in this section, this plays a minor role. ‘The biaxial deformation tensor reads oe 0 |, cme ° 0 0 Cpe en ‘Thus, ¢>0 (¢ <0) means tension (compression) in the xy plane. ‘A, Energy bands ‘The conduction band (C) and the split-off band (SO) behave smoothly under strain, The band edges without strain (k=0,¢=0), Ec(0,0)=E,, and Ey9(0,0)=—A are shifted by AB c(0,¢)=a,(2+Cp)e and, in the linear regime (\el<1%), by AB so(0.e)=(1/3)l,+2m,12 +Cy le, respectively. The effective masses, me and miso, change only little. The strain-induced coupling between $0 and LH affects the SO band only for rather high strain values (say, |e|> 1%). ‘Therefore, we concentrate on the Ty valence-band complex, the heavy- (HH) and light-hole states (LH). Figure 1 illustrates characteristic features of the band structure for different strains. The splitting of the band edges and the drastic changes of the effective masses for k directions not parallel to the 2 axis are typical for this class of semiconductors. For tensile strain (Fig. 1(@), the LH states are HHlike in all k directions, except kilz, and vice versa. This “ex- change of effective masses” between HH and LH reflects a repulsive interaction between both bands, which is due to the symmetry reduction. For compressive strain [Fig. 1], the HH effective mass in the x direction is smaller than that of the LH. But the repulsive interaction just ‘mentioned causes an anticrossing of both bands. We will return to this interaction when discussing the symmetry of the HH and LH states. Therefore, under stress, the notions HH and LH as used in Fig. 1 lose their original meaning, since the effective masses of both bands strongly depend on the k direction. For biaxial stress with the stress axis in the 2 direction [ef. Eq. (9), the conventional band curvature is Parameter Fundamental gap wv Conduction-band mass mo Heavy-hole mass in (100) mit mo Heavy-hole mass in [111] mana ma Light-hole mass in [100] direction ms mo Spin-orbit split-off energy a 0.267 wv Deformation potential of conduction band ae -36n wv Deformation potent lence band a +1179 ev. Shear deformation potential ° 1.695 ev Elastic constant Cy kem=? Elastic constant Ca 52x10" Kem Lattice constant, nominally unstrained a 0.5867 nm "Reference 26.16700 P. ENDERS, A. BARWOLFF, M. FIG. 1. Heavy- (HH,) and light: (LH,) hole bands in strained and unstrained bulklike 13m-In,Ga,-,As,P,-, on InP; conventional band labeling (Gee text). (a) tensile strain: 1%; both bands cross. at Ke. .0,k,), Em, (ks) Ey, (k,)=E,5 (0) no strain: ¢=0; (¢) compressive strain ~ 1%, The cases k\[010}, k][110}, ee. are similar to kif. Crystal momentum, &; in units of yg ~7/ap (ap is the lattice constant of unstrained crystal). Such high strain values have been chosen in order to depict clearly the qualitative features preserved only in the z direction. In order to indicate this change, we will label these states HH, and LH,, respec tively. Actually, these bands energies are not analytical functions of crystal momentum and strain at e=0 (ef. Refs. 4, 11, and 12). This singularity is also found in the strain dependence of the effective masses.” In principle, there are two possibilities to label the Ty valence-band complex. {i) According to their symmetry at the F point, k=0. This is the most usual form and ‘will be referred to as conventional ordering. Tt has been used in Fig. 1. It involves that the HH, energy, Ey, de- creases with increasing ¢ (tensile strain in the xy plane), while the LH, energy, Fix, increases. For kijz, the HH, band separates. For k not parallel to z, the bands are deliberately assigned as follows: Eyyy < Ex for €>0 and Eun Ey for e$0, respectively. (i) Strict energetic level ordering. This will be referred to as energetic order- ing. The energetically higher/lower states will be grouped into the upper/lower valence band (UVB/LVB) Evy = Ey for e>0 and for €<0. Both schemes have their advantages and disadvantages. For the isoenergy surfaces, the energetic ordering is ap- propriate. For no and compressive strain and sufficiently, small k values, the isoenergy surfaces are warped ellip- soids, cf. Ref. 10. For tensile strain, however, the UVB and LVB touch in the HH,-LH, crossing points WOERNER, AND D. SUISKY k,(€)=(0,0,:k;(e)); see Fig. Ua). Figure 2 displays the UVB and LVB isoenergetic surfaces for energies near and equal to the energy E,~Eyy(k,(€))=Exy(k,(e)]. Note the similarity to the indicatrix of a birefringent biaxial crystals, cf. Ref. 15. 