WorksPhysics and Astronomy
Surface potential-driven surface states in 3D topological photonic crystals
Abstract
Surface potential in a topological matter could unprecedentedly localize the waves. However, this surface potential is yet to be exploited in topological photonic systems. Here, we demonstrate that photonic surface states can be induced and controlled by the surface potential in a dielectric double gyroid (DG) photonic crystal. The basis translation in a unit cell enables tuning of the surface potential, which in turn regulates the degree of wave localization. The gradual modulation of DG photonic crystals enables the generation of a pseudomagnetic field. Overall, this study shows the interplay between surface potential and pseudomagnetic field regarding the surface states. The physical consequences outlined herein not only widen the scope of surface states in 3D photonic crystals but also highlight the importance of surface treatments in a photonic system.
regarding the surface states. The physical consequences outlined herein not only widen the scope of surface states in 3D photonic crystals but also highlight the importance of surface treatments in a photonic system.
Main text:
The discovery of topological insulators opened exciting new realms of physics[1, 2], and extensive efforts have been made to understand topological physics over the last decade. At this field's core, it is important to note zero-dimensional degeneracies such as Dirac[3, 4, 5, 6, 7, 8, 9] or Weyl points[10, 11] [12, 13, 14, 15, 16, 17, 18, 19] and one-dimensional degeneracies such as nodal lines20-28. These band degeneracies have been theoretically or experimentally realized using metals21,29, semimetals30-36, phononic crystals37-39, electrical circuits40, and photonic crystals[10, 11] [18, 41, 42, 43, 44, 45, 46, 47, 48, 49]. Based on the definition of topological insulators, many studies propose several boundary states, such as one-way surface/edge states39,50,51, drumhead surface states40,52-55, and Fermi arcs47,48,56-59.
Beyond them, one can adjust the boundary state with surface/edge potential60-67 and pseudomagnetic field13,37,68. In condensed matters, the surface/edge potential60-64 arises from the passivation on the surface/edge of a given material65-67. The surface/edge potential60-63 has been successfully implemented in two-dimensional (2D) materials including graphene65, boron nitride66, and semimetals;60,64,67 consequently, the excitation and manipulation of surface/edge states came to the fore. However, such driving and using surface potential are yet to be exploited in topological photonic crystals.
Meanwhile, the pseudomagnetic field is a virtual effective field stemming from a spatial reconfiguration of the crystal lattice without a real magnetic field69-72. Like a real magnetic field enabling the quantum Hall effect73,74, the pseudomagnetic field was found to be a gold vista for topologically nontrivial surface/edge states in semimetals75-79, photonic crystals80-82 and phononic crystals37,70,71,83. Nevertheless, the surface state’s profiles along the zeroth Landau level on a 3D Weyl photonic crystal, driven by the pseudomagnetic field, were out of reach.
Here, we demonstrate photonic surface waves that arise from the interplay between surface potential and pseudomagnetic field in photonic systems. First, we investigate the effect of the surface potential in a system governed by the Weyl equation with a pseudomagnetic field. The pseudomagnetic field is switched on by the Weyl points that vary with the unit cell positions along the boundary-boundary direction in the system. Then, we apply the effective Hamiltonian description to a photonic system based on double gyroid (DG) photonic crystals17. The DG photonic crystals are found to exhibit Weyl points due to the geometrical perturbation, which appears as a defect-like shape. To realize a pseudomagnetic field, we constitute a photonic array of the DG unit cells with a spatial gradient of the perturbation, i.e., the degree of the defect varies with the position. We then compute quantized Landau levels and eigenstates along the zeroth Landau level to quantitate the asymmetric localization of photonic waves on the surfaces. This observed asymmetry in wave localization evidences the existence of the surface potential. Then, we tune the translation of the basis of the unit cell to tame the surface termination. This tuning exquisitely controls the surface potential so that the degree of wave localization varies with respect to the tuning. Finally, we implement such results to the evasion behaviors of a photonic wave to observe the interplay between the surface potential and pseudomagnetic field.
The attraction of waves by the surface potential in the Weyl system
First, let us describe the pseudomagnetic field’s effects using the Weyl Hamiltonian. For this, we consider a system periodic along the - and -directions for an orthogonal coordinate system. The system consists of unit cells along the -direction (the horizontal rightward direction in Figure 1a-Figure 1b), and surface boundaries are parallel to the - and -directions. We assume that this system is governed by that describes a Weyl point at , where is the identity matrix, and () are the Pauli matrices. We decompose the Weyl point’s location into .

