Higher-order topological insulator in cubic semiconductor quantum wells.

The search for exotic new topological states of matter in widely accessible materials, for which the manufacturing process is mastered, is one of the major challenges of the current topological physics. Here we predict higher order topological insulator state in quantum wells based on the most common semiconducting materials. By successively deriving the bulk and boundary Hamiltonians, we theoretically prove the existence of topological corner states due to cubic symmetry in quantum wells with double band inversion. We show that the appearance of corner states does not depend solely on the crystallographic orientation of the meeting edges, but also on the growth orientation of the quantum well. Our theoretical results significantly extend the application potential of topological quantum wells based on IV, II-VI and III-V semiconductors with diamond or zinc-blende structures.

Introduction

Since the discovery of topological insulators (TIs)[1,2], topological phases and materials have been extensively explored in two- (2D) and three-dimensional (3D) systems[3-5]. The recent classification[6-8] of TIs with given crystalline symmetry has led to the discovery of a new type of topological phases, the higher-order topological insulators (HOTIs)[9-15]. The 3D HOTIs are gapped in the bulk and on all surfaces, but they have one-dimensional (1D) gapless modes along “hinges”, where two surfaces meet. These hinge states were experimentally observed in bismuth[16], BiBr[17] and WTe crystals[18,19] and theoretically predicted for strained SnTe[12], transition metal dichalcogenides [20] and antiperovskites[21].

In 2D HOTIs, pioneering works[9,10,13] suggest the presence of zero-dimensional (0D) corner states inside the insulating edge and bulk band-gap of certain materials. Recently, Peterson et  al.[22] have shown that 0D corner states may reside either in the bulk band-gap or fully within the bulk bands of a HOTI, depending on the material’s details. HOTIs that fall into the latter case do not host 0D corner states within their bulk band-gap and, as such, cannot be distinguished from trivial insulators by their spectrum alone. Nevertheless, even in this case the higher-order topology can be still identified via a fractional corner anomaly[22,23].

So far, a large part of the experimental study of 2D HOTIs has been performed in engineered metamaterials[24-34], while only a few candidates have been theoretically predicted in solids, including black phosphorene[35], graphdiyne[36], bismuthene[37] and twisted bilayer graphene at certain angles[38]. Although twisted bilayer graphene can be indeed a realistic candidate to probe 2D HOTI state experimentally, it is still highly desirable to identify controllable and widely accessible higher-order topological materials.

Nowadays, many technologically important semiconductors with the most developed molecular-beam-epitaxy growth hold a cubic crystal structure, including the diamond structure for the group-IV elements, and the zinc-blend structure for the III–V and II–VI compounds. For instance, the time-reversal-invariant 2D TI state, also known as quantum spin Hall insulator (QSHI)—was first discovered in the cubic semiconductor QWs with an inverted band structure: HgTe/CdHgTe QWs[2,39] and broken-gap InAs/GaSb QW bilayers[40,41]. Later, InN/GaN QWs[42], Ge/GaAs QWs[43] and InAsBi/AlSb QWs[44] were also predicted to be 2D TIs. Such a list can be obviously extended by including variety of type-II broken-gap QW heterostructures[45-48] (similar to the InAs/GaSb QWs) on the basis of III–V semiconductors and their alloys.

This work shows that in addition to “conventional” first-order 2D TI state, cubic IV, III–V and II–VI semiconductors are also promising for the implementation of time-reversal-invariant 2D HOTI. Starting from realistic multi-band kp Hamiltonian[49], we directly derive an effective 2D low-energy Hamiltonian preserving the cubic symmetry of the semiconductors. Then, by applying open boundary conditions, we obtain an effective 1D Hamiltonian for the edge states and demonstrate the existence of the corner states in the QWs with double band inversion. As two prototype 2D systems, we consider three-layer InAs/GaInSb QWs[50] and double HgTe/CdHgTe QWs[51] grown along (0mn) crystallographic orientations (where m and n are integers).

Results

Insulating state with double band inversion

Let us first explore the possibility for double band inversion in the prototype QWs (see Fig. 1). Since these QWs can be considered as two tunnel-coupled HgTe QWs[2,39] or InAs/GaInSb QW bilayers[40,41], each of which features a single band inversion, the appearance of double band inversion is not surprising for such multi-layer systems. In Fig. 1C,D, we provide a phase diagram for three-layer InAs/GaInSb and double HgTe/CdHgTe QWs with different layer thicknesses. Both QWs are supposed to be grown on the (001) crystallographic plane.

