Arxiv:2011.11041V1 [Cond-Mat.Mes-Hall] 22 Nov 2020 in Semiconductor Nanostructures of Diﬀerent Size, Shape, a Perturbation

Electron, hole and exciton eﬀective g-factors in semiconductor nanocrystals

M. A. Semina, A.A. Golovatenko and A. V. Rodina Ioffe Institute, 194021, St.-Petersburg, Russia. (Dated: November 24, 2020) We review the existing and present the new results of k · p calculations of the electron, hole, and exciton effective g-factors in semiconductor nanocrystals of different shape and symmetry. We propose a simple yet accurate method for calculation of electron g-factor size dependence in bare nanocrystals within the eight-band Kane model. Using the spherical approximation for Luttinger Hamiltonian we find the dependence of hole g-factor on light to heavy hole effective mass ratio in semiconductor nanostructures with spherical, axial, and cubically symmetric shape. We show that the non-equidistant Zeeman splitting of the four-fold degenerate hole state may take place in cube and spheroidal nanocrystals. We present a comparison of the calculated hole g-factors in nanostructures based on II-VI and III-V semiconductors for different sets of the Luttinger parameters and analyze the main effects contributing to the g-factor renormalization with the respect to the bulk value. We discuss different approaches to the definition of the hole and exciton g-factors which should be taken into account during the analysis of the experimental data and compare our results of g-factor calculations with the experimental data for semiconductor spherical nanocrystals and thin nanoplatelets available in the literature.

I. INTRODUCTION tocharging, when electron or hole is trapping to the surface from photoexcited exciton. The subsequent act of For several decades the synthesis of semiconductor photoexcitation of charged nanocrystal results in the for- nanocrystals by colloidal chemistry has passed into the mation of a trion, charged three-particle exciton complex. state of a mature technology. It provides precise con- Both in the case of exciton and trion the knowledge of trol over the size, shape, and composition of nanocrys- hole g-factor becomes crucial. However, much less atten- tals [1–3]. Today colloidal semiconductor nanocrystals tion was devoted to the properties of the hole and exciton (NCs) are widely demanded by the market. The area of g-factors in semiconductor nanostructures as compared NCs application includes solar cells, displays, photode- with the electron g-factor [28, 29]. tectors, and molecular sensors [4–10]. All these applica- In present paper we consider nanocrystals based on di- tions are mostly based on emission or absorption of light rect gap semiconductors with bandgap in Γ point of the by spatially confined electron-hole pairs. Spin properties Brillouin zone with Γ6 conduction band (electron spin of confined charge carriers in colloidal nanocrystals are S = 1/2,Sz = ±1/2), Γ8 topmost valence band (elec- of great interest from both experiment and theory due to tron spin J = 3/2,Jez = ±1/2, ±3/2) and Γ7 spin-split the prospect of their use in spintronics and quantum com- valence band (electron spin J = 1/2,Jez = ±1/2). The puting devices [11–15]. The control over the spin state corresponding energy bands are shown schematically in of carriers requires knowledge about their g-factors, spin Fig.1 (a) in the electron representation. It could be, for coherence times, the fine structure of their states, and example, zinc blende modifications of AIIBVI or AIIIBV states of their complexes. compounds, such as zb-CdSe and GaAs. Spin projections The g-factor determines the response of the electrons, Jez = ±3/2 and Jez = ±1/2 describe so-called heavy and holes, or their complexes to the external magnetic field light hole subbands, respectively. In wurtzite modifica- including the Zeeman energy splitting between spin sub- tion of CdSe (wz-CdSe), which is widely used for fabri- levels. The comprehensive experimental and theoretical cation of colloidal nanocrystals, the effective crystal field studies are being performed in this direction. Using dif- splits the Γ8 topmost valence subband into two Γ9 (heavy ferent magneto-optical techniques g-factors of electrons, holes) and Γ7 (light holes) subbands, see scheme in Fig. holes and excitons were measured [16–21]. Theoretical 1 (b). The heavy and light hole splitting ∆cr is typically understanding of experimental data on electron g-factor small enough (25 meV in bulk wz-CdSe) to be treated as

arXiv:2011.11041v1 [cond-mat.mes-hall] 22 Nov 2020 in semiconductor nanostructures of different size, shape, a perturbation. For these semiconductors dependencies and dimensionality were developed within multiband k·p of hole g-factor on type and dimensionality of confining theory [22–25] and tight-binding calculation [26, 27]. The potential are calculated with respect to the ratio of light surface-induced contribution to the electron g-factor in hole and heavy hole masses. We discuss the relationship core/shell nanocrystals was studied as well [24]. between various formalisms in defining the hole g-factor, Quite often researchers extract from experiments g- common in various communities. The relation between factor of excitons [18, 19], i.e. electron-hole pairs cou- electron, hole, and exciton g-factors in different types of pled by strong direct and exchange Coulomb interaction. nanostructures is also considered. We make the calcu- Moreover, colloidal nanocrystals can be in a charged state lations and compare the results with the experimental due to intentional chemical doping, or as a result of pho- data for colloidal nanocrystals assuming high potential 2 barriers for the electrons and holes at the NC surface. ±1/2 on the direction of magnetic field. The scheme of the electron energy levels splitting is shown in Fig.2. (a) (b)

(a) (b)

Electron energy Electron

Energy, electron representation electron Energy, Hole energy Hole

FIG. 1. The band structure of the AIIBVI or AIIIBV typi- FIG. 2. The scheme of Zeeman energy levels splittings of cal semiconductor with (a) zinc-blende; (b) wurtzite crystal electrons and Γ8 holes in bulk zinc-blende semiconductor (a) structure. Note, that for graphic reasons Eg, ∆SO and ∆cr in electron and (b) in mixed electron-hole representations. are shown not to scale. Eh and Ee are the hole and electron energy levels in zero magnetic field calculated from the top of the valence band, Ev = 0, and the bottom of the conduction band, Ec = Eg, The rest of the paper is organized as follows: In Sect.II respectively. The order of electron and hole levels is shown we revisit the electron g-factor calculation in the frame- for positive g-factors both for electrons and holes: ge > 0, work of the eight-band k · p model and compare results gh > 0. The circular polarized optical transitions are denoted with the results of the tight-binding calculations as well by σ+ and σ−. as with experimental data. In Sect.III we discuss the definition of hole g-factor (IIIA), present semi-analytical In bulk semiconductors the spin-orbit interaction is effective mass methods and results of calculation of hole usually parametrically strong and ge differs significantly g-factors in spherical and cube NCs, quantum wires and from the value of free electron g-factor g = 2 [30]. For planar two-dimensional nanoplatelets (NPLs) (IIIB). We 0 example, ge = −0.44 in GaAs [31], ge = 0.42 in zb-CdSe compare the results of the hole g-factor calculations in [32], g = 0.68 in wz-CdSe [33], g = −1.66 in CdTe different semiconductor nanostructures with published e e [34] and ge = 1.2 in InP [34]. Spatial localization of experimental data and discuss the dependence of hole g- electrons in nanostructures leads to additional renormal- factor on its size in Sect. (IIIC). In Sect.IV we discuss ization of g-factors and even to their spatial anisotropy the definition and size dependence of g-factor of different [22, 23, 25, 35, 36]. The main mechanisms of conduction exciton states in low-dimensional structures. We sum- band electrons g-factor renormalization are well-known, marize our results in Sect.V. The additional details of and the theoretical calculations quantitatively well de- calculations are presented in AppendicesA,B andC. scribe the experimental data [23–25, 27, 36]. The value of electron g-factor at the bottom of the conduction band in typical semiconductors with zinc blende II. ELECTRON g-FACTOR IN THE crystal structure can be calculated within the second- EIGHT-BAND k · p -MODEL order k · p theory. It was first made to account the contributions from the Γ8 and Γ7 valence subbands in In the presence of an external magnetic field, electron Ref. [30]: states with different spin projections on the magnetic field direction split. In the low field regime this split- 2Ep 1 1 ge ≈ g0 − − . (2) ting is linear on magnetic field strength and it is defined 3 Eg Eg + ∆SO by electron effective g-factor or Lande factor ge. Corre- 2 sponding Zeeman part of the electron Hamiltonian in the Here Ep = 2|hX|pˆx|Si| /m0 is the Kane energy expressed low magnetic field limit can be written as: via the interband momentum matrix element, Eg is the bandgap, ∆SO is the spin-orbit splitting of the valence (e) E1/2 − E−1/2 band. More accurate calculations require consideration HbZ = µBge (SeB) , ge = . (1) µBB of the remote conduction band contributions [37]. In most cases, however, it can be accounted for by a fitting e Here µB = ~ is Bohr magneton, m0 is free electron 2m0c parameter grb added to g0 in Eq. (2) in order to obtain mass, e = |e| is the absolute value of electron charge, and the experimental value of the bulk electron g-factor ge E±1/2 are energies of states with spin projection Sez = when the other parameters Ep, Eg and ∆SO are known. 3

The eight-band Kane or k · p model allows one to ac- 2.0 count for the eﬀect of the electron localization in low- dimensional structures on the g-factor value [24, 38, 39]. 1.5 InP It results in the dependence of the electron g-factor on

1.0 the electron size quantization energy Ee counted from wz-CdSe the bottom of the conduction band as -factor g 0.5 Z GaAs c 2 3 ge(Ee) = g0 + (˜ge(Ee) − g0)|Ψe(r)| d r + gsur , (3) 0 Electron

-0.5

CdTe 2Ep 1 1 g˜e(Ee) ≈ g0 +grb − − . -1.0 3 Eg + Ee Eg + ∆SO + Ee (4) 1 2 3 4 5 6 7 8 9 10 c Here Ψe(r) describes the conduction band contribution Nanocrystal radius (nm) to the eight-band electron envelope function Ψe(r) and gsur describes the surface/interface contribution propor- FIG. 3. Size dependencies of the electron g-factor in differ- c 2 ent semiconductors calculated within tight-binding approach, tional to the squared value |Ψe|sur taken at the surface of the nanostructure or at the interface between two Ref. [27], (symbols) and using the eight-band k · p, Eq. (5), semiconductors in the heterostructure. Importantly, the with the same sets of parameters listed in Table A1. R 2 3 normalization condition |Ψe(r)| d r = 1 [23, 24, 40] R c 2 3 implies |Ψe(r)| d r = A(Ee) < 1 even in bare core nanocrystals. Generally, the renormalization constant In Fig.4 the dependencies of the electron g-factor on A(Ee) depends on the electron quantization energy Ee the radius of wz-CdSe and zb-CdSe nanocrystals calcu- and can be found from the dependence of electron effec- lated according to Eq. (5) are shown by the green and tive mass on its energy me(Ee) reflecting the nonparabol- violet solid lines, respectively. Calculations are made icity of the conduction band energy dispersion (see for with the set of parameters from Refs. [33 and 42] and details the AppendixA). As it was shown in Ref. [41], grb = 0.02 for wz-CdSe and from Ref.32 and grb = −0.125 for Ee Eg the renormalization of the conduction band for zb-CdSe, correspondingly, and with gsur = 0. Dashed contribution can be expressed as A(Ee) ≈ me/me(Ee) lines show the calculations according to Eq. (4). One where me ≡ me(Ee = 0) is the bulk electron effective can see that the difference between solid, ge(Ee), and mass. Combining together the renormalization of the dashed,g ˜e(Ee), lines increases with decreasing of NC size conduction band contribution in Eqs. (3) and (4) we related to the A(Ee) effect while the difference between arrive to: g-factors calculated for zb- and wz-CdSe decreases. We show also the values of the g-factors measured by the ge(Ee) = g0(1 − A(Ee)) +g ˜e(Ee)A(Ee) + gsur (5) pump-probe Faraday rotation experiment in Refs. [17 and 43] for wz-CdSe (g1 and g2 values) and in Ref. [44] for the bare core nanostructures. The distinctive feature for zb-CdSe. Calculated ge(Ee) dependence for wz-CdSe of Eq. (5) is that it allows one to calculate the electron g- is in a good agreement with the experimental data (g1 factor in nanostructures knowing the bulk semiconductor values) from Refs. [17 and 43] shown by filled red and parameters Ep,Eg, ∆SO, me, ge and the surface contribu- blue circles and attributed the ground state electron g- tion gsur. factor. Note, that a good agreement of the calculated size As it follows from Eqs. (4) and (A3), additional lo- dependence of electron g-factor with experiment can be calization in nanostructures changes the electron g-factor achieved with more than one set of k · p model parame- noticeably only if the size-quantization energy Ee is com- ters. In Refs. [17, 24] another set of parameters was used parable with the bandgap Eg. Such a situation is realized together with the general boundary condition assuming in small nanocrystals and, especially, in nanoplatelets non-vanishing electron wave function at NC surface re- where the localization length of charge carriers, at least sulting in gsur 6= 0. in one dimension, is about 1 or 2 nm (see Fig. A1). As a It is noteworthy, that the second set of experimental result, the contribution of the valence band to the elec- data for wz-CdSe, g2 values, given by open symbols in tron g-factor decreases withg ˜e(Ee) tending to g0 + grb Fig.4(a), as well as the data for zb-CdSe, can not be de- and A(Ee) decreasing, so that the ge(Ee) values become scribed by the calculated dependence for electron g-factor closer to g0. Importantly, the eight-band k·p calculations at lowest quantum size level, 1Se. There were different of the electron g-factor, according to Eq. (5) provide a assignments of these values to the exciton g-factor [17] very good agreement with results of tight-binding calcu- and surface-localized electron g-factor [43] for wz-CdSe lations [27] for spherical NCs made from wz-CdSe, CdTe, and to the electron g-factor for zb-CdSe [44]. Wherein, GaAs, and InP, see Fig.3. We used the same set of bulk g1 values corresponding to the electron g-factor size de- parameters for the k · p calculations in Fig.3 obtained pendence in zb-CdSe were not observed in spherical NCs. by the tight-binding calculations in [27], see Table A1. Interestingly, the values larger than g1 and very close to 4

(5) for electrons in the NPLs with parabolic and box-like

(a) 2.0 conﬁning potentials, respectively (see AppendixA). One can see that the parabolic potential better describes the

1.6 g

2 experimental data while the measured g-factors are about 10 % larger than calculated for the box-like potential.

1.2 However, the discrepancy can be related to the eﬀect of -factor g g 1 g-factor anisotropy in the NPL neglected in the presented

0.8 calculations as well as to the eﬀect of the additional in- plane localization of the resident electron. Therefore, we

Electron will examine further the eﬀect of both conﬁning poten- 0.4 Gupta2002

Hu2019 tials on the hole and exciton g-factors and compare the Zhang2014 calculated values with the experimental data.