Figure 1. Effects of surface potential on the wave localization by pseudomagnetic fields. a, b, Schematic illustrations of two systems with opposite . The systems are finite along -direction (the horizontal direction) and periodic along - and -direction. In each panel, the varying is schematically represented as the size of yellow circles. Both systems are described by equation (1). c, Comparisons of their eigenstates’ distributions along -direction, exhibiting the mutual symmetric dispersions. d, Schematic plot of , the scalar coefficient of the surface potential. e, Comparisons of the eigenstates’ distributions of a and b when is applied. In c and e, gray dotted lines are the symmetric curves with respect to the center for comparisons of the red and blue plots.
is a constant while depends on the unit cell’s index . We assume that linearly varies with a specific value (henceforth, this value is referred to as perturbation strength) which also linearly varies with :
where is a proportional constant (refer to Figure 1a-Figure 1b). Then, the pseudomagnetic field can be written as , which is proportional to .
Let us suppose two systems with different signs of slopes, i.e., one system described by (Figure 1a) and the other described by (Figure 1b) where is positive. Their are schematically illustrated as the size of yellow circles in Figure 1a-Figure 1b. The pseudomagnetic fields for these two are in the same magnitudes and in opposite directions. Their wave localization of the zeroth Landau level at a specific point in the momentum space shows mutually symmetric characteristics for these opposite fields, as shown in Figure 1c.
Now, let us implement a surface potential as follows:
where is nonzero only around the boundaries60,61,64-66, like Figure 1d. When we include the surface potential term with , further detailed in Methods, eq. (1) is rewritten as:
By applying the same for the two systems in Figure 1a-Figure 1b, we can observe mutually asymmetric localization, as shown in Figure 1e. The two localized states commonly show a slight shift in the left direction, implying that the surface potential attracts them along that direction.
Applying surface potential in a finite-sized array
Although there can be several methods to drive a surface potential into a photonic crystal, we herein adjust the surface terminations in a finite-sized array. In a fully periodic crystal, the fields or eigenvectors of a propagating mode are invariant under discrete translation, not depending on a specific position of the basis in a unit cell. On the contrary, a finite-sized array has terminations at the boundaries. A propagating wave profile depends on the terminations according to the variations of the basis’s positions in the unit cells84, as shown in Figure 2a. Thus, the effect of surface potential can be quantitated with respect to adjusting the surface terminations.

Figure 2. Schematics on the realization of surface potential. a, Inducing different surface potentials by different surface terminations. The surface terminations are adjusted by the basis’s positions in a unit cell. b, Adding pseudomagnetic field on the systems in a. Generating surface localization of a photonic wave is performed by the pseudomagnetic field, and the degree of localization depends on the surface potential.
Here, we remark on the followings: (i) Eqs. (1) to (3) do not consider a detailed geometry in each unit cell. Thus, observation of the relation between the surface termination and surface potential should be phenomenologically carried out using a real array structure. (ii) A photonic band structure for a finite-sized array displays projected bands, called folding of bands. Without the pseudomagnetic field, the zeroth Landau level adheres to the bulk bands so that surface states cannot be obtained. Thus, we should use the pseudomagnetic field, as shown in Figure 2b. Our study shows the interplay between the surface potential and pseudomagnetic field; the pseudomagnetic field gives the surface localization of a photonic wave, and the surface potential tunes the degree of the localization.
To build the array in Figure 2b and to induce both the surface potential and pseudomagnetic field, we use a DG photonic crystal that exhibits Weyl points17. We adjust the surface terminations by tuning the translation of the basis along the normal direction to the boundaries.
Double gyroid photonic array exhibiting pseudomagnetic field
To drive the surface potential by the surface termination, we use a photonic array that exhibits a pseudomagnetic field by in eq. (3). We consider the DG, as shown in Figure 3a. The yellow and blue single gyroids (SGs) are given by a set of such that and , respectively, where and are the level-set values that determine the volume fraction of each SG. The yellow SG has selectively the perturbation term to break the inversion symmetry and to generate Weyl points in momentum space. Due to this perturbation, the yellow SG exhibits a defect-like narrow region on the arm passing the surface, as shown in Figure 3a. The higher perturbation strength induces the deeper defect-like shape; this corresponds to the yellow circle’s size in Figure 1a-Figure 1b and Figure 2. (See Methods for the detailed explanations about the DG crystal.) When this DG photonic crystal is periodic along all three lattice vector directions, it can exhibit four Weyl points, as shown in Figure 3b-Figure 3c. The Weyl points and (marked with blue and red solid points in Figure 3c, respectively) have positive and negative topological charges, respectively.

