betaFeedback

WorksPhysics and Astronomy

Circulators Based on Coupled Quantum Anomalous Hall Insulators and Resonators

Abstract

Integrated plasmonics is advancing rapidly, enabling a wide range of functionalities to be incorporated onto a single chip. Applications span information processing, computation, quantum sensing, and dark-matter detection. This progress has driven the development of integrated non-reciprocal devices, which are essential for preventing unwanted feedback that can degrade system performance. While non-reciprocal devices have been realized in edge magnetoplasmon materials via classical interference effects, their operation is often limited by the input power range. Here, we demonstrate that topological circulators utilizing asymmetric coupling offer improved input power range, isolation, and insertion loss. In this configuration, we demonstrate the coupling between a chiral edge magnetoplasmonic resonator and a pair of LC resonators is well described by an effective non-Hermitian two-site Hatano-Nelson model with asymmetric directional couplings, resulting in nonreciprocal behavior. The coherent photon-plasmon interaction enables a circulator with up to 50 dB of isolation across a broad range of excitation power. These results suggest that magnetic topological insulators provide a promising platform for realizing asymmetric non-Hermitian couplings at radio frequencies and for exploring regimes of strong directional suppression and possible exceptional-point physics. More broadly, they highlight the potential of topological-material-based microwave devices for future integration with superconducting quantum information platforms.

Introduction

The chiral nature of the edge magnetoplasmon makes it a promising candidate for advancing chip-scale non-reciprocal devices at quantum-classical interfaces [1, 2, 3]. Edge magnetoplasmon (EMP) propagation has been extensively studied in quantum Hall (QH) systems [4, 5, 6]. Their coupling to external devices in the non-Hermitian regime has been investigated, while the possibility of accessing strongly asymmetric transport and exceptional-point-related phenomena remains largely unexplored. Recent studies have demonstrated that non-Hermitian systems, particularly those of the Hatano–Nelson type, offer potential for realizing isolators and switches because their intrinsically asymmetric coupling strengths can induce exceptional points (EPs), leading to highly non-reciprocal transport phenomena [7]. In this context, EMPs can serve as non-reciprocal links that asymmetrically couple radio-frequency (RF) resonators, enabling circulation or isolation behavior.

Compared to quantum Hall (QH) systems, quantum anomalous Hall (QAH) materials offer distinct advantages, such as maintaining a stable number of edge channels across a broad range of applied magnetic fields, including zero magnetic field [8, 9]. Moreover, QAH states can be engineered in multilayer magnetic topological insulators or heterostructures, providing design flexibility that can be tailored to meet specific circuit requirements [10, 11]. This makes them well-suited for integration with superconducting circuits, where magnetic fields can be detrimental to performance, and semiconductor devices, which can tolerate or even require moderate magnetic fields for operation. The QAH insulator, Cr-doped (BixSb1−x)2Te3(\mathrm{Bi}_{x}\mathrm{Sb}_{1-x})_{2}\mathrm{Te}_{3}, provides a promising platform for realizing strong plasmon-photon interactions. This is due to its high-impedance plasmon mode, with an impedance of Z=h/Ce2∼25.8 kΩZ = h/Ce^{2} \sim25.8\ \mathrm{k}\Omega and a Chern number C=1C = 1. Because the zero-point voltage fluctuation scales as Z\sqrt{Z}, this large ZZ markedly amplifies the electric-dipole coupling between the plasmon and nearby quantum emitters, making high-impedance edge magnetoplasmonic resonators ideal for achieving strong asymmetric plasmon-photon coupling and opening up exciting opportunities in non-reciprocal electronics.

In this work, we investigate a coupled plasmon-photon system, in which an on-chip QAH plasmon resonator is asymmetrically coupled to a pair of LC resonators. While Mahoney et al. studied a similar architecture in a QH system, they did not explore the connection to Hatano-Nelson-type couplings [1]. In our hybrid system, we observe a pronounced nonreciprocal transmission dip accompanied by a positive π\pi phase shift. By leveraging the chiral nature of the EMP modes, we interpret the observed behavior as being consistent with an effective two-site Hatano–Nelson–type plasmonic system [7, 12, 13], in which asymmetric coupling may drive the system towards a strongly nonreciprocal non-Hermitian transport regime. Furthermore, the system achieves an average isolation of 20 dB over an approxi-

mately 50 MHz bandwidth and a peak isolation exceeding 50 dB at 543.8 MHz for an input microwave power of −119 dBm. The circulator remains operational down to an input power of −149 dBm. These results suggest a new mechanism for harnessing the intrinsic unidirectionality of QAH edge states to engineer non-reciprocal devices, with promising implications for scaling superconducting quantum processors and improving axion haloscope sensitivity.

Results

Circulator Configuration and Effective Hatano-Nelson-Type Coupling

The configuration of the three-port circulator device is shown in Figure 1a, which consists of an EMP resonator with a radius of 100 μm\mu\mathrm{m} and three lumped element inductors, each with an inductance L=180 nHL = 180\ \mathrm{nH}. The inductor and the stray capacitor CstrayC_{\mathrm{stray}} (Figure 1b) form an LC resonator, whose resonance frequency ωLC=1/LCstray\omega_{\mathrm{LC}} = 1/\sqrt{LC_{\mathrm{stray}}} is set to closely match the EMP resonance frequency ωm=[σxy2πϵ0ϵeff(ln⁡2∣q∣w+1)+vD]q\omega_m = \left[\frac{\sigma_{xy}}{2\pi\epsilon_0\epsilon_{\mathrm{eff}}}\left(\ln\frac{2}{\lvert q\rvert w}+1\right)+v_D\right]q, where qq is the wave vector, σxy=Ce2/h\sigma_{xy} = Ce^2/h the Hall conductivity, ϵeff\epsilon_{\mathrm{eff}} the effective dielectric constant, vDv_D the drift velocity of the 1D edge state, and ww the physical width of the EMP mode at the edge. The value of ωm\omega_m can be directly extracted from experimental characterization of the circulator without requiring an impedance-matching circuit. At each port of the circulator, the EMP resonator is capacitively coupled to an LC resonator through an edge capacitor CedgeC_{\mathrm{edge}}, which is formed between the Au pad and the EMP resonator (Figure 1b). The microwave transmission coefficients S31S_{31} and S32S_{32} are characterized using a measurement scheme described in Appendices A and B. All cryogenic measurements were performed in a dilution refrigerator at a temperature of approximately 10 mK.