Figure 1

Band structure of three-layer InAs/GaInSb and double HgTe/CdHgTe QWs. (A) Schematic representation of symmetrical three-layer InAs/GaInSb QW confined by outer AlSb barriers[50]. Here, d and t are the thicknesses of InAs and GaInSb layers, respectively. The QW is supposed to be grown on (001) GaSb buffer. (B) Schematic representation of double HgTe/CdHgTe QW. Here, d is the thickness of HgTe layers and t is the middle CdHgTe barrier thickness. The double QW is assumed to be grown on (001) CdTe buffer. The Hg content in all the barriers is chosen to be 0.3[51]. (C,D) The phase diagrams for different d and t. The left-hand and right-hand solid curves correspond to the crossing between E1–H1 subbands and E2–H2 subbands, respectively. These curves divide the plane into three parts with trivial band ordering corresponding to band insulator (BI, see the left-hand white region), single band inversion (grey and blue regions) and double-band inversion (right-hand white region). The striped region defines a semimetal (SM) phase with vanishing indirect band-gap[50,51]. (E,F) Band structure calculated on the basis of effective 2D low-energy Hamiltonian for the QWs with the layer thicknesses marked by the red symbols in (B) and (C). The blue and red curves represent band dispersion of electron-like and hole-like subbands, respectively. The wave vector is oriented along (100) crystallographic direction. The dotted curves represent the calculations based on realistic multi-band kp Hamiltonian[49] .

In the diagrams, the left-hand black solid curve describing the crossing between the first electron-like (E1) and hole-like (H1) subbands divides the d-t plane into a white region, corresponding to band insulator (BI) with trivial band ordering, and a grey region of QSHI with inverted band structure. If the middle barrier is thick enough, in addition to QSHI, the double HgTe/CdHgTe QWs also hold a specific state with a band structure similar to the one of bilayer graphene (BG) (see the blue region in Fig. 1D). A detailed discussion of the “bilayer graphene” state can be found in[51].

Further increasing of d and t results in the band crossing between the second electron-like (E2) and hole-like (H2) subbands, which is shown by the right-hand black solid curve in the diagram. This curve, in its turn, separates the grey and blue regions with single band inversion from the right-hand white region corresponding to the double band inversion, when two electron-like E1 and E2 levels lie below two hole-like H1 and H2 subbands.

Finally, at certain d and t values, corresponding to the striped region, the so-called semimetal (SM) phase is implemented. The semimetal phase is characterized by a vanishing indirect band-gap when the side maxima of the valence subband exceed in energy the conduction subband bottom[50,51]. Thus, by varying the layer thicknesses in the prototype QWs, one can indeed realize band insulator, QSHI (or “bilayer graphene” state), semimetal phase and the insulator state with a double band inversion. In this work, we identify a double-band-inversion insulator state 2D HOTI with the corner states arising due to cubic symmetry of II–VI and III–V semiconductors.

To describe double band inversion at the point of the Brillouin zone, we derive an effective 2D low-energy Hamiltonian taking into account E1, E2, H1 and H2 subbands and preserving the cubic symmetry of the prototype QWs. Starting from a realistic multi-band kp Hamiltonian for (0mn)-oriented cubic semiconductor QWs[49] and following expansion procedure described in the Supplementary Materials, in the basis , , , , , , , , the effective 2D Hamiltonian in the vicinity of the point has the form:where the asterisk denotes complex conjugation. The diagonal blocks of are split into isotropic and anisotropic parts:The isotropic part is written as[52,53]:whereHere, , and are the momentum components in the QW plane, and , , , , , and are isotropic structure parameters being defined by the QW geometry, the growth orientation and the materials. The Hamiltonian has a block-diagonal form because we keep the inversion symmetry (see Supplementary Materials) by neglecting the terms resulting from the anisotropy of chemical bonds at the QW interfaces[54] and the bulk inversion asymmetry of the unit cell of zinc-blende semiconductors[55]. The latter is absent for diamond-like semiconductors.

