A metasurface, composed of a thin layer of subwavelength structures, has a distinct capability to modulate light, particularly in hybridizing various degrees of freedom (DoFs) of light, such as wavelength, polarization, and orbital angular momentum (OAM).1^–6 Recent advancements have significantly expanded the applications of metasurfaces toward the quantum regime, encompassing high-dimensional entangled photon state generation, quantum interference manipulation, quantum imaging, quantum state tomography, and quantum algorithm implementation.7^–21 A pivotal feature of quantum metasurfaces is their hybridization capability, facilitating the precise manipulation of quantum entanglement, which is a cornerstone of quantum information science.22^–24 Specifically, metasurfaces enable effective generation of quantum hybrid entanglement between different DoFs of photon pairs, such as entanglement between polarization and OAM or between polarization and path11^,25^,26 However, the exploration of quantum entanglement involving the spatial holographic field, one of the most complex DoFs of light, has been limited due to the challenging implementation of such entanglement using conventional bulky optical components. Metasurfaces become an ideal candidate to overcome this challenge, making spatial holographic entanglement accessible.27^,28

In this work, we utilize a metasurface and entangled photon pairs (signal and idler photons) to generate quantum holograms based on polarization–hologram entanglement. The metasurface in the signal arm generates two distinct holographic states of the signal photons, which are entangled with two orthogonal polarization states of the idler photons, respectively. This polarization–hologram entanglement enables us to remotely control the quantum holograms of the signal photon by adjusting the idler polarization. Specifically, we demonstrate a selective erasure of signal holographic contents when we insert a polarizer in the idler arm. This erasure results from the interference between the two tailor-made holographic fields generated by the metasurface, with the interference controlled remotely through the idler polarizer. Notably, this process can also be interpreted as a quantum eraser experiment.29^–36 The polarization of the idler photon acts as a path marker, revealing which hologram-path information of the signal photon (and thus its particle nature) leads to the absence of interference between two holograms. On the contrary, inserting a polarizer (eraser) in the idler arm removes this path marker, erasing the path information and revealing the wave nature of the signal photon, which restores interference between the two holographic paths. This erasing action is visualized as the selective erasure of holographic contents through remote control via different idler polarization selections. A notable benefit of utilizing holograms, recognized as high-dimensional spatial states, is the straightforward visualization of the quantum-erasing action. In addition, expanding quantum entanglement to incorporate spatial holograms holds promise for enhancing channel complexity and counterfeiting resilience in quantum communication.37^–39 Furthermore, such a metasurface-integrated approach to investigate quantum holograms marks a notable expansion for a metasurface to explore fundamental quantum concepts—entanglement and quantum erasers. In addition to broadening the practical applications of metasurfaces, this work also enhances our insights into the core principles underlying quantum optics and information processing.