Figure 3. Design of 3D photonic crystal for pseudomagnetic field. a, A DG photonic crystal. The SGs are defined by that satisfies the inequalities denoted in the figure. The yellow SG’s inequality has an additional term to impose a defect-like shape on this structure (see the bottom-right arm). b, Resulting photonic band structure exhibiting four Weyl points, the band degeneracies. c, Weyl points marked in the three-dimensional first Brillouin zone. They are on a single plane. d, Movements of Weyl points on the plane with the perturbation strength . Weyl points in c are also marked here with red and blue solid points and denoted by and , respectively. e, Deviations of the Weyl points’ movements with from the positions of and , shown in the right and left enlargements in d, respectively.
Varying of the DG shifts the positions of the Weyl points, as shown in Figure 3d. We denote the moved positions of the Weyl points and in momentum space by and , respectively. Increasing makes the Weyl points shift away from -point on the single plane ((001)-plane). For the narrow range of around the central value , the traces of all Weyl points exhibit linear shapes so that we can write their positions as and where and are overall directions of the traces (see Figure 3d). The deviations of shifted Weyl points, and , also show linear relations with , as shown in Figure 3e.
To generate the pseudomagnetic field, we design a photonic array made of DGs whose linearly varies with respect to the position in the array13,70,71,85,86. First, we assume a DG-array that consists of DGs with unit cells along the -direction between two boundaries, as shown in Figure 4a. The array is periodic along the - and -directions. Next, we apply the non-uniform geometry on this array using the perturbation strength that linearly varies along the :

Figure 4. Design of a photonic array and resulting pseudomagnetic fields. a, Schematic geometry which consists of cells with spatially linearly varying along -direction. This gradient is imposed only on the yellow structure. b, Bulk and surface Brillouin zones. c-d, Direction vectors of the pseudomagnetic field around each Weyl point by the above design, denoted by , and , respectively. These are calculated from the traces around , , , and marked in Figure 3d, respectively.
like the inset linear plot in Figure 4a. Here, is a coordinate along the , is a distance between the planes at , and , and is a central value of . From the information in Figure 4a-Figure 4b, the pseudomagnetic field has the form , which are parallel to the array boundaries (see Figure 4c-Figure 4d). As a result, the waves can be localized at the boundaries (see Methods for detailed explanations about the array and pseudomagnetic field calculations).
Surface potential by surface termination and resulting photonic wave localization
Plugging the design in the previous section to the array of primitive cells allows us to generate Landau spectra. The detailed explanations of the Landau levels related to the pseudomagnetic field are in Section 1, Supplementary Information.
Here, we extract and compare the zeroth Landau levels by two arrays with and , as illustrated in Figure 5a. Although the two arrays have different internal geometries due to the different values, their overall translation status are identical, i.e., they generally use the formulae denoted in Figure 3a. Thus, we consider that their surface terminations are identical. The photonic band structure in Figure 5a shows that the resulting zeroth Landau levels accessible from the two arrays are not equal. Furthermore, the eigenstate intensity distributions for the two cases reveal mutual asymmetric (Figure 5b-Figure 5c). The distributions in Figure 5d-Figure 5e are biased toward the from the symmetric curve (the gray dotted lines). From the fact that the results in Figure 5d-Figure 5e exhibit the same tendency as with Figure 1e, we can conclude that there exists a surface potential in these arrays.

Figure 5. Photonic wave localization and evidence of surface potential. a, Comparison of Landau levels by the opposite perturbation fields. The left and right cases correspond to and , respectively. b, c, Normalized eigenstates respectively along the blue- and red-colored Landau levels shown in a to see surface states localized on the boundaries. d, e, Comparisons of the wave localization along the vertical lines in b and c. Data with the same symbols are overlapped in the same plot. Gray dotted lines are the symmetric curves with respect to for comparisons of the red and blue plots.
Now, we replace as to apply overall translation of the DG by along the -direction to verify the surface termination’s effect, as illustrated in Figure 6a. Note that the profile of does not move (see the upper right plot in Figure 6a). For a specific point on the zeroth Landau levels (marked in Figure 6b), we calculate the field intensity for several values, as shown in Figure 6c. The discussions so far are about that the surface potential can adjust the degree of localization of surface states generated by the pseudomagnetic field. The results in Figure 6c indicate that the localization degree can be adjusted by the surface termination. The zeroth Landau level in the photonic band structure also can be moved to another position by changing . Due to the relatively small , the overall configurations of the zeroth Landau level and the surrounding bulk bands remains almost intact, as shown in Figure 6b.