Schematic of the three-port circulator device, equivalent circuit diagrams, Hatano–Nelson chain, and circulator configuration

Figure 1. Device and Hatano-Nelson model. a, Schematic of the three-port circulator device, consisting of a QAH mesa with an edge conductance of e2/he^{2}/h, and three 180 nH lumped-element wire-loop inductors are wire-bonded between the mesa and the transmission line of the printed circuit board. Edge magnetoplasmon resonant modes are excited and propagate along the edge of the circular mesa. The inset shows the optical image of the circulator, with a 100-μ\mum-radius QAH mesa separated from the gold pad that defines the three signal ports by a 15 μ\mum gap. b, Equivalent circuit diagrams illustrating parasitic capacitance CpC_{\mathrm{p}}, stray capacitance CstrayC_{\mathrm{stray}}, and edge capacitance CedgeC_{\mathrm{edge}}. Components aa and bb (blue pads) represent the LC resonators at ports 1 and 3, respectively. m1m_{1} and m2m_{2} represent the anticlockwise transport of the EMPs from resonator aa to bb, and from bb to aa, respectively. c, Schematic of a one-dimensional (1D) Hatano-Nelson chain, where asymmetric couplings λab\lambda_{\mathrm{ab}} and λba\lambda_{\mathrm{ba}} between cells give rise to non-reciprocal transport in a non-Hermitian system. d, Sketch of the circulator configuration composed of two LC resonators, aa and bb, coupled non-reciprocally through the interference of EMP and the parasitic capacitive pathways, leading to asymmetric coupling λab\lambda_{\mathrm{ab}} and λba\lambda_{\mathrm{ba}}.

The circulator architecture, consisting of two LC resonators (aa and bb) and the two directional EMP channels, is described by a minimal non-Hermitian coupled-mode matrix that assumes idealized non-reciprocity propagation [7, 14, 15, 16]:

Heffℏ=(ωa−iκag000ωm1−iκm1g000ωb−iκbgg00ωm2−iκm2)(1)\frac{H_{\mathrm{eff}}}{\hbar} = \begin{pmatrix} \omega_a - i\kappa_a & g & 0 & 0 \\ 0 & \omega_{m1} - i\kappa_{m1} & g & 0 \\ 0 & 0 & \omega_b - i\kappa_b & g \\ g & 0 & 0 & \omega_{m2} - i\kappa_{m2} \end{pmatrix} \tag*{(1)}

where κa\kappa_a and κb\kappa_b are the decay rates of the LC resonators aa and bb, respectively. κm1\kappa_{m_1} and κm2\kappa_{m_2} are the decay rates of the EMP resonators along two propagation paths, m1m_1 and m2m_2, as illustrated in Figure 1b. ωa,b,m1,m2\omega_{a,b,m_1,m_2} are the bare resonance frequencies of the LC and EMP modes, respectively, and gg denotes the coupling strength between the LC resonators and the EMP modes. In the frequency-domain coupled-mode treatment (see Appendix § 0.9), the response coefficients are written as αj(ω)=−iω+iωj+κj=κj−i(ω−ωj)\alpha_j(\omega)=-i\omega+i\omega_j+\kappa_j=\kappa_j-i(\omega-\omega_j).

By eliminating the intermediate EMP modes, Eq. (1) can be reduced to an effective two-site Hatano–Nelson-type model (Figure 1c) with unequal direction-dependent couplings between neighboring unit cells (see Appendix § 0.9),

λab(ω)=g2αm1,λba(ω)=g2αm2,(2)\lambda_{ab}(\omega)=\frac{g^2}{\alpha_{m_1}},\qquad\lambda_{ba}(\omega)=\frac{g^2}{\alpha_{m_2}}, \tag*{(2)}

where αm1(ω)\alpha_{m_1}(\omega) and αm2(ω)\alpha_{m_2}(\omega) describe the frequency-domain complex response of the two EMP propagation paths. Because these paths have different lengths and dissipation, αm1≠αm2\alpha_{m_1}\ne\alpha_{m_2} in general, leading to asymmetric effective coupling. In our effective model, the non-reciprocal inter-resonator couplings, λab\lambda_{ab} and λba\lambda_{ba}, arise from destructive and constructive interference between the EMP and parasitic capacitive pathways (Figure 1b and Figure 1d). As a result, they are determined by the EMP propagation lengths and depend sensitively on both the input microwave power and the device operating temperature. Within this effective model, sufficiently strong suppression of the one directional coupling could in principle drive the reduced two-site system towards an EP, marked by eigenvalue and eigenvector coalescence [17].

Circulator Characterization

We begin by examining the transmission in an EMP resonator, both without and with LC resonators, across a wide spectral range. Figure 2a–Figure 2c show the transmission through the EMP resonator, without coupling to an LC resonator. Owing to the constructive and destructive interference between the EMP mode (red traces), coupled via CedgeC_{\mathrm{edge}}, and the port-to-port RF signal transmitted through the parasitic capacitance CpC_p (Figure 1b), S31S_{31} and S32S_{32} exhibit a non-reciprocal response with a bandwidth of approximately 100 MHz (Figure 2 d–f). The isolation magnitude, ∣S32∣−∣S31∣|S_{32}|-|S_{31}|, reaches a maximum of 20 dB over an input-power range spanning approximately 5 dBm.

Transmission responses for circulators without and with coupling to LC resonators

Figure 2. Transmission response for circulators without and with coupling to LC resonators. a–c, Power-dependent microwave transmission responses ∣S31∣|S_{31}|, ∣S32∣|S_{32}|, and ∣S32∣−∣S31∣|S_{32}|-|S_{31}| for a 100 μ\mum-radius circulator without coupling to LC resonators. The signal is excited from ports 1 and 2, and measured at port 3. d–f, Frequency line cuts from panels a and b, shown for three representative input power levels. g–i, Power-dependent microwave transmission responses ∣S31∣|S_{31}|, ∣S32∣|S_{32}|, and ∣S32∣−∣S31∣|S_{32}|-|S_{31}| for a 100 μ\mum-radius circulator coupled to LC resonators. j–l, Frequency line cuts from panels g and h, shown for three representative input power levels.