1B. Symmetry of the wave functions ‘and character ofthe states ‘A more detailed analysis of the band dispersions de- ppicted in Fig. 1 requires calculation of the symmetry of wave functions involved. In particular, a band anticross- ing is accompanied with a change of the symmetry of the involved states. The symmetry of the band wave func- tions is determined by the contributions (portions) of the atomiclike basis functions (3). For 2 the quantization axis, the wave functions at ‘can be written in a comprehensive form as (ef, €.8., Ref. 9) (10a) so; Characteristic features are the isotropy of C and SO states and lack of a z component in HH states. The C and SO wave functions are only less affected by the strain (9); therefore, their k dependence is very similar to the strainsree case. For them, one obtains essentially the same resulls as Kane? did for InSb. Consequently, we will concentrate on the strain and k dependence of the P', ‘group states. For the understanding of the symmetry of these states, iis essential to note thatthe eigenfunctions for x being the quantization axis, (106)ty kp THEORY OF ENERGY BANDS, WAVE FUNCTIONS, AND 09-09, perl band £110 eon ka | BE, o 10 censin) FIG. 2. Isoenergetic surfaces of the upper (UVB) and lower (LVB) ‘valence thole) bands ink space: Evys(k) Exya(k st. Tensile strain: 0.196. (a) £=0.98,; (0) E=E,. k values in units of k,. E, is the energy of the HH,-LH, crossing point k, =(0,0,k,), ef. Fig. 1a) and for y being the quanti «|? HH, LH, 7 (106) are also eigenfunctions of the (unstrained) Hamiltonian, Hy+H,, and of the angular momentum. The functions (10a)=(10e) are the limiting cases of the wave functions for k +0 and kz, k[x, and Ky, respectively. Due to the degeneracy of HH and LH at k=O, there is no ‘unique limit for k—-0, but a dependence on the direction, oft. Inthe unstrained crystal, the k-p interaction polarizes the states along the direction of k.“* Within the basis X, Y, and Z for the valence bands [see Eq. (3)], the HH wave function has no component in the direction of k. As K rotates, the composition of the valence-band. states changes accordingly. Note that this is even necessary for 16701 the bands to be optically isotropic in the cubic crystal For small k values, the composition of the LH wave func tions is determined by the spin-orbit interaction. For large k values, it becomes, aside from phase factors, equal to that of the HH wave functions. Accordingly, both en- ergy bands become parallel, ef. Fig. 1b). In a strained crystal, this polarization by the direction of quasiparticle motion may compete with the preference direction imposed by the strain. The interplay of strain, spin-orbit interaction, and k-p interaction in the symme- ity of the HH and LH wave functions is depicted in Fig 3. For small k values, the composition is determined by the strain. The z axis exhibits fourfold rotation symme- try, while the x and y axes exhibit only twofold. This makes the 2 axis the axis of quantization. The wave func- tions are given by Eq. (10a). In the unstrained case, as, here kijx, x is the axis of quantization, and the wave func~ tions are given by Eq. (10b). As the k, value increases, the X component decreases in Figs. 3a), 3(d), and 3(0, and the states become HH- 0.1% compression ous, cae, porters ofX.