Non-Hermitian funneling in anisotropic media

Yuan Tian1, *, Nankun Gao1, *, Xiujuan Zhang1, †, Ming-Hui Lu1, 2, 3, ‡, and Yan-Feng
Chen1, 3, §
1State Key Laboratory of Solid State Microstructures and Department of Materials Science and Engineering,
Nanjing University, Nanjing 210093, China
2Jiangsu Key Laboratory of Artificial Functional Materials, Nanjing 210093, China
3Collaborative Innovation Center of Advanced Microstructures, Nanjing University, Nanjing 210093, China

*These authors contributed equally.

†‡§Email: xiujuanzhang@[Link]; luminghui@[Link]; yfchen@[Link]

Abstract

Non-Hermitian skin effect (NHSE) has emerged as a distinctive phenomenon

enabling non-Bloch wave manipulation. However, it has been limited to discrete lattices
requiring fine-tuned onsite gain/loss or asymmetric couplings. Here, moving beyond
these discrete models, we realize novel NHSE in uniform media by leveraging
anisotropy of non-Hermitian density tensors. Experiments based on an acoustic
anisotropic metamaterial demonstrate that enabled by the NHSE, wave energy can be
directed toward and collected at specific boundaries, exhibiting broadband and wide-
angle characteristics. This intriguing phenomenon is termed non-Hermitian wave
funneling, which, remarkably, occurs under uniform non-Hermitian modulations, free
of fine-tuning. Furthermore, we identify a second-order NHSE, enabling wave
funneling toward corners. Our work establishes a paradigm for exploring NHSE in
uniform media, advancing the fundamental understanding of non-Hermitian physics
and providing novel mechanisms for non-Bloch wave control in metamaterials or even
natural materials without delicate tuning.

Introduction

Quantum mechanics, traditionally built on Hermitian Hamiltonians for closed

systems, faces challenges in describing real-world open systems that exchange energy
with their environment. Non-Hermitian Hamiltonians resolve this by naturally
accommodating gain and loss through complex eigenvalues and eigenstates [1,2],
extending quantum frameworks to nonconservative regimes. This formalism reveals
remarkable phenomena like parity-time (PT) phase transitions and exceptional
points [3–6], enabling unidirectional wave propagation [7], coherent perfect
absorption [8], and enhanced sensitivity [9–11].

Recently, non-Hermiticity has been brought forward to interplay with topological

physics, leading to the discovery of non-Hermitian skin effect (NHSE) [12–20]. It
describes under specific non-Hermitian modulation, a unique point-gap topology
emerges, enabling non-Bloch wave control like unidirectional amplification and
dynamical properties such as self-healing [21], self-acceleration [22] and edge
burst [23,24]. While NHSE has been demonstrated in a variety of systems including
photonics [25,26], quantum walks [27], phononics [28–31], mechanics [32–35] and
electric circuits [36–39], these implementations are confined to tight-binding discrete
lattices with fine-tuned asymmetric couplings or complex gain/loss modulations.

Here, we transcend discrete models by realizing NHSE in uniform media. Inspired

by the theoretical links between anisotropy and non-Hermiticity [40,41], we design an
acoustic metamaterial with subwavelength multilayers, which can be perceived as an
anisotropic uniform medium. Integrated with a non-Hermitian sponge layer, this
medium features an anisotropic complex density tensor that directly facilitates NHSE.
Enabled by this NHSE, wave energy can be funneled along designed paths and localized
at targeted boundaries across broadband frequencies and wide-angle incidences, as
schematically illustrated in Fig. 1. By direct generalization, we can also realize second-
order NHSE that funnels waves to corners. NHSE in uniform media represents a novel
bulk-boundary relation beyond periodicity, opening a new playground for non-
Hermitian physics. The minimal tuning requirements also make real applications highly
feasible, not only in metamaterials well-known for their unprecedented material
parameters, but also in natural materials where uniformity is readily accessible.
 FIG. 1. Schematic of NHSE in anisotropic uniform medium, enabling non-Hermitian

wave funneling along designed paths and localization at targeted boundaries, exhibiting

broadband and wide-angle characteristics.