1 2 3 4 5 6 Besides the renormalization of the ge-factor controlling

Nanocrystal radius (nm) the Zeeman splitting of spin sublevels, electron localization in the nanostructure also leads to the renormaliza-

(b) 2.0 tion of the orbital effective g-factor (orbital magnetic mo- Kudlacik2020 mentum) [24, 47]. The orbital magnetic field effect comes 2 1.6 from the electron kinetic energy term k /2me, where the e electron wave vector k is replaced by k + c A, where A is

1.2 the vector potential of the applied magnetic ﬁeld. By this -factor g term, one takes into account the velocity generated by the envelope wave function. As it was shown in [47], there 0.8 is also the second contribution to the orbital magnetic

Electron momentum generated by velocity coming from the Bloch 0.4 functions. We restrict our consideration hereafter by the account only by the envelope velocity contribution. The

1 2 3 4 5 6 size dependence of the orbital effective g-factor is then controlled by the m /m (E ) ratio [24]. However, the Nanoplatelet thickness (nm) e e e envelope orbital magnetic field effect does not influence the electron states in bulk semiconductors, and the low- FIG. 4. Size dependencies of the electron g-factor: (a) in spherical wurtzite (green lines) and zinc blende (purple lines) est electron states with zero orbital momentum in most CdSe nanocrystals. Solid (dashed) lines are calculated for the nanostructures. As we show below, the situation is sig- ground 1Se state of the infinite box-like potential according nificantly different for the holes where the orbital contri- to Eq. (5) (Eq. (4) ) with parameters from Refs. [33 and 42] bution leads to the substantial variation of hole g-factors and Ref. [32] for wz-CdSe and zb-CdSe, respectively. Symbols between different nanostructures. correspond to experimental g-factors observed via the pump- probe Faraday rotation in wz-CdSe NCs, Refs. [17] (blue circles), [43] (red circles) and zb-CdSe NCs [44] (black triangles); III. HOLE g-FACTOR (b) in zb-CdSe nanoplatelets calculated for the 1Se state in the infinite box-like (dashed line) and parabolic (solid line) A. Definition confining potential. Black squares are localized resident electron g-factors in zb-CdSe NPLs obtained via the SFRS in Ref. [45] In the present paper, we consider the hole from the four-fold degenerate Γ8 valence subband for semiconductors with relatively large spin-orbit splitting (for CdSe, ZnSe ∆ ∼ 400 meV). In these materials, the spin-split g2 in small NCs can be obtained from Eq. (5) for elec- SO valence subband can be safely neglected from considera- trons localized at the excited 2Se size-quantization level. While the calculations of the surface-localized electron tion. We also neglect the cubic symmetry of the crystal g-factor are beyond the scope of the present paper, we environment (valence band warping) as it affects the presented results only quantitatively but not strongly. With will discuss the possibility to assign the g2 values to the exciton g-factor in Sect.IV. these approximations the valence band holes in the ex- Recently, g-factor of the localized resident electron was ternal magnetic field can be described by the following measured in zb-CdSe nanoplateletes with thickness of 3, Hamiltonian, which we rewrite in the hole representation 4 and 5 monolayers (ml) by spin-flip Raman scattering as the sum of three contributions: (SFRS) [45]. The attribution of the obtained g-factors (h) Hb = HbL + HbZ + HbB. (6) to the resident electron (not exciton) is justified by the polarization selection rules [45, 46]. In Fig.4(b) the mea- Here 2 sured results are given by black squares. The solid and ~ 5 2 HbL = γ1 + γ k − γ{kαkβ}{JαJβ} , (7) dashed lines are results of the calculations according Eq. 2m0 2 5 is the Luttinger Hamiltonian [48, 49] in hole representa- Here we use the following definition of the hole effective tion in zero magnetic field describing hole kinetic energy g-factor [52, 53]: in spherical approximation. Operator J is the hole inter- E − E nal angular momentum operator, for Γ8 valence subband E−Jhz − E+Jhz −3/2 +3/2 gh = = = J = 3/2, γ1 and γ = (2γ2 + 3γ3)/5 are Luttinger param- 2JhzµBB 3µBB eters related to the bulk light-hole, mlh = m0/(γ1 + 2γ), E−1/2 − E+1/2 and heavy- hole, mhh = m0/(γ1 − γ), effective masses. = , (10) µBB Following the classical approach introduced by Lut- tinger [48], the Zeeman part of free hole Hamiltonian, where Jhz is the hole spin projection on the magnetic (h) field direction (which is chosen along the z-axis). The Hb , has the form: Z positive g-factor corresponds to the hole ground state with positive spin projection Jhz. It is convenient to (h) write the effective Zeeman Hamiltonian (8) as HbZ = −2µBκ (JhB) = +2µBκ (JeB) , (8) (h) HbZ = −µBghJhzB. (11) where κ is the magnetic Luttinger parameter [48], which value can be estimated by perturbation theory as [30]: Hamiltonian (11) describes the splitting of the otherwise four-fold degenerate hole ground state in bulk crystals as κ ≈ −2/3 + 5γ/3 − γ1/3. (9) well in spherically-symmetric structures having the same g-factor for light and heavy holes. For example, in bulk Here Jh is the hole spin operator in hole representation semiconductors, the Zeeman effect for both light (Jhz = (in what follows J ≡ Jh) and Je is the electron spin ±1/2) and heavy (Jhz = ±3/2) holes is characterized bulk operator in valence band in the electron representation. by the same g-factor gh ≡ gh = 2κ. Note that the Both in the electron and hole representations, the sign actual Zeeman splitting of heavy holes (Jhz = ±3/2) is of gh and the relative sign between the Zeeman term three times larger than that of light holes (Jhz = ±3/2) and kinetic energy remains the same as the change of for the same g-factor: ∆E3/2 = 3∆E1/2. The scheme of representation results in sign inversion of both energy hole energy levels splitting both in the electron and hole and spins. representations for non-interacting electron and holes is The second, orbital, contribution from magnetic field shown in Fig.2. The definition Eq. (10) and Hamiltonian Eq. (11) are to the hole Hamiltonian, HbB, comes from the hole wave vector k being replaced by k − e A. The explicit form widely used in the physics of colloidal nanocrystals with c J being changed by the total angular momentum pro- of H can be found in Ref. [50] (in the units of the hz bB jection M [21,53–57]. We will use hereafter the definition heavy hole effective Rydberg energy) and in Appendix of Eqs. (10),(11). B (in the units of µ B). In semiconductors with the B It has to be noted, that another definition of hole cubic lattice symmetry, one can also separate the cubi- g-factor with the opposite sign as compared with (10) cally symmetric contribution to the Zeeman part of hole and (11), is also widely used in literature, see, for ex- Hamiltonian, originating from the valence band warp- ample, Refs. [58–61]. Moreover, especially for struc- ing, ∝ q(B J 3 + B J 3 + B J 3). In most semiconductors x x y y z z tures with strong light and heavy holes splitting, the the cubic Luttinger parameter q is small, and the cubic heavy hole Zeeman splitting is often defined as ∆E = contribution to the hole Zeeman splitting is much smaller 3/2 (h) E+3/2 − E+3/2 = µBghhB with ghh = −3gh = −6κ then the isotropic contributions HbZ and HbB [51]. Cubic [58, 59] describing the whole Zeeman splitting of heavy contribution plays an important role only if symmetric holes. While all definitions of hole g-factor follow from contributions lead to the absence of hole Zeeman split- the same Zeeman contribution to hole Hamiltonian Eq. ting. For example, such a situation is realized for heavy (8) and describe the same splitting of hole states in the holes in in-plane symmetric quantum well-like structures magnetic field, one should carefully consider the cho- with the magnetic field applied in the structure plane. sen definition when comparing of the calculated g-factors In what follows, we consider the low field regime, and with experimentally evaluated data. only linear on magnetic field terms are taken into account. In that case, for a bulk hole, the orbital contribution HbB disappears. However, for the localized hole, the B. Hole effective g-factor in low-dimensional states with different orbital momenta are mixed, and HbB structures becomes important. Also, in high magnetic fields, the contributions from quadratic on magnetic field terms ap- The symmetry considerations have proven themselves pearing from HbB become important even for bulk holes. very helpful in the analysis of hole states in the com- It results in a diamagnetic shift and strong mixing of hole plex valence band.[62–66] For localized holes, their in- states with Jhz = ±3/2 and Jhz ± 1/2 and a consequent ternal angular momentum and its projections are not possible magnetic field dependence of the hole effective good quantum numbers anymore. For example, in spher- g-factor. ically symmetric systems, the states could be classified by 6 the hole total angular momentum.[49, 64, 65, 67] In axi- 1. Hole g-factor in spherically symmetric potentials ally symmetric structures, the good quantum number is the total angular momentum projection on the symmetry The spherically symmetric structures represent the axis.[65, 66] In structures with an inversion center, one special case of the highest possible symmetry. In spher- can use a state ”parity” as a quantum number to classify ically symmetric external potential V (r) ≡ V sph(r), hole states.[62, 66] The states with odd and even parity ext hole states can be classified by its total angular are an analog of electron spin-up and spin-down states momentum.[49, 64, 65, 67] Its wave function can be writ- and are degenerate in zero magnetic field. ten following [49] as For a localized hole Hamiltonian takes a form (h) Hb = HbL + HbZ + HbB + Vext(r), (12) p X l−3/2+M l ΨM = 2j + 1 (−1) (i) Rjl(r)× where the potential Vext(r) is a quantum structure po- l tential acting on the hole. For a hole in a free exciton or X l 3/2 j × Y u , (14) an acceptor, Vext(r) is the Coulomb potential. The pro- m µ −M l,m µ file, as well as the symmetry of Vext(r), strongly affect m+µ=M the structure of the hole wave function and, therefore, its g-factor. Here j = Jh + l is the hole total angular momentum In the general case, the problem of hole g-factor cal- with M being its z-axis projection (in considered case culation can not be solved analytically or even semi- j = 3/2, Jh = 3/2), l is the hole orbital momentum, Ylm i k l analytically. Although, in some special cases, the sym- are spherical harmonics [71], m n p are 3j Wigner symmetry of the system allows one to simplify the calcu- bols, uµ are the Bloch functions of Γ8 valence band with lation and obtain the expression for hole g-factor in spin z-axis projection µ [72] (for details see AppendixB, closed semi-analytical form. In this section, we present Eq. (B1)). The hole ground state is four-fold degener- methods for the hole g-factor calculation in such spe- ate SD-like state and consists of functions with l = 0 cial cases. The main effect of the Vext(r) on the hole and l = 2. For simplicity, we denote the respective ra- g-factor is the additional mixing of the states from the dial functions as R0 and R2. Taking into account the heavy and light-hole subbands depending on the mass spherical symmetry of the structure one can simplify the ratio β = mlh/mhh = (γ1 − 2γ)/(γ1 + 2γ). The ad- Schrödingerequation and reduce it to the system of two ditional spin-orbit interaction induced by the external equations for R0 and R2 [49]. potential makes the hole internal angular momentum Jh In external magnetic field the four-fold degenerate not a good quantum number. We assume in the following ground state splits into four equidistant levels with M = that the hole states in each nanostructure are character- ±3/2, ±1/2. As in a bulk semiconductor, such a splitting ized by the total angular momentum projection M on is characterized by the single g-factor gh,3/2 = gh,1/2 = its symmetry axis, which is parallel to the z-direction, g ≡ gsph according to (13). The expression for g ≡ gsph and consider the magnetic field along z as well. In this h h h h was first obtained in Ref. [52] for the hole bound to the case, one can use the definitions similar to Eqs. (10) and acceptor Coulomb potential. We rewrite it as (11) for the hole effective g-factor and Zeeman term, respectively, with Jhz replaced by M. However, except for the spherically symmetric confining potential, the Zee- sph 4 g = 2κS(β) + γ1I(β) , (15) man splitting of the light and heavy subbands might be h 5 controlled by the different effective g-factors depending 4 1 − β 2β S(β) = (1 − Ig) ,I(β) = Ig + Ig . on |M|, so that it is instructive to define 5 2 1 + β 1 1 + β 2

E−M − E+M (h) gh,|M| = , HbZ = −µBgh,|M|MB. (13) Function S(β) describes the renormalization of the spin 2MµBB ˆ h Zeeman contribution HZ , function I(β) describes the or- The first correction to the hole effective g-factor as bital contribution to g-factor correction stemming from compared with bulk comes from the renormalization of ˆ g g HB. Integrals I1 and I2 were introduced in Ref. [52]: the Zeeman term ∝ κ of Eq. (8). This renormalization ˆ hΨM |Jhz|ΨM i/M is a function of β, which particular ∞ ∞ form is controlled by the shape and type of nanostruc- Z dR (r) Z Ig = r3R (r) 0 dr, Ig = r2R2(r)dr . (16) ture potential as it causes the mixing of holes states with 1 2 dr 2 2 different spin projections on the magnetic field direction. 0 0 The second one is related to the orbital contribution ∝ γ1 and γ and is also controlled by some function of β. For It is worth noting, that if the hole wave function does not each semiconductor there are several parametrizations, vanish on the surface of the nanocrystal or in the case of ones we use are taken from [68–70]. For our estimations, the hole confined in the core-shell spherical nanocrys- we considered materials with zinc blende crystal struc- tal constructed from the semiconductors with strongly ture with the inclusion of wurtzite modification of CdSe. different values of Luttinger parameters, one has to use 7

g another expression for the I1 in each material: 0 . 8 Z " g 1 3 dR0(r) dR2(r) I1 = r R2(r) − R0(r) − 2 dr dr 0 . 4 S ( β) ) β

# ( I

2 , − 3r R0(r)R2(r) dr . (17) ) 0 . 0 β ( β

S I ( ) g g - 0 . 4 The exact values of I1 and I2 , and, therefore, S(β) and I(β), depend only on profile of the confining potential and on light to heavy hole effective mass ratio β but - 0 . 8 not on the characteristic size of the localized hole wave 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 function. The expression (15) is obtained for low mag- β netic fields, when only linear on magnetic field terms are taken into account by perturbation theory. Considering FIG. 5. Quantities S(β) and I(β) for the g-factor calculation quadratic on magnetic field terms requires numerical so- in spherical potentials: parabolic, weak and strong Gaussian lution of Schrödingerequation. (black), infinite spherical box (red), Coulomb (green), zero Now we consider g-factor of a hole in spherical NC in radius (blue). sph more detail. For the box-like infinite potential Vbox (r) radial wave functions R0 and R2 can be found analytically, see Refs. [73 and 74]. For smooth parabolic or smaller that hole wave function localization length.[75] Gaussian potentials V sph (r), although the solution Although, for box-like potential with abrupt infinite bar- p,Gauss g g can not be found analytically, the variational method can rier I1 and I2 can be quite different. The g-factors of be used. It allows one to obtain the hole wave function hole in exciton and acceptor are different as well due to for any light to heavy hole effective mass ratio in a sim- renormalization of Luttinger parameter γ1 in exciton.[67] ple analytical form, which demonstrates a good agreement with numerical calculations not only for the energy of the ground state but also for the hole wave func- 2. Hole g-factor in nanocrystals of spheroidal shape tion itself.[75] Thereby, the integrals (16) contributing to the hole g-factor for parabolic and Gaussian potentials Now we consider axially symmetric NCs with the shape can be easily calculated as well. We have demonstrated of spheroid (ellipsoid of revolution). We assume that g g that integrals I1 and I2 almost the same dependence on the symmetry axis, which is the short axis in oblate (or β for both mentioned types of smooth potentials [75]. long axis in prolate) NC, is directed along z. The de- The wave functions of the hole localized on an accep- viation from the spherical shape is assumed to be small tor with the Coulomb potential of a charged center also and can be treated as a perturbation of the spherical can be found by variational method [76] or numerically NC. Such axially-symmetric perturbation splits the hole [67]. For resting exciton (with the vanishing momentum ground state into two Kramers doublets with spin pro- of the center of masses), the solution is the same as for jections M ± 3/2 and M ± 1/2 on the crystal axis sim- the acceptor with the renormalized Luttinger parameter ilar to the effect of the internal crystal field in wurtzite 0 γ1 = γ1 + m0/me, with me being electron effective mass. semiconductors.[54, 75, 77, 78] The total anisotropic Due to the renormalization, which makes the effective β splitting of the hole states is given by ∆ = vcr(β)∆cr + closer to unity as the electron usually is much lighter than ∆sh with the crystal field-induced splitting vcr(β)∆cr be- the heavy hole, the effects of the complex band structure ing independent of a and the shape anisotropy induced 2 are substantially weaker for a hole in an exciton than for contribution increasing as ∆sh ∝ vsh(β)µ/a in small a hole in an acceptor. NCs. The proportionality coefficient vsh(β) is different The dependencies of functions S(β) and I(β) for box- for the abrupt box-like and smooth parabolic potentials like, parabolic, Gaussian, zero-radius, and Coulomb po- and may have opposite signs for β < 0.14.[75] For the tentials are shown in Fig.5. Note, that the sign of S(β) vast majority of Luttinger parameters sets in oblate NCs remains always the same and its absolute value decreases the ground state is a state with the total angular mo- up to 3/5 of the bulk value for β = 0. In opposite, func- mentum projection M = ±3/2 on NC axis (it will be re- tion I(β) is negative and may result in the change of ferred as heavy holes as it consists mostly of Bloch states bulk gh sign as compared with bulk hole g-factor gh . For with Jhz = ±3/2), and M ± 1/2 (light holes) in prolate NCs with parabolic and Gaussian potential profiles the NCs.[50, 75, 77] Note, that changing ∆sh one can com- g g values of I1 and I2 are almost the same for given mate- pensate crystal field related splitting and even reverse the rial parameters.[75] Here we denote the strong Gaussian sign of the total splitting. For example, for wz-CdSe with potential as potential, which characteristic size is com- β = 0.28, the ”quasi-spherical” situation with ∆ = 0 and parable with hole wave function localization length. The degenerate hole state can be realized for both types of characteristic size of weak Gaussian potential is much the confining potentials with the prolate shape. Below 8 we study the effect of such nanocrystal anisotropy on the heavy and light hole g-factors, g and g , corre- ) h,3/2 h,1/2 d - 0 . 7 0 i spondingly. o ( a ) r - 0 . 7 2 o b l a t e p r o l a t e e