The proposed scheme to generate quantum holograms based on hybrid polarization–hologram entanglement utilizes a polarization-entangled photon pair and a geometric metasurface, as shown in Fig. 1. The polarization state of the entangled photon pair (idler and signal photons) is |ϕ⟩=(|L⟩i|L⟩s−|R⟩i|R⟩s)/2, where the |L⟩ (|R⟩) stands for left (right)-handed circular polarization (LCP/RCP), and the subscript i (s) represents idler (signal) photon. The metasurface in the signal arm generates two polarization-dependent holograms at the observed plane with complex holographic field distribution ψL(r) and ψR(r) under the LCP and RCP light incidence, respectively. These LCP and RCP holograms are assumed to be planar (2D) for simplicity and be scalar in specification, i.e., projected to a particular output polarization: horizontal polarization in this work. These two holograms, as seen in Fig. 1(a), are purposely designed with identical amplitudes, featuring the four letters “HDVA,” but differ in phase, which is specifically designed to enable the erasure of desired elements through interference. Representing the basis holographic states with ket notation as |ψL⟩ and |ψR⟩, the metasurface can be effectively described as a spin–orbit coupling operation, denoted as M^s=|ψL⟩s⟨L|s+|ψR⟩s⟨R|s, of which the detailed operation will be described later. Consequently, the resultant quantum state for the photon pair is |Φ⟩=M^s|ϕ⟩=12(|L⟩i|ψL⟩s−|R⟩i|ψR⟩s),(1)indicating the construction of a polarization–hologram entangled state, referred to as the quantum hologram. We note that this represents a form of hybrid entanglement, involving the entanglement of two different DOFs of photons: polarization and spatial mode. Hybrid entanglement can take various forms, including entanglement between photon-number states and the phase of classical light,40^,41 as well as entanglement between different physical systems.42 In the current work, “hybrid entanglement” refers to the entanglement of polarization and the complex spatial modes (holograms), which are different DOFs of the quantum photons.43^–45 Hereafter, we simply refer to this as polarization–hologram entanglement. To reveal the properties of the quantum hologram, we consider observing the signal arm with and without a polarizer in the idler arm. A direct heralded measurement on the signal photon’s hologram without any polarization selection of idler photon (polarizer removed) reveals the incoherent mixture of the two holographic states as Tri[|Φ⟩⟨Φ|]=(|ψL⟩⟨ψL|+|ψR⟩⟨ψR|)/2, giving an intensity distribution as (|ψL|2+|ψR|2)/2, showing all four letters [Fig. 1(b)]. On the other hand, inserting a polarizer (eraser) in the idler arm to select polarization collapses the signal photon to the corresponding superposition of holographic states. For example, as shown in Fig. 1(b), by projecting the idler photon to the linear polarization state of the angle ϕi: |ξ⟩i=(|L⟩i+ei2ϕi|R⟩i)/2, the signal photon collapses to ⟨ξ|Φ⟩i=(|ψL⟩−e−i2ϕi|ψR⟩)/2, which has an intensity field distribution of |ψL−e−i2ϕiψR|/4. With polarizers along different angles ϕi, the resultant signal holographic field is targeted to illustrate different contents, with the corresponding selected letter removed in the current example.

To construct the quantum hologram, we use a geometric-phase metasurface to generate the LCP and RCP holograms with a common amplitude profile but different phase profiles at the image plane. The phase difference between the two holograms at various locations plays an important role in manipulating the interfered holographic result when a polarizer is inserted in the idler arm. Specifically, we adopt a modified Gerchberg–Saxton (GS) algorithm for designing the transmission phase profile of a metasurface. Without losing generality, an image with the letters “HDVA” is picked as our common hologram amplitude |ψL/R|, and the phase difference between the two holograms, Arg(ψL/ψR), is picked to show four discrete values 0, 3π/2, π, and π/2 for the regions of letters “H,” “D,” “V,” and “A,” respectively, as shown in the left-hand side of Fig. 2(a). Then, the modified GS algorithm takes these as input and yields two-phase masks, φL and φR, such that FT(UeiφL/R)=ψL/R, where FT represents the Fourier transform and U is the uniform source amplitude. The only modification of the GS algorithm is an additional constraint to ensure the phase difference Arg(ψL/ψR) complies with specifications. Up to this point, it seems that we need two metasurfaces to implement these two phase masks as transmission phase profiles. However, if we exploit the fact that these two holograms are for two orthogonal circular polarizations and we only collect the real image at a fixed distance f (called focal length here) from the metasurface, one can actually combine the two phase masks to be implemented by the transmission phase profile, let’s say L to R polarization, for a single geometric metasurface using Arg(tRL)=Arg(eiφLe−i2πλ(r2+f2−f)+e−iφRei2πλ(r2+f2−f)),(2)where the wavelength λ is designed at 810 nm and the focal length f is designed at 1000μm. In this case, the LCP hologram is generated as a real image at the distance f as ψL for LCP incidence by noticing the converging phase profile e−i2πλ(r2+f2−f) is added to it, whereas the RCP hologram is a virtual image by the diverging lens profile ei2πλ(r2+f2−f) and is not collected at the image plane. On the contrary, for RCP incidence, as the metasurface is assumed to be composed of only geometric phase elements having tLR=tRL*, the second term in Eq. (2) is turned into eiφRe−i2πλ(r2+f2−f), meaning a real image ψR at the image plane as the designed RCP hologram, whereas the LCP hologram becomes diverging and is not collected. Such a scheme in designing the transmission phase of the metasurface from the two phase masks is summarized in Fig. 2(b). As a result, the metasurface operation can be effectively described as M^1,s=|R,ψL⟩s⟨L|s+|L,ψR⟩s⟨R|s, where the first slot of the ket represents the output polarization state and the second slot represents the spatial mode as a holographic state. Furthermore, to implement it using geometric phase elements, Fig. 2(c) shows the scanning electron microscope (SEM) image of our fabricated metasurface. Each unit cell with a period of 0.7μm consists of geometric-phase grating with a width of 105 nm and a rotation angle being half of the required transmission phase profile. We note that there is a residual part in the copolarization channel, but it is not captured at the image plane because the converging lens phase profile only applies to the cross-polarization channel. For completeness, we have actually a final horizontal polarizer just after the metasurface; the operation now becomes M^s=|H⟩s⟨H|sM^1,s=(1/2)(|H,ψL⟩s⟨L|s+|H,ψR⟩s⟨R|s), which can be simply written as M^s=|ψL⟩s⟨L|s+|ψR⟩s⟨R|s with global factor and signal polarization state omitted for convenience, coupling the incident light polarization and the generated hologram. Hereafter, we call this the operation of metasurface for simplicity.