Theoretical model

We consider a two-dimensional (2D) uniform medium with an anisotropic density

𝜌𝜌𝑥𝑥𝑥𝑥 𝜌𝜌𝑥𝑥𝑦𝑦
tensor 𝜌𝜌 = �𝜌𝜌 𝜌𝜌𝑦𝑦𝑦𝑦 � and bulk modulus 𝐾𝐾. Acoustic wave propagation is governed
𝑦𝑦𝑦𝑦

by the following wave equation [42]

𝜔𝜔 2 𝑝𝑝
𝛻𝛻 ∙ (−𝜌𝜌−1 𝛻𝛻𝛻𝛻) − 𝐾𝐾
= 0, (1)

and dispersion relation

2
𝜔𝜔w/o 2 2
= 𝜂𝜂𝑥𝑥𝑥𝑥 �𝑘𝑘𝑥𝑥 + 𝑞𝑞𝑥𝑥 𝑘𝑘𝑦𝑦 � + 𝜂𝜂𝑦𝑦𝑦𝑦 �𝑘𝑘𝑦𝑦 + 𝑞𝑞𝑦𝑦 𝑘𝑘𝑥𝑥 � − �𝜂𝜂𝑥𝑥𝑥𝑥 𝑞𝑞𝑥𝑥2 𝑘𝑘𝑦𝑦2 + 𝜂𝜂𝑦𝑦𝑦𝑦 𝑞𝑞𝑦𝑦2 𝑘𝑘𝑥𝑥2 �, (2)
𝐾𝐾

where 𝑝𝑝 represents the pressure field, 𝜔𝜔 (𝜔𝜔w/o) the angular frequency, 𝑘𝑘𝑥𝑥 and 𝑘𝑘𝑦𝑦

the wave vectors along the x- and y-directions, respectively. 𝜌𝜌−1 is replaced by a new
notation 𝜂𝜂 defined as

𝜂𝜂𝑥𝑥𝑥𝑥 𝜂𝜂𝑥𝑥𝑦𝑦 1 𝜌𝜌𝑦𝑦𝑦𝑦 −𝜌𝜌𝑥𝑥𝑥𝑥

𝜂𝜂 ≡ 𝜌𝜌−1 = �𝜂𝜂 𝜂𝜂𝑦𝑦𝑦𝑦 � = �
𝜌𝜌𝑥𝑥𝑥𝑥 𝜌𝜌𝑦𝑦𝑦𝑦 −𝜌𝜌𝑥𝑥𝑥𝑥 𝜌𝜌𝑦𝑦𝑦𝑦 −𝜌𝜌𝑦𝑦𝑦𝑦 𝜌𝜌𝑥𝑥𝑥𝑥 �, (3)
𝑦𝑦𝑦𝑦

with material anisotropy characterized by 𝑞𝑞𝑥𝑥 and 𝑞𝑞𝑦𝑦 , obeying

𝜂𝜂𝑥𝑥𝑥𝑥 +𝜂𝜂𝑦𝑦𝑦𝑦 𝜂𝜂𝑥𝑥𝑥𝑥 +𝜂𝜂𝑦𝑦𝑦𝑦

𝑞𝑞𝑥𝑥 = , 𝑞𝑞𝑦𝑦 = . (4)
4𝜂𝜂𝑥𝑥𝑥𝑥 4𝜂𝜂𝑦𝑦𝑦𝑦

Detailed derivations can be found in Supplementary Material [45].

 Equation (2) reveals that the 𝑞𝑞𝑥𝑥 -term (mediated by 𝑘𝑘𝑦𝑦 ) deflects 𝑘𝑘𝑥𝑥 , which creates

an effective gauge field [43], driving asymmetric wave propagation in the x-direction.
Under non-Hermitian modulation, this gauge field becomes complex, inducing non-
Bloch wave dynamics characterized by directional amplification or attenuation. When
encountering boundaries, wave localizations occur, precisely manifesting the NHSE.
Similarly, the 𝑞𝑞𝑦𝑦 -term mediated by 𝑘𝑘𝑥𝑥 leads to non-Bloch wave dynamics in the y-

direction, which, cooperating with the 𝑞𝑞𝑥𝑥 -term, generates higher-order NHSE.