As an example we consider the parabolic conﬁning po- h

p - 0 . 7 4 s tential. Following Ref. [75], we introduce the shape c

i - 0 . 7 6 anisotropy of quantum dot as anisotropy of its potential: l o

b - 0 . 7 8 a

κ κ an ρ 2 z 2 r Vext(r) ≡ Vp (ρ, z) = ρ + z (18) a - 0 . 8 0 2 2 p (

l i g h t h o l e s κ 2 an r - 0 . 8 2 o

= r + ∆Vp (ρ, z, µ) , t h e a v y h o l e s 2 c - 0 . 8 4 a f 1 -

an 2 2 g - 0 . 8 6 ∆V (ρ, z, µ) = −κµ z − r . p 3 0 . 2 ( b )

Here κ = (2κρ +κz)/3 is the average stiﬀness of parabolic 0 . 0 o b l a t e p r o l a t e

potential and µ is the anisotropy parameter: ) e b

(κ − κ ) 2 4 u - 0 . 2

ρ z c (

µ = , κρ ≈ κ 1 + µ , κz ≈ κ 1 − µ . 2κ 3 3 r o - 0 . 4 t

c l i g h t h o l e s a

The dependence of the light-heavy hole energy splitting f - - 0 . 6 h e a v y h o l e s ∆Ean = E1/2 − E3/2 on µ was studied in Ref. [75]. At g small enough µ this dependence is linear and can be cal- - 0 . 8 culated within the first-order perturbation theory. The - 0 . 2 - 0 . 1 0 . 0 0 . 1 0 . 2 slope ∆Ean/(µE) depends only on effective mass ratio µ β. The dependencies of light and heavy holes g-factors on the anisotropy parameter µ calculated for zinc blende CdSe are shown in Fig.6(a). In Fig.6(b) the simi- FIG. 6. The dependencies of the light, gh,1/2, and heavy lar dependencies for cubic NCs are shown, for details see hole, gh,3/2, g-factors on the anisotropy parameter µ in (a) spheroidal NCs with parabolic potential and (b) cuboid NCs section IIIB3. with box-like infinite potential. The calculations are done One can see, that at small enough µ, the dependencies for zb-CdSe parameters given in TableI with β = 0.22 and shown in Fig.6(a) are also linear and can be approxi- gsph = −0.78. mated as h

sph sph α gh,3/2 = gh − α3/2µ, gh,1/2 = gh − α1/2µ, (19) 0 . 5 1 / 2 α where α and α are some constants. From Fig.6(a) 0 . 4 3 / 2

3/2 1/2 e one can see, that slopes α and α have not only op- p 3/2 1/2 o 0 . 3 l s

posite signs but also diﬀerent absolute values. In other r 0 . 2 words, in anisotropic nanocrystals the light and heavy o t c holes g-factors would be diﬀerent. The numerically cal- a 0 . 1 f -

culated dependencies of α3/2 and α1/2 on the light to g heavy hole mass ratio β are shown in Fig.7. 0 As our analysis showed, the coefficients α3/2 and α1/2 - 0 . 1 can be calculated by perturbation theory as first-order - 0 . 2 corrections coming from HB by using wave functions, 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 which already take into account the shape anisotropy in β zero magnetic field. Alternatively, one can use a coordinates substitution FIG. 7. The dependencies of the anisotropy slopes for the p p p light, α , and heavy, α , holes g-factors on the effective x → x κ/κρ, y → y κ/κρ, z → z κ/κz, which makes 1/2 3/2 an mass ratio β in spheroidal nanocrystals with parabolic poten- a quantum dot potential spherical so that ∆Vp = 0 but result in additional terms in the kinetic energy Luttinger tial. an ˆ an Hamiltonian HL → HL +HL [76] and HB → HB +HB . an ˆ an The HL can be found in [75] and AppendixB, the HB the first-order of perturbation theory and mutual correc- is given by: an tions from the HB and HL in the second-order perturba- ˆ an tion theory. The derivation of the analytical expressions HB = −2γµµBB(JyJz + JzJy)xkz . (20) for αhh and αlh is, therefore, the problem for a separate To calculate the corrections to hole g-factors due to the paper. anisotropy in this method, one has to calculate the cor- The method of the coordinates substitution x → xb/a, rections to the energy splittings ∆EM from Eq. (20) in y → yb/a, z → zc/a can be also used for the spheroidal 9 nanocrystals with abrupt box potential at the surface

2 2 2 2 2 2 x /b + y /b + z /c = 1. In this case, the small 0 . 8 anisotropy parameter is given by µ = c/b − 1 with | M | = 3 / 2 ( E 0 ) ( a ) | M | = 1 / 2 ( E 0 ) c ≈ a(1 − 2µ/3), b ≈ a(1 + µ/3) [77]. Similarly, this 0 . 4 | M | = 3 / 2 ( E 1 ) method can be applied to treat the anisotropy in the | M | = 1 / 2 ( E 1 ) NCs of square cuboid shape with abrupt box-like po- )

β 0 . 0 tential and dimensions 2b and 2c. The coordinate sub- ( 1

stitution transforms them into the NCs of cubic shape Q with the cube edge 2a. Such nanocrystals with the cubic - 0 . 4 shape are considered in the next subsection IIIB3 and AppendixC. - 0 . 8 It is worth noting, that in contrast to the potential 1 . 2 shape anisotropy, the effective crystal field perturbation ( b ) ∝ ∆cr in wurtzite semiconductors does not affect the hole 0 . 8 g-factors (while magnetic field is weak enough). Thus, the crystal field and magnetic field perturbations can be 0 . 4 )

taken into account separately, and both light and heavy β

( 0 . 0 hole Zeeman splittings in spherical wurtzite NCs can be 2 sph Q described by gh . In contrast, on needs two g-factors, - 0 . 4 gh,3/2 6= gh,1/2, for ”quasi-spherical” wurtzite NCs. - 0 . 8

3. Hole g-factor in nanocrystals of cubic shape 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 β

Here we consider nanocrystals with the infinite potential of cubic shape or cube NCs (see details in Appendix FIG. 8. The dependencies (a) Q1,|M|(β) and (b) Q2,|M|(β) C). Note, that we still neglect the effects of cubic sym- for the light (|M| = 1/2) and heavy (|M| = 3/2) hole two metry of crystal lattice. In this case, the system under first quantum states E0 and E1 in nanocrystal of cubic shape calculated numerically. study loses its spherical and even axial symmetry. The total angular momentum and its projection on magnetic field direction (unless the magnetic is directed along one The hole effective g-factors for the light (|M| = 1/2) of the cubic axes) are not good quantum numbers. Im- and heavy (|M| = 3/2) hole states can be written as portantly, the hole ground state remains four-fold degenerate in zero magnetic field because cubic anisotropy does gh,|M| = 2κQ2,|M|(β) + γ1Q1,|M|((β), (21) not split states with total angular momentum less than 5/2 (as it would be in spherically symmetric NC) [79]. where functions Q1,|M|(β) and Q2,|M|(β) depend only on We have shown before [80], that in spherical NCs light to heavy hole effective mass ratio β and describe the with box-like or parabolic potential the energies of the orbital contribution to hole g-factor and the renormaliza- two lowest hole states with the total angular momen- tion of the spin contribution, respectively. In contrast to tum j = 3/2 but different orbital momenta (SD-like with the spherical case, these functions are now different for l = 0, 2 and PF -like with l = 1, 3) become close to each the heavy and light hole g-factors. other at small values of β as well. However, the SD- The dependencies of the functions Q1,|M|(β) and like state remains the lowest energy state for all β values (Q2,|M|(β) for two lowest quantum states of cube NC cal- in spherically symmetric potentials. The striking feature culated numerically are shown in Fig.8. It is interesting, of cube nanocrystals is that there is a range of β val- that even in cube nanocrystal, where the hole ground ues, where the hole ground state is a PF -like (E1) state and first excited states in zero magnetic field are four- instead of SD-like (E0) one. The dependencies of the en- fold degenerate, the g-factors of heavy and light holes ergies of two lowest hole states, E0 and E1, on β in zero are different and at some band parameters can have even magnetic field are shown in Fig. C1 (a) in AppendixC. opposite signs. Such a situation is demonstrated in Fig. In the range of β where the intersection occurs, the 6 (b), there the dependencies of light and heavy hole energy of both states are close to each other. This is ground state g-factors are shown as functions of cuboid reflected in the significantly different contributions from CdSe NC anisotropy parameter µ. This effect is the man- different Bloch states to the E0 state comparing to the ifestation of the breaking of spherical symmetry and can range of β values, where the E0 and E1 energies are far be qualitatively understood by considering a cubic con- 3 3 3 from each other (see Fig. C1(b) in AppendixC). This tribution Q(BxJx + ByJy + BzJz ) to the spherically- fact exhibits itself in the values of hole g-factors as func- symmetric hole Hamiltonian with a large Q (unlike ne- tions of β even for the magnetic field directed along the glected term with a small q in bulk semiconductors and symmetry axis z of the cubic NC. spherical nanocrystals, which originates from cubic sym- 10

1 . 0 Here W ( β) 2 2(3 + β) 1 − β √ W (β) = F (β) − 3H(β), (24) 1 1 + β 1 + β ) 0 . 5 β

( W2(β) = 1 − 4F (β), 2 ∞ ∞ W

Z Z

, 2 dψ1(ρ) 2 ) 0 . 0 F (β) = ψ2(ρ)ρdρ, H(β) = 2 ψ2(ρ) ρ dρ. β dρ ( 1 0 0 W - 0 . 5 Function W1(β) describes the orbital contribution to g- ˆ β factor correction stemming from HB while the function W 1 ( ) W2(β) describes the renormalization of the spin Zeeman - 1 . 0 contribution Hˆ h . Note, for the cylindrical core/shell het- 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 Z β erostructure or nonvanishing hole wave function at the cylindrical surface, more general expression for each material is: FIG. 9. Functions W1(β) and W2(β) for g-factor in wire-like ∞ potentials: parabolic and strong Gaussian (solid line) and Z dψ (ρ) dψ (ρ) weak Gaussian (dotted line). H(β) = ψ (ρ) 1 − ψ (ρ) 2 ρ2dρ− 2 dρ 1 dρ 0 ∞ metry of crystal lattice itself). If one would include Z − 2 ψ (ρ)ψ (ρ)ρdρ. (25) such a contribution in Hamiltonian (12) with a spheri- 1 2 cal nanocrystal potential, the resulting g-factors for light 0 and heavy holes will be different. Functions W1,2(β) for parabolic and strong Gaussian potentials (solid line) and weak Gaussian potential (dotted line) are shown in Fig.9. The accuracy of Eq. (23) 4. Hole g-factor in cylindrical quantum wires was checked with the numerical calculation. For infinite box-like potential both F (β) and H(β) are close to zero If the symmetry of nanocrystal potential is lower than due to negligible mixture of hole states and ψ2 in (22) spherical or cubic, heavy and light holes states are split being small. even in zero magnetic field. For cylindrical quantum wires, in which the structure size along the symmetry axis is much larger than in its cross-section, the hole ground 5. Hole g-factor in a nanoplatelet state is characterized by the total angular momentum projection M = ±1/2 and constructed mostly from the In a quantum well-like nanocrystal, nanoplatelet, light hole states [50] with Jhz = ±1/2. However, this where the thickness is much smaller than in-plane dimen- state has a small admixture of the valence band states sions, the heavy and light hole states are strongly split in with Jhz = ∓3/2 and orbital momentum projection ±2 zero magnetic field (about 150 meV in 4 ml NPL [82]). even in the 1D limit of the nanowire with infinite length In the in-plane isotropic NPL, the hole ground state is and zero potential along the wire axis. The correspond- characterized by the total angular momentum projection ing wave functions are:[50] M = ±3/2 on the symmetry axis directed perpendicular to the NC plane (z-axis). In the limiting case of an isotropic NC structure with infinite in-plane size, the hole wire 1 ΨM=± 1 = ψ1(ρ)|Jhz = ± i+ ground state is composed only of the heavy hole valence 2 2 3 band states with Jhz = ±3/2 with a vanishing admix- + ψ2(ρ) exp(∓2iϕ)|Jhz = ∓ i, (22) ture of light holes. Therefore, in the limiting case of a 2 two-dimensional (2D) structure with the vanishing width where ρ and ϕ are polar coordinates in the quantum wire to in-plane size ratio there is no renormalization of the cross section, ψ1(ρ) and ψ2(ρ) are in-plane radial wave spin Zeeman effect. However, as it is was shown in Ref. functions, depending on β and quantum wire potential. [83], the orbital corrections to the heavy hole g-factor 2D In-plane wave functions ψ1(ρ) and ψ2(ρ) can not be found gh,3/2 ≡ gh,3/2 coming from the HB in the first order analytically in general case, but can be calculated, for perturbation theory for magnetic field directed along the example, by the variational method.[81] symmetry axis are present due to the quantization of kz Magnetic field lifts Kramers degeneracy, and the hole and magnetic field induced heavy and light hole mixing wire ground state g-factor gh,1/2 ≡ gh,1/2 was derived in Ref. (see details in AppendixB): [50]. Similarly to (15) we rewrite it : 2 ∞ ˆ 2 X |hlh2n|γkz|hh1i| g2D = 2 − 4 ~ . (26) wire h,3/2 κ m E − E gh,1/2 = 2κW2(β) + γ1W1(β) . (23) 0 n=1 lh2n hh1 11

as well as on crystallographic orientation of the structure

2 . 5 [84]. In the analysis of the experimental data, however,

3 it is possible to take into account the anisotropy of the 2 . 0 gh,3/2 via the coeﬃcient cos Θ, where Θ is an angle 2 between the magnetic ﬁeld direction and the anisotropic

axis [21, 85]. In this approach, the Zeeman splitting of 1 . 5 1 the heavy-hole is zero for the magnetic ﬁeld directed in )

( the NPL plane unless the eﬀects of the qubic terms or 1 . 0 - 5 - 5 G 0 . 0 2 . 0 x 1 0 4 . 0 x 1 0 additional light-hole to heavy-hole mixing induced by the in-plane anisotropy are taken into account [50]. 0 . 5

0 . 0 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 β C. Hole g-factor in diﬀerent semiconductor

nanostructures: comparison with experimental data.