Figure 2 g–i show the transmission spectra as a function of frequency and input power when the EMP and LC resonators are coupled. Notably, the magnitude of S32S_{32} remains nearly constant over an input power range spanning more than 50 dBm. The circulator exhibits ∼20\sim20 dB of non-reciprocity over an input power range from −150-150 to −90-90 dBm, with an approximate bandwidth of 50 MHz (Figure 2i). This level of isolation and power handling is particularly important for quantum information science, where protecting qubits from amplifier back-action is critical, and for axion dark matter searches like ADMX, which require ultra-low-noise, broadband signal routing at cryogenic temperatures. The ability to achieve strong non-reciprocity at such low input powers and moderate bandwidths supports scalable readout chains and noise-resilient detection architectures in both domains [18, 19].

The spectral response at varying input power levels for the LC-resonator-coupled device is presented in greater detail in Figure 2 j–Figure 2l. At an input power of P=−149P = -149 dBm, S32S_{32} exhibits a Lorentzian-like resonance with the quality factor Q=21Q = 21, while S31S_{31} shows a subtle resonance peak (Figure 2j). Compared to the EMP circulator without coupling to an LC resonator, the LC-resonator-coupled configuration achieves significantly higher transmission, with an insertion loss of approximately 6 dB at the transmission peak (see Appendix § 0.8). This enhancement can be attributed to the impedance matching between the port impedance and that of the QAH edge state. Transmission can be enhanced to near-unity levels through optimized impedance matching and improved circuit fabrication.

The broadband non-reciprocal response in the 450–650 MHz frequency range, observed in both the LC-coupled and uncoupled circulators, arises from interference between the phase-sensitive plasmonic pathway and the parasitic capacitance pathway. For the shorter edge path (port 2 to 3), two LC resonators are capacitively coupled to the EMP resonator, leading to constructive interference with the parasitic transmission. In contrast, for the longer edge path (port 1 to 3), the same coupling occurs; however, the increased path length induces an out-of-phase interaction with the parasitic transmission. This phase mismatch results in destructive interference between the plasmonic and parasitic paths, thereby reducing overall transmission.

For the LC-coupled-circulator, we find that increasing the input microwave power leads to the gradual emergence of a sharp transmission dip near 543.8 MHz, which reaches its minimum at P=−119P = -119 dBm (Figure 2 k and Figure 2l). In addition to the broadband suppression of transmission from port 1 to port 3, this narrow resonance feature produces exceptionally high isolation, exceeding 50 dB within a bandwidth of approximately 5 MHz. In contrast, this pronounced and narrow transmission feature is absent when the circulator operates without coupling to the LC resonators (Figure 2 a–Figure 2f).

Next, we further characterize the non-reciprocal response at f=543.8f = 543.8 MHz and investigate its physical origin. In particular, we examine how the transmission and phase evolution of S31S_{31} depend on the input power. Figure 3 a and Figure 3b show the magnitude and phase of S31S_{31} and S32S_{32}, respectively, over a narrow frequency range. Figure 3 c and Figure 3d illustrate the evolution of the ∣S31∣\lvert S_{31}\rvert and phase difference ∠S31−∠S32\angle S_{31} - \angle S_{32} as a function of input power. A particularly striking observation is that the sharpest transmission contrast appears only within a narrow input-power window (Figure 3c). Within this regime, S31S_{31} undergoes an abrupt phase transition from negative to positive, exhibiting an approximately 180° phase shift relative to S32S_{32} (Figure 3d, red curve). This pronounced nonreciprocal transmission, accompanied by a corresponding π\pi phase shift, is consistent with strongly asymmetric coupling between the two LC resonators, where the effective coupling becomes highly suppressed in one propagation direction [20, 21, 22].

Circulator transmission measurement

Figure 3. Circulator transmission measurement. a, b, Experimental results for the transmission magnitude and phase for S31S_{31} and S32S_{32}. The data were taken at an input power of −120 dBm, corresponding to the estimated RF power at the circulator input ports, where the strongest asymmetric transport behavior is observed. c, d, Measured S31S_{31} and S32S_{32} magnitudes and their relative phase at input power levels below, near, and above the condition of maximum nonreciprocal transmission.

We attribute the emergence of the anomalous transmission dip to an effective, power-dependent phase shift in the EMP propagation pathway, which dynamically modulates the coupling between the two LC resonators. As the input microwave power increases, the EMP mode experiences a shift in both velocity and dissipation, altering the accumulated phase along the edge-transport path. The phase and dissipation shifts with increasing power may arise from the presence of charge puddles in proximity to the edge states. The photon-excited charge puddles could slow edge-plasmon propagation and provide dissipation channels through the bulk states [23]. Because the EMP-mediated coupling and the parasitic capacitive coupling interfere either constructively or destructively depending on their relative phase, even a small power-induced change in the EMP phase can significantly modify the net interaction strength. At low powers, the plasmonic and parasitic pathways partially cancel, producing moderate suppression of transmission. As the power increases toward (P∼−119P \sim-119 dBm), the phase accumulated along the longer EMP path shifts such that the plasmonic signal becomes nearly out of phase with the parasitic contribution. This condition leads to strong destructive interference and substantial suppression of the effective coupling in one propagation direction. The resulting asymmetric transport behavior produces the sharp, unidirectional transmission dip observed in the measurements and is qualitatively consistent with the strongly asymmetric transport regime of the non-Hermitian coupled plasmon-photon system.