Figure 6. Adjustment of surface potential by surface termination. a, Schematic illustration of three surface terminations. Observation of the surface potentials is carried out by adjusting the basis’s positions in each unit cell. The upper left three structures are enlargements of the bottom right arrays. The lower left three figures are respectively the schematics of the upper left structures, and they clearly show the translation of the structures along the -direction. The right upper plot indicates that the perturbation versus the position is the same for all three cases. b-c, Zeroth Landau levels (b) and eigenstate intensities (c) at the point marked in b for three different surface terminations tuned by translation along the -direction when and are used.
Evasion behavior of photonic wave
We embody the evasion behavior of a photonic wave in a DG structure using oppositely graded . Adjusting the propagation path of a photonic or phononic wave has been an interest in photonics or wave mechanics. These have been proved using materials with negative-refractive indices16,87-89 or Dirac/Weyl crystals exhibiting one-way propagations of waves[3] [12, 13, 44, 90]. Such proofs are related to invisibility-cloaking. Here, we perform the behavior by pulling a photonic wave that was originally propagating on one boundary toward the opposite boundary.
DG arrays that consist of 8 cells with , 0, , , and are prepared, based on the array in Figure 4a, (). (The number of cells is counted based on the BCC primitive cell.) The reason for using a smaller number of cells than in the previous case is to broaden intervals between Landau levels. We classify these arrays into groups with and the others with . The directions of for these two groups are opposite to each other (see the inset of Figure 7a and Figure 7b). The band structures for and (plotted in Figure 7a and Figure 7b, respectively) exhibit Landau levels along -direction. From these plots, we consider as an optimized frequency for the evasion behavior because this frequency between the bulk bands and all zeroth Landau levels commonly meets this frequency only once. We then constitute a DG array, as shown in Figure 7c, using the 8-cells arrays depicted in the inset of Figure 7a-Figure 7b. Each system consists of 8 and 32 cells along - and -directions, respectively, and it is assumed to be infinitely periodic along -direction. The blocks and use negative-valued , and the central region between the two sections has positive-valued .

Figure 7. Evasion behavior of photonic waves using heterogeneous blocks. a, b, Zeroth-Landau levels by the 8-block system with several perturbation fields, where . The perturbation fields used in a and b are opposite. c, Heterogeneous DG system. The PEC is applied to the pink and green boundaries. In pink and green colored bounded blocks, the perturbation fields’ directions are opposite as marked. The system is periodic only along -direction. d-g, Evasion behavior of photonic waves with several values of the central blocks with fixing values of the blocks around both ends. Incident points are marked as the star symbols.
The DG array reveals the evasion behavior of photonic waves. At , a photonic wave is localized in the section of around the boundary at . We then input an incident photonic wave around the localization region, as marked by the star symbols in Figure 7d-Figure 7g. If the blocks were the same as the other blocks, the localization would continue in the central region. We simulate here the situations that the central region’s is respectively , , , and . The results show the gradual attraction of the photonic waves onto the boundary at with increasing (see Figure 7d-Figure 7g).
In addition, we observe the attraction of waves towards the boundaries. In the blocks and , the waves are localized on the boundary. On the contrary, the localized waves in the blocks exhibit biased behaviors. Then, we can conclude that this arises from surface potential and surface termination. If we adjust the surface termination, we may observe the bias towards the opposite direction. All these show the interplay between surface potential and pseudomagnetic field.
Discussion
We have demonstrated the asymmetric localization of photonic waves via an interplay between surface potential and pseudomagnetic field using a DG photonic crystal. The pseudomagnetic field has induced the surface states of the photonic waves and the surface potential adjusted the degree of localization. The pseudomagnetic field was formed by the graded location of Weyl points, and the surface potential was applied using surface termination by tuning unit cells’ basis translation. We have observed the shift of the surface states as a result of the surface potential.
In the case of graphene65, boron nitride66, and semimetals60,64,67, surface potentials can be applied using materials’ surface/edge passivation. Although this study used surface termination by crystal’s basis translation, we believe that there could be several types of photonic passivation, for example, a thin material doping. Then, this study and the follow-up studies will open the various possibility of using topological photonic crystals. Meanwhile, there is no example of the detailed analysis of the surface states along the zeroth Landau level by the Hall effect and pseudomagnetic field in three-dimensional Weyl photonic crystals. Therefore, this study will fill this gap thereby this study will positively affect the other studies on the three-dimensional quantum (spin) Hall effect91-96.
Methods
Eigenstates localization by Weyl Hamiltonian. We apply eq. (1) or (3) to the finite system illustrated in Figure 1a-Figure 1b to observe the wave localization like Figure 1c or e. Instead of deriving the Landau level analytically using ladder operators33,97, we use a traveling wave solution to consider the finite array. The traveling wave solution can be rewritten as the product of a state that depends on only and an exponent that depends on only other variables: . The surface localization is obtained by substituting this into equation (1) or (3). To consider the surface potential in eq. (2), we used the plot in Figure 8 as . Detailed derivations, explanations, and additional results related to Figure 1 are given in Section 2, Supplementary Information.