FIG. 10. G(β) for calculation of g-factor in well-like potentials for parabolic and strong Gaussian (solid line), weak Gaussian In Table I and Fig. 11 the hole g-factors for discussed (dash-dotted line), and box-like infinite (dotted line) poten- above cases in II-VI and III-V semiconductors with differ- tials. Thin dashed line shows asymptotics for box-like poten- ent parametrizations for Luttinger parameters are shown. tial for β → 0, thin dash-dotted line shows asymptotics for One can see in the Fig. 11 that in spherically symmetric parabolic potential (for details see AppendixB). The inset potentials, wire potential, and cube potential for a heavy shows G(β) for parabolic potential in vicinity of β = 0. hole, the hole g-factor is almost the same for all the studied semiconductors and Luttinger parameters. Contrary, for NPLs with parabolic or box-like potential, and cube Here |hh1i is the envelope wave function of heavy hole NCs for the light hole, hole g-factors show distinct de- ground state of the quantization along z-axis, |lh2ni are pendence on the semiconductor material and Luttinger wave functions of even excited states of a light hole, Ehh1 parameters. Worth to note, that for wz-CdSe localiza- and Elh2n are the corresponding energies. After summation of a hole results in a minor renormalization of the tion Eq. (26) for 2D hole g factor can be written as bulk hole g-factor. In contrast, in zb-CdSe, InP, and γ GaAs nanostructures the hole g-factor experiences strong g2D = 2 − 1 G(β), (27) h,3/2 κ 3 renormalization up to sign inversion. This strong renormalization is caused by the orbital contribution ∝ γ1. where G(β) depends only on β and the type of the local- In the literature the following experimental hole g- ization potential along the z direction. In a similar way, factors were reported: gh = −1.04 and gh = −0.76 one may obtain the 2D-limiting case expression for the in bare core wz-CdSe NCs with diameters 2.5 and 1.9 light-hole g-factor, g2D , at the lowest light-hole size- h,1/2 nm, respectively [16]; gh = −0.6 in CdSe/ZnS NCs [19]; quantization level, Elh1 , (see Eq. (B14)). In this case, the gh = −0.54 in CdSe/CdS NCs with thick shell [21]. In corrections to the bulk value 2κ come from the admix- zb-CdSe based nanoplatelets, gh = −0.4 was reported in ture of the even excited states of the heavy-hole, Ehh2n . CdSe/CdS NPLs with thick shell (changing to gh = −0.7 Importantly, a proximity of the lowest light-hole state with increase of magnetic field) [85], and gh ≈ −0.1 (in energy with one of the energies Ehh2n of the heavy-hole the range of -0.03 and -0.2) in bare-core CdSe NPLs may result in a giant enhancement of the light-hole Zee- [55, 56]. These values were deduced from the circular po- man splitting [59]. Even in the absence of such a giant larized emission of the negative trions. Interestingly, the enhancement, for example in the case of thin CdSe NPLs, hole g-factors in CdSe/CdS NPLs, gh = −0.4 and −0.7, 2D 2D gh,1/2 6= gh,3/2. are close to the values calculated for a hole in spheri- The dependencies of parameter G(β) on β for parabolic cally or cubically symmetric potential, while gh = −0.1 and strong Gaussian, weak Gaussian, and box-like infi- is close to the value calculated for a hole in 2D parabolic nite potentials are shown in Fig. 10. In the figure, one (not box-like) potential (see Table A1). A smooth con- can see the much stronger dependence for smooth poten- fining potential instead of an abrupt one may be induced tials due to smaller distances between hole energy levels in thin NPLs due to the dielectric confinement effect. In- as compared with box-like potential. Also here there is deed, the repulsion of the carries from the image-charge almost no difference between parabolic and strong Gaus- results in the additional repulsing potential ∼ 1/d, where sian potentials and a very slight difference with weak d is the distance to the surface. Gaussian potentials. The expressions above for hole g-factors in spherical, Here we discussed the case of the magnetic field cube, quasi-1D and ideal 2D structures, Eqs. (15), (21), directed perpendicular to the structure plane. The hole (23) and (26), are written neglecting the quadratic on g-factor in NPLs and other planar nanostructures is magnetic field terms. Taking it into account results in strongly dependent on magnetic field direction [36, 60] the hole g-factors nontrivial dependence on the mag- 12

parabolic box-like parabolic box-like parabolic bulk

FIG. 11. Hole g-factors for different II-VI and III-V semiconductor model nanostructures. Confinement potentials notation: sph.par. - spherical parabolic potential, sph.box - spherical infinite box potential, QW par. - NPL with parabolic potential along the z-axis, QW box - NPL with infinite box potential, wire par. - nanowire with parabolic potential in the cross-section plane, cubic gh,3/2 - cubic potential for heavy hole, cubic gh,1/2 - cubic potential for light hole.

netic field. Indeed, the magnetic field substantially mod- dence. Recently, gh = −1.9 was reported in InP/ZnSe ifies light and heavy hole mixing as Zeeman splitting spherical NCs [112], which is much larger than the theo- becomes comparable with distances between hole size- retically calculated values for two parameterizations (see quantization energy levels [66, 104]. This effect could be TableI). Besides the corrections coming from the admix- enhanced if the structure is anisotropic in the plane per- ture of Γ7 hole states, the anisotropic corrections to heavy pendicular to the magnetic field [105]. The nonlinear de- and light hole g-factors caused by NC shape could be also pendence of hole Zeeman splitting on the magnetic field important. Indeed, it was argued in Ref. [112] that for was also observed in quantum wells (e.g. [104, 106, 107]). zinc-blende InP with β = 0.15 the splitting ∆ between In the calculations presented in the TableI and in Fig. the heavy and light hole energy levels in the spheroidal 11 the renormalization of Luttinger parameters γ1, γ and NCs is very small as vsh(β = 0.14) = 0. Therefore, one κ caused by the quantum confinement of holes which sim- may deal with the degenerate hole state (and as con- ilar the to effect of the non-parabolic energy dispersion sequence, with the isotropic exciton, see the next sec- for electrons [108] was neglected. Such a renormalization tion) even in prolate and oblate InP NCs. We have can be taken into account within the 8-band Kane model shown above, however, that even in this case the shape [58, 109] and results in the size-dependence of the hole anisotropy results in the corrections to the gh,3/2 6= gh,1/2 g-factor similar to the effect for electron [23, 36, 58]. This which should be taken into account in the analysis of the effect might be important for small NCs and thin NPLs experimental data. Such a situation can be also expected when the hole energy Eh becomes comparable with Eg. for the spheroidal NCs based on GaAs where β = 0.14. For quantum wells, the size dependence of the hole g- As we have shown above, a similar situation can be real- factor was widely studied experimentally [36, 104, 110]. ized in the case of the hole state in cube NCs based on For quantum wires and spherical NCs with small ra- zinc-blende with degenerate hole state (∆ = 0) in zero dius a, hole g-factors may depend on a because of the magnetic field and equidistant levels splitting in the ex- admixture of different bands by spin-orbit interaction ternal magnetic field with gh,3/2 = gh,1/2. A similar situ- [104, 111]. ation is also expected for the wz-CdSe ”quasi-spherical” Importantly, if the hole quantization energy becomes NCs, where the anisotropic splitting caused by their prolate shape compensate exactly the crystal field splitting. comparable with the spin-orbit energy ∆SO, the admixture of Γ7 spin-orbit split valence band states (as ex- Above, we considered the ideal model nanostructures, pected in InP where ∆SO = 100 meV) must be taken into for example, 2D quantum well or quantum wires with an account as well and may result in additional corrections infinite in-plane area or length, respectively. For such to the hole g-factor and, consequently, to its size depen- structures as well as for spherical and cube NCs con- 13

TABLE I. Hole eﬀective g-factors in nanostructures. bulk sph sph 2D 2D wire gh gh , gh , gh,3/2, gh,3/2, gh,1/2, cube cube ∗ Material γ1 γ κ β gh,3/2 gh,1/2 Refs. = 2κ par. box par. box par. zb-CdSe 5.51 1.78 0.46 0.93 0.22 -0.78 -0.94 -0.11 0.36 -1.36 -0.78 0.15 [68] zb-CdSe 3.27 1.33 0.46 0.93 0.1 -0.83 -1.12 -0.13 0.46 -1.17 -0.81 0.36 [69] zb-CdSe 3.8 1.65 0.81 1.62 0.07 -0.83 -1.05 0.17 1.03 -1.24 0.49 1.34 [69] wz-CdSe 2.04 0.58 -0.38 -0.76 0.28 -1.16 -1.12 -1.05 -0.93 -1.3 -1.1 -0.95 [86] wz-CdSe 2.52 0.83 -0.12 -0.25 0.2 -0.98 -1.05 -0.74 -0.5 -1.21 -0.96 -0.73 [87] wz-CdSe 1.7 0.4 -0.57 -1.13 0.36 -1.32 -1.29 -1.3 -1.23 -1.42 -1.29 -1.21 [88] wz-CdSe 2.1 0.55 -0.45 -0.9 0.31 -1.23 -1.19 -1.16 -1.05 -1.37 -1.18 -1.03 [42] wz-CdSe 1.67 0.56 -0.29 -0.58 0.2 -1.04 -1.09 -0.92 -0.76 -1.17 -1 -1 [89] zb-ZnSe 3.94 1.31 0.21 0.41 0.2 -0.86 -0.94 -0.38 -0.004 -1.26 -0.87 -0.31 [68] zb-ZnSe 3.77 1.5 0.57 1.14 0.11 -0.79 -1.16 0.003 0.63 -1.22 -0.81 0.31 [90] ZnTe 4 1.11 -0.15 -0.29 0.28 -0.56 -1.03 -0.26 -0.03 -1.45 -1.03 -0.55 [91] ZnTe 3.8 1.14 -0.04 -0.08 0.25 -1 -0.98 -0.69 -0.42 -1.36 -0.95 -0.44 [92] ZnTe 3.9 1.11 -0.11 -0.22 0.27 -1.07 -0.97 -0.79 -0.55 -1.4 -0.99 -0.5 [93] ZnTe 3.8 1.07 -0.15 -0.31 0.28 -1.1 -1.01 -0.84 -0.62 -1.42 -1.09 -0.62 [94] CdTe 4.14 1.4 0.3 0.6 0.19 -0.82 -1.04 -0.27 0.15 -1.24 -0.84 0.23 [95] CdTe 5.3 1.88 0.7 1.4 0.17 -0.69 -1.1 0.17 0.78 -1.3 -0.84 -0.04 [92] CdTe 4.11 1.6 0.63 1.26 0.12 -0.76 -1.2 0.08 0.72 -0.71 -0.78 -0.24 [96] GaAs 6.98 2.63 1.39 2.77 0.14 -0.57 -1.3 0.91 1.88 -1.38 -0.69 0.02 [97] GaAs 6.85 2.58 1.35 2.7 0.14 -0.58 -1.3 0.87 1.83 -1.37 -0.69 0.02 [98] GaAs 7.17 2.89 1.77 3.54 0.11 -0.60 -1.38 1.28 2.53 -1.45 -0.68 0.31 [96] GaAs 6.79 2.38 1.03 2.06 0.18 -0.61 -1.36 0.53 1.28 -1.37 -0.78 0.41 [99] GaAs 6.8 2.4 1.06 2.12 0.17 -0.60 -1.09 0.56 1.33 -1.37 -0.79 0.38 [100] GaAs 7.2 2.69 1.42 2.83 0.15 -0.55 -1.3 0.95 1.92 -1.39 -0.92 0.1 [92] GaAs 7.1 2.55 1.22 2.44 0.16 -0.57 -1.74 0.75 1.6 -1.37 -0.8 0.39 [101] InP 5.05 1.68 0.45 0.9 0.2 -0.77 -0.94 -0.11 0.66 -1.31 -0.89 0.09 [102] InP 4.6 1.68 0.6 1.2 0.15 -0.73 -1.2 -0.06 0.64 -1.25 -0.9 -0.35 [102]

∗ References are given for the γ1, γ2 and γ3 Luttinger parameters. We use relations γ = (2γ2 + 3γ3)/5 [103] and κ ≈ −2/3 + 5γ/3 − γ1/3 [30].

sidered in the framework of Luttinger Hamiltonian, hole each direction α = x, y, z as Lα = 4Lh,α, where Lh,α is g-factors are independent of the characteristic size of the the oscillator length along respective direction calculated system unless the effect of the renormalization of Lut- with the heavy hole effective mass. One can see from the tinger parameters is considered. The realistic colloidal Fig. 12, that in the limit L = Lz → 0 for both potentials NPLs, however, have a finite thickness and a finite in- all curves tend to the same limit, corresponding to the 2D plane size [82, 113, 114]. We discuss below the effect of ideal 2D quantum well: gh,3/2 = −0.11 for the parabolic the finite in-plain size of NPL on the heavy hole g-factor. 2D potential and gh,3/2 = 0.36 for the abrupt box-like po- Even conventional semiconductor quantum wells, while tential. Dashed line on both panels corresponds to the their in-plane size could be quite large, have a finite thick- value of the heavy hole g-factors in cubic (spherical) NC ness. The finiteness of the ratio of the NPL thickness to which are occasionally the same in the case of zb-CdSe: in-plane size results in non-zero light and heavy holes sph cube gh ≈ gh,3/2 ≈ −0.78. With an increase of L, gh,3/2 mixing even in zero magnetic field. This leads to the starts to depend on the NC in-plane size and thickness nonzero first order perturbation renormalization of the even without account of the energy dependence of the spin Zeeman term as well as orbital correction to the Luttinger parameters. Overall, the smaller the in-plane hole effective g-factor. As a consequence, the hole ground cross-section, Lx × Ly, the stronger is the dependence. 2D 2 state g-factor is somewhere in between the 2D, gh,3/2, For the smallest cross-section, 8 × 8 nm , at L = 8 nm sph the case of cubic (spherical) NCs takes place, so at L > 8 and spherical, gh limits. The in-plane anisotropy of the structure results in additional corrections to g-factor nm we have the prolate NC with the ground state being values as well. In Fig. 12 we show the dependencies of the light hole. It leads to the discontinuity of the ground the hole ground state g-factor on nanocrystal width L state g-factor in cube NC and to the knee in g-factor de- for different in-plane sizes in the case of (a) box-like in- pendence for parabolic NC (see Fig.6). On both panels finite and (b) parabolic potentials. The calculations are one can see, that for realistic in-plane size of the NPL, done for the zb-CdSe parameters given in TableI with the dependence of hole g-factor on its thickness could be β = 0.22. For the sake of comparison, we define that size important. Also the values of hole g-factors even in thin Lα of the nanocrystals with parabolic potential along NPL can differ from value for 2D-limit because of the 14

treated as independent non-interacting quasi-particles. )

c 0 . 4 In Fig.2 the energy levels and allowed transitions are i l ( a ) 3 0 x 3 0 o 8 x 1 6 shown schematically for the 4-fold degenerate hole en- b 0 . 2 2 a L x L ( n m ) 5 x 2 4 r x y

a 1 8 x 1 8 ergy level in zero ﬁeld Eh,3/2 = Eh,1/2 = Eh and

p 0 . 0

l 8 x 8 gh,3/2 = gh,1/2 = gh realized in the bulk zinc blende a

d - 0 . 2 i semiconductors or spherical nanocrystals. o s

p - 0 . 4 i l l e

( - 0 . 6 E − E E − E r σ+ σ− F =+1 F =−1

o g = = , (28)

t ex,1

c - 0 . 8 µBB µBB a f -

g - 1 . 0 where F = Jhz + Sz = ±1 or F = M + Sz denotes the ) 0 . 4 r

a total exciton spin projection on magnetic ﬁeld direction. l

u 0 . 2

g From Eq. (28), the exciton eﬀective g-factors for the

n ( b )

a hh lh t 0 . 0 bright exciton with heavy, g , and light, g , hole

c ex,1 ex,1 e r

- 0 . 2 are: e k i

l - 0 . 4 - hh lh

x gex,1 = −(ge + 3gh,3/2), gex,1 = ge − gh,1/2 . (29) o

b - 0 . 6 (

r The dark exciton states with F = ±2 are not optically o - 0 . 8 t c

a active in one-photon processes in dipole approximation. f

- - 1 . 0

g However, for such excitons one can use the deﬁnition of 2 4 6 8 1 0 1 2 n a n o p l a t e l e t t h i c k n e s s L ( n m ) g-factor similar to Eq. (28):