Simulation

We now turn to simulations of the transmission magnitude and the associated positive phase shift using coupled-mode theory. The system is modeled by the non-Hermitian coupled-mode matrix in Eq. (1), which assumes ideal non-reciprocity [7, 14, 15, 16]. Given the almost identical resonance frequencies of the input and output LC resonators (Figure 5c), we set ωa=ωb=ω0\omega_a = \omega_b = \omega_0 and κa=κb=κ0\kappa_a = \kappa_b = \kappa_0. In the EMP resonator, asymmetric coupling emerges due to its chiral nature, leading to differences in propagation time and loss for the m1m_1 and m2m_2 modes (despite ωm1=ωm2=ωm\omega_{m_1} = \omega_{m_2} = \omega_m). These differences depend on the path length and the input microwave power. By applying coupled-mode theory to the steady-state driven response of the non-Hermitian four-mode model in Eq. (1), and then eliminating the intermediate EMP modes as described in Appendix § 0.9, the transmission spectrum can be expressed as,

Figure 5. Experimental configuration and the transmission spectra of the device without and with the inductors measured at room temperature. a, Experimental setup showing the circulator mounted on the mixing-chamber plate of a dilution refrigerator, which cools the device to below 10 mK. b, Complete six-way transmission responses of the 100-μ\mum-radius circulator without inductors measured at room temperature. c, The transmission spectra SijS_{ij} for the 100-μ\mum-radius circulator with inductors, showing a frequency response resulting from the LC impedance-matching circuits.

t(ω)=−2κ0αmg2α02αm2−g4(3)t(\omega)=-\frac{\sqrt{2}\kappa_{0}\alpha_{m}g^{2}}{\alpha_{0}^{2}\alpha_{m}^{2}-g^{4}} \tag*{(3)}

where α0≡−iω+iω0+κ0\alpha_{0}\equiv-\mathrm{i}\omega+\mathrm{i}\omega_{0}+\kappa_{0} characterizes the LC resonator and αm≡−iω+iωm+κm\alpha_{m}\equiv-\mathrm{i}\omega+\mathrm{i}\omega_{m}+\kappa_{m} characterizes the EMP resonator. Since the longer propagation length of mode m1m_{1} is approximately twice that of m2m_{2}, we adopt the phenomenological approximation αm1≈2αm2\alpha_{m_{1}}\approx2\alpha_{m_{2}} near the operating point of strongest asymmetry. This compactly captures the path-dependent phase accumulation and dissipation, and gives αm=αm1αmm2≈2αm2\alpha_{m}=\sqrt{\alpha_{m_{1}}\alpha_{m_{m_{2}}}\approx\sqrt{2}\alpha_{m_{2}}}.

We find that the measured transmission spectra are qualitatively in agreement with the transmission response derived from the effective two-site Hatano-Nelson-type model, indicating that the circulation is consistent with a mechanism dominated by asymmetric coupling between the two LC resonators mediated by the EMP mode (Figure 4 a and Figure 4b). At low excitation power, the measured transmission closely matches the model, with fitted parameters: g/2π=5.3g/2\pi= 5.3 MHz, κ0/2π=3.0\kappa_0/2\pi= 3.0 MHz, and κm/2π=1.0\kappa_m/2\pi= 1.0 MHz (Figure 4c). As the input power increases, the fitted coupling strength gg rises to 6.9 MHz, while κ0/2π\kappa_0/2\pi remains at 3.0 MHz and κm/2π\kappa_m/2\pi increases to 3.3 MHz, respectively, at the point of maximum isolation (Figure 4d).

Comparison of data and simulation

Figure 4. Comparison of data and simulation. a, b, Close-up views of the experimental (a) and simulated (b) spectral responses as a function of frequency and input power. The white (orange) dashed line indicates the input power corresponding to the measurements shown in c, d. c, d, Measured ∣S31∣|S_{31}| magnitude (blue dots) as functions of frequency at excitation powers of P=−134P=-134 and −119-119 dBm, respectively. The solid red curves represent fits from the theoretical model in Eq. (3). e, Measured (blue) phase difference between S31S_{31} and S32S_{32} at an excitation power of approximately P=−113P=-113 dBm, and calculated (red) phase response based on Eq. (3) using parameters extracted from the fit presented in d. See text for details.

This effective model also captures the phase behavior observed in the experimental data, particularly in the regime of strongest asymmetric transport, as shown in Figure 4e. It is worth noting that the phase data in Figure 4, a and b are not sampled densely enough to enable a reliable comparison between experiment and simulation. To address this, we took a high-resolution data scan and plotted the phase response of S31S_{31} after subtracting the phase of S32S_{32}, thereby removing the line delay. This allows for a direct comparison with the theoretical result. The residual mismatch between experiment and theory at higher powers is likely due to power-dependent plasmon damping and coupling effects that lie outside the minimal model.

Conclusion

In this work, we demonstrated a novel circulation mechanism in a topological circulator comprising a QAH edge magnetoplasmonic resonator coupled to LC resonators. Within an input power range of −150-150 to −120-120 dBm, the circulator maintains a relatively stable bandwidth, passband center frequency, and isolation, indicating robust performance under low-power operating conditions. This power independence suggests that the device is well-suited for integration into superconducting quantum information science platforms and high-energy physics experiments. If fully harnessed, this mechanism could facilitate the realization of a topological circulator with near-unitary forward transmission and exceptionally high isolation in the reverse direction.

In the present device geometry, the operating frequency is primarily determined by the edge magnetoplasmon velocity and the dimensions of the topological insulator mesa. The demonstrated operation near 0.5 GHz serves as a proof of principle for realizing effective non-Hermitian coupling and nonreciprocal microwave transport using magnetic topological insulators. Furthermore, we note that compact on-chip nonreciprocal devices operating in the sub-GHz frequency range may find applications in superconducting quantum circuits that employ low-frequency signals for flux control, biasing, and tunable coupling. In particular, the frequency range demonstrated here overlaps with that of fluxonium-based architectures, where control-line-limited relaxation and flux-controlled interactions have recently attracted significant interest [24, 25, 26, 27]. Looking forward, the operating frequency can be extended to higher microwave frequencies through device scaling and QAH material engineering of the edge plasmon properties.

In conclusion, our results highlight the significant potential of chiral plasmons to provide enhanced control over signal transmission and isolation across a wide dynamic range.

Acknowledgments

We would like to thank Yaniv Rosen, Kristin Beck, and Enrico Rossi for helpful discussions. This work was performed under the auspices of the US Department of Energy by Lawrence Livermore National Laboratory under Contract No. DE-AC52-07NA27344. The project was supported by the Laboratory Directed Research and Development (LDRD) programs of LLNL (21-ERD-014 and 26-ERD-010) and HEP Laboratory Detector Research and Development program. The authors also acknowledge support from NSF Convergence Accelerator program under Grant 2040737.