Figure 3(a) shows our experimental setup to generate and demonstrate the quantum hologram based on polarization–hologram entanglement. Polarization-entangled photon pairs are first produced through spontaneous parametric downconversion (SPDC) with a 405-nm pump laser (beam in blue color) on a type-II β-barium borate (BBO) crystal. The generated photon pairs are separated with a prism into the idler (upper) and signal (lower) arms. In each arm, a half-wave plate (HWP) with a 45-deg optical axis and a BBO with half the thickness of the main BBO are employed to compensate for the translational and longitudinal walk-off effects.46 The generated state is conveniently expressed in terms of circular polarization as |ϕ⟩=(|L⟩i|L⟩s−|R⟩i|R⟩s)/2. The signal photon is further directed through a 10-m-long fiber for delay and then recollimated to free space. The use of the single-mode fiber serves two purposes. First, it acts as a spatial filter, selecting only those photon pairs traveling in the specific directions that show quantum entanglement. This is important because SPDC inherently produces photons in a range of spatial modes. By focusing on the photon pairs at the intersections of their emission cones, we ensure that the selected pairs are indeed entangled and thus useful for our experiments. Second, the fiber functions as an alignment-free delay line, removing the need for a complex free-space delay setup that would involve multiple carefully aligned optical components for constructing an extended optical path. This reduces the experimental complexity. A similar arrangement is also employed in investigations of two-photon interference.15 A quarter-wave plate (QWP) and an HWP before the metasurface are used to correct the minor polarization change due to the long-fiber transmission. With a lens and 10× objectives for imaging, the incident signal photon interacts with the metasurface and is imaged by a single-photon avalanche diode (SPAD) camera at a distance f and with a horizontal polarizer placed before the camera. This setup results in the final quantum hologram, or the polarization–hologram-entangled state for the photon pair at the dashed line in Fig. 3(a), being expressed as Eq. (1). Next, the property of the quantum hologram can be probed by collapsing the idler photon, which is used to herald the SPAD camera for imaging. The idler photon is detected by a single-photon countering module (SPCM) with or without a polarizer in front of the detector. For the case without the polarizer in the idler arm, Fig. 3(b) shows the heralded hologram of the signal photons, where all four letters are visible. The intensity of the heralded hologram, (|ψL|2+|ψR|2)/2, is obtained by simply adding the intensities for the two hologram channels without any interference. On the other hand, inserting a linear polarizer in the idler arm effectively introduces interference between the two holographic states. Expressing the idler polarization state as |ξ⟩i=(|L⟩i+ei2ϕi|R⟩i)/2, where ϕi represents the idler arm’s linear polarizer angle relative to the horizontal axis, we can observe the heralded holographic state in the signal arm as |ξ⟩s=⟨ξ|Φ⟩i=12(|ψL⟩s−e−i2ϕi|ψR⟩s).(3)