EF =+2 − EF =−2 gex,2 = = ge − 3gh,3/2 . (30) FIG. 12. Size dependencies of hole ground state g-factor µBB for (a) ellipsoidal nanocrystals with parabolic potential and (b) cuboid nanocrystals with box-like infinite potential. The The definition (28),(30) is used, for example, in Refs. nanocrystal width L is the width along the nanocrystal axis [53, 54, 60]. Note, however, that depending on the z being the direction of the external magnetic field and the chosen hole g-factor definition, the resulting expressions curves of different colors correspond to different NC in-plane for the exciton g-factors may differ. In addition, an- 2 sizes shown as Lx × Ly in nm . The calculations are done other definition of the exciton g-factors with gex,|F | = for zb-CdSe parameters given in TableI with β = 0.22. The (EF − E−F )/2F µBB can be used. Often in the litera- dashed horizontal lines correspond the g-factor value in (a) ture, for example in [55, 115], the dark exciton g-factor spherical and (b) cube CdSe NCs: gsph ≈ gcube ≈ −0.78. h h,3/2 is denoted as gF ≡ gex,2. For the correct description of exciton Zeeman splitting the fine structure of exciton energy needs to be taken into strong nonparabolicity effect. account even for excitons confined in the nanocrystals in a strong confinement regime. First, we consider exciton fine energy structure in zero magnetic field. In the 2D IV. EXCITON EFFECTIVE g-FACTOR case of a nanoplatelet, the strong quantization along the symmetry axis splits the heavy and light hole states by The knowledge of electron and hole effective g-factors the large energy ∆. The further structure of the exciton in nanostructures allows one to describe energy split- states formed with heavy and light holes is determined tings of allowed optical transitions in a low external mag- by the electron-hole exchange interaction with character- netic field and to calculate respective bright exciton ef- istic energy ∆exch = 4η > 0 which is much smaller than fective g-factors. For excitons formed with heavy holes ∆ and results in the energy splitting between states with Jhz = ±3/2 (or M = ±3/2 in nanocrystals), the allowed different values of |F | (Fig. 13 (a)). In the notation of transitions in σ± polarization are: the exciton states, apart from the total projection of the exciton spin F , the letters U and L are used for the up- 1 3 per and lower levels with the same F . In the 2D case Eσ± = Eg + Ee + Eh,3/2 ∓ geµBB ∓ gh,3/2µBB. 2 2 in Fig. 13 (a), lower excitons are the heavy-hole excitons and the upper are formed with the light holes. The For transitions with light hole excitons, J = ±1/2 or hz ground exciton ±2 is dark and the heavy-hole bright ex- M = ±1/2 one has L citon ±1 is shifted up by ∆EAF = 3η ≈ 5 in 4 ml 1 1 CdSe NPL [82]. We assume that the magnetic field en- Eσ± = Eg + Ee + Eh,1/2 ± geµBB ∓ gh,1/2µBB. ergy µ B is much smaller than ∆ and ∆ . Then the 2 2 B exch Zeeman energy splitting of the exciton states with F 6= 0 Here we neglected both the direct Coulomb and ex- is highly anisotropic (depends on the angle between the change electron-hole interaction and electron and hole are anisotropic axis and external magnetic field Θ) and it is 15

c c ϴ ϴ B B B

(a) nanoplatelet (cubic CdSe) (b) spherical nanocrystal (cubic CdSe) (c) spherical nanocrystal (wurtzite CdSe)

+1 +1 +1

-1 -1

-1

Energy +1 Energy Energy +2

-1 -1 +2 +1 +2 -2 -2

-2

FIG. 13. Scheme of the exciton energy levels in magnetic ﬁeld. (a) light and heavy hole splitting ∆ is large as compared with exciton splitting due to exchange interaction ∆exch; (b) ∆ = 0; (c) ∆ ∆exch, all resulting g-factors are assumed to be positive, Θ is the angle between anisotropy axis and magnetic ﬁeld. The order of energy levels is sketched for ge = 1.7, sph gh = −0.7, gh,3/2 = −0.1, gh,1/2 = 0.1.

L hh U lh controlled by gex,1 = g1,ex and gex,1 = g1,ex from Eqs. tween levels changes with NC size. The level structure (29) for bright and Eq. (30) for dark excitons, corre- in Fig. 13 (c) is shown for the case 0 < ∆ ∆exch. spondingly. Note, that the 0L is a dark state and the In external magnetic field the Zeeman splitting of the excitation probability of the 0U is strongly dependent on dark exciton state with |F | = 2 is, for any nanostructure, light propagation direction [54]. sph still given by Eq. (30) with gh,3/2 = gh in spherical In opposite, for small spherical NCs of zinc blende or nanocrystals. However, g-factors of bright exciton states wurtzite semiconductors exchange interaction plays a key with |F | = 1 are modified due to the mixing of the light role in exciton energy structure. The electron-hole ex- and heavy hole states and are given in general case as change interaction in the spherically symmetric struc- gL,U = |C±|2ghh + |C∓|2glh . (31) tures of zinc-blende semiconductor split the 8-fold de- ex,1 ex,1 ex,1 generate by the total exciton projection F = M + Sz Here C± are the heavy to light hole mixing coefficients exciton ground state into two states: 3-fold degenerate introduced in Ref. [53, 54]: state with total momentum F = 1 and 5-fold degenerate state with F = 2 (Fig. 13 (b)) [53, 54, 78]. The ! ± 2 1 ∆ − 2η exchange energy splitting ∆exch = E1 − E2 = 4η > 0 in- |C | = 1 ± . (32) p 2 2 creases in small nanocrystals as η ∝ 1/a3. On the other 2 (∆ − 2η) + 12η hand, one has to consider the total anisotropic splitting ∆ = ∆ + ∆ of the hole states caused by the crystal Again, as ∆ and η scales with a differently, the values of cr sh |C+|2 and |C−|2 = 1 − |C+|2 are varying with nanocrys- field in wurtzite semiconductors and shape anisotropy + 2 as discussed in Sect. IIIB2. As it was already men- tal size from |C | = 1 for large nanocrystals with ∆ > 0 similar to the QW case (or |C+|2 = 0 for ∆ < 0 similar tioned, for wz-CdSe, ”quasi-spherical” nanocrystals with + 2 isotropic excitons (∆ = 0) shown in Fig. 13 (b) can be re- to the nanowire case) to |C | = 1/4 in small nanocrystals. In the latter case and for spherical nanocrystals alized for the prolate shape depending on the NC radius sph and the anisotropy parameter µ [17]. with gh,3/2 = gh,1/2 = gh , the effective g-factors can be obtained from Eq. (31) as [54]: In the general case, the exciton fine structure in zero magnetic field is caused by the joint effect of the exchange sph sph L ge − 3gh U ge + 5gh interaction and the uniaxial anisotropy. The energy split- gex,1 ≈ , gex,1 ≈ − . (33) L 2 2 tings, including ∆EAF between dark ±2 and ±1 bright excitons depended on both, ∆ and η energy parameters. The energy level splitting for this limit are shown in Fig. As ∆ and η scales with a differently, the splitting be- 13(b) for zb-CdSe (isotropic splitting) and Fig. 13(c) for 16

( a ) 5 . 0 ∆ = 2 3 m e V 2 . 0 4 . 5 1 . 5 1 . 0 4 . 0 0 . 5 s r s 3 . 5 r o t

o 0 t c

c 3 . 0 a a f - 0 . 5 f - - g g 2 . 5 - 1 . 0 2 . 0 - 1 . 5 1 . 5 - 2 . 0 - 2 . 5 1 . 0 - 3 . 0 ( b ) 2 . 0 ∆ = 0 m e V 1 2 3 4 5 6 N a n o p l a t e l e t t h i c k n e s s ( n m ) 1 . 8 FIG. 15. Size dependence of g-factors for the dark exciton

s 1 . 6 r (magenta line) and the lower bright exciton (red lines) in zb- o t

c CdSe NPLs. Calculations are made for zb-CdSe (see TableI a 1 . 4 f

- for β = 0.22) with magnetic ﬁeld directed perpendicular to g the platelet plane. Green lines show the calculated electron 1 . 2 g-factor ge. Dash lines correspond to the parabolic potential G u p t a 2 0 0 2 ( g = −0.1 in the 2D limit) and solid lines to the box 1 . 0 H u 2 0 1 9 h,3/2 Z h a n g 2 0 1 4 potential ( gh = 0.36 in the 2D limit) for the in-plane size L × L = 30 × 30 nm2. Symbols show the experimental g 0 . 8 x y e 1 2 3 4 5 6 data from Ref. [45], the same as in Fig.4. N a n o c r y s t a l r a d i u s ( n m )

FIG. 14. (a) Size dependence of g-factors for dark exciton on the field direction with respect to the crystallographic (magenta line) and upper and lower bright excitons (blue and axes. This effect is rather weak and we neglect it. If one red lines) in spherical wz-CdSe NCs. Solid lines are calcu- considers nanocrystals with cubic shape a similar term in sph lated with gh = −1.1 (see TableI for β = 0.28), and dashed exchange interaction will appear (as it happens for hole sph lines with gh = −0.73 from Ref. [17]. Symbols show the g-factors) and in that case, the effect might be strong. experimental g2 values from Refs. [17, 43, 44], the same as L U We will discuss this problem in a separate paper. in Fig.4(a). (b) Size dependencies of gex,1 and gex,1 with sph All above consideration of the exciton g-factor in NCs gh = −0.73 calculated according to Eq. (33) in quasi- spherical (∆ = 0) wz-CdSe NCs (solid lines) and zb-CdSe is valid if the magnetic field is directed along the crystal NCs (dash lines). axis c for wz-CdSe. If it is not the case, effective exciton g-factors in Eqs. (29), (30), (33) will obtain the factor cos Θ as it is shown in Fig. 13. In real ensemble of ran- domly oriented nanocrystals the observable lines would wz-CdSe (anisotropic splitting depending on the angle be broadened. In the following Fig. 15 (a) it is assumed, Θ). Note, that for spheroidal nanocrystals with the shape that magnetic field is parallel to NC c-axis (cos Θ = 1) anisotropy contribution ∆, one has to use Eq. (31) with and perpendicular to NPL plane in Fig. 15 (b). g 6= g . h,3/2 h,1/2 The dependence of exciton and electron g-factors on The states with F = 0 do not split in the magnetic nanocrystal diameter for spherical nanocrystals without L field. However, strong fields B k c may mix the 0 with shell and based on wurtzite CdSe are shown in Fig. 15 0U states as well as ±1L states with ±1U [116], while the (a). Hole g-factor is gh = −1.12, see TableI for β = 0.28, perpendicular component of the magnetic field ∝ B sin Θ (solid lines) and gh = −0.73 [17] (dashed lines), crystal mixes the state with F differing by ±1 [53]. Such a mix- field splitting is ∆ = 23 meV and exchange interaction ing should be taken into account in the analysis of exci- 3 parameter η(a/aB) = 0.1 meV corresponds to the ac- ton energy level structure in strong magnetic fields with count of short-range interaction only [118]. One can see µBB comparable or larger than the zero-field fine struc- the significant difference between the exciton and single ture splittings (see, for example, [112].) electron g-factors. So measuring g-factors in experiment If one takes into account the cubic symmetry of the can in the ideal situation help one to distinguish the state crystal lattice additional term in exchange interaction ∝ of NC being observed. Symbols show the experimentally 3 3 3 (σe,xJh,x +σe,yJh,y +σe,zJh,z) appears [117]. It splits the measured g2 values from Refs. [17, 43, 44], the same as exciton state with F = 2 into 2-fold degenerate in zero in Fig.4(a). As it was discussed in Sect.II, these g2 field state and the 3-degenerate state. The effect of the values can not be ascribed to ground state of the res- external field in that case would be strongly dependent ident electron neither in wz-CdSe nor in zb-CdSe. As 17 it is shown in Fig. 15 (a), these values indeed could be in nanostructure as compared with the bulk value comes L possibly ascribed to g1,ex exciton in the ”quasi-spherical” from the orbital effect. It is more substantial in semicon- wz-CdSe and spherical zb-CdSe with ∆ = 0 (see dash- ductors with small β and a large value of γ1 Luttinger dotted lines) as it was suggested in Ref. [17]. However, parameter. Our original results for cube NCs demon- this good fit does not prove the origin of these g2 val- strate the consequence of breaking the rotational sym- ues as they can be also fit for the electron excited states metry and call for further experimental studies of cube or assigned to surface-localized electron states as it was zb-CdSe which can be probably synthesized by the cation suggested in Ref. [43]. Further experimental studies of exchange from the cube PbSe NCs. We have predicted the bright exciton g-factor size-dependence in spherical the non-equidistant Zeeman splitting of the hole ground NCs by different experimental techniques, for example by state which is four-fold degenerate in the isotropic cube the SFRS, are needed to clarify this situation. The avail- NCs in zero magnetic field. We have also considered the able data for the dark exciton g-factor in CdSe NCs were effect of the NC shape on the heavy and light hole g- obtained from the polarized photoluminesence in mag- factors and predicted their difference in the prolate and netic field [115, 119, 120] or from the Zeeman splitting oblate NCs even if the hole ground state degenerate in in single NC [19] and correspond to the calculations with zero field, for example in ”quasi-spherical” wz-CdSe or sph ”isotropic” InP NCs. |gh | ≤ 0.73 (dashed line in Fig. 15 (a)) rather than with gh = −1.1 (solid line). Finally, we have investigated the size-dependences of In Fig. 15 (b) we show the size dependence of g-factors dark and bright exciton g-factors in spherical NCs and for the dark exciton (magenta line) and the lower bright planar NPLs and discussed the attribution of the exper- L hh exciton gex,1 = gex,1 (Eq. (29), red lines) in CdSe NPLs. imentally observed multiply g-values in spherical CdSe Calculations are made for zb-CdSe (see TableI for β = NCs. We have shown that the experimental data for the 0.22). Solid lines correspond to the parabolic potential electron, hole, and exciton g-factors in thin CdSe NPLs (gh,3/2 = −0.1 in the 2D limit) and dashed lines to the can be well described by assuming that electron and hole box potential ( gh = 0.36 in the 2D limit) for the in- are confined in a smooth parabolic potential. 2 plane size Lx × Ly = 30 × 30 nm . The respective size dependencies of gh,3/2 can be found in Fig. 12. Green lines show the electron g-factor. The comparison with ACKNOWLEDGMENTS the experimental data from Ref. [55], not only gh,3/2, but also the dark exciton g-factor, g calculated for the ex,2 We thank D.R. Yakovlev, M.M. Glazov, P. Sercel and parabolic confining potential correspond to the observed Al.L. Efros for valuable discussions. This work was values g ≈ −0.1 and g ≈ 2 for 4 ml NPL. The h,3/2 ex,2 funded by the Russian Science Foundation (Grant No. main difference for the exciton g-factors calculated for 20-42-01008). two confining potentials comes from the hole g-factor, while the difference for the electron g-factor is not so significant. However, one can see in Fig. 15 (b) that the electron g-factor size dependence calculated for the Appendix A: Size dependence of the electron energy parabolic potential perfectly describes the experimental level and effective g-factor in the eight-band Kane data for the resident electron g-factor measured by the model SFRS in Ref. [45]. In the eight-band k · p model, the normalization condition for the total electron wave function Ψe = c v c v V. CONCLUSION Ψe + Ψe , where Ψe(r) and Ψe (r) describe the conduction and valence band contributions, respectively, reads R |Ψ (r)|2d3r = R (|Ψc(r)|2 +|Ψv(r)|2)d3r = 1. Using the To summarize, we have revisited the problem of def- e e e expression of Ψv(r) via Ψc(r)[40, 41] one can rewrite the inition and calculation of electron, hole, and exciton g- e e normalization condition as factors in II-VI and III-V-based nanocrystals. We have calculated the electron g-factor within 8-band k·p-model Z Z 2 3 c 2 −1 3 in spherical and planar bare NCs and demonstrated a |Ψe(r)| d r = |Ψe(r)| A (Ee)d r = 1 , good agreement with the results of the tight-binding cal- −1 culations as well as with the experimental data. ∂me(Ee) Ee A(Ee) = 1 + . (A1) We have presented the semi-analytical methods for cal- ∂Ee me(Ee) culation of hole g-factor in NCs of different shapes and symmetry: spherical, axial (planar, spheroid and wire), Here me(Ee) is the electron effective mass at the energy and cube in a full range of heavy and light hole effective Ee calculated from bottom of the conduction band: mass ratio β. The results allow one to calculate the hole g-factor for all studied structures just knowing the set " !#−1 Ep 2 1 of Luttinger parameters. We have shown that the main me(Ee) = m0 γrb+ + , (A2) 3 ˜ ˜ contribution to the renormalization of the hole g-factor E Eg + ∆SO 18

TABLE A1. Parameters for calculation of electron g-factor according to Eq. (5). Eg is the energy gap, Ep is the interband coupling matrix element in the Kane model, ∆SO is the spin orbit energy, ge and me/m0 are the electron g-factor and eﬀective mass in bulk semiconductor. For a comparison with tight-binding results for size dependence of electron g-factor in wz-CdSe, CdTe, GaAs, InP from Ref. [27] we use set of parameters from Ref.[27], except electron eﬀective masses. The second set of parameters for wz-CdSe is taken from Refs. [33 and 42]. For zb-CdSe we use set of parameters from Ref. [32].