As a hallmark of NHSE, the bulk-boundary correspondence is fundamentally

broken. To demonstrate this, we truncate the medium along the x-direction by hard
boundaries, as illustrated in Fig. 1(a). Its dispersion relation is given as (see
Supplementary Material [45])

𝜋𝜋𝜋𝜋 2
2
𝜔𝜔w/B
= 𝜂𝜂𝑥𝑥𝑥𝑥 � 𝐿𝐿 � + 𝜂𝜂𝑦𝑦𝑦𝑦 𝑘𝑘𝑦𝑦2 − 4𝜂𝜂𝑥𝑥𝑥𝑥 𝑞𝑞𝑥𝑥2 𝑘𝑘𝑦𝑦2 , 𝑛𝑛 = 0, 1, 2, 3, … (5)
𝐾𝐾

Here, 𝜔𝜔w/B denotes eigenfrequencies with boundaries, in contrast to 𝜔𝜔w/o in Eq. (2)

representing eigenfrequencies in the bulk. Figure 1(b) visualizes their difference. 𝜔𝜔w/o

(red-blue) forms open loops for 𝑘𝑘𝑦𝑦 ≠ 0, where forward (𝑘𝑘𝑥𝑥 > 0) and backward (𝑘𝑘𝑥𝑥 <

0 ) waves experience asymmetric amplification/attenuation rates, generating

unidirectional energy flow. Upon encountering boundaries, energy becomes localized,

collapsing 𝜔𝜔w/B (grey) into straight arcs.

We point out that the inconsistency of dispersion relations with and without
boundaries is a smoking-gun signature of NHSE in uniform media, similar to the broken
bulk-boundary correspondence in discrete lattices. But differently, due to homogeneity,
the bulk dispersion loops for uniform media are not closed, posing a curiosity on the
applicability of point-gap topology in such systems. In fact, the literature have
suggested modified winding numbers by tracing open spectral loops, which may
promise novel topological properties [40,41].
 FIG. 2. (a) Sketch of a slice of uniform medium with hard boundaries in the x-direction.

(b) Comparison of dispersions between an infinite medium (𝜔𝜔w/o, red-blue surface) and

the finite medium in (a) (𝜔𝜔w/B , grey surface). The material parameters are set as 𝜌𝜌𝑥𝑥𝑥𝑥 =

𝜌𝜌𝑦𝑦𝑦𝑦 = 3.08 − 0.43𝑖𝑖 kg/m3, 𝜌𝜌𝑥𝑥𝑥𝑥 = 𝜌𝜌𝑦𝑦𝑦𝑦 = 1.12 − 0.27𝑖𝑖 kg/m3, and 𝐾𝐾 = 222 kPa. (c)

Cross-sections of (b) at selected 𝑘𝑘𝑦𝑦 . Arrows trace 𝑘𝑘𝑥𝑥 -trajectory from negative to positive.

(d) Pressure field amplitudes for the first five 𝜔𝜔w/B eigenmodes. (e) An acoustic

multilayer metamaterial implementing the uniform medium. Its feasibility is verified by

simulated (f) 𝜔𝜔w/o-𝜔𝜔w/B dispersions and (g) pressure field amplitudes, which closely

match those of the uniform medium. Geometric parameters: 𝐿𝐿 = 200√2 mm.

The dispersion inconsistency is further quantified for selected cases in Fig. 2(c).

Crucially, 𝑘𝑘𝑦𝑦 governs both the trajectory of 𝜔𝜔w/o and its winding direction (as 𝑘𝑘𝑥𝑥

sweeps from negative to positive). This is directly aligned with the description of
effective gauge field, which is induced by anisotropy but mediated by wave vectors [see
Eq. (2)]. This 𝑘𝑘𝑦𝑦 -dependence enables flexible control of NHSE. For the finite medium

in Fig. 2(a), the pressure field amplitude follows (see Supplementary Material [45]).