¥, 2 no strain pertane aX, ¥,2 0.1% tension | muy, portions of %Y,2 hs hoe FIG. 3. Relative portions of the basis functions X, Y, and Z: in the heavy- (HH,) and lighthole (LH,) wave functions Unn(Rav€) and Wuy(,,€) (Conventional band labeling, ef. Fig Dh ky=k,=0. The wave functions wyy(0,0) and wy (0,0) have ‘been deliberately st t0 guarantee continuity. (a),(6) c= —10" (compression in the xy plane), heavy-hole/light-hole band. (@\ld) e=0 (no strain), heavy-hole/light-hole band. (e)(0) £=+10™ (tension in the xy plane), heavy-hole/light-hole band. [Note that the S portion in the HH, and LH, wave functions remains very small forall k, and c values, even for the medium ik, range. Thus, the exchange of 3- and prlike basis states be- tween the conduction- and the valence-band states caused by the k:p interaction is carried almost completely by the SO band alone.16702 like. In contrast, the states in Figs. 3(b) and 3(c) become LH-like with respect to the axis of quantization x and the spin-orbit interaction being the dominant interaction. In the case of tensile strain and medium k values, knot parallel to 7, the standard labeling is misleading [k,=10-—10"" kp in Figs. 3(e) and 310). For large k values, the k-p interaction dominates. Be- sides the factor (1—e) introduced by the Pikus-Bir trans- formation, it is isotropic. Correspondingly, in all cases, shown in Fig. 3, the symmetry is HH-like. "As the small differences between the Y and the Z portions are caused mainly by terms of first order in the strain, the wave functions in Fig. 3c) (3(9] are LH-(HH)-ike, in contrast to the standard labeling. Consequently, for small k values, the standard labeling is appropriate, while for larger k values, k not parallel to 2, the energetic labeling is. A similar behavior has been found for biaxially strained semimagnetic semiconductors in an external magnetic field Bix.” The symmetry of the states determines the optical selection rules. Within the eight-band mode, the oscilla- tor strength for interband transitions and an electromag- netic field being linearly polarized in, say, the x direction is determined by the S component of the wave function of cone band state and the X component of the other one. Thus, we obtain dichroism for optical transitions near the band gaps and isotropy for larger photon energies. C, Momentum matrix and oscillator strengths ‘The “irregular” behavior of the symmetry of the heavy- and light-hole wave functions in dependence of € ‘and k should be reflected in the dispersion of the oscilla- tor strengths, In order to quantify the importance of a consistent treatment of the momentum operator, we have calculated numerically the interband oscillator strengths, . 2 2 Muoeks|pualvpek,)!? Fike ng 2 Eyal kt)~ Elke) (u=C; v=HH, LH; p,o=1,2, the Kramers degenerate subband indices). A ‘projection operator technique has ‘been used, which cizeumvents the evaluation of the wave functions (see the Appendix). In Fig. 4, some results ob- tained by means of the exact formula (6b) are compared with those of two common approximations for the momentum matrix. In the first one, the k-dependent terms {second matrix in p4,, Eq. (7) are neglected. The second one isthe effective-mass approximation, Ee(O,€) Fite #7 0 ap Exel Pawflin®) Poul) 5 yg) for allowed transitions. ‘Obviously, the deviations of both approximations from the exact results are small in unstrained materials, while the effective-mass approximation (12) fails principally in, strained crystals. The reason for that is the following. Since the momentum operator has a defined parity (odd), the oscillator strength becomes sensitive to detect the symmetry of the involved states. The band mixing effects discussed above result in nonmonotonous compositions of P. ENDERS, A. BARWOLFF, M. WOERNER, AND D. SUISKY 3 heavy holes (HH) light holes (LH,) a” * ' : i 10 . (a) | fey(hs0 al ol 000 ob oBa abo abe Obs 0.1 % compression fs ® i @ 121 & ° fa(k-10%) 10 fenlkt0%) 7 000 0b Ob 000 Oa Oba Keg FIG. 4. Oscillator strengths, (ke), of transitions C-HH, and CLH, a$ functions of k (K in units of Ry, =2/a9). (a) C- HE,, €=0; ) CLH,, €0; (e) CHH,, = 10"? (compres- sion); (€) C-LH,, c= ~10~* (compression). Full lines: con- sistent definition ofthe momentum matrix elements and numeri- cally exact diagonalization of the eight-band Hamiltonian; dot ted lines: effective-mass approximation, Eq, (12); dashed lines: ‘omission of the K° terms in the momentum matrix, ef. Eq. (7) ‘The oscillator strengths shown are values averaged over the k- space directions x, y, and z a8 fia(kest)=[finlkeo®) 4 Seathyo€ +S lkno6)]/3. This neglects the angular disper- sion ofthe energy bands, but the purpose here isto demonstrate the qualitative failure of the effective-mass approximation for strained materials. the hole wave functions and, therefore, in nonmonoto- nous dispersion of the oscillator strength. This behavior ‘cannot be reproduced by Eq. (12), since the gaps are monotonous functions of € and k. Consequently, extra- polations from the F point to the case k >0 are by no ‘means a valid approximation, especially in strained erys- tals, IV, SUMMARY AND CONCLUSIONS. We have reconsidered the k-p theory for homogene- ‘ously strained semiconductors with particular attention to the second-order k terms and the momentum operator. Our main results are the following. (j) The second-order, terms have to be multiplied with (1—e):(1—e), in order to obey basic requirements of symmetry, although this, procedure introduces contributions which are of second, order in the strain e. (i) The second-order k terms are to be included in the framework of a consistent treatment of the momentum operator. (ii) There is a heavy-hole and,st k-p THEORY OF ENERGY BANDS, WAVE FUNCTIONS, AND light-hole band mixing and anticrossing, except along ‘some higher-symmetric directions. Consequently, extra- polations of k-dependent parameters from the I’ point to values of k>0 are not straightforward. Moreover, the notions of light-hole band and heavy-hole band can be- ‘come misleading for k >O when considering the symme- try of these states, since the latter changes dramatically as a consequence of state mixing. This is important to recognize for the interpretation of the optical properties of strained materials. ‘Numerical results for the band structure have shown bulklike 1.3m-In,Ga,_,As,P,-, nearly lattice matched to InP as a representative of this class of semi. ‘conductors. Hence, our results for the momentum opera- tor and oscillator strengths should apply to other materi- als of this type as well. These are (i) that the neglection of the &? terms can result in significant numerical devia tions for practically relevant cases, and that (ii) the ceffective-mass approximation for the k dependence of the ‘momentum matrix elements is applicable in special cases only. In unstrained materials, the effective-mass approxi mation produces, in general, a too weak k dependence of the oscillator strengths; while in strained material, it is not able to reproduce their nonmonotonous behavior at all. Consequently, a realistic model of momentum- related properties (e-g., optical properties, momentum re- laxation) requires (i) a consistent definition of the momen- tum operator within the representation considered, such as Eqs. (6) and (7) and ii) the calculation of its energy representation by means of an exact method (cf. also Ref- 4) SSS 16703, ‘Thus, we have presented band structure, wave func- tion, and oscillator strength calculations for 1.3. um—In,Ga,_,As,P,_, on InP using an eight-band k-p ‘model and accounting for the von Neumann symmetry principle and consistency of the momentum operator. ‘These results can be useful also as asymptotic values for investigations on quantum wells of varying thickness, and the formulas used may serve as