Semiconductor Eg, eV Ep, eV ∆SO, eV ge ge (exp.) me/m0 wz-CdSe 1.8174 21.40 0.3871 0.633 0.68[33] 0.13 [68] wz-CdSe 1.84 17.5 0.42 0.68 [33] 0.11 zb-CdSe 1.764 18.3 0.47 0.42 0.42 [32] 0.13 CdTe 1.611 19.57 0.8221 -1.236 -1.66 [34] 0.09 [68] GaAs 1.519 25.34 0.3399 -0.065 -0.44 [31] 0.067 [121] InP 1.424 20.45 0.108 1.22 1.2[34] 0.08 [121]

where E˜ = Eg +Ee and γrb takes into account the contri- N a n o p l a t e l e t t h i c k n e s s ( n m ) bution of remote bands and me ≡ me(Ee = 0) is the elec- 1 2 3 4 5 6 1 . 8 0 . 9 tron eﬀective mass at the bottom of the conduction band. ) z b - C d S e

V I n P Note, that in the case of semiconductor heterostructure, e

( 1 . 5 0 . 8 the electron eﬀective mass in (A1) as well as the electron e E

y effective g-factorg ˜e(Ee) in (3) can be different in differ- 1 . 2 ) e g 0 . 7 r ent materials. In bare NCs, A does not depend on the E ( e

n 0 . 9

coordinate and can be directly used as the renormaliza- e

tion constant for the conduction band contribution. It n 0 . 6 o

r 0 . 6 can be also written explicitly as t c e l 0 . 3 0 . 5

−1 E A(Ee) = [1 + αp(Ee)Eeme(E)/m0] , (A3) ! 0 0 . 4 Ep 2 1 1 2 3 4 5 6 αp(Ee) = + . 2 2 N a n o c r y s t a l r a d i u s ( n m ) 3 E˜ (E˜g + ∆SO)

At small energies Ee Eg, one can approximate FIG. A1. Size dependencies of the electron energy Ee (solid m−1(E) ≈ m−1 − m−1α | E and A(E ) ≈ lines, left axis) and renormalization constant A(Ee) (dashed e e 0 p Ee=0 e e lines, right axis) in zb-CdSe and InP nanocrystals. Open m /m (E)[41]. e e circles show the results of the tight-binding calculations for As it was discussed in Ref. [24], the surface contribu- zb-CdSe NPLs from [122]. tion gsur to the electron g-factor can be nonzero even in the case of the bare semiconductor nanostructure with the infinite potential barrier at the surface. For simplicity, we consider below the case of vanishing electron conduction band wave function component at the surface ge(a) Eqs. (5) and (A4). Fig. A1 shows the size depen- c Ψe(r)|r=s = 0 corresponding to gsur = 0. In this case, dencies of the electron energy Ee (solid lines, left axis) equation (5) gives the universal dependence of the elec- and renormalization constant A(Ee) (dashed lines, right tron effective g-factor on the electron energy Ee (not the axis) in zb-CdSe and InP nanocrystals. The electron en- optical transition energy) within eight-band k · p model ergies calculated according to Eq. (A4) are in very good valid for any nanostructure potential shape. The size agreement with the results of the tight-binding calcula- dependence of the electron g-factor can be found by es- tions for zb-CdSe NPLs from [122]. One can see that tablishing the correspondence between the electron quan- in small NCs the renormalization A(Ee) can be as small tization energy Ee and the size of the nanostructure of a as 0.5 resulting in additional corrections to the electron particular shape. g-factor, up to 10% as compared with Eq. (4). It is also For spherical semiconductor nanocrystals of the radius possible to calculate the electron g-factor for the excited 2 a, the equation for the ground state, 1Se, electron quan- nSe states in spherical NCs using Eq. (5) and factor n tization energy under assumption of zero boundary con- in Eq. (A4). The size dependencies of the electron en- c dition Ψe(a) = 0 reads: ergy level, effective mass and effective g-factor in a QW or NPL with the thickness L can be obtained after the ~2π2 replacement of nanocrystal radius a by the well width L Ee = 2 . (A4) 2me(Ee)a in Eq. (A4).

It can be solved numerically for Ee at the a given a to We consider also spherical NCs and 2D nanoplatelets 2 obtain ge(Ee(a)). Alternatively, one can use the elec- with parabolic confining potential Vext(r) = κr /2 and 2 tron energy Ee to obtain the parameterized dependence Vext(r) = κz /2, respectively. For such potentials, the 19 electron energy level can be found form: 3 X − iY |Γ8, − i =↓ √ . 2 2 2 ~ n Each of functions B1 have momentum 3/2 and its pro- Ee = . (A5) p 2 2 meme(Ee)Le jection on z-axis +3/2, +1/2, −1/2, and −3/2. Luttinger Hamiltonian of Eq. (7) can be written in a where n = 3 for NC and n = 1 for NPL is the number in matrix form in the basis (B1) as follows [72]: 1/2 1/4 of spatial directions and Le = ~ /(κme) is the potential oscillator length with electron mass m taken  FHI 0  e 2 ∗ at the bottom of conduction band (E = 0). In order to ˆ ~ H G 0 I e HL =  ∗  , calculate the size dependence of the electron g-factor in 2m0  I 0 G −H  ∗ ∗ such NCs one needs to define the NC radius a or NPL 0 I −H F 2 2 2 thickness L, in other words, relate it to L0. For example, F = (γ1 − 2γ)kz + (γ1 + γ)(kx + ky), the choice L = πLe allows one to have in thick NPLs 2 2 2 G = (γ1 + 2γ)kz + (γ1 − γ)(kx + ky), (where Ee Eg and the energy dependence of the √ √ 2 electron effective mass in Eq. (A5) can be neglected) the H = −2 3γkz(kx − iky),I = − 3γ(kx − iky) . (B2) electron energy Ee, and consequently ge(Ee), the same as in the box-like potential. In thin NPLs, however, the Anisotropic correction to Luttinger Hamiltonian has electron energy level and respective electron g-factor the form [75, 77]: in the parabolic potential becomes larger than for the 2 " box-like potential due to the weaker effect of the electron ˆ an 2µ ~ 5 ˆ2 ˆ2 HL = − (γ1 + γ)(k − 3kz ) − (B3) effective mass energy dependence. The size dependence 3 2m0 2 of the electron g factor in CdSe NPLs is plotted in Fig. # 15(b) for L = 4Le corresponding to 99.5% of the electron 2γ[(kJˆ )2 − 3{(kJˆ )kˆ J }] . density inside the NPL with boundaries at z = ±L/2. z z

Note, that here we use the opposite sign of µ as compared with Refs. [75 and 77]. In matrix form it can be written as:

Appendix B: Magnetic field and anisotropic  F an Han Ian 0  2 an∗ an an perturbation to the hole Hamiltonian ˆ an 2µ ~ H G 0 I HL = −  an∗ an an  , 3 2m0  I 0 G −H  0 Ian∗ −Han∗ F an The orbital Bloch functions of the top of valence band an 2 2 2 in studied semiconductors are of p-like symmetry and F = −2(γ1 − 2γ)kz + (γ1 + γ)(kx + ky), (B4) are often designated as X, Y , and Z, each having two an 2 2 2 G = −2(γ1 + 2γ)kz + (γ1 − γ)(kx + ky), possible spins, ↑ and ↓ [72]. Taking into account spin- √ √ an an 2 orbit interaction leads to the splitting of the valence band H = 3γkz(kx − iky),I = − 3γ(kx − iky) . with the following set of functions being the basis of the In the presence of external magnetic field, the hole topmost Γ8 subband [72]: eA wave vector k has to be renormalized as k → k − c , where A is the vector potential of external magnetic~ 3 X + iY field. The case B k z in the Landau gauge correspond to |Γ8, + i = − ↑ √ (B1) 2 2 A = (0, Bx, 0). The B-linear correction to the Luttinger Hamiltonian (7) is [50]:

HˆB = µBB {− (2γ1 + 5) xky+ (B5) r 1 2 X + iY 2 i |Γ8, + i = ↑ Z− ↓ √ 2γ 2Jy xky + {JyJz}xkz + {JxJy} xkx − . 2 3 6 2 Here we neglect ∝ B2 corrections leading to diamagnetic shift as we consider a low ﬁeld regime. It can also be 1 r2 X − iY rewritten in the matrix form as B-linear corrections to |Γ8, − i = ↓ Z+ ↑ √ the matrix Luttinger Hamiltonian (B2): 2 3 6 20

 √ √ √  −2xk√y(γ1 + γ) −2i 3γxkz −2i 3γx(kx − iky) + 3γ √ 0 √ ˆ  2i 3γxkz −2xky(γ1 − γ) 0 −2i 3γx(kx − iky) + 3γ  HB = µB B  √ √ √  .  2i 3γx(kx + iky) + 3γ √ 0 √ −2xk√y(γ1 − γ) 2i 3γxkz  0 2i 3γx(kx + iky) + 3γ −2i 3γxkz −2xky(γ1 + γ) (B6)

Now we obtain Eq. (26) from the main text as a ﬁrst Jz = +3/2 and Jz = +1/2: order correction from term Hˆ in the hole Hamiltonian B √ (12). The orbital contribution from the magnetic ﬁeld (1) X E 3 , 1 = −2i 3µBBγ hΨ0|kz|ΨnihΨn|x|Ψ0i = ˆ 2 2 HB is more convenient to consider using its matrix form n in Eq. (B6). We are interested in energy splitting of 2 2 6~ X |hΨ0|γkz|Ψni| heavy hole states (J = ±3/2) in thin quantum well-like = µBB , (B11) z m E − E structures, where the spitting of heavy and light holes is 0 n n 0 large, so far, their mixing is negligible and hole ground √ 3 ˆ state can be considered as a pure heavy hole. Now we where px = − ~ γkz from (B9) for the H 3 1 compo- m0 L, 2 , 2 calculate the contributions to the energy of hole with nent is used. Jz = +3/2 coming from (B6). Firstly comes the diagonal Similarly, term, −2µBB(γ1 + γ)xky: 2 2 (1) 6~ X |hΨ0|γkz|Ψni| E− 3 ,− 1 = −µBB . (B12) (1) X 2 2 m0 En − E0 E 3 3 = −2µBB(γ1 + γ) hΨ0|ky|ΨnihΨn|x|Ψ0i, (B7) n 2 , 2 n (1) In the same way we obtain that corrections E 3 1 = 2 ,− 2 where Ψ0 is the hole ground state wave function, n de- (1) E 3 1 do not contribute to hole g-factors. Finally, the notes all possible intermediate hole states described by − 2 , 2 wave functions Ψn. We next use relations [123] orbital contribution to heavy hole Zeeman splitting is

∆E = E(1) − E(1) = ~ hΨ0|px|Ψni − 3 ,− 1 3 , 1 hΨ0|x|Ψni = i , 2 2 2 2 m E − E 2 2 0 n 0 12 X |hΨ0|γkz|Ψni| = −µ B ~ (B13) ~ hΨ0|py|Ψni B m E − E hΨ0|y|Ψni = −i , (B8) 0 n n 0 m0 En − E0 and the heavy hole g-factor g2D is given by Eq. (26) with E being ground state energy and E being inter- h,3/2 0 n in the main text. As the only contribution to Eq. (B) mediate state energy in order to express the coordinate is coming from heavy and light hole mixing, n in Eq. matrix elements throw the momentum operator p matrix (26) denotes even light hole states, which have non-zero elements and the relation matrix elements hΨ0|γkz|Ψni with the ground state of the heavy hole. m ∂Hˆ p = 0 L (B9) As the in-plane envelope wave functions are the com- ∂k ~ plete basis and in Eq. (B) only operator kz is present, the summation over in-plane wave functions gives unity. in order to relate p with the wave vector k. Note, Therefore, in Eq. (B) as well as in Eq. (26) of the main that Luttinger Hamiltonian HˆL is acting in space of text, Ψ0 denotes the envelope wave function of heavy 4-component wave functions and each case one has to hole ground state and Ψ are envelope wave functions of ˆ n take corresponding component of HL. Eq. (B8) and excited light hole states quantization along z axis. px = (γ1 + γ)kx from Eq. (B9) for the Hˆ 3 3 compo- ~ L, 2 , 2 The function G(β) describing the orbital contribution nent allow us to rewrite (B7) as to the heavy-hole g-factor for NPLs with box-like and parabolic potentials can be expressed analytically 2 (1) 2 2i~ X hΨ0|ky|ΨnihΨn|kx|Ψ0i ∞ 2 2 E 3 , 3 = −µBB(γ1 + γ) . 192(β − 1) n 2 2 m E − E X 0 n n 0 G(β) = − 2 2 2 2 (B10) n=1 π (β + 1) (4n − 1) (β − 4n ) The corresponding correction for hole with Jz = −3/2 is the same as (B10), so the diagonal matrix elements of for box-like potential and ˆ HB do not contribute to heavy hole g-factor. √ √ √ ∞ 4 2n 1−2n 1 X 24 β β − 1 β + 1 Γ n + Now we calculate√ the contribution from the matrix G(β) = − √ √ 2 π(β + 1) β − 4n + 1 Γ(n) element −2i 3µBBγxkz, which mixes hole states with n=1 21

for parabolic potential with Γ(x) being the Euler gamma function. In the limit β → 0 the expression (B14) can be 7 ( a ) expanded into a series and summarized analytically: E 0 6 E 2 3 4 L 1

G(β) ≈ 0.568 − 1.568β + 1.897β − 1.82β + 1.839β . E

, 5 y g r

The corresponding dependence is shown by thin dashed e

n 4 e

line in Fig. 10 of the main text. It can be seen, that e l asymptotic (B14) demonstrates a good applicability for o 3 β . 0.5, corresponding to materials listen in the Table 1. h For parabolic potential, the expression for G(β) behaves 2 1/4 as β with β → 0, although this asymptotic works only 1 . 0

n ( b ) f 3 / 2 ( E 0 ) at non-physically small β, see the inset in Fig. 10 of the o

i f ( E )

t 1 / 2 0 main text. For realistic values of β the empirical asymp- u f ( E ) 0 . 8 3 / 2 1 b

i f ( E )

r 1 / 2 1

totic t

n 0 . 6 o c

−0.107 2

G(β) ≈ 1.1524β + 9.1β − 5.714β s e

l 0 . 4 o

−4 h can be written. It works for 10 β 0.3 and shown . . 2 / 0 . 2 in Fig. 10 of the main text by thin dash-dotted line. 3 = | z