The most important quantities in are two mass parameters and describing the band inversion between E1–H1 subbands and E2–H2 subbands, respectively. The trivial band insulator corresponds to positive values of and . The QSHI and “bilayer graphene” states arise if and , and the difference between these states is defined by the gap between H1 and H2 subbands, which is zero in the case of “bilayer graphene” state[51]. The insulator state with double band inversion is defined by the negative values of and . We note that since the semimetal phase represented by the striped areas in the diagrams is formed by non-local overlapping of the valence and conduction subbands, it cannot be described within the low-energy Hamiltonian for the small values of and .

The isotropic term preserves the rotational symmetry in the QW plane, therefore it is independent of the orientation of x and y axis. In contrast, the form of the anisotropic term in Eq. (2), resulting from the cubic symmetry of diamond and zinc-blende semiconductors, depend not only on the QW growth orientation but also on the orientation of x and y axis (see Supplementary Materials). For (0mn)-oriented QWs, has the formwhere and are cubic structure parameters, which depend on the QW geometry and materials. In Eq. (5), is the angle defining the QW growth orientation, while the angle is the angle between the x axis and the (001) crystallographic direction.

As clear from above, depending on the structure parameters, the effective 2D Hamiltonian in Eq. (1) describes QSHI, “bilayer-graphene” state, trivial band insulator or insulator with double band inversion. Note that thirteen parameters involved in cannot take arbitrary values. Since their calculation is based on the wave-functions of the multi-band kp Hamiltonian at (for details, see Supplementary Materials), the set of parameters corresponding to specific topological state is determined by the thicknesses and materials of the QW layers, its growth orientation, and the buffer on which the QW is grown. The latter is crucial for taking into account the effect of lattice-mismatch strain on the band structure in the QW. Thus, all of the structure parameters are the functions of d, t, m, n and the QW layer and buffer materials. Further, we perform the calculations for two sets of structure parameters involved in , which correspond to the prototype QWs with the layer thicknesses marked by the red symbols in Fig. 1C,D. Figure 1E,F compare the band structure calculations based on realistic multi-band kp Hamiltonian[49] and in Eq. (1). The isotropic and cubic structure parameters of for the insulator state with double band inversion are given in the Supplementary Materials.

Anisotropic edge states

Let us now analyze the edge states arising in the insulator with double band inversion. Since in Eq. (1) has block-diagonal form, we further focus on the upper block only, while the calculations for the lower block can be performed in the similar manner.

To derive an effective 1D low-energy edge Hamiltonian, we split into two parts so that the first part represents two independent BHZ-like models[2] with and for the pairs of E1–H1 subbands and E2–H2 subbands, while the second part includes the rest isotropic and cubic terms describing the inter-pairs mixing. Then, assuming open-boundary conditions in a semi-infinite plane , we solve the eigenvalue problem for the independent BHZ-like blocks to find the edge wave-functions at . In this case, the edge orientation represented by the x axis is defined by the angle measured from (100) crystallographic direction. Finally, we construct a low-energy edge Hamiltonian by projecting onto the obtained set of the basis edge functions (see Supplementary Materials).

The projection of two independent BHZ-like blocks[2] with non-zero leads towhere with and 2 corresponding to the pairs of E1-H1 and E2-H2 subbands, respectively. One can see that describes the linear edge dispersion due to the inversion of the subband pairs – and – in the absence of their mixing[56]. The two energy branches cross at :that allows the representation of in the form of the Hamiltonian of ‘tilted” 1D massless Dirac fermions:where , is a identity matrix, is one of the Pauli matrices, and is a constant corresponding to the energy of the crossing point at . In Eq. (8), and are written asNote that the crossing of other Kramer’s partners occurs at .

The projection of the rest terms of representing the mixing between the pairs – and – results in anti-diagonal mass terms describing the band-gap opening. After straightforward calculations with the details provided in the Supplementary Materials, we finally obtain the low-energy 1D edge Hamiltonian:where , , , , , include the angle dependence on and :Here, and are defined by the matrix elements and , respectively, both calculated by using the basis edge functions at (see Supplementary Materials). In Eqs. (10), we have also introduced the edge isotropic (, ) and cubic (, ) parameters:The similar calculations for the block in Eq. (1) results in . The parameters , , , , , as a function of the edge orientation for the prototype QWs grown along (001), (011) and (013) crystallographic directions are provided in the Supplementary Materials.