Experimental configuration and device fabrication

The transmission of the circulator is probed by measuring the S21S_{21} transmission parameters for various powers. The experimental configuration is plotted in Figure 5a. The S21S_{21} measurements were performed on two ports of the circulator with an RF switch located at room temperature. The input microwave signals are transmitted through coaxial cables with a 20 dB attenuator at room temperature (not shown in Figure 5a) and a 20 dB attenuator at 4 K. The total attenuation from the vector network analyzer down to the sample is about −64 dB. The power values expressed in decibels in this letter represent the estimated power at the input port of the circulator. The circulator output goes through a cryogenic isolator at the still stage and is amplified by a HEMT amplifier (LNF-LNC0.3-14A, Low Noise Factory) with a noise temperature of ∼2\sim2 K at the 4 K stage. At room temperature, the transmission signal is further amplified by a low-noise room temperature amplifier with a total gain of 20 dB (ZX60-4016E-S+, Mini-Circuits).

Topological insulators are materials with an insulating bulk and topologically protected conducting surface or edge states arising from strong spin–orbit coupling and a nontrivial electronic band topology [28, 29, 30]. Magnetic topological insulators are topological insulators in which magnetic order breaks time-reversal symmetry, leading to gapped surface states and enabling exotic phenomena such as the QAH effect. The 6 quintuple layers of magnetic topological insulator Cr0.12(Bi0.26Sb0.62)2Te3\mathrm{Cr}_{0.12}(\mathrm{Bi}_{0.26}\mathrm{Sb}_{0.62})_{2}\mathrm{Te}_{3} thin films were grown using ultrahigh vacuum molecular beam epitaxy. Epi-ready semi-insulating (ρ>106 Ω cm\rho> 10^{6}\ \Omega\ \mathrm{cm}) GaAs (111) B substrates were pre-annealed in the chamber at up to 600 °C under Te-rich environment. During the co-evaporated growth, high-purity Bi (99.9999%), Te (99.9999%), Cr (99.99%) were evaporated from Knudsen effusion cells while Sb (99.999%) was evaporated from a cracker cell. Upon material growth, the circulator devices were fabricated via standard photolithography. The mesa disk was first patterned by photolithography and then etched by CF4\mathrm{CF}_{4}/Ar using reactive ion etching. The coplanar waveguides and ground planes were patterned by photolithography, and 10 nm of Ti and 100 nm of Au were deposited using electron beam deposition, followed by a liftoff process. The Hall characterization of the same batch of magnetic topological insulator wafer can be found in reference [31]. All measurements presented in this work were performed under an external magnetic field of approximately 40 mT, generated by attaching a permanent magnet plate underneath the QAH mesa.

Room temperature device characterization

The circulator sample was initially characterized without a matching inductor and its impedance was expected to be on the order of the Hall resistance, ρxy=25 kΩ\rho_{xy} = 25\,\mathrm{k}\Omega. Since the impedance of the measurement circuit was 50 Ω50\,\Omega, the lack of a matching circuit at the input results in approximately 99% of the power being reflected at the input port of the circulator. As a control, we first measured all microwave SS parameters before mounting the inductors, observing that all ports are equivalent with a transmission magnitude approximately −50 dB-50\,\mathrm{dB} at 530 MHz, as shown in Figure 5b.

We then implemented a matching network at the input ports of the circulator by utilizing 3 lumped element 180 nH inductors in series (Figure 1a). The inductors coupled with the effective capacitance of the circuit serve as an impedance matching network for the 50 Ω50\,\Omega coaxial lines and the high impedance of the circulator. Physically, the matching process utilizes the resonance created by the inductors and the effective stray capacitance of the circuit. The resonance is tuned to coincide with the circulator’s edge magnetoplasmonic resonance frequency, effectively increasing the energy coupled into the circulator. Increasing the energy coupled into the circuit increases the signal-to-noise ratio. In practice, better coupling improves the insertion loss associated with reflections arising from impedance mismatch. The appropriate inductor value was determined by testing different inductors on a similar sample and collecting transmission measurements at room temperature. Figure 5b and Figure 5c present the six-way transmission spectra measured without and with the inductors, respectively. At room temperature, the LC network exhibits a resonant frequency of f=566f = 566 MHz and a quality factor of Q=21Q = 21.

Power dependent transmission measurements for S31S_{31}

Figure 6 shows the magnitude of S31S_{31} plotted as a function of frequency for various input power levels at temperatures below 10 mK. When the signal is applied at port 1, both the center frequency and bandwidth exhibit a clear dependence on excitation power. In contrast, transmission from port 2 shows no significant power dependence.

Transmission spectra from port 1 to 3 at various input power levels

Figure 6. Transmission spectra from port 1 to 3 at various input power levels. The transmission dip shows a power-dependent frequency shift.

The phase difference between S31S_{31} and S32S_{32} at different excitation powers is illustrated in Figure 7. At the maximum S31S_{31} transmission dip (P=−119P = -119 dBm), a phase shift of approximately 180∘180^\circ is observed (Figure 7a). The slope of phase change decreases when PP exceeds −119-119 dBm (Figure 7b).

Measured phase difference between $S_{31}$ and $S_{32}$ at various input power levels

Figure 7. The measured phase difference between S31S_{31} and S32S_{32} at various input power levels. a, ΔPhase=∠S31−∠S32\Delta\mathrm{Phase} = \angle S_{31} - \angle S_{32} at an input power of P=−119P = -119 dBm. b, The phase difference at various input power levels above the EP condition.

Figure 8 presents the non-reciprocity ∣S32∣−∣S31∣\lvert S_{32}\rvert- \lvert S_{31}\rvert, the resonant frequency shift, and the

Parameters extracted from the transmission spectra as a function of input power

Figure 8. Parameters extracted from the transmission spectra as a function of input power. The non-reciprocity, defined as the difference between ∣S32∣|S_{32}| and ∣S31∣|S_{31}|, along with the frequency dip in ∣S31∣|S_{31}| and the 3 dB bandwidth, are plotted versus input power.

3 dB bandwidth as functions of input power. Notably, within the input power range of −145 to −125 dBm, both the circulator’s bandwidth and center frequency remain relatively stable, indicating a power-independent regime. This suggests that the circulator can operate reliably within power levels relevant to superconducting quantum information systems and high-energy physics experiments.