For example, in Fig. 3(c), the heralded hologram result erases the letter “H” when ϕi=0, i.e., a horizontal (H) linear polarizer is inserted. This occurs because the area corresponding to “H” shares the same phase in both holographic paths, as indicated by the red region in Arg(ψL/ψR) in Fig. 2(a). Thus, the heralded holographic state |ψL⟩s−|ψR⟩s is manifested with an erased letter “H” because of destructive interference. Here, for simplicity, the global factor 1/2 is omitted in the presentation. Similarly, with ϕi=π/4 [diagonal (D) polarizer in the idler arm], the heralded hologram in the signal arm becomes |ψL⟩s+i|ψR⟩s. In this case, the letter “D” in the LCP hologram |ψL⟩ exhibits a 3π/2 phase shift over the same area in the RCP hologram |ψR⟩, resulting in the erasure of the letter “D,” as shown in Fig. 3(d). Figures 3(e) and 3(f) display the corresponding results when ϕi is taken as π/2 and 3π/4 with a vertical (V) and an antidiagonal (A) polarizer, respectively, which erases the letters “V” and “A.” Compared with the result obtained without the polarizer [Fig. 3(b)], the results with the polarizer exhibit selective erasure of corresponding letters. We note that we have purposely designed the heralded hologram shown up as selective erasure. We call the polarizer in the idler arm “eraser.” The erased letters show a drop in intensity by −13.8dB on average compared to the results with the eraser off, indicating effective erasure. Conversely, the remaining letters show an average Pearson correlation of 0.64 compared with the heralded hologram without erasure, with an average contrast of 7.5 dB (see details in Sec. 3 in the Supplementary Material). This suggests that despite the selective erasure of certain letters, the remaining holographic content retains sufficient clarity and coherence, facilitating effective interpretation and analysis. Indeed, it is noteworthy that when a specific letter is erased due to total destructive interference, the opposite letter, which signifies the orthogonal polarization to the erased one, experiences enhancement with constructive interference as per the tailor-made design. Meanwhile, the other two letters are expected to display similar amplitudes, indicative of intermediate interference levels. For instance, in Fig. 3(c), the letter “H” is erased, whereas the letter “V” appears brightest, with the letters “D” and “A” exhibiting lower and approximately the same intensity.

Our current setup requires 10 min (6000 frames at 100 ms per frame) to collect one raw hologram image, with an additional 10 min needed for background characterization of the whole set of experiments. These relatively long measurement times are due to several loss channels in our metasurface, including its low transmission efficiency, copolarization transmission that does not carry the desired information, and a design scheme that allows diverging photons to lose power. Although these losses do not affect the fundamental operation of the metasurface—because we use postselection and record only events where both signal and idler photons are detected—their presence nevertheless increases the overall time required to image the quantum holograms. To improve the loss situation from the current transmission efficiency of 75%, several approaches can be implemented: using metasurfaces with titanium dioxide (TiO2) materials can achieve transmission efficiency up to 89%27 and employing both propagation and geometric phases for LCP and RCP hologram generation47^–49 would enable independent control of each polarization state. This independent control eliminates the need to simultaneously generate diverging and converging holograms, therefore minimizing photon losses associated with the diverging lens profile in Eq. (2) and copolarization loss. In this work, we implemented a more straightforward design approach to demonstrate proof-of-concept functionality and facilitate fabrication feasibility.

Interestingly, the above experiment in revealing the quantum hologram can also be interpreted as a quantum eraser now toward the holographic level or called a quantum holographic eraser. Here, the polarization–hologram hybrid entanglement enables the marking of which-hologram-path information of the signal photon by the polarization of idler polarization. Such a which-hologram marker’s presence makes the two terms on the right-hand side of Eq. (1) orthogonal, and thus the detection of the signal photon alone captures the particle nature of the signal photon with a density of state (|ψL⟩⟨ψL|+|ψR⟩⟨ψR|)/2. It is noted that no cross term is in the result, indicating that the interference between |ψL⟩ and |ψR⟩ is destroyed because of the availability of which-hologram information. On the contrary, the measurement of the idler photon with a linear polarizer (eraser) inserted erases the which-hologram information and thus restores the interference. This is interpreted as the wave nature of the signal photon with a superposition state of the two holographic states. Here, we have promoted the meaning of the quantum eraser to the hologram level and have purposely visualized the eraser action of the particle nature of photon as a selective erasure of the holographic content (the letters) via different polarization directions of the eraser.

To further consolidate the existence of the interference between the two holographic paths, we now rotate the polarizer in the signal arm, which was previously fixed horizontally, with and without the polarizer (eraser) insertion in the idler arm. Figure 4(a) shows that with an H eraser in the idler arm, rotating the polarizer in the signal arm results in changing intensities for individual letters due to interference. On the contrary, without eraser insertion, the polarizer rotation in the signal arm would not affect the letters’ intensities (as depicted in Fig. S1 in the Supplementary Material), indicating the absence of an interference effect. Figure 4(b) evaluates the average intensity of individual letters with signal polarizer at different angles when the eraser is on and off. When the eraser is on, the intensities appear as sinusoidal curves (solid lines), indicating the presence of interference between the two holograms. On the other hand, when the eraser is off, the intensities remain constant (dashed lines) because of the lack of interference.