The similar expression as Eq. (B) gives the orbital J | 0 . 0 corrections to the Zeeman splitting of the hole at the 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 at the lowest light-hole quantum size level, E0 = Elh1 , β coming from the admixture of the even excited states of the heavy-hole, Ehh2n [59]. In this case, Ψ0 in Eq. denotes the envelope wave function of light hole ground FIG. C1. (a) The dependence of first two hole states energies E0 and E1 in cube nanulcrystals on the effective mass ratio state and Ψn are envelope wave functions of excited even β in zero magnetic field. Energies are given in units EL = heavy hole states quantization along z axis. The resulting 2 2 2 ~ π /mhhL , L being nanocrystal size (cube width); (b) The expression for the light-hole g-factor reads heavy hole contributions as function of β for the hole states originating from the ground and first excited states. 2 2 12 X |hhh2n|γkz|lh1i| g2D = 2 − µ B ~ . (B14) h,1/2 κ B m E − E 0 n lh1 hh2n the hole can not be simplified, compare with Ref. [49]. We numerically diagonalize hole Hamiltonian matrix cal- Appendix C: Electron and hole energy levels in the culated on the 4-component basis of eigenfunctions of in- nanocrystals with cubic shape finite rectangular quantum well along coordinate axes:

nx,ny ,nz nx ny nz Ψ (x, y, z) = φ (Jz)φ (y)φ (z), (C2) Here we consider nanocrystals with the cubic shape Jz Jz Jz Jz r and edges directed along cubic axes of the crystal. The 2 πn L φnα (α) = sin α + , potential of such cube nanocrystal is described by a rect- Jz L L 2 angular potential with infinite barrier: where Jz = ±3/2, ±1/2, nx, ny, nz = 1...N, and α = L x, y, z. All matrix elements are calculated analytically cube 0, |x|, |y|, |z| ≤ 2 V (x, y, z) = L , (C1) and basis size is N = 16, which is more than enough to ∞, |x|, |y|, |z| > 2 obtain convergence for several lowest hole states at any where x, y, z are electron or hole coordinates. The con- reasonable value of light to heavy effective mass ratio β. duction band electron√ energy levels can be found from The dependencies of the energies of the two lowest hole (A4) with a → L/ 3. With this substitution, the size states on β are shown in Fig. C1(a). The interesting point dependence of the energy levels, effective mass and effec- is the crossing of the levels E0 and E1 leading to level E1 tive g-factors can be seen in Figs. A1 and4. being the ground state in some range of β. This leads to For holes situation is more complicated as the top of va- the significant change of the hole ground state structure lence band is 4-fold degenerate and hole wave function is in the vicinity of the crossing point. On fig. C1 (b) we a four component one. In order to calculate hole states in show the heavy hole (|Jz| = 3/2) contributions to hole cubic nanocrystal we developed numerical method. The wave function, f|M|, as function of β for states, corre- hole kinetic energy is described by Luttinger Hamilto- sponding to energy levels E0 and E1. For rather large nian, Eq. (7) and nanocrystal potential is taken from β and far from crossing the ground state is S-like state Eq. (C1). Potential (C1) mixes hole states with differ- E0 and 3/2 hole states are mostly formed from heavy ent spin projections Jz and, unlike spherically symmetric holes and 1/2 state are mostly light holes. On the op- case, due to lower symmetry the Schrödingerequation of posite, for smaller β, in crossing range or near it levels 22

E0 and E1 are close to each other and, consequently, are and light holes to the ground state sublevels are compa- strongly mixed. As a result, the contributions of heavy rable leading to the speciﬁc features in dependencies of hole g-factors on β, see Fig.8 of the main text.

[1] Celso de Mello Donegá,“Synthesis and properties of Rodina, “Spin dynamics in semiconductor nanocrys- colloidal heteronanocrystals,” Chem. Soc. Rev. 40, 1512 tals,” Phys. Rev. B 66, 125307 (2002). (2011). [18] H. Htoon, S.A. Crooker, M. Furis, S. Jeong, Al. L. Efros, [2] S. Ithurria and B. Dubertret, “Quasi 2d colloidal cdse and V.I. Klimov, “Anomalous circular polarization of platelets with thicknesses controlled at the atomic photoluminescence spectra of individual CdSe nanocrys- level,” J. Am. Chem. Soc. 130, 16504 (2008). tals in an applied magnetic field,” Phys. Rev. Lett. 102, [3] M. V. Kovalenko, L. Manna, A. Cabot, Z. Hens, D. V. 017402 (2009). Talapin, C. R. Kagan, V. I. Klimov, A. L. Rogach, P. [19] L. Biadala, Y. Louyer, Ph. Tamarat, and B. Lounis, Reiss, D. J. Milliron, Ph. Guyot-Sionnnest, G. Kon- “Band-edge exciton fine structure of single CdSe/ZnS stantatos, W. J. Parak, T. Hyeon, B. A. Korgel, C. B. nanocrystals in external magnetic fields,” Phys. Rev. Murray, and W. Heiss, “Prospects of nanoscience with Lett. 105, 157402 (2010). nanocrystals,” ACS Nano 9, 1012 (2015). [20] M. J. Fernée,C. Sinito, Y. Louyer, C. Potzner, T. L. [4] D. Graham-Rowe, “From dots to devices,” Nature Pho- Nguyen, P. Mulvaney, P. Tamarat, and B. Lounis, tonics 3, 307 (2009). “Magneto-optical properties of trions in non-blinking [5] S. E. Lohse and C. J. Murphy, “Applications of colloidal charged nanocrystals reveal an acoustic phonon bottle- inorganic nanoparticles: From medicine to energy,” J. neck,” Nature Commun. 3, 1287 (2012). Am. Chem. Soc. 134, 15607 (2012). [21] F. Liu, L. Biadala, A. V. Rodina, D. R. Yakovlev, [6] R. Freeman and I. Willner, “Optical molecular sensing D. Dunker, C. Javaux, J. P. Hermier, Al. L. Efros, with semiconductor quantum dots (QDs),” Chem. Soc. B. Dubertret, and M. Bayer, “Spin dynamics of nega- Rev. 41, 4067 (2012). tively charged excitons in CdSe/CdS colloidal nanocrys- [7] P. V. Kamat, “Boosting the efficiency of quantum dot tals,” Phys. Rev. B 88, 035302 (2013). sensitized solar cells through modulation of interfacial [22] E. L. Ivchenko and A. A. Kiselev, Sov. Phys. Semicond. charge transfer,” Acc. Chem. Res. 45, 1906 (2012). 26, 827 (1992). [8] G. H. Carey, A. L. Abdelhady, Z. Ning, S. M. Thon, [23] A. A. Kiselev, E. L. Ivchenko, and U. Rössler,“Elec- O. M. Bakr, and E. H. Sargent, “Colloidal quantum tron g factor in one- and zero-dimensional semiconduc- dot solar cells,” Chem. Rev. 115, 12732 (2015). tor nanostructures,” Phys. Rev. B 58, 16353 (1998). [9] E. Lhuillier, M. Scarafagio, P. Hease, B. Nadal, H. [24] A. V. Rodina, Al. L. Efros, and A. Yu. Alekseev, “Effect Aubin, X. Z. Xu, N. Lequeux, G. Patriarche, S. Ithur- of the surface on the electron quantum size levels and ria, and B. Dubertret, “Infrared photodetection based electron g factor in spherical semiconductor nanocrys- on colloidal quantum-dot films with high mobility and tals,” Phys. Rev. B 67, 155312 (2003). optical absorption up to THz,” Nano Lett. 16, 1282. [25] I. A. Yugova, A. Greilich, D. R. Yakovlev, A. A. Kiselev, (2016). M. Bayer, V. V. Petrov, Yu. K. Dolgikh, D. Reuter, [10] R. Wang, Y. Shang, P. Kanjanaboos, W. Zhou, Z. Ning, and A. D. Wieck, “Universal behavior of the electron g- and E. H. Sargent, “Colloidal quantum dot ligand engi- factor in GaAsAlxGa1−xAs quantum wells,” Phys. Rev. neering for high performance solar cells,” Energy Envi- B 75, 245302 (2007). ron. Sci. 9, 1130 (2016). [26] Joshua Schrier and K. Birgitta Whaley, “Tight- [11] D.Loss and D.P. DiVincenzo, “Quantum computation bindingg-factor calculations of CdSe nanostructures,” with quantum dots,” Phys. Rev. A 57, 120 (1998). Phys. Rev. B 67 (2003). [12] A. Imamoglu, D. D. Awschalom, G. Burkard, D. P. Di- [27] A. Tadjine, Y.-M. Niquet, and C. Delerue, “Universal Vincenzo, D. Loss, M. Sherwin, and A. Small, “Quan- behavior of electron g-factors in semiconductor nanos- tum information processing using quantum dot spins tructures,” Phys. Rev. B 95, 235437 (2017). and cavity QED,” Phys. Rev. Let. 83, 4204 (1999). [28] P. Chen and K. B. Whaley, “Magneto-optical response [13] S. Nadj-Perge, S. M. Frolov, E. P. A. M. Bakkers, and of CdSe nanostructures,” Physical Review B 70 (2004). L. P. Kouwenhoven, “Spin-orbit qubit in a semiconduc- [29] D. Csontos, P. Brusheim, U. Zülicke, and H. Q. tor nanowire,” Nature 468, 1084 (2010). Xu, “Spin-3/2 physics of semiconductor hole nanowires: [14] Richard J. Warburton, “Single spins in self-assembled Valence-band mixing and tunable interplay between quantum dots,” Nat. Mater. 12, 483 (2013). bulk-material and orbital bound-state spin splittings,” [15] G. Cao, H.-O. Li, G.-D. Yu, B.-Ch. Wang, B.-B. Chen, Physical Review B 79 (2009). X.-X. Song, M. Xiao, G.-C. Guo, H.-W. Jiang, X. Hu, [30] L. M. Roth, B. Lax, and S. Zwerdling, “Theory of and Guo-Ping Guo, “Tunable hybrid qubit in a GaAs optical magneto-absorption effects in semiconductors,” double quantum dot,” Phys. Rev. Lett. 116, 086801 Phys. Rev. 114, 90 (1959). (2016). [31] C. Weisbuch and C. Hermann, “Optical detec- [16] M. Kuno, M. Nirmal, M. G. Bawendi, A. Efros, and tion of conduction-electron spin resonance in GaAs, M. Rosen, “Magnetic circular dichroism study of CdSe Ga1−xInxAs, and Ga1−xAlxAs,” Phys. Rev. B 15, 816 quantum dots,” J. Chem. Phys. 108, 4242 (1998). (1977). [17] J. A. Gupta, D. D. Awschalom, Al. L. Efros, and A. V. [32] O. Z. Karimov, D. Wolverson, J. J. Davies, S. I. 23

Stepanov, T. Ruf, S. V. Ivanov, S. V. Sorokin, C. B. semiconductor nanostructure,” Phys. Rev. Lett. 112 O’Donnell, and K. A. Prior, “Electrong-factor for cu- (2014). bic Zn1−xCdxSe determined by spin-flip Raman scat- [48] J. M. Luttinger, “Quantum theory of cyclotron reso- tering,” Phys. Rev. B 62, 16582 (2000). nance in semiconductors: general theory,” Phys. Rev. [33] W W Piper, “II-VI semiconducting compounds,” in In- 102, 1030 (1956). ternational Conference on II-VI Semiconducting Com- [49] B. L. Gel’mont and M. I. D’yakonov, “Acceptor levels pounds (1967: Brown University), edited by D. G. in diamond-type semiconductors,” Soviet Physics. Semi- Thomas (WA Benjamin, New York, 1967). conductors 5, 2191 (1971). [34] M. Oestreich, S. Hallstein, A. P. Heberle, K. Eberl, [50] M. A. Semina and R. A. Suris, “Holes localized in nanos- E. Bauser, and W. W. Rühle,“Temperature and den- tructures in an external magnetic field: g-factor and sity dependence of the electron landégfactorin semicon- mixing of states,” Semiconductors 49, 797 (2015). ductors,” Phys. Rev. B 53, 7911 (1996). [51] X. Marie, T. Amand, P. Le Jeune, M. Paillard, [35] V.K. Kalevich and V.L. Korenev, “Electron g-factor P. Renucci, L. E. Golub, V. D. Dymnikov, and E. L. anisotropy in asymmetric GaAs/AlGaAs quantum Ivchenko, “Hole spin quantum beats in quantum-well well,” JETP Lett. 56, 253 (1992). structures,” Phys. Rev. B 60, 5811 (1999). [36] A. A. Sirenko, T. Ruf, M. Cardona, D. R. Yakovlev, [52] B. L. Gel’mont and M. I. D’yakonov, “g-factor of ac- W. Ossau, A. Waag, and G. Landwehr, “Electron and ceptors in semiconductors with the diamond structure,” hole g-factors measured by spin-flip Raman scattering in Soviet Physics. Semiconductors 7, 2013 (1973). CdTe/Cd1−xMgxTe single quantum wells,” Phys. Rev. [53] Al. L. Efros, M. Rosen, M. Kuno, M. Nirmal, D.J. Nor- B 56, 2114 (1997). ris, and M. Bawendi, “Band-edge exciton in quantum [37] C. Hermann and C. Weisbuch, “k · p perturbation the- dots of semiconductors with a degenerate valence band: ory in III-V compounds and alloys: a reexamination,” Dark and bright exciton states,” Phys. Rev. B 54, 4843 Phys. Rev. B 15, 823 (1977). (1996). [38] E.L. Ivchenko, A.A. Kiselev, and M. Willander, “Elec- [54] Al. L. Efros, “Fine Structure and Polarization Proper- tronic g factor in biased quantum wells,” Solid State ties of Band-Edge Excitons in Semiconductor Nanocrys- Comm. 102, 375 (1996). tals; Chapter 3,” in Semiconductor and Metal Nanocrys- [39] A. A. Kiselev, E. L. Ivchenko, and U. Rössler,“Elec- tals: Synthesis and Electronic and Optical Properties, tron g factor in one- and zero-dimensional semiconduc- edited by V. I. Klimov and M. Dekker (New York, 2003) tor nanostructures,” Phys. Rev. B 58, 16353 (1998). pp. 103–141. [40] I. A. Merkulov and A. V. Rodina, “Exchange interac- [55] E. V. Shornikova, A. A. Golovatenko, D. R. Yakovlev, tion between carriers and magnetic ions in quantum size A. V. Rodina, L. Biadala, G. Qiang, A. Kuntzmann, heterostructures,” in Introduction to the physics of di- M. Nasilowski, B. Dubertret, A. P., I. Moreels, and luted magnetic semiconductors, edited by J. Kossut and M. Bayer, “Surface spin magnetism controls the polar- J. A. Gaj (Springer, 2010) Chap. 3, pp. 65–101. ized exciton emission from CdSe nanoplatelets,” Nature [41] A. V. Rodina and A. Yu. Alekseev, “Theory of intrinsic Nanotechnology 15, 277 (2020). electric polarization and spin Hall current in spin-orbit- [56] E. Shornikova, D. Yakovlev, L. Biadala, S. Crooker, V. coupled semiconductor heterostructures,” Phys. Rev. Belykh, M. Kochiev, A. Kuntzmann, M. Nasilowski, B. B 78, 115304 (2008). Dubertret, and M. Bayer, “Negatively charged excitons [42] A. I. Ekimov, F. Hache, M. C. Schanne-Klein, D. Ri- in cdse nanoplatelets,” Nano Let. 20, 1370 (2020). card, Ch. Flytzanis, I. A. Kudryavtsev, T. V. Yazeva, [57] E. Shornikova, D. Yakovlev, D. Tolmachev, V. Ivanov, A. V. Rodina, and Al. L. Efros, “Absorption and I. Kalitukha, Victor Sapega, D. Kudlacik, Y. Kusrayev, intensity-dependent photoluminescence measurements A. Golovatenko, S. Shendre, S. Delikanli, H. Demir, and on CdSe quantum dots: assignment of the first elec- M. Bayer, “Magneto-optics of excitons interacting with tronic transitions,” JOSA B 10, 100 (1993). magnetic ions in cdse/cdmns colloidal nanoplatelets,” [43] R. Hu, D. R. Yakovlev, P. Liang, G. Qiang, C. Chen, ACS Nano 14, 9032 (2020). T. Jia, Zh. Sun, M. Bayer, and D. Feng, “Origin of [58] A. A. Kiselev and L.V. Moiseev, Physics of the Solid two larmor frequencies in the coherent spin dynamics State 38, 866 (1996). of colloidal CdSe quantum dots revealed by controlled [59] M. V. Durnev, M. M. Glazov, and E. L. Ivchenko, charging,” J. Phys. Chem. Lett. 10, 3681 (2019). “Giant zeeman splitting of light holes in GaAs/AlGaAs [44] Zh. Zhang, Z. Jin, H. Ma, Y. Xu, X. Lin, G. Ma, and X. quantum wells,” Physica E 44, 797 (2012). Sun, “Room-temperature spin coherence in zinc blende [60] H. W. van Kesteren, E. C. Cosman, W. A. J. A. van der CdSe quantum dots studied by time-resolved faraday Poel, and C. T. Foxon, “Fine structure of excitons in ellipticity,” Physica E 56, 85–89 (2014). type-II GaAs/AlAs quantum wells,” Physical Review B [45] D. Kudlacik, V. F. Sapega, D. R. Yakovlev, I. V. Kali- 41, 5283 (1990). tukha, E. V. Shornikova, A. V. Rodina, E. L. Ivchenko, [61] A. V. Rodina, M. Dietrich, A. Göldner, L. Eckey, G. S. Dimitriev, M. Nasilowski, B. Dubertret, and M. A. Hoffmann, Al. L. Efros, M. Rosen, and B. K. Meyer, Bayer, “Single and double electron spin-flip raman scat- “Free excitons in wurtzite GaN,” Phys. Rev. B 64, tering in CdSe colloidal nanoplatelets,” Nano Lett. 20, 115204 (2001). 517 (2019). [62] D. A. Broido and L. J. Sham, “Effective masses of holes [46] A. V. Rodina and E. L. Ivchenko, “Theory of single and at gaas-algaas heterojunctions,” Phys. Rev. B 31, 888 double electron spin-flip raman scattering in semicon- (1985). ductor nanoplatelets,” arXiv:2010.10385 (2020). [63] D. A. Broido and L. J. Sham, “Erratum: Effective [47] J. van Bree, A. Yu. Silov, P. M. Koenraad, and M. masses of holes at gaas-algaas heterojunctions,” Phys. E. Flatté,“Spin-orbit-induced circulating currents in a Rev. B 31, 6831(E) (1985). 24