Insertion loss characterization

The insertion loss of the circulator was measured in a different Bluefors dilution refrigerator after the sample, with the LC resonators attached, had been stored for more than two years at ambient temperature in a low-vacuum desiccator. The characterization diagram is shown in Figure 9. The circulator was cooled to below 30 mK. A network analyzer was used to perform S21S_{21} transmission measurements through the system. To determine the insertion loss of the circulator, a transmission measurement was made through the transmission port (I1I_1) and the proportion of power transmitted through the receiver was measured at the output (O1O_1). Two cryogenic switches (Radial R585433210) enabled swapping between the circulator ports and a short cable. In addition, a second transmission line (I2I_{2}) was used to measure the isolation of the circulator. On the output line (O1O_{1}), the signal was amplified by a HEMT amplifier mounted to the 4 K stage, followed by a room temperature amplifier. An isolator before the HEMT amplifier was used to reduce standing waves within the receiver.

Figure 9. Experimental setup for insertion loss measurements. The circulator is cooled to below 30 mK. A network analyzer is used to measure transmission through the receiver and two RF switches enable bypassing the circulator. Measurement of transmission with and without the circulator enable a direct measurement of the insertion loss of the circulator.

Broadband transmission measurements were made of both the S32S_{32} and S31S_{31} of the circulator. At approximately 10 mK, we observe a ∼2\sim2 dB difference in transmission between input lines I1I_{1} and I2I_{2}.

To determine the insertion loss of the circulator, transmission measurements were performed with the circulator both in the receiver circuit and bypassed, while applying a range of input powers. Any additional transmission losses observed when the circulator was switched in were attributed to its insertion loss. As shown in Figure 10, the insertion loss of the circulator was found to be approximately 5.9±25.9 \pm2 dB with an isolation of ∼20\sim20 dB in the low power regime.

Insertion loss of the circulator at approximately 10 mK

Figure 10. Insertion loss of the circulator at approximately 10 mK. a, Transmission of the circulator and the short bypass cable at the VNA power of PVNA=−85P_{VNA} = -85 dBm. b, An insertion loss of 5.9 dB is observed at the transmission peak of ∣S32∣\lvert S_{32}\rvert with an isolation ∼20\sim20 dB.

Effective Hatano-Nelson description and derivation of the transmission model

In this Appendix, we show how the four-mode model in Eq. (1) reduces to an effective two-resonator description with asymmetric directional couplings, and how the same reduction yields the transmission formula in Eq. (3). Equation (1) is a minimal device-level coupled-mode model for the two LC resonators and the two directional edge-magnetoplasmon (EMP) propagation channels.

We begin from the non-Hermitian four-mode matrix in the basis Ψ=(a,m1,b,m2)T\Psi= (a,m_{1},b,m_{2})^{T},

Heffℏ=(ωa−iκag000ωm1−iκm1g000ωb−iκbgg00ωm2−iκm2)(4)\frac{H_{\mathrm{eff}}}{\hbar} = \begin{pmatrix} \omega_{a} - i\kappa_{a} & g & 0 & 0 \\ 0 & \omega_{m_{1}} - i\kappa_{m_{1}} & g & 0 \\ 0 & 0 & \omega_{b} - i\kappa_{b} & g \\ g & 0 & 0 & \omega_{m_{2}} - i\kappa_{m_{2}} \end{pmatrix} \tag*{(4)}

where aa and bb denote the two LC resonators, m1m_{1} and m2m_{2} denote the two directional EMP channels, and gg is the LC-EMP coupling rate. In the experiment, the resonators are nearly identical, so we take ωa=ωb=ω0\omega_{a} = \omega_{b} = \omega_{0} and κa=κb=κ0\kappa_{a} = \kappa_{b} = \kappa_{0}. The off-diagonal structure of Eq. (4) describes the directional sequence

a→m1→b→m2→a.(5)a \rightarrow m_{1} \rightarrow b \rightarrow m_{2} \rightarrow a. \tag*{(5)}

The two EMP channels correspond to inequivalent chiral propagation paths around the QAH mesa, and therefore generally acquire different phase shifts and dissipation.

For a probe frequency ω\omega, we define

αj(ω)≡−iω+iωj+κj,j∈{a,b,m1,m2}.(6)\alpha_j(\omega) \equiv-i\omega+ i\omega_j + \kappa_j,\qquad j \in\{a,b,m_1,m_2\}. \tag*{(6)}

It follows that the steady-state coupled-mode equations for a coherent drive, sins_{\mathrm{in}}, incident on resonator aa are

αaa+igm1=sin,(7)\alpha_a a + igm_1 = s_{\mathrm{in}}, \tag*{(7)}
αm1m1+igb=0,(8)\alpha_{m_1}m_1 + igb = 0, \tag*{(8)}
αbb+igm2=0,(9)\alpha_b b + igm_2 = 0, \tag*{(9)}
iga+αm2m2=0.(10)iga + \alpha_{m_2}m_2 = 0. \tag*{(10)}

Eliminating the EMP amplitudes (m1,m2)(m_1,m_2) gives

αaa+g2αm1b=sin,(11)\alpha_a a + \frac{g^2}{\alpha_{m_1}}b = s_{\mathrm{in}}, \tag*{(11)}
αbb+g2αm2a=0.(12)\alpha_b b + \frac{g^2}{\alpha_{m_2}}a = 0. \tag*{(12)}

The two resonators are therefore coupled by the effective directional couplings

λabeff(ω)=g2αm1(ω),λbaeff(ω)=g2αm2(ω),(13)\lambda_{ab}^{\mathrm{eff}}(\omega) = \frac{g^2}{\alpha_{m_1}(\omega)},\qquad\lambda_{ba}^{\mathrm{eff}}(\omega) = \frac{g^2}{\alpha_{m_2}(\omega)}, \tag*{(13)}

and because the circulation paths have different lengths and dissipation, αm1≠αm2\alpha_{m_1} \ne\alpha_{m_2}, leading to asymmetric coupling. Hence, Eq. (1) reduces to a two-site Hatano-Nelson-type model with effective matrix