In this case, we need to consider the polarizer in the signal arm. Taking the intermediate state before the signal polarizer |Φ˜⟩=M^1,s|ϕ⟩=(1/2)(|L⟩i|R,ψL⟩s−|R⟩i|L,ψR⟩s), we detect the idler photon to horizontal polarization and project the signal photon to different linear polarizations. The resultant holographic state can be expressed as |ξ˜⟩s=(1/2)(⟨L|s+e−i2ϕs⟨Rs|)i⟨H|Φ˜⟩=(1/(22))(e−i2ϕs|ψL⟩s−|ψR⟩s), where ϕs is the signal polarizer angle. Recalling that the two hologram fields have a spatial-dependent phase relation with each other, which can be expressed as |ψL(r)⟩=eiθ(r)|ψR(r)⟩, where θ(r)=Arg(ψL/ψR) is shown in Fig. 2(a), we can obtain the intensity distribution of the resultant heralded holographic state in the signal arm as I(r)∝|ψ(r)|2|e−i2ϕseiθ(r)−1|2/4=|ψ(r)|2sin2(ϕs−θ(r)/2), where |ψ(r)|=|ψL/R(r)| as shown in Fig. 2(a). We can further identify that there are four distinct regions in θ(r) showing different phases as θH=θ(r∈“H”)=0 (in the region of the “H” letter), and similarly θD=3π/2, θV=π, θA=π/2, where the subscript describes the corresponding regional letter. Consequently, the region with the letter “H” displays the intensity as IH∝sin2(ϕs−θH/2)=sin2(ϕs), as confirmed by the first row in Fig. 4(a) and the blue solid line with a visibility of 80% in Fig. 4(b), increasing from ϕs=0 to π/2 and decreasing afterward. Similarly, the intensity for the letter “D” is obtained as ID∝sin2(ϕs−θD/2)=sin2(ϕs−3π/4), as shown by the second row in Fig. 4(a) and the yellow solid line with a visibility of 73% in Fig. 4(b), reaching its maximum and minimum at ϕs=π/4 and 3π/4. Furthermore, for the letters “V” and “A,” the intensities are IV∝sin2(ϕs−π/2) and IA∝sin2(ϕs−π/4), shown as the third and fourth rows in Fig. 4(a), and evaluated with the green and red solid lines in Fig. 4(b), with visibilities of 62% and 71%, respectively. It is noted that the fitting results show a global shift of around 4.6 deg for the four visibility curves due to the experimental errors from the angle of polarizers. On the other hand, when the eraser is off, rotating the signal polarizer results in the same image as (|ψL|2+|ψR|2)/2, indicating unchanged intensity for all four letters. This is evaluated with four dashed lines in Fig. 4(b), with raw images shown in Fig. S1 in the Supplementary Material. It is worth noting that we have used a GS algorithm modified to design and capture tailor-made interference between two holograms. Alternatively, it is also possible to resort to machine-learning techniques to further improve the quality of the holograms with fewer speckles.50

We have demonstrated the significant potential of metasurfaces in quantum optics by generating and manipulating quantum holograms through polarization–-hologram hybrid entanglement. By employing a metasurface, we created a polarization-hologram entanglement between the holographic state of a signal photon and the polarization state of an idler photon. The properties of the resultant quantum hologram can be revealed via the idler polarization selection, leading to selective erasure of holographic contents through interference in the signal arm. Although our experiments primarily demonstrate quantum holograms with a focus on their visual effects, constructing quantum holograms also opens up opportunities in quantum information processing. This is largely due to the possible high dimensionality of quantum holographic states that can serve as a versatile basis for various protocols.51^,52 For example, in quantum communication, it becomes possible to establish a high-dimensional BB84 quantum key distribution (QKD) protocol using quantum holograms (with implementation details and analysis provided in Sec. 4 in the Supplementary Material). In this case, quantum holograms serve as a high-dimensional basis,53 demonstrating enhanced flexibility for BB84 protocols through their hybrid entanglement of polarization and spatial DOFs. Our estimates show that a four-dimensional QKD scheme can be established using the holograms to achieve a quantum bit error rate of ∼1.50%, well below the 18% security threshold54 and yields a mutual information of 1.86 bits per measurement—surpassing traditional two-dimensional protocols.55 Extension to even higher dimensions can be realized by incorporating both copolarization and cross-polarization conversion channels or by utilizing a metasurface lens array to introduce additional path DOFs.56 Thus, the capability of generating quantum holograms enables a possible pathway to have greater flexibility in constructing the bases for high-dimensional QKD, leading to benefits including enhanced data capacity, improved security thresholds, and more robust QKD implementations in quantum communication while maintaining security even under higher quantum bit error rates. Other than enabling more complex and secure communication channels, these structured photons also exhibit state robustness against perturbative transmission and generate resilient quantum entanglement.57^–59