a starting point for com- puting the polarization and strain dependence of optical properties of this material. ACKNOWLEDGMENTS ‘The authors thank Professor T. Elsisser for helpful di cussions and a critical reading of the manuscript. This ‘work was supported by the Deutsche Forschungsgemei schaft (DFG). APPENDIX: PROJECTION OPERATOR METHOD FOR ‘CALCULATING THE OSCILLATOR STRENGTH ‘WITHOUT WAVE FUNCTION, ‘The projection operator Pygltit = E volta an projects on the subspace of the eigenstates Yu(r) of the Hamiltonian H to the band v and crystal momentum k (p degeneracy index): Hise Weal = Evanl LS Foal | £ veo Enadgalt) > Pulte Worl) E Vout ff ¥onll Yael = (a2) BuoDronltD (aa) ‘The optical matrix elements entering the oscillator strength can be expressed as follows: Kuyoklplvpk) P= ff J Yuorle)* reales fff vault Pv youll de? (P,pP.p) ‘The trace opens alternative calculational options, since it is representation independent. The crucial point is that there is another formula for the projection operator,”* which is independent of the eigenfunctions of the Hamil tonian, (as) (1 is the unit operator). The proof is straightforward by using (A3) and (A2). Therefore, Py and p can be calcu lated most conveniently within the basis uy; the cigen- functions of H are not needed. The formalism even simplifies, when both Hand p block diagonalize. Its ap- wa ee plication to the calculation of higher-order transition probabilities entering the calculation of nonlinear optical properties is in progress. Finally, we remark that from Eq, (A1), it follows im- mediately that Prova (a6) EM ol Yaga! ‘This means the portion of the basic state u,, in the eigen- state Yypx equals the corresponding diagonal matrix ele- ‘ment of the projection operator onto the eigenstate Wy in the representation of these basic states {17}16704 4, M. Luttinger and W. Koha, Phys. Rev. 97, 869 (1955). 2E_ O. Kane, in Handbook on Semiconductors, edited by T. S. “Moss (North-Holland, New York, 1982), Vol, Chap. 44. SHUR. Trebin, U. Réssler, and R. Ranvaud, Phys. Rev. B 20, 686 1979) 4G._L. Bir and G. E. Pikus, Symmetry and Strain-Induced ‘Effects in Semiconductors (Wiley, New York, 1974). ST. B. Bahder, Phys. Rev. B 41, 11992 (1990) 6G. E. Pikus and G. L- Bir, Fiz. Tverd. Tela (Leningrad) 1, 1642 (1959) [Sov. Phys. Solid State 1, 1502 (1960) 7A. P. Cracknell, Angewandte Gruppenthcorie (Akademie- Verlag, Berlin, 1971), p. 49. SE. . Kane, J. Phys. Chem. Solids 1, 249 (1987) 98. L. Chuang, Phys. Rev. B 43, 9649 (1991), 10K. Suzuki and J.C, Hensel, Phys. Rev. B9, 4184 (1974) 1, C, Hensel and K. Suzuki, Phys. Rev. B9, 4219 (1974). RB, H, Pollak and M. Chandrasekhar, Phys, Rev. B 45, 2127 a7. DW. Brauer and H.-W. Streitwolf, Theoretische Grundlagen der Halbleterphysik (Akaderaie-Verlag, Berlin, 1973), Appendix A MIM. Born and E. Wolf, Principles of Optics, 2nd ed. (Pergamon, P. ENDERS, A. BARWOLFF, M. WOERNER, AND D. SUISKY 51 New York, 1964), Fig. 149; A. Yariv and P. Yeh, Optical Waves in Crystals (Wiley, New York, 1984), Fig. 4.1. '5W. Shockley, Phys. Rev. 78, 173 (1950) ‘6p. Enders, Phys. Status Solidi B 187, 541 (1995) "G. Jones and E. P. O'Reilly, IEEE J. Quantum Electron. 29, 1344 (1993). "SP. Enders (unpublished) Semiconductors. Intrinsle Properties of Group IV Elements ‘and IV, ILIV, and I-VI Compounds, Landolt-Bérastein: New Series, Group III, Vol. 22, P. a Springer, Heidelberg, 1982) 285, Adachi, J. Appl. Phys. 83, 8775 (1982) 21M. P. C. Krijn, Semicond. Sci. Technol. 627 (1991), 22, M. Thidilkovski?’, Band Structure of Semiconductors (Nav- ka, Moscow, 1978), p. 74 (in Russian. %c. Santos, D. Suisky, P. Enders, F. Neugebauer, M. Hart- ‘mann, J. Réseler, and W. Heimbrodt (unpublished) MP, Enders, A. Birwolf, R. Miller, and T. Elsisser (unpub- lished). 30. Ziep (private communication). 2G. P. Agrawal and N. K. Dutta, Long-Wavelength Semicon- ductor Lasers (Van Nostrand, New York, 1986).