Meff(ω)=(αa(ω)λabeff(ω)λbaeff(ω)αb(ω)).(14)M_{\mathrm{eff}}(\omega) = \begin{pmatrix} \alpha_a(\omega) & \lambda_{ab}^{\mathrm{eff}}(\omega) \\ \lambda_{ba}^{\mathrm{eff}}(\omega) & \alpha_b(\omega) \end{pmatrix}. \tag*{(14)}

The effective asymmetric coupling between the two resonators can be interpreted as a two-site reduction of the canonical Hatano-Nelson Hamiltonian,

H/ℏ=∑n(ω0a^n†a^n+λaba^n+1†a^n+λbaa^n†a^n+1).(15)H/\hbar= \sum_n \left(\omega_0 \hat{a}_n^\dagger\hat{a}_n + \lambda_{ab}\hat{a}_{n+1}^\dagger\hat{a}_n + \lambda_{ba}\hat{a}_n^\dagger\hat{a}_{n+1}\right). \tag*{(15)}

where ω0\omega_0 is the resonant frequency of the nnth resonator; a^n†\hat{a}_n^\dagger and a^n\hat{a}_n are the bosonic creation and annihilation operators at the nnth resonator; λab\lambda_{ab} and λba\lambda_{ba} represent non-Hermitian coupling between adjacent resonators. In our system, the effective couplings are set by the EMP propagation paths and are generally unequal, λab≠λba\lambda_{ab} \ne\lambda_{ba}. Thus, the device realizes the minimal two-site analogue of Hatano-Nelson physics in an open, driven, and dissipative system.

The same reduction also yields the transmission formula used to fit the data. Solving Eqs. (11) and (12) gives

bsin=−g2αm1αaαbαm1αm2−g4.(16)\frac{b}{s_{\mathrm{in}}}=-\frac{g^2\alpha_{m_1}}{\alpha_a\alpha_b\alpha_{m_1}\alpha_{m_2}-g^4}. \tag*{(16)}

For nearly identical LC resonators, αa=αb≡α0\alpha_a=\alpha_b\equiv\alpha_0, where

α0=−iω+iω0+κ0,(17)\alpha_0=-i\omega+i\omega_0+\kappa_0, \tag*{(17)}

so that

bsin=−g2αm1α02αm1αm2−g4.(18)\frac{b}{s_{\mathrm{in}}}=-\frac{g^2\alpha_{m_1}}{\alpha_0^2\alpha_{m_1}\alpha_{m_2}-g^4}. \tag*{(18)}

Introducing αm\alpha_m through

αm2≡αm1αm2,(19)\alpha_m^2\equiv\alpha_{m_1}\alpha_{m_2}, \tag*{(19)}

and using αm1=2αm2\alpha_{m_1}=2\alpha_{m_2} near the operating point (see main text), we obtain Eq. (3);

t(ω)=−2κ0αmg2α02αm2−g4,(20)t(\omega)=-\frac{\sqrt{2}\kappa_0\alpha_m g^2}{\alpha_0^2\alpha_m^2-g^4}, \tag*{(20)}

where the factor κ0\kappa_0 comes from the canonical input–output transmission amplitude normalization for a resonator coupled to a transmission line.