Figure 2 shows the edge state picture of the prototype QWs with double band inversion based on low-energy 1D edge Hamiltonian for the edge oriented along (100) crystallographic direction, i.e. . One can see that the edge dispersion of each Kramer’s partner mimics the dispersion of “tilted” 1D massive Dirac fermions. However, an important distinctive property of the edge states shown in Fig. 2 is that their gap is described simultaneously by two mass and parameters that prevent the gap vanishing at specific edge orientation (cf. Refs.[35-38]). Indeed, if one neglects the cubic terms of , resulting in in Eq. (10), the mass parameter vanishes, while becomes independent of the edge orientation. Nevertheless, even such complex structure of the edge states yields the corner states in prototype QWs with double band inversion.

Figure 2

Energy dispersion of the bulk and edge states in the prototype QWs with double band inversion. The (A) and (B) presents the calculations based on low-energy 1D and 2D Hamiltonians for the (001)-oriented three-layer InAs/GaInSb and double HgTe/CdHgTe QWs, respectively. The layer thicknesses of the prototype QWs are marked by the red symbols in Fig. 1B,C. Bulk states being described by in Eq. (1) are shown in black. Red and blue curves correspond to different Kramer’s partners of the edge states. The dashed curves are the eigenvalues of in Eq. (8). The wave vector and the edge are oriented along (100) crystallographic direction.

0D corner states and boundary conditions

To calculate the energy of the corner states, we apply linear approximation of the effective low-energy 1D edge Hamiltonian . Before going further, we should make a certain remark simplifying the calculations. With the parameters provided in the Supplementary Materials, one can verify that both and are significantly lower than and for any orientation of the edges (also see Fig. 3). Therefore, one can neglect these terms in the first approximation and take them into account by means of perturbation theory.

Figure 3

Orientation of the corner edge of (0mn)-oriented prototype QWs. (A) Schematic of two edges with a common corner with respect to main crystallographic axes in the QW plane. The sample and external vacuum are shown in grey and white, respectively. External and internal angles differ by the parameter and (see the text). (B–D) Dependence of , and on the edge orientation for the (001)-oriented three-layer InAs/GaInSb (in red) and double HgTe/CdHgTe QW (in blue) with the layer thicknesses marked by the red symbols in Fig. 1B,C. The edge independent parameters (, ) equal to (36.6, 68.1) meV nm and (, 200.3) meV nm for the three-layer InAs/GaInSb and double HgTe/CdHgTe QW, respectively. This means a good approximation of for both QWs.

Thus, by applying the unitary transformation, the linear part of Eq. (9) can be written asIt is clear that Eq. (12) represents a 1D Dirac Hamiltonian with the mass , which changes its sign with (see Fig. 3), modified by the presence of “tilted” term and second mass term .

Let us now define the coordinate x along the curved edge so that corresponds to the meeting corner as shown in Fig. 3A. The latter means that , in Eq. (12) are the function of x, and . Note that is defined in disjoint regions out of . To fully define the 1D system, one needs to specify the boundary conditions that the wave functions must satisfy in the vicinity of .

As shown in the Supplementary Materials, the general linear boundary condition, conserving the probability current along the curved edge, can be written in the formwhere and are the wave-functions defined from different sides of the corner, while and are real parameters lying in the range from to . A physical interpretation of these parameters will be discussed later.

On the basis of Eq. (13), it is convenient to introduce a new Hamiltonian , whose eigenfunctions are continuous at :where and arewith also being a function of x.

In view of the above, the Schrödinger equation for the corner states takes the formAn exact solution of Eq. (16) can be found for the case in which and are proportional to each otherwhere and m are real constants defined asNote that Eq. (17) is exact for the sharp corner shown in Fig. 3, for which and are the step-like functions of x.

As shown in the Supplementary Materials, under the condition of Eq. (17), the wave-function can be presented in the formwhere is the spin part of the wave function satisfying equationwith eigenvalues .

Then, by introducing a new variable and representing in the formwe arrive at the equation for the coordinate part:where (the sign of coincides with those for ), and and are defined asEquation (21) possesses a special symmetry that corresponds to the formulation of supersymmetric quantum mechanics[57] with the supersymmetric potential . If the signs of the asymptotes and are opposite, i.e.Eq. (21) always has a localized solution with .

Thus, by means of Eqs. (20)–(23), the wave function of the corner state is written as:where C is the normalization constant andNote that in Eqs. (24) and (25), the sign of should be chosen in accordance with normalized condition of . If , , while for , . These two cases correspond to the internal and external corners at the same orientations of the two edges.

Let us make few remarks concerning the results obtained above. First, Eq. (23) can be also written in equivalent formwhich is reduced to the well-known condition for the existence of the bound state in 1D Dirac system if one of the mass parameters or is absent.

Second, depends on the values of and as seen from Eq. (15). Nevertheless, sincetakes place for any values of and , the corner state energy always lies in the band-gap of the edge states as soon as Eq. (26) is fulfilled. One can show that may formally achieve the energies of the 1D band edges at certain values of and , corresponding toThe latter however represents the moment, when the corner state becomes delocalized.

Finally, we take into account the small terms proportional to and previously neglected in Eq. (12). The straightforward calculations on the basis of (see Supplementary Materials) lead to the first-order energy shift:whereOne can verify that indeed for both prototype QWs. The calculations for another Kramer’s pair of the edge states described by results in the same energy of the corner state.

Trivial corner states

Let us now discuss the physical origin of the localized 0D corner state found above. Since it is clear from Eq. (15), the existence condition of Eq. (23) (or equivalently Eq. (26)) is fulfilled for different functions , and . The function characterizes the corner itself, while and include the characteristic of the entire system. Further, we show that, in the most general case, arising of 0D corner state simultaneously depends on the corner boundary conditions and the cubic symmetry of the system.

For a better understanding of the existence condition of Eq. (26), we first neglect the terms arising due to the cubic symmetry. In this case, one should set parameters and of 2D Hamiltonian in Eq. (5) to zero also resulting in the zero values of and in Eq. (11). The latter, in turn, leads to the vanishing of , and in Eq. (10), while , and become independent of x. Thus, in the absence of the cubic symmetry, by means of Eqs. (15) and (18), the existence condition of Eq. (26) is reduced towhere and chosen in accordance with Eq. (13). Since it is easy to see, the localized corner state exists if . These parameters can be given a precise physical interpretation.

Let us consider a -function electrostatic potential at . Then, if we integrate the Schrödinger equation with (where is a real parameter) in the range of (where is a positive infinitesimal quantity), we directly obtain Eq. (13) with . Therefore, can be interpreted as the presence of -like electrostatic potential localized at the corner. Thus, the corner states arising in this case are topologically trivial, since they are independent of the cubic symmetry of the system. We also note their independence of the sign of , as clearly seen from Eq. (30).

Topological corner states

We now focus on the opposite case, when there is no potential barrier at the corner i.e. . As seen from Eq. (13), these parameters can be both set to zero without loss of generality, which leads to and . In this case, the presence of the corner states is governed by the cubic symmetry of 2D system, represented by the non-zero values of and in Eq. (11). The latter means that the picture of the symmetry-protected corner states should strongly depends not only on crystallographic orientations of the meeting edges but also on the QW growth direction.

Figure 4A,B present the calculations for the prototype QWs with double band inversion grown along (001) crystallographic orientation (see the diagrams in Fig. 1). For both QWs, the angles and defining the corner orientation in Fig. 3A are chosen to be and , while in Eq. (10) is assumed to be zero. One can verify that the condition of existence of Eq. (26) is fulfilled for these angles, and, therefore, the corner states arise in the system. As seen from Fig. 4A,B, due to the relationship between and , the localized energies for internal and external corners are very close to the extrema of 1D dispersion of the edge states.