𝜋𝜋𝜋𝜋
|𝑝𝑝(𝑥𝑥, 𝑦𝑦)| = 2𝑒𝑒 −Im(2𝑞𝑞𝑥𝑥 𝑘𝑘𝑦𝑦 )𝑥𝑥 𝑒𝑒 Im(𝑘𝑘𝑦𝑦 )𝑦𝑦 �cos � 𝑥𝑥��. (6)
𝐿𝐿

As visualized in Fig. 2(d), Im�𝑞𝑞𝑥𝑥 𝑘𝑘𝑦𝑦 � > 0 produces left-boundary localization, while
Im�𝑞𝑞𝑥𝑥 𝑘𝑘𝑦𝑦 � < 0 yields right-boundary localization. At Im�𝑞𝑞𝑥𝑥 𝑘𝑘𝑦𝑦 � = 0 (where the

effective gauge field vanishes), the Bloch condition is restored, featuring extended
standing waves. Notably, here 𝑘𝑘𝑦𝑦 is taken as purely real for clarity. With complex 𝑘𝑘𝑦𝑦 ,

non-Bloch wave manipulation can be extended to the y-direction, enabling wave

funneling along designed paths and localization at targeted boundaries.

To realize the anisotropic uniform medium, we design an acoustic metamaterial

consisting of alternating layers of air and acoustic rigid materials (e.g., photosensitive
resin, a 3D printable material), as illustrated in Fig. 2(e). This multilayer design exhibits
effective anisotropic mass density (𝜌𝜌∥ , 𝜌𝜌⊥ ), with non-Hermiticity introduced by adding
a uniform layer of sponge (an acoustic lossy material) at the bottom of the metamaterial.
Details of the effective parameter calculations are provided in Supplementary Material
[45].

With anisotropic (𝜌𝜌∥ , 𝜌𝜌⊥ ), complex gauge fields 𝑞𝑞𝑥𝑥 and 𝑞𝑞𝑦𝑦 can be generated by

rotating the metamaterial, yielding the density tensor

𝜌𝜌⊥ cos2 𝜃𝜃 + 𝜌𝜌∥ sin2 𝜃𝜃 (𝜌𝜌⊥ − 𝜌𝜌∥ )cos𝜃𝜃sin𝜃𝜃

𝜌𝜌 = � �, (7)
(𝜌𝜌⊥ − 𝜌𝜌∥ )cos𝜃𝜃sin𝜃𝜃 𝜌𝜌∥ cos2 𝜃𝜃 + 𝜌𝜌⊥ sin2 𝜃𝜃

where 𝜃𝜃 denotes the rotation angle, which controls the strength of the gauge fields. As
an example, we consider 𝜃𝜃 = 45∘ , which presents effective material parameters
consistent with the uniform medium, as verified by matched dispersions shown in Fig.
2(f). Simulated pressure field amplitudes in the metamaterial at different 𝑘𝑘𝑦𝑦 further

confirm the emergence of NHSE [see Fig. 2(g)].

It is important to note that the specific example in Figs. 2(f-g) is used solely to
validate the data in Figs. 2(b-d). In fact, NHSE occurs for broad (even random) material
parameters, as long as Im(𝑞𝑞𝑥𝑥 𝑘𝑘𝑦𝑦 ) ≠ 0 (see Supplementary Material [45]). This is

precisely due to our novel mechanism of generating effective complex gauge fields
using anisotropic material parameters, which significantly relaxes non-Hermitian
modulations compared to the fine-tuned discrete models. In our design, a simple
uniform layer of lossy material is sufficient for the required non-Hermiticity. This
makes the practical applications of NHSE highly feasible, especially for on-chip
integrations, where the ease of gain/loss control can substantially reduce the fabrication
complexity.