[64] K. J. Vahala and P. C. Sercel, “Application of a total- [83] Th. Wimbauer, K. Oettinger, Al. L. Efros, B. K. Meyer, angular-momentum basis to quantum-dot band struc- and H. Brugger, “Zeeman splitting of the excitonic ture,” Phys. Rev. Lett. 65, 239 (1990). recombination in InxGa1−xAs/GaAs single quantum [65] P. C. Sercel and K. J. Vahala, “Analytical formalism wells,” Phys. Rev. B 50, 8889 (1994). for determining quantum-wire and quantum-dot band [84] M. Kubisa, K. Ryczko, and J. Misiewicz, “Spin split- structure in the multiband envelope-function approxi- ting of holes in symmetric GaAs/GaxAl1−xAs quantum mation,” Phys. Rev. B 42, 3690 (1990). wells,” Phys. Rev. B 83, 195324 (2011). [66] L. G. C. Rego, P. Hawrylak, J. A. Brum, and A. Wojs, [85] E. V. Shornikova, L. Biadala, D. R. Yakovlev, D. Feng, “Interacting valence holes in p-type SiGe quantum disks V. F. Sapega, N. Flipo, A. A. Golovatenko, M. A. Sem- in a magnetic field,” Phys. Rev. B 55, 15694 (1997). ina, A. V. Rodina, A. A. Mitioglu, M. V. Ballottin, [67] A. Baldereschi and N. O. Lipari, “Spherical model of P. C. M. Christianen, Y. G. Kusrayev, M. Nasilowski, shallow acceptor states in semiconductors,” Phys. Rev. B. Dubertret, and M. Bayer, “Electron and hole g- B 8, 2697 (1973). factors and spin dynamics of negatively charged excitons [68] Sadao Adachi, Handbook on physical properties of semi- in cdse/cds colloidal nanoplatelets with thick shells,” conductors (Springer US, 2004). Nano Lett. 18, 373 (2018). [69] S. Zh. Karazhanov, “Ab initio studies of the band pa- [86] D.J. Norris, Al. L. Efros, M. Rosen, and M.G. Bawendi, rameters of III–v and II–VI zinc-blende semiconduc- “Size dependence of exciton fine structure in CdSe quan- tors,” Semiconductors 39, 161 (2005). tum dots,” Phys. Rev. B 53, 16347 (1996). [70] P. Horodyská,P. Nˇemec,D. Sprinzl, P. Malý,V. N. [87] H. Fu, L.-W. Wang, and A. Zunger, “Applicability of Gladilin, and J. T. Devreese, “Exciton spin dynamics thek·pmethod to the electronic structure of quantum in spherical CdS quantum dots,” Physical Review B 81 dots,” Physical Review B 57, 9971 (1998). (2010). [88] A. B. Kapustina, B. V. Petrov, A. V. Rodina, and [71] A. R. Edmonds, Angular momentum in Quantum R. P. Seisyan, “Magnetic absorption of hexagonal crys- mechanics (Princenton University Press, Princenton, tals CdSe in strong and weak fields: Quasi-cubic ap- 1957). proximation,” Phys. of Sol. State 42, 1242 (2000). [72] E. L. Ivchenko, Optical Spectroscopy of Semiconductor [89] J. Berezovsky, M. Ouyang, F. Meier, D. D. Awschalom, Nanostructures (Alpha Science, Harrow UK, 2005). D. Battaglia, and X. Peng, “Spin dynamics and level [73] Al. L. Efros, “Luminescence polarization of cdse micro- structure of quantum-dot quantum wells,” Phys. Rev. crystals,” Phys. Rev. B 46, 7448 (1992). B 71 (2005). [74] Al. L. Efros and A. V. Rodina, “Band-edge absorption [90] P. Lawaetz, “Valence-band parameters in cubic semi- and luminescence of nonspherical nanometer-size crys- conductors,” Physical Review B 4, 3460 (1971). tals,” Phys. Rev. B 47, 10005 (1993). [91] R.A. Stradling, “Cyclotron resonance from thermally [75] M. A. Semina, A. A. Golovatenko, and A. V. Rodina, excited holes in ZnTe,” Solid State Comm. 6, 665 “Ground state of the holes localized in II-VI quantum (1968). dots with Gaussian potential profiles,” Phys. Rev. B [92] M. Said and M. A. Kanehisa, “Excited acceptor states 93, 045409 (2016). in II–VI and III–V semiconductors,” phys. stat. sol. (b) [76] A. V. Rodina, “A+-center and exciton bound to neutral 157, 311 (1990). acceptor in diamond-like semiconductors,” Solid State [93] Y. Oka and M. Cardona, “Resonant spin-flip raman Commun. 85, 23 (1993). scattering on donor and acceptor states in ZnTe,” Phys. [77] Al. L. Efros and A. V. Rodina, “Band-edge absorption Rev. B 23, 4129 (1981). and luminescence of nonspherical nanometer-size crys- [94] H.P. Wagner, S. Lankes, K. Wolf, W. Kuhn, P. Link, tals,” Phys. Rev. B 47, 10005 (1993). and W. Gebhardt, “Spectroscopic investigations of [78] A. V. Rodina and Al. L. Efros, “Band-edge biexciton in donor and acceptor states in n- and p-doped ZnTe epi- nanocrystals of semiconductors with a degenerate va- layers,” Journal of Crystal Growth 117, 303 (1992). lence band,” Phys. Rev. B 82, 125324 (2010). [95] T. Friedrich, J. Kraus, M. Meininger, G. Schaack, and [79] A. Baldereschi and Nunzio O. Lipari, “Cubic contribu- W. O. G. Schmitt, “Zeeman levels of the shallow lithium tions to the spherical model of shallow acceptor states,” acceptor and band parameters in cadmium telluride,” J. Phys. Rev. B 9, 1525 (1974). Phys.: Condens. Matter 6, 4307 (1994). [80] A. A. Golovatenko, M. A. Semina, A. V. Rodina, and [96] Ch. Neumann, A. Nöthe, and N. O. Lipari, “Two- T. V. Shubina, “Excitons and biexcitons in spheroidal photon magnetoabsorption of ZnTe, CdTe, and GaAs,” quantum dots A2B6,” Phys. of Sol. State 60, 1510 Physical Review B 37, 922 (1988). (2018). [97] M S Skolnick, A K Jain, R A Stradling, J Leotin, and [81] M. A. Semina and R. A. Suris, “Effect of localization in J C Ousset, “An investigation of the anisotropy of the quantum wells and quantum wires on heavy-light hole valence band of GaAs by cyclotron resonance,” J. Phys. mixing and acceptor binding energy,” Semiconductors C 9, 2809 (1976). 45, 917 (2011). [98] “Semiconductors. group IV elements, IV-IV and III-V [82] E. V. Shornikova, L. Biadala, D. R. Yakovlev, V. F. compounds. lattice properties,” in Landoldt-Börnstein Sapega, Y. G. Kusrayev, A. A. Mitioglu, M. V. Ballot- Tables, Vol. New Series, Group III, 22a, edited by tin, P. C. M. Christianen, V. V. Belykh, M. V. Kochiev, O. Madelung (Springer, Berlin, 1982). N. N. Sibeldin, A. A. Golovatenko, A. V. Rodina, N. A. [99] L. W. Molenkamp, R. Eppenga, G. W. ’t Hooft, P. Daw- Gippius, A. Kuntzmann, Y. Jiang, M. Nasilowski, B. son, C. T. Foxon, and K. J. Moore, “Determination of Dubertret, and M. Bayer, “Addressing the exciton fine valence-band effective-mass anisotropy in gaas quantum structure in colloidal nanocrystals: the case of CdSe wells by optical spectroscopy,” Phys. Rev. B 38, 4314 nanoplatelets,” Nanoscale 10, 646 (2018). (1988). 25

[100] B. V. Shanabrook, O. J. Glembocki, D. A. Broido, Delikanli, and H. V. Demir, “Lateral size-dependent and W. I. Wang, “Luttinger parameters for gaas spontaneous and stimulated emission properties in col- determined from the intersubband transitions in loidal cdse nanoplatelets,” ACS Nano 9, 5041 (2015). GaAs/AlxGa1−xAs multiple quantum wells” Phys. [114] S. Ayari, M. T. Quick, N. Owschimikow, S. Rev. B 39, 3411 (1989). Christodoulou, G. H. V. Bertrand, M. Artemyev, I. [101] N. Binggeli and A. Baldereschi, “Determination of the Moreels, U. Woggon, S. Jaziri, and Alexander W. hole effective masses in GaAs from acceptor spectra,” Achtstein, “Tuning trion binding energy and oscilla- Phys. Rev. B 43, 14734 (1991). tor strength in a laterally finite 2d system: CdSe [102] W. Hackenberg, R. T. Phillips, and H. P. Hughes, “In- nanoplatelets as a model system for trion properties,” vestigation of the luttinger parameters for inp using hot- Nanoscale 12, 14448 (2020). electron luminescence,” Phys. Rev. B 50, 10598 (1994). [115] G. Qiang, A. A. Golovatenko, E. V. Shornikova, D. R. [103] N. O. Lipari and A. Baldereschi, “Angular momentum Yakovlev, A. V. Rodina, E. A. Zhukov, I. V. Kalitukha, theory and localized states in solids. Investigation of V. F. Sapega, V. K. Kaibyshev, M. A. Prosnikov, P. shallow acceptor states in semiconductors,” Phys. Rev. C. M. Christianen, A. A. Onushchenko, and M. Bayer, Lett. 25, 1660 (1970). “Polarized emission of cdse nanocrystals in magnetic [104] R. Kotlyar, T. L. Reinecke, M. Bayer, and A. Forchel, field: the role of phonon-assisted recombination of the “Zeeman spin splittings in semiconductor nanostruc- dark exciton,” arXiv:2011.02786 (2020). tures,” Phys. Rev. B 63, 085310 (2001). [116] A. V. Rodina and Al. L. Efros, “Radiative recombi- [105] S. Kapoor, J. Kumar, and P.K. Sen, “Magneto-optical nation from dark excitons in nanocrystals: Activation analysis of anisotropic CdZnSe quantum dots,” Physica mechanisms and polarization properties,” Phys. Rev. E 42, 2380 (2010). B 93, 155427 (2016). [106] N. J. Traynor, R. T. Harley, and R. J. Warburton, [117] K. Cho, S. Suga, W. Dreybrodt, and F. Willmann, “Zeeman splitting and g factor of heavy-hole excitons “Theory of degenerate 1s excitons in zinc-blende-type in InxGa1−xAs/ GaAs quantum wells,” Phys. Rev. B crystals in a magnetic field: Exchange interaction and 51, 7361 (1995). cubic anisotropy,” Phys. Rev. B 11, 1512 (1975). [107] J. Jadczak, M. Kubisa, K. Ryczko, L. Bryja, and [118] M. Nirmal, D.J. Norris, M. Kuno, M.G. Bawendi, Al. L. M. Potemski, “High magnetic field spin splitting of ex- Efros, and M. Rosen, “Observation of the ”dark excitons in asymmetric GaAs quantum wells,” Phys. Rev. citon” in CdSe quantum dots,” Phys. Rev. Lett. 75, B 86, 245401 (2012). 3728–3731 (1995). [108] P. Pfeffer and W. Zawadzki, “Five-levelk·pmodel for the [119] E. Johnston-Halperin, D. D. Awschalom, S. A. Crooker, conduction and valence bands of GaAs and InP,” Phys. Al. L. Efros, M. Rosen, X. Peng, and A. P. Alivisatos, Rev. B 53, 12813 (1996). “Spin spectroscopy of dark excitons in CdSe quantum [109] Al. L. Efros and M. Rosen, “Quantum size level struc- dots to 60 T,” Phys. Rev. B 63, 205309 (2001). ture of narrow-gap semiconductor nanocrystals: Effect [120] A. Granados del Aguila,´ G. Pettinari, E. Groeneveld, of band coupling,” Phys. Rev. B 58, 7120 (1998). C. de Mello Donegá,D. Vanmaekelbergh, J. C. Maan, [110] M. J. Snelling, E. Blackwood, C. J. McDonagh, R. T. and P. C. M. Christianen, “Optical spectroscopy of dark Harley, and C. T. B. Foxon, “Exciton, heavy-hole, and and bright excitons in CdSe nanocrystals in high mag- electron g factors in type-I GaAs/AlxGa1−xAs quantum netic fields,” The J. Phys. Chem. C 121, 23693 (2017). wells,” Phys. Rev. B 45, 3922 (1992). [121] Sadao Adachi, Properties of Group-IV, III-V and II-VI [111] M. Bayer, V. B. Timofeev, T. Gutbrod, A. Forchel, Semiconductors (John Wiley & Sons, Ltd, 2005). R. Steffen, and J. Oshinowo, “Enhancement of spin [122] R. Benchamekh, N. A. Gippius, J. Even, M. O. Nestok- splitting due to spatial confinement in InxGa1−xAs lon, J.-M. Jancu, S. Ithurria, B. Dubertret, Al. L. Efros, quantum dots,” Phys. Rev. B 52, R11623 (1995). and P. Voisin, “Tight-binding calculations of image- [112] A. Brodu, V. Chandrasekaran, L. Scarpelli, J. Buhot, F. charge effects in colloidal nanoscale platelets of CdSe,” Masia, M. V. Ballottin, M. Severijnen, M. D. Tessier, Phys. Rev. B 89, 035307 (2014). D. Dupont, F. T. Rabouw, P. C. M. Christianen, C. [123] G Wang, L Bouet, M M Glazov, T Amand, E L de Mello Donega, D. Vanmaekelbergh, W. Langbein, Ivchenko, E Palleau, X Marie, and B Urbaszek, and Z. Hens, “Fine structure of nearly isotropic bright “Magneto-optics in transition metal diselenide mono- excitons in InP/ZnSe colloidal quantum dots,” J. Phys. layers,” 2D Materials 2, 034002 (2015). Chem. Lett. 10, 5468 (2019). [113] M. Olutas, B.k Guzelturk, Y. Kelestemur, A. Yeltik, S.