Figure 4

Topological corner states in (001)-oriented prototype QWs. (A,B) 1D dispersion of the edge states (in blue) for the boundary oriented at and the corner states energy (in red) formed by the second boundary oriented at . The (A) and (B) correspond to the (001) three-layer InAs/GaInSb (A) and double HgTe/CdHgTe QWs, respectively. The layer thicknesses of the QWs are marked by the red symbols in Fig. 1B,C. (C,D) The phase diagram for the presence (grey region) and absence (white region) of topological corner states in the three-layer InAs/GaInSb and double HgTe/CdHgTe QWs as a function of the two edge orientations. In the grey regions, the existence condition of Eq. (26) at is fulfilled. The vertical and horizontal thin dotted lines represent the angles when the edges of 2D system, which are actually the faces of (001)-oriented QWs, coincide with [110], , [100] and [010] crystallographic planes.

Figure 4C,D represent the angle diagram showing the edge orientations in the (001)-oriented prototype QWs, for which the existence condition of Eq. (26) is valid. It is seen that the values of and yielding the corner states are qualitatively represented as the grey regions elongated along the lines defined byexcluding the points lying at verticals and horizontals corresponding to and , where , and are integers. As seen from Fig. 3A, the 2D system edges at these angles, coincides with [110], , [100] or [010] crystallographic planes, being the faces of (001)-oriented QWs in this case. Since we have neglected the terms resulting from possible breaking of inversion symmetry in the system, the (001)-oriented prototype QWs preserve the mirror symmetry about the mentioned planes. Therefore, one concludes that if one of the QW edge coincides with the mirror symmetry planes, the corner states are absent in the system.

In light of the above, Eq. (31) describes the corners, whose bisector corresponds to the one of the mirror symmetry planes. The latter results in reflection symmetry of the angle diagram with respect to the lines described by Eq. (31). As expected from the four-fold rotational symmetry of (001)-oriented QWs with respect to the growth direction, the angle pattern in Fig. 4C,D also possess a -periodicity in and .

So far, we have discussed the case of (001)-oriented QWs, by whose example we have shown a direct relationship between topological corner states and the symmetry elements of the QW. As the inversion symmetry is preserved in our model, the (001)-oriented QWs have the point group origin from the point group of bulk multi-band kp Hamiltonian (see Supplementary Materials). Obviously, if one reduces the symmetry of 2D system, the picture of the corner states should change as well. Further, we consider -symmetric QWs grown along (013) crystallographic direction inspired by recent experimental investigations of double HgTe QWs of the same orientation[58-60].

Figure 5A,B provide the angle diagram of topological corner states in the prototypes three-layer InAs/GaInSb and double HgTe/CdHgTe QWs oriented along (013) crystallographic direction. The calculations are performed by assuming in Eq. (10) with (m, n) = (1, 3). It is seen that the values of and , for which the existence condition is fulfilled, represent significantly different picture of the corner states than the one for (001)-oriented QWs.

Figure 5

Topological corner states in (013)-oriented prototype QWs. (A,B) The phase diagram for the presence (grey region) and absence (white region) of 0D topological corner states in the three-layer InAs/GaInSb (A) and double HgTe/CdHgTe QWs (B) as a function of the two edge orientations. The vertical and horizontal thin dotted lines represent the angles when the edges of 2D system coincide with [100] and crystallographic planes. (C) The function representing the existence condition of Eq. (26) at in the ranges marked with colored rectangles in (B). The dotted line marks the zero value.

We first note -periodicity in and of the angle pattern due to the two-fold rotational symmetry of (0mn)-oriented QWs in contrast to the -periodicity of (001)-oriented 2D systems. Second, due to the symmetry lowering, (013)-oriented QWs have only two [100] and mirror planes perpendicular to the QW plane. These planes are represented by dotted verticals and horizontals at and in Fig. 5A,B. It is seen that the corner states are absent if one of the QW edge coincides with the mirror symmetry planes. Finally, in contrast to (001)-oriented QWs, (0mn)-oriented QWs possess the corner states if one of the edges is at ° with respect to the (100) crystallographic axis. The latter is clearly seen in Fig. 5C, which shows the negative values of the function illustrating the existence condition of Eq. (26) for the corner states.

Discussion

To summarize, we have investigated the existence conditions for 0D corner states in cubic semiconductor QWs with double band inversion. We have demonstrated that 0D corner states in such 2D system can appear either due to the presence of electrostatic potential localized in the corner, or due to the crystalline symmetry of the QW. The former case can be physically interpreted as the impurity potential resulting in trivial corner states. The latter corresponds to the symmetry-induced corner states inherent for 2D HOTI state in cubic semiconductor QWs with double band inversion. We have shown that the corners hosting 0D topological states depend not only on crystallographic orientations of the meeting edges but also on the growth orientation and parameters of the QWs.