Figure 8. Plot of the surface potential’s scalar coefficient . This is the discretized function from Figure 1d. Plots in Figure 1c and e were obtained using this .
Preparations of DG photonic crystal. In the following, we give detailed explanations of the DG photonic crystal and array. In this study, we consider the DG, reported in ref. 17. The body-centered cubic (BCC) primitive cell of this structure is defined by the lattice vectors , where . The mathematical formulae of , , and denoted in Figure 3a are , , and , respectively. Here, the local coordinates are given by and , where is a lattice constant, is the shift coefficient that describes the translation of the yellow SG, and is the translation direction. For the inequalities marked in Figure 3a, the level-set values and are used that determine the volume fraction of each SG. The refractive indices of the two SGs are commonly 4.0, and the outside region of them is filled with air whose refractive index is 1.0.
The perturbation strength plays an important role in our work. If becomes zero, the space group of this DG is (no. 230), and this does not exhibit Weyl points18. A DG with appropriate nonzero-valued exhibits four Weyl points in the momentum space between the fourth and fifth bands17. Due to the inversion symmetry of the momentum space, the positions of these four Weyl points are inversion symmetric to -point. When is used, we can get the four Weyl points, as shown in Figure 3b-Figure 3c. They are on a single -plane passing and where the reciprocal primitive vectors are defined by . These Weyl points are also denoted as and in Figure 3d. The Chern numbers of the Weyl points and are and , respectively (see § 12.1, Supplementary Information).
DG photonic array for pseudomagnetic field. First, we assume 48 DGs along the -direction between two parallel perfect electric conductor (PEC) boundaries, where is normal to the boundaries (see Figure 4a). The array is periodic along and -directions, where and . The boundaries are parallel to and , while the -plane shown in Figure 4b is parallel to and (see Figure S4 in § 13, Supplementary Information). Thus, the boundaries and -plane are neither parallel nor perpendicular. The Weyl points and on the -plane is partially conserved by projecting it onto the boundary, as shown in Figure 4b. Then, the perturbation strength is given by the position-dependent form, i.e., . It linearly varies with the distance from the boundary at along the , i.e., , and it equals at the midplane between the boundaries, as schematically illustrated in the inset of Figure 4a. The DGs in the primitive cells in Figure 4a exhibit different shapes. Especially, the yellow parts show a stronger defect-like shape with larger .

(a)

(b)
Figure S4. Berry phase around the equifrequency Weyl points (a) and (b) shown in Figure 1e in the main text. The Chern numbers of and are and , respectively.
To calculate Figure 5 and Figure 6, we use where . The details not mentioned here are the same as the explanations in § 11, Supplementary Information.
Pseudomagnetic field by DG photonic array. Let us assume that the effective Hamiltonian around a Weyl point is expressed as where is the Weyl points locations ( or ), is the velocity tensor, and are the Pauli matrices 13,15.
can be decomposed into where the superscripts N or H of all terms are omitted. Only the last term relies on the real space coordinate-dependent , i.e., . From the information in Figure 4a-Figure 4b, the resulting pseudomagnetic field is, therefore, written as
where and are the unit vectors of and , respectively, placed on the boundary (see Figure 4a). The components and are determined by the trajectories in Figure 3d-Figure 3e. Meanwhile, the proportional constant is the gradient of geometrical non-uniformity and is defined by the ratio of the change to the distance between these two boundaries (see the inset in Figure 4a). (Detailed derivations and explanations of this result are given in § 13, Supplementary Information.) Because the length between the two boundaries can be written with the lattice constant , the proportional constant can be expressed in terms of . The pseudomagnetic field in equation (5) has the linear combination form of and , parallel to the boundaries, so the field is parallel to the boundaries, as marked in Figure 4c-Figure 4d. In other words, the common perpendicular direction of and coincides with the surface normal to the boundaries. As a result, we can quantitate the overall Hall effect driven by the pseudomagnetic field and the resultant wave localization around boundaries.
Evasion of the photonic wave. Like the other photonic band structure calculations in this study, Figure 7a-Figure 7b is calculated using the array that consists of DG primitive cells, as illustrated in Figure 9. The perturbation strength slope is counted by , e.g., . Figure 7d-Figure 7g are calculated using the array that consists of big cells whose lattice vectors are , , and , where , , and , respectively. We regard that the one big cell coincides with two DG primitive cells because of the projection of and onto the plane of is half of and , respectively. Thus, the array marked with and in Figure 7c consists of 16 and 4 big cells along and -directions, respectively. Along with , only one cell layer was used with periodic boundary conditions. To get Figure 7d-Figure 7g, the ‘Frequency Domain’ solver was used with an input frequency of , marked in Figure 7a and Figure 7b in the main text.

Figure 9. DG primitive cells array that was input to get Figure 7a and Figure 7b in the main text. The array consists of 8 primitive cells with spatially linearly decreasing or increasing along -direction.
Simulation details. All photonic structure simulations were performed using COMSOL Multiphysics{sup}®`. To input periodicity, the Floquet periodic boundary condition was imposed on all periodic boundaries. All band structures were obtained using the ‘Eigenfrequency’ solver.
Data availability
The datasets generated during this study are available from the corresponding author on reasonable request.
Code availability
The custom codes used in this study are available from the corresponding author on reasonable request.
Acknowledgement
This work was supported by Samsung Research Funding & Incubation Center for Future Technology of Samsung Electronics (Project Number SRFC-MA1801-04). We also acknowledge the support of the European Regional Development Fund (ERDF) via the Welsh Government (80762-CU145 (East)). H.P. acknowledges the research professor fellowship, supported from by a Korea University grant.
Author contributions
H.P. and S.L. conceived this topic. H.P. discovered the change of Weyl points’ positions with . H.P. also predicted and simulated the existence of pseudomagnetic fields, asymmetric localization, and the evasion behavior of the photonic wave. H.P. and S.S.O. simulated the surface potential with the Weyl equation. S.S.O. reviewed all theoretical formulations and simulations. S.L. supervised all the work. All authors contributed to a discussion of the data and the manuscript.
Competing interests
The authors declare that they have no competing interests.
Additional information
Correspondence and requests for materials should be addressed to S.S.O. and S.L.
Supplementary Information for “Surface potential-driven surface states in 3D topological photonic crystals”
Haedong Park; Sang Soon Oh; Seungwoo Lee
*Email: [email protected], [email protected]
Photonic Landau levels and photonic wave localization
Applying the design in Figure 4a in the main text to the array of 48 primitive cells generates Landau levels and Landau plateau near (Figure S1). The perturbation strength linearly increases from to . We display all these Landau spectrums along with the - (Figure S1-Figure S1) and -directions (Figure S1-Figure S1) ( and are marked in the inset in Figure S1). The width of the Landau plateau is a scale of the magnetic lengths, and , and marked in Figure S1 and d, respectively. We show the results only on positive - and -directions because all eigenfrequencies and eigenstates are even at the -point, due to time-reversal symmetry. We investigate the eigenstates along the zeroth Landau level (red curves in Figure S1 and d) to see surface states localized on the boundary. The process of collecting eigenstates is as follows: first, for a given , we gather the average normal values of magnetic field eigenstates in each unit cell. These values are normalized with the maximum among these. We then collect all these sets for the along the - and -axes, as shown in Figure S1 and e, respectively. Localization aspects vary significantly by the or . For and , the eigenstates are localized on the lower boundary (which corresponds to the leftmost boundary in Figure 4a in the main text). The confinement length on the lower boundary is a scale of the magnetic length. Around and , eigenstates vary dramatically with increasing . For and , we observe the eigenstates localized on the upper boundary (which corresponds to the rightmost boundary in Figure 4a in the main text). These results show that and , the Weyl points with , are the phase transition points. We carried out the same computations with several values of . The zeroth Landau level flattens over a larger momentum interval as the pseudomagnetic field increases (Figure S1 and f).

Figure S1. Photonic Landau levels and eigenstates in a pseudomagnetic field by a spatial gradient of . a and d, Photonic band structures along - (a) and -directions (d), which exhibit Landau plateaus at around and , respectively. Here, is used where . b and e, Normalized eigenstates along the zeroth Landau levels (red curves in a and d, respectively) to see surface states localized on the boundary. c and f, Zeroth Landau levels according to several magnitudes of pseudomagnetic fields. The and -directions are marked in the inset of d.
Surface states by Weyl equation
Based on equation (3) in the main text, we give detailed derivation of the formula exhibiting surface states. For an orthogonal coordinate system by , , and , we consider a system periodic along with - and -directions, and finite along -direction, as mentioned in the main text. Thus, between the surface boundaries parallel to - and -directions, there are cells along -direction. One boundary’s coordinate is considered as , and the other is , where is the -directional lattice constant. A point in th grid from the first boundary is discretized as . The traveling wave solution can be written as
where . Substituting this solution in the equation (3) in the main text gives
where is the Weyl point’s location, which can be decomposed as . Thus, depends on the grid index . The scalar coefficient of the surface potential also varies with . For , is numerically converted as
Thus, equation (7) becomes
where
and
Here, has a subscript due to varies with . Around the boundaries i.e., and , we use following numerical differentials:
and
By substituting equations (S7) and (S8) into (S2), we have
and
respectively. Equations (S4), (S9), and (S10) are summarized as follows:
For a system with , we simply use and . We set the constant term of the Weyl point’s location as . The varying term is written as , where is the Weyl point’s shift direction with varying , and is the distance between the midplane and the grid.
We first investigate the surface localization when the surface potential is not considered (i.e., is zero for all the grids). When and were respectively used (), the zeroth Landau levels shown in Figure S2-Figure S2 are identical. We collect the (normalized) eigenstates intensities along with the zeroth Landau levels to observe surface states localized on the boundaries. As shown in Figure S2-Figure S2, the eigenstate amplitudes are mutually symmetric.

Figure S2. Symmetric localization of eigenstates by Weyl equation when pseudomagnetic field is applied. a, b, Band structures and the zeroth Landau levels calculated by equation (16). Each case corresponds to and , respectively, where . passes the projected Weyl point’s location. c, d, Normalized eigenstates along the zeroth Landau levels shown in a and b to see surface states localized on the boundaries. e, f, Comparisons of the localized wave intensities along the vertical lines marked in c and d. Data with the same symbols are overlapped in the same plot. Gray dotted lines are the symmetric curves with respect to for comparisons of the red and blue plots.
On the contrary, when we consider the , the mutual symmetric distributions of wave intensities for the two cases are broken. We assume that is non-zero only on the first/last one or two grids around the boundaries, as shown in Figure 8 in the main text; , , and . Although several studies[1, 2, 3, 4, 5] have used the Dirac delta function or a similar function as , we impose finite values as we consider the finite and discretized system. The overall shape of is similar to Figure 1d in the main text. The resulting surface localization is biased towards a specific boundary, as shown in Figure S3.

Figure S3. Asymmetric localization of eigenstates by Weyl equation when pseudomagnetic field and surface potential are applied. The explanations for Figure S2 can be used again for all the panels. The main point in here is that the wave intensities are biased towards the boundary due to the application of the surface potential.
Berry phase of Weyl points at equifrequency
We introduced a double gyroid (DG) photonic crystal in Figure 3 in the main text. The DG exhibits four point-degeneracies marked as and in Figure 3d in the main text. To see if these are Weyl points, we calculate the Berry phase using the Wilson Loop method[6, 7, 8, 9, 10]. The Berry phases of the lower and upper bands connected to point exhibit decreasing and increasing by , respectively (see Figure S4). Therefore, its Chern number is . Likewise, the lower and upper bands connected to point reveals increasing and decreasing by , respectively (see Figure S4). Therefore, its Chern number is . All these are nonzero Chern numbers. Thus, these points are Weyl points, and all surface states in this study are topologically nontrivial by the Weyl points.
Derivation of pseudomagnetic field in DG array
Here, we use , , and . The normal vector of the perfect electric conductor (PEC) boundary at shown in Figure 4a in the main text is . We assume that this plane passes the origin.
Therefore, the distance between a point and this plane is or . The perturbation strength , plotted in the inset of Figure 2c in the main text, is given by
where is a proportional constant, and is a distance between the planes at and , marked in the inset of Figure 4a in the main text. The constant is a central value of (see the inset of Figure 4a in the main text) or a perturbation strength used for the default photonic crystal to generate Figure 3a-Figure 3c in the main text. If the number of unit cells between these boundaries is , and the perturbation strengths at these boundaries has been determined, the proportional constant is calculated as follows:
In other words, this quantity reflects the design of the DG array, such as the number of unit cells and the gradient of the perturbation strength.
As shown in Figure 3d in the main text, Weyl points are confined to -plane, irrespective of . The normal direction of this plane coincides with the normal direction of the plane spanned by and (see Figure S5). If we transform the coordinate system by

Figure S5. DG primitive cells array that was input to get Figure 4-Figure 6 in the main text. The array consists of cells with spatially linearly changing along -direction. Several directions significantly dealt with in this study are also marked.
or
the directions of and become equivalent to and , respectively, so that we can easily write a Hamiltonian for this system. (The relation between the directions of , , and are marked in Figure S5.) The with respect to the new coordinate system is given by
The effective Hamiltonian around a Weyl point is expressed as [9, 11]
where is the anisotropic velocity tensor, is the wave vector, and is the Pauli matrices. is or . Here, the superscripts N or H of all terms are omitted. If the Weyl points behave like Figure 3d in the main text, can be decomposed into a constant term, , and a varying term, , i.e., . For the Weyl points marked in Figure 3d in the main text, the vector potential and pseudomagnetic field for the new coordinate system are respectively given by
and
where is a constant term. Proportional constants and are obtained from the traces in Figure 3d-Figure 3e in the main text. Thus, unlike in equation (S13), the constants and do not reflect the design of the array, but they are about the nature of a three-dimensional DG Weyl photonic crystal itself. The pseudomagnetic field with respect to the original coordinate system is written as follows:
where and are respectively
and and are the unit vectors of and , respectively, placed on the boundary (see Figure 4a in the main text). The number of cells and the difference of at both boundaries are reflected only in , as mentioned in equation (S13). Therefore, the direction of the pseudomagnetic field is always fixed, regardless of the system scale and the gradient of .
The values of each variable in the above derivations are summarized in Table S1. All these values are about the pseudomagnetic fields derived by Weyl points around and for . Both pseudomagnetic fields at the antipodes of and are and , respectively, due to the signs of and are flipped.
| Information of pseudomagnetic fields derived by Weyl points around and for 48 cells. All these are calculated by . The magnetic length is given by . |
| (in -coordinate) |
| (in -coordinate) |
| (in -coordinate) |
| (in -coordinate) |
Table S1. Information of pseudomagnetic fields derived by Weyl points around and for 48 cells. All these are calculated by . The magnetic length is given by .
References
- [1]Potter, A. C., Kimchi, I. & Vishwanath, A. Quantum oscillations from surface Fermi arcs in Weyl and Dirac semimetals. Nature Communications 5, 5161 (2014).
- [2]Bhowmick, S. & Shenoy, V. B. Weber-Fechner type nonlinear behavior in zigzag edge graphene nanoribbons. Physical Review B 82, 155448 (2010).DOI
- [3]Shtanko, O. & Levitov, L. Robustness and universality of surface states in Dirac materials. Proceedings of the National Academy of Sciences 115, 5908-5913 (2018).DOI
- [4]Dou, Z. et al. Imaging Bulk and Edge Transport near the Dirac Point in Graphene Moiré Superlattices. Nano Letters 18, 2530-2537 (2018).
- [5]Dongre, N. K. & Roychowdhury, K. Effects of surface potentials on Goos-Haenchen and Imbert-Fedorov shifts in Weyl semimetals. arXiv preprint arXiv:2106.04573 (2021).arxiv.org/abs/2106.04573
- [6]Park, H. & Lee, S. Double Gyroids for Frequency-Isolated Weyl Points in the Visible Regime and Interference Lithographic Design. ACS Photonics 7, 1577-1585 (2020).DOI
- [7]Yu, R., Qi, X. L., Bernevig, A., Fang, Z. & Dai, X. Equivalent expression of Z2 topological invariant for band insulators using the non-Abelian Berry connection. Physical Review B 84, 075119 (2011).DOI
- [8]Wang, Q., Xiao, M., Liu, H., Zhu, S. & Chan, C. T. Optical Interface States Protected by Synthetic Weyl Points. Physical Review X 7, 031032 (2017).arxiv.org/abs/1708.05664
- [9]Soluyanov, A. A. et al. Type-II Weyl semimetals. Nature 527, 495 (2015).DOI
- [10]Yang, Z. et al. Weyl points in a magnetic tetrahedral photonic crystal. Opt. Express 25, 15772-15777 (2017).
- [11]Jia, H. et al. Observation of chiral zero mode in inhomogeneous three-dimensional Weyl metamaterials. Science 363, 148-151 (2019).DOI