Another potential application arising from our study is in the realm of anticounterfeiting. Our experimental results also vividly illustrate the quantum eraser concept, demonstrating the interplay between the wave and particle nature of photons. We envision that the wave–particle duality can serve as an additional layer of protection in anticounterfeiting applications. Our metasurface incorporates two layers of security that leverage both the particle-like and wave-like characteristics of the holograms. In the first layer, we produce the intensity hologram without using the eraser. By verifying the generated intensity hologram, which does not depend on polarization, we highlight the particle-like nature, as no interference occurs between the two holograms. Even at this stage, the complexity involved in designing and fabricating such a metasurface provides a baseline level of protection against counterfeiting. In the second layer, we introduce a hidden information channel that counterfeiters are likely to overlook. This additional information is encoded as a relative phase profile between the two holograms, represented as Arg(ψL/ψR). Instead of a simple configuration—such as four square patches with successive π/2 steps, as shown in Fig. 2(a)—one can adopt a fully controlled, more general pattern with varied phase values and region shapes, potentially appearing random. By rotating the polarizer in the idler arm (the eraser), this hidden phase signature is revealed through interference patterns, thereby demonstrating the wave-like nature of the holograms and providing an additional layer of security. Thus, the capability of generating holograms with arbitrary relative-phase information, again from the high spatial dimensionality of the holograms, supports a two-layer anticounterfeiting protocol: first by examining the polarization-independent intensity hologram (particle nature) and then by examining the interfered hologram (wave nature) to uncover the relative-phase profile. Because this profile would appear meaningless or random to a counterfeiter, the dual-layer approach leads to a more secure anticounterfeiting scheme.

Our interpretation of the quantum holographic eraser also further opens new avenues for studying fundamental quantum phenomena, such as entanglement, nonlocality, and the role of information in quantum optics. In a broader context, the integration of metasurfaces in quantum optics paves the way for compact, integrated quantum devices, crucial for the practical implementation of quantum technologies. Our study significantly advances the understanding and application of quantum metasurface holograms, indicating the transformative potential of metasurfaces in quantum optical applications.

Hong Liang received his BS degree from the Kuang Yaming Honors School at Nanjing University in 2020, and his PhD from the Hong Kong University of Science and Technology in 2024 under the supervision of Prof. Jensen Li. He was awarded the Hong Kong PhD Fellowship during his PhD. His research interests include quantum imaging, communication, and entangled state generation with metasurfaces. Hong is passionate about building optical platforms from scratch and is currently working in industry.

Wai Chun Wong received his BS and MPhil degrees from the Hong Kong Polytechnic University and his PhD from the Hong Kong University of Science and Technology. He is currently a postdoctoral research associate under Prof. Jensen Li. His research interests lie in non-Hermitian physics in optics, such as phenomena related to exceptional points and non-reciprocal behavior. He is also interested in extending the study of exceptional points from classical systems to quantum systems.

Tailin An received his BS and MS degrees from Xidian University, China, and University of Birmingham, UK. He is currently a PhD student, supervised by Prof. Jensen Li, in the Physics Department at the Hong Kong University of Science and Technology. His current research focuses on nano fabrication and applications of metasurfaces in quantum optics.

Jensen Li is a professor of Computational Engineering and Metamaterials at the University of Exeter, having relocated from the Hong Kong University of Science and Technology, where he earned his PhD in 2004. His research focuses on multiple wave domains, including electromagnetic and acoustic metamaterials, with a current emphasis on non-Hermitian, time-varying, and quantum optical regimes. He is an elected member of the Hong Kong Young Academy of Sciences, a 2025 Optica fellow, and a recipient of the Croucher Senior Research Fellowship in 2022.