Let us now make a few remarks concerning the approximation used in this work. As it is clear from the above, theoretical investigations were performed on the basis of multi-band kp Hamiltonian[49] for the envelope Bloch functions, which is indeed valid on scales much larger than the unit cell. In this case, all functions that are changed on a scale comparable to the unit cell are considered as step-like or delta-functions. In this sense, a meeting point of the corner shown in Fig. 3A represents a 2D unit cell at the corner of an effective 2D system in the QW plane. The special arrangement of atoms in the vicinity of the corner preserving symmetry of the bulk may lead to the fractional charge of the corner states, known as a fractional corner anomaly[22,23,61-63]. We note that since fractional corner anomaly requires a theoretical description on the unit cell scale, it cannot be treated on the basis of multi-band kp Hamiltonian.

Throughout the work, we have neglected the inversion symmetry breaking terms resulting from anisotropy of chemical bonds at the QW interfaces[54] and possible bulk inversion asymmetry of the unit cell[55]. The former is known as an interface inversion asymmetry[54], while the latter causes the difference between and point groups of diamond and zinc-blende semiconductors. Taking into account these additional terms in bulk kp Hamiltonian leads to the non-diagonal blocks of the effective 2D Hamiltonian in Eq. (1). Since the inversion symmetry breaking terms do not affect the very fact of double band inversion, their presence will modify only the edge state parameters , , , , , keeping the form of 1D edge Hamiltonian in Eq. (9).

For instance, taking into account the interface and bulk inversion asymmetry in (001)-oriented zinc-blende semiconductor QW reduces the point symmetry from to [54], which, in their turn, also changes the periodicity of , , , , , in Eq. (9) from to . In this case, the symmetry-induced corner states will arise at the corner with the meeting edges, whose orientations differ from those shown in Fig. 4. For the prototype InAs/GaInSb and HgTe/CdHgTe QWs considered in this work, the terms due to interface and bulk inversion asymmetry are small and induce therefore only a slight modification of the phase diagrams.

We emphasize that since the presence of the corner states depends on the mutual ratio between and , one may find the QW orientation and strength of inversion-symmetry breaking terms, for which the existence condition of Eq. (26) cannot be fulfilled. The study of all possible cases of violation of Eq. (26) is however out the scope of our work first considering the QWs based on IV, II–VI and III–V semiconductors, in which inversion-symmetry breaking terms are usually small.

Finally, we stress the importance of theoretical results obtained in this work in view of possible applications and their impact on further experimental investigations. After the tremendous interest in II–VI and III–V semiconductor QWs induced by prediction and observation of QSHI[39-41], our work shows the importance of cubic semiconductor QWs for the realization of high-order topological states as well. In view of mature growth of IV, II–IV and III–V semiconductor QWs on Si-wafers[64,65] as well as device fabrication technology, our results provide an important first step to future realistic electronics operating on the basis of higher-order topological states including higher-order topological superconductors in hybrid devices[66,67].

Methods

Band structure calculations were performed by using multi-band kp Hamiltonian[49], which directly takes into account the interactions between , , and bands in bulk materials. This model well describes the electronic states in a wide range of narrow-gap semiconductor QWs, particularly in the InAs/GaInSb[46,48] and HgTe/CdHgTe QWs[58-60]. In the multi-band kp Hamiltonian, we also took into account the terms, describing the strain effect arising because of the mismatch of lattice constants in the buffer, QW layers, and barriers. The calculations had been performed by expanding the eight-component envelope wave functions in the basis set of plane waves and by numerical solution of the eigenvalue problem. Details of calculations and the form of the Hamiltonian can be found in the study of Krishtopenko et al.[49]. Parameters for the bulk materials and valence band offsets for the the InAs/GaInSb and HgTe/CdHgTe QWs used in the calculations are provided in Ref.[50] and Ref.[49], respectively. To derive effective 2D Hamiltonian valid in the vicinity of the point from the multi-band kp Hamiltonian, we implied the procedure proposed by Bernevig et al.[2] and described in details in the Supplementary Materials.

Supplementary Information.