NHSE in uniform media allows wave guiding with a very distinctive mechanism.
As Eq. (6) indicates, wave amplitude is governed by non-Bloch wave dynamics in the
x- and y-directions, characterized by the amplification/attenuation rates Im(𝑞𝑞𝑥𝑥 𝑘𝑘𝑦𝑦 ) and
Im(𝑘𝑘𝑦𝑦 ), respectively. By modulating the effective gauge field and the wave vector, both

wave trajectory and localization can be precisely engineered. This guiding principle is
termed non-Hermitian wave funneling. Of the two governing parameters, while 𝑞𝑞𝑥𝑥 is
determined by material anisotropy, 𝑘𝑘𝑦𝑦 relies on excitation conditions. In the preceding
analyses, we focused on the discussions of eigenspectra and eigenmodes with real 𝑘𝑘𝑦𝑦

(for simplicity), yielding complex 𝜔𝜔w/B, as described by Eq. (5). Alternatively, this

eigenvalue problem can be reformulated by imposing real 𝜔𝜔w/B, enabling complex 𝑘𝑘𝑦𝑦

as

𝜔𝜔2
w/B 𝜋𝜋𝜋𝜋 2
−𝜂𝜂𝑥𝑥𝑥𝑥 � �
𝑘𝑘𝑦𝑦 = ±� 𝐾𝐾 𝐿𝐿
2 , 𝑛𝑛 = 0, 1, 2, 3, … (8)
𝜂𝜂𝑦𝑦𝑦𝑦 −4𝜂𝜂𝑥𝑥𝑥𝑥 𝑞𝑞𝑥𝑥

We emphasize that both real and complex 𝑘𝑘𝑦𝑦 are eigenvalues of Eq. (5), manifesting
the NHSE. Crucially, the complex 𝑘𝑘𝑦𝑦 not only regulates the effective gauge field along

the x-direction via Im�𝑞𝑞𝑥𝑥 𝑘𝑘𝑦𝑦 �, but also contributes to the non-Bloch wave dynamics

along the y-direction (see more discussions in Supplementary Material [45]). Moreover,

complex 𝑘𝑘𝑦𝑦 corresponds to real 𝜔𝜔w/B , making it more suitable for wave control

applications where excitations typically operate at real frequencies.

Experimental realization

For demonstration, we fabricate an acoustic metamaterial sample as shown in Fig.

3(a). A point source with real excitation frequency is placed at the bottom of the sample

to launch upward-propagating waves, corresponding to Im�𝑞𝑞𝑥𝑥 𝑘𝑘𝑦𝑦 � > 0. Figure 3(b)

presents the experimentally measured wave trajectory (see set-up and measurement
details in Supplementary Material [45]). It shows that despite being semi-directional,
wave energy from the point source is entirely funneled leftward and eventually
localized at the left-boundary. Conversely, when the source is placed at top of the

metamaterial, corresponding to downward propagation with Im�𝑞𝑞𝑥𝑥 𝑘𝑘𝑦𝑦 � < 0 , wave

funneling occurs toward the right boundary [Fig. 3(c)]. These observations are well
aligned with above theoretical analyses. Notably, the point source excites a wide range
of wave vector components, revealing collective funneling behavior. By tuning
individual incident angle or excitation location, the funneling path can be customized
(see Supplementary Material [45]). Beyond wide-angle operation, the non-Hermitian
wave funneling persists under broadband excitation, as demonstrated in Fig. 3(d).

FIG. 3. (a) Fabricated metamaterial sample demonstrating the non-Hermitian wave

funneling, where upward-propagating waves from a point source (blue) are guided toward

the left-boundary, while their downward counterparts (green) are guided toward the right-

boundary. Experimental validation is presented in (b) for leftward and in (c) for rightward

funneling. Measured wave intensities are normalized at each y-scanning step, with cut-

line plots highlighting the intensity evolution along the funneling paths. (d) Cut-line plots

for leftward funneling under broadband excitation (see full intensity distributions in

Supplementary Material [45]).

 By direct generalization, we next demonstrate the realization of higher-order
NHSE. Consider a piece of anisotropic medium with hard boundaries in both the x- and
y-directions. As illustrated in Fig. 4(a), the effective gauge fields 𝑞𝑞𝑥𝑥 and 𝑞𝑞𝑦𝑦 jointly

manifest, inducing non-Bloch wave localization at corners, known as the corner skin
modes, which hallmark the higher-order NHSE. To reveal the NHSE, we numerically
calculate the eigenspectra for the finite medium (using identical material parameters to
Fig. 2). The results are presented in Fig. 4(b) (left panel). For comparison, spectra
without boundaries are projected onto the Re(𝜔𝜔)-Im(𝜔𝜔) plane. The observed spectral
inconsistency with versus without boundaries unambiguously confirms NHSE. Further
eigenmode analysis in Fig. 4(b) (right panel) verifies wave localization at corners.
Owing to inversion symmetry of the density tensor [Eq. (7)], the corner skin modes
emerge simultaneously at either upper-left/lower-right corners (along the 𝜌𝜌∥ direction,
modes #1 and 2) or lower-left/upper-right corners (along the 𝜌𝜌⊥ direction, modes #3
and 4). Notably, as a result of the complex gauge fields, modes #1 and 2 exhibit lower
decay rates than modes #3 and 4. This is because acoustic waves preferentially
propagate through air layers along the 𝜌𝜌∥ direction. Conversely, along the 𝜌𝜌⊥
direction, rigid resin layers block waves and divert them through the sponge layer,
increasing the decay rates. Additionally, modes #1 and 2 are closer to the real axis in
the complex frequency spectrum, and thus couple more efficiently to real-frequency
excitations [44].

Asymmetric decay rate as a key signature of the higher-order NSHE enables non-
Hermitian wave funneling to specific corners. To demonstrate this, we fabricate a
metamaterial sample with hard boundaries in both the x- and y-directions [see Fig. 4(c)].
According to the eigenmode analysis, wave energy from a point source will be guided
and collected at the upper-left and lower-right corners along the 𝜌𝜌∥ direction due to
the low decay rate, as indicated by the arrow trajectories. This is indeed confirmed
experimentally in Fig. 4(d), revealing a broadband corner funneling. This unique wave
control mechanism, arising from the synergy between material anisotropy and non-
Hermiticity, offers exceptional flexibility compared to the fine-tuned discrete lattice
models. Beyond tolerance to wide incident angles, broadband frequencies, and relaxed
material constraints, this mechanism accommodates diverse boundary types and shapes,
free from lattice restrictions (see Supplementary Material [45]).

FIG. 4. (a) Illustration of higher-order NHSE in anisotropic medium with hard boundaries

in both the x- and y-directions. Corner skin modes emerge from the combined action of

gauge fields 𝑞𝑞𝑥𝑥 and 𝑞𝑞𝑦𝑦 . (b) (Left) Eigenspectral inconsistency with (color-coded

scatterers) and without (grey shading) boundaries confirms the NHSE. The color indicates

the summed intensity of the upper-left and lower-right corners (see calculation details in

Supplementary Material [45]). (Right) Intensity distributions for selected eigenmodes,

revealing two types of corner skin modes: modes #1/2 (along the 𝜌𝜌∥ direction) with low

decay rates and modes #3/4 (along the 𝜌𝜌⊥ direction) with high decay rates. Such

asymmetric decay rates enable wave funneling to the upper-left and lower-right corners.

(c) Fabricated metamaterial sample demonstrating the corner funneling. (d)

Experimentally measured funneling paths under point-source excitation at 6, 8, and 10

kHz. The intensity is normalized along the radial direction from the source.

Conclusions

In conclusion, we have proposed a novel mechanism for NHSE based on

anisotropic media with uniform non-Hermitian modulations. This approach goes
beyond previous reliance on fine-tuned discrete models and enables a unique non-
Hermitian wave funneling, which can guide wave energy along designed paths toward
targeted boundaries. As experimental demonstrations, we have implemented wave
funneling using an acoustic multilayer metamaterial, showcasing its broadband and
wide-angle performance. The collaboration between material anisotropy and non-
Hermiticity introduces a new paradigm for NHSE, not only broadening our
understanding of this unique phenomenon, but also providing a novel and practical
mechanism for non-Hermitian wave control. Crucially, by eliminating the need for fine-
tuned non-Hermitian modulations, this mechanism significantly reduces the fabrication
complexity and enhances the feasibility for real-world applications such as energy
harvesting and on-chip wave funneling.

Acknowledgments

We acknowledge support from the National Key R&D Program of China (Grants No.
2023YFA1407700 and No. 2023YFA1406904), the National Natural Science
Foundation of China (Grant No. 12222407) and the Key R&D Program of Jiangsu
Province (Grant No. BK20232015). Y.T. thanks the supports from the China
Postdoctoral Science Foundation (Grant No. 2023M731614 and Grant No.
2023T160298), the Jiangsu Funding Program for Excellent Postdoctoral Talent (Grants
No. 2023ZB114).

Reference

[1] Y. Ashida, Z. Gong, and M. Ueda, Adv. Phys. 69, 249 (2020).
[2] E. J. Bergholtz, J. C. Budich, and F. K. Kunst, Rev. Mod. Phys. 93, 15005
(2021).
[3] R. El-Ganainy, K. G. Makris, M. Khajavikhan, Z. H. Musslimani, S. Rotter,
and D. N. Christodoulides, Nat. Phys. 14, 11 (2018).
[4] K. Ding, C. Fang, and G. Ma, Nat. Rev. Phys. 4, 745 (2022).
[5] L. Feng, R. El-Ganainy, and L. Ge, Nat. Photonics 11, 752 (2017).
[6] M. A. Miri and A. Alù, Science (80-. ). 363, eaar7709 (2019).
[7] Y. Huang, Y. Shen, C. Min, S. Fan, and G. Veronis, Nanophotonics 6, 977
(2017).
[8] C. Wang, W. R. Sweeney, A. D. Stone, and L. Yang, Science (80-. ). 373, 1261
(2021).
[9] H. Hodaei, A. U. Hassan, S. Wittek, H. Garcia-Gracia, R. El-Ganainy, D. N.
Christodoulides, and M. Khajavikhan, Nature 548, 187 (2017).
[10] W. Chen, Ş. K. Özdemir, G. Zhao, J. Wiersig, and L. Yang, Nature 548, 192
(2017).
[11] W. Mao, Z. Fu, Y. Li, F. Li, and L. Yang, Sci. Adv. 10, 14 (2024).
[12] X. Zhang, T. Zhang, M. H. Lu, and Y. F. Chen, Adv. Phys. X 7, (2022).
[13] S. Yao, F. Song, and Z. Wang, Phys. Rev. Lett. 121, 136802 (2018).
[14] J. Zhong, K. Wang, Y. Park, V. Asadchy, C. C. Wojcik, A. Dutt, and S. Fan,
Phys. Rev. B 104, 1 (2021).
[15] S. Yao and Z. Wang, Phys. Rev. Lett. 121, 86803 (2018).
[16] K. Kawabata, N. Okuma, and M. Sato, Phys. Rev. B 101, 195147 (2020).
[17] N. Okuma, K. Kawabata, K. Shiozaki, and M. Sato, Phys. Rev. Lett. 124,
86801 (2020).
[18] F. Song, S. Yao, and Z. Wang, Phys. Rev. Lett. 123, 170401 (2019).
[19] Z. Yang, K. Zhang, C. Fang, and J. Hu, Phys. Rev. Lett. 125, 226402 (2020).
[20] X. Zhang, F. Zangeneh-Nejad, Z. G. Chen, M. H. Lu, and J. Christensen,
Nature 618, 687 (2023).
[21] S. Longhi, Phys. Rev. Lett. 128, 157601 (2022).
[22] S. Longhi, Phys. Rev. B 105, 1 (2022).
[23] J. Zhu, Y. L. Mao, H. Chen, K. X. Yang, L. Li, B. Yang, Z. Da Li, and J. Fan,
Phys. Rev. Lett. 132, 203801 (2024).
[24] L. Xiao, W.-T. Xue, F. Song, Y.-M. Hu, W. Yi, Z. Wang, and P. Xue, Phys.
Rev. Lett. 133, 70801 (2023).
[25] Z. Lin, W. Song, L.-W. Wang, H. Xin, J. Sun, S. Wu, C. Huang, S. Zhu, J.-H.
 Jiang, and T. Li, Phys. Rev. Lett. 133, 73803 (2024).
[26] Y. G. N. Liu, Y. Wei, O. Hemmatyar, G. G. Pyrialakos, P. S. Jung, D. N.
Christodoulides, and M. Khajavikhan, Light Sci. Appl. 11, (2022).
[27] S. Longhi, Light Sci. Appl. 13, (2024).
[28] L. Zhang, Y. Yang, Y. Ge, Y. J. Guan, Q. Chen, Q. Yan, F. Chen, R. Xi, Y. Li,
D. Jia, S. Q. Yuan, H. X. Sun, H. Chen, and B. Zhang, Nat. Commun. 12, 6
(2021).
[29] X. Zhang, Y. Tian, J. Jiang, M. Lu, and Y. Chen, Nat. Commun. 12, 5377
(2021).
[30] Z. Gu, H. Gao, H. Xue, J. Li, Z. Su, and J. Zhu, Nat. Commun. 13, 1 (2022).
[31] Q. Zhou, J. Wu, Z. Pu, J. Lu, X. Huang, W. Deng, M. Ke, and Z. Liu, Nat.
Commun. 14, 1 (2023).
[32] A. Ghatak, M. Brandenbourger, J. Van Wezel, and C. Coulais, Proc. Natl.
Acad. Sci. 117, 29561 (2020).
[33] W. Wang, X. Wang, and G. Ma, Nature 608, 50 (2022).
[34] W. Wang, M. Hu, X. Wang, G. Ma, and K. Ding, Phys. Rev. Lett. 131, 207201
(2023).
[35] W. Wang, M. Hu, X. Wang, G. Ma, and K. Ding, Phys. Rev. Lett. 131, 1
(2023).
[36] T. Hofmann, T. Helbig, F. Schindler, N. Salgo, M. Brzezińska, M. Greiter, T.
Kiessling, D. Wolf, A. Vollhardt, A. Kabaši, C. H. Lee, A. Bilušić, R.
Thomale, and T. Neupert, Phys. Rev. Res. 2, 1 (2020).
[37] T. Helbig, T. Hofmann, S. Imhof, M. Abdelghany, T. Kiessling, L. W.
Molenkamp, C. H. Lee, A. Szameit, M. Greiter, and R. Thomale, Nat. Phys. 16,
747 (2020).
[38] D. Zou, T. Chen, W. He, J. Bao, C. H. Lee, H. Sun, and X. Zhang, Nat.
Commun. 12, 1 (2021).
[39] S. Liu, R. Shao, S. Ma, L. Zhang, O. You, H. Wu, Y. J. Xiang, T. J. Cui, and S.
Zhang, Research 2021, (2021).
[40] T. Yoda, Y. Moritake, K. Takata, K. Yokomizo, S. Murakami, and M. Notomi,
 ArXiv 1 (2023).
[41] K. Yokomizo, T. Yoda, and Y. Ashida, Phys. Rev. B 115115, 1 (2023).
[42] D. Torrent and J. Sánchez-Dehesa, New J. Phys. 10, (2008).
[43] F. Liu and J. Li, Phys. Rev. Lett. 114, 103902 (2015).
[44] S. Bid and H. Schomerus, Phys. Rev. Res. 7, 23062 (2025).
[45] See Supplemental Material for theoretical derivations, simulation methods,
experimental setup, and additional results including effective parameter retrieval, real-
𝜔𝜔 skin modes, and robustness analysis.