References

References

  1. [1]A. C. Mahoney, J. I. Colless, S. J. Pauka, J. M. Hornibrook, J. D. Watson, G. C. Gardner, M. J. Manfra, A. C. Doherty, and D. J. Reilly, On-chip microwave quantum Hall circulator, Phys. Rev. X 7, 011007 (2017).DOI
  2. [2]A. C. Mahoney, J. I. Colless, L. Peeters, S. J. Pauka, E. J. Fox, X. Kou, L. Pan, K. L. Wang, D. Goldhaber-Gordon, and D. J. Reilly, Zero-field edge plasmons in a magnetic topological insulator, Nat. Commun. 8, 1836 (2017).DOI
  3. [3]L. A. Martinez, G. Qiu, P. Deng, P. Zhang, K. G. Ray, L. Tai, M.-T. Wei, H. He, K. L. Wang, J. L. DuBois, and D.-X. Qu, Edge magnetoplasmon dispersion and time-resolved plasmon transport in a quantum anomalous hall insulator, Phys. Rev. Res. 6, 013081 (2024).arxiv.org/abs/2308.15665
  4. [4]R. C. Ashoori, H. L. Stormer, L. N. Pfeiffer, K. W. Baldwin, and K. West, Edge magnetoplasmons in the time domain, Phys. Rev. B 45, 3894 (1992).pubmed.ncbi.nlm.nih.gov/10001992
  5. [5]N. Kumada, H. Kamata, and T. Fujisawa, Edge magnetoplasmon transport in gated and ungated quantum hall systems, Phys. Rev. B 84, 045314 (2011).DOI
  6. [6]N. Kumada, P. Roulleau, B. Roche, M. Hashisaka, H. Hibino, I. Petkovic, and D. C. Glattli, Resonant edge magnetoplasmons and their decay in graphene, Phys. Rev. Lett. 113, 266601 (2014).
  7. [7]O. E. Örsel, J. Noh, P. Zhu, J. Yim, T. L. Hughes, R. Thomale, and G. Bahl, Giant non-reciprocity and gyration through modulation-induced hatano-nelson coupling in integrated photonics, Phys. Rev. Lett. 134, 153801 (2025).DOI
  8. [8]C.-Z. Chang, J. Zhang, X. Feng, J. Shen, Z. Zhang, M. Guo, K. Li, Y. Ou, P. Wei, L.-L. Wang, Z.-Q. Ji, Y. Feng, S. Ji, X. Chen, J. Jia, X. Dai, Z. Fang, S.-C. Zhang, K. He, Y. Wang, L. Lu, X.-C. Ma, and Q.-K. Xue, Experimental observation of the quantum anomalous hall effect in a magnetic topological insulator, Science 340, 167 (2013).DOI
  9. [9]C.-Z. Chang, C.-X. Liu, and A. H. MacDonald, Colloquium: Quantum anomalous hall effect, Rev. Mod. Phys. 95, 011002 (2023).DOI
  10. [10]Y.-F. Zhao, R. Zhang, R. Mei, L.-J. Zhou, H. Yi, Y.-Q. Zhang, J. Yu, R. Xiao, K. Wang, N. Samarth, et al., Tuning the chern number in quantum anomalous hall insulators, Nature (London) 588 (2020).DOI
  11. [11]Y.-F. Zhao, R. Zhang, J. Cai, D. Zhuo, L.-J. Zhou, Z.-J. Yan, M. H. Chan, X. Xu, and C.-Z. Chang, Creation of chiral interface channels for quantized transport in magnetic topological insulator multilayer heterostructures, Nat. Commun. 14 (2023).DOI
  12. [12]N. Hatano and D. R. Nelson, Localization transitions in non-hermitian quantum mechanics, Phys. Rev. Lett. 77, 570 (1996).DOI
  13. [13]N. Hatano and D. R. Nelson, Non-hermitian delocalization and eigenfunctions, Phys. Rev. B 58, 8384 (1998).DOI
  14. [14]H. K. Lau and A. A. Clerk, Fundamental limits and non-reciprocal approaches in non- hermitian quantum sensing, Nat. Commun. 9, 4320 (2018).
  15. [15]Y.-Y. Wang, S. van Geldern, T. Connolly, Y.-X. Wang, A. Shilcusky, A. McDonald, A. A. Clerk, and C. Wang, Low-loss ferrite circulator as a tunable chiral quantum system, Phys. Rev. Appl. 16, 064066 (2021).
  16. [16]B. Peng, Ş. K. Özdemir, F. Lei, F. Monifi, M. Gianfreda, G. L. Long, S. Fan, F. Nori, C. M. Bender, and L. Yang, Parity-time-symmetric whispering-gallery microcavities, Nat. Phys. 10, 394 (2014).DOI
  17. [17]W. D. Heiss, The physics of exceptional points, Journal of Physics A Mathematical General 45, 444016 (2012).
  18. [18]F. Mallet, F. R. Ong, A. Palacios-Laloy, F. Nguyen, P. Bertet, D. Vion, and D. Estève, Single-shot qubit readout in circuit quantum electrodynamics, Nat. Phys. 5, 791 (2009).DOI
  19. [19]C. Bartram et al. (ADMX Collaboration), Axion dark matter experiment: Run 1b analysis details, Phys. Rev. D 103, 032002 (2021).DOI
  20. [20]M.-A. Miri and A. Alù, Exceptional points in optics and photonics, Science 363, eaar7709 (2019).DOI
  21. [21]A. Li, H. Wei, M. Cotrufo, W. Chen, S. Mann, X. Ni, B. Xu, J. Chen, J. Wang, S. Fan, C.-W. Qiu, A. Alù, and L. Chen, Exceptional points and non-hermitian photonics at the nanoscale, Nat. Nanotechnol. 18, 706 (2023).DOI
  22. [22]F. U. J. Klauck, M. Heinrich, A. Szameit, and T. A. W. Wolterink, Crossing exceptional points in non-hermitian quantum systems, Sci. Adv. 11, 8275 (2025).DOI
  23. [23]T. Röper, H. Thomas, D. Rosenbach, A. Uday, G. Lippertz, A. Denis, P. Morfin, A. A. Taskin, Y. Ando, and E. Bocquillon, Propagation, dissipation, and breakdown in quantum anomalous hall edge states probed by microwave edge plasmons, Phys. Rev. B 110, L161403 (2024).DOI
  24. [24]L. B. Nguyen, Y.-H. Lin, A. Somoroff, R. Mencia, N. Grabon, and V. E. Manucharyan, High-coherence fluxonium qubit, Phys. Rev. X 9, 041041 (2019).DOI
  25. [25]D. A. Rower, L. Ding, H. Zhang, M. Hays, J. An, P. M. Harrington, I. T. Rosen, J. M. Gertler, T. M. Hazard, B. M. Niedzielski, M. E. Schwartz, S. Gustavsson, K. Serniak, J. A. Grover, and W. D. Oliver, Suppressing counter-rotating errors for fast single-qubit gates with fluxonium, PRX Quantum 5, 040342 (2024).arxiv.org/abs/2406.08295
  26. [26]H. Zhang, C. Ding, D. Weiss, Z. Huang, Y. Ma, C. Guinn, S. Sussman, S. P. Chitta, D. Chen, A. A. Houck, J. Koch, and D. I. Schuster, Tunable inductive coupler for high-fidelity gates between fluxonium qubits, PRX Quantum 5, 020326 (2024).
  27. [27]K. Azar, L. Ateshian, M. T. Randeria, R. D. Piñero, J. M. Gertler, J. An, F. Contipelli, L. Ding, M. Gingras, K. Grossklaus, M. Hays, T. M. Hazard, J. Kim, B. M. Niedzielski, H. Stickler, K. L. Tiwari, H. Zhang, J. A. Grover, J. L. Yoder, M. E. Schwartz, W. D. Oliver, and K. Serniak, Characterization and comparison of energy relaxation in fluxonium qubits (2026), arXiv:2603.23636.arxiv.org/abs/2603.23636
  28. [28]D. X. Qu, Y. S. Hor, J. Xiong, R. J. Cava, and N. P. Ong, Quantum oscillations and hall anomaly of surface states in the topological insulator Bi₂Te₃, Science 329, 821 (2010).DOI
  29. [29]D.-X. Qu, N. E. Teslich, Z. Dai, G. F. Chapline, T. Schenkel, S. R. Durham, and J. Dubois, Onset of a two-dimensional superconducting phase in a topological-insulator–normal-metal Bi₁−ₓSbₓ/Pt junction fabricated by ion-beam techniques, Phys. Rev. Lett. 121, 037001 (2018).DOI
  30. [30]D.-X. Qu, Y. S. Hor, and R. J. Cava, Quantum oscillations in magnetothermopower measurements of the topological insulator Bi₂Te₃, Phys. Rev. Lett. 109, 246602 (2012).pubmed.ncbi.nlm.nih.gov/23368356
  31. [31]G. Qiu, P. Zhang, P. Deng, S. K. Chong, L. Tai, C. Eckberg, and K. L. Wang, Mesoscopic transport of quantum anomalous hall effect in the submicron size regime, Phys. Rev. Lett. 128, 217704 (2022).DOI

Paper details

vlnc:W_k4jh59q6
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents