--- title: 8-Cell Microwave Resonator for Axion Searches url: https://www.emergentmind.com/topics/8-cell-microwave-resonator type: topic --- # 8-Cell Microwave Resonator for Axion Searches Searching arXiv for the cited paper and closely related work on multi-cell cavity haloscopes. Tool call: arxiv_search({"query":"id:2507.04344 OR \"Probing KSVZ Axion Dark Matter near 5.9 GHz Using a 8-Cell Cavity Haloscope\"","max_results":5,"sort_by":"submittedDate","sort_order":"descending"}) An 8-cell microwave resonator is a multicell cylindrical cavity architecture in which a single cylindrical shell is subdivided azimuthally into eight identical wedge-shaped cells to support an axion-sensitive, TM$_{010}$-like in-phase mode at substantially higher frequency than a conventional single-cell cavity of the same bore, while preserving a large detection volume. In the implementation reported by Ahn et al., the device functions as the core resonant element of a cavity haloscope operating near 5.9 GHz, where it is combined with an 8 T solenoidal field, a flux-driven Josephson parametric amplifier (JPA) operating near the quantum noise limit, and a sideband-summing analysis to search for axion dark matter near the KSVZ benchmark [2507.04344]. ## 1. Geometric configuration and physical realization The reported resonator is a single cylindrical shell of inner diameter $D=128$ mm and length $L=270$ mm, subdivided azimuthally into eight identical wedge-shaped cells by thin copper partitions. Its total detection volume is approximately $V\approx 3.1$ L, chosen to fill the 165 mm bore of the 8 T superconducting solenoid. The cavity therefore realizes high-frequency operation within a magnet geometry that would ordinarily favor a much lower-frequency single-cell mode [2507.04344]. Eight-fold rotational symmetry is central to the design. It supports coherent coupling of the TM$_{010}$-like mode across all cells, with the axion-sensitive mode identified as the lowest TM-type mode in each cell. In this mode, the electric field $\mathbf E$ is predominantly parallel to the solenoidal field $\mathbf B_0\parallel\hat z$, which is the required field orientation for axion-photon conversion in a standard haloscope configuration [2507.04344]. A common misconception is that multicell subdivision necessarily fragments the relevant search mode into unusable local resonances. In the reported device, the opposite design goal is realized: strong inter-cell coupling through the central gap retains a single well-resolved search mode with adequate spacing to neighboring modes, avoiding mode crossings across the tuning range. This is a specific property of the implemented geometry rather than a generic guarantee for all multicell cavities [2507.04344]. ## 2. Electromagnetic mode structure The resonator can be described in a coupled-resonator picture with nearest-neighbor coupling coefficient $\kappa$. The band of eight resonances follows approximately $$ f_m \approx f_0 \sqrt{\,1 + 2\,\kappa\cos\!\bigl(\tfrac{m\pi}{N+1}\bigr)\,} \quad (m=1,\dots,8), $$ where $f_0$ is the single-cell frequency without coupling. The axion search uses the $m=1$ in-phase band [2507.04344]. Finite-element COMSOL simulations were used to obtain the field profiles. When the alumina rods are near the cell center, corresponding to the low-frequency end of the tuning range, the electric field is concentrated centrally. As the rods rotate toward the cell wall, corresponding to the high-frequency end, the electric field shifts toward the periphery. This field migration is part of the tuning phenomenology and directly affects the form factor across the scan band [2507.04344]. The frequency-scaling motivation for the 8-cell architecture is explicit. A single-cell cavity of diameter $D=128$ mm has a fundamental TM$_{010}$ mode near 2 GHz. By subdividing the cylinder into eight cells, each cell’s effective diameter is reduced, pushing the TM$_{010}$-like band up to approximately 6 GHz, corresponding to an $\gtrsim 3\times$ frequency extension for the same bore. The total volume remains $V=3.1$ L independent of tuning. This directly addresses the usual tension between higher resonant frequency and usable cavity volume in haloscope design [2507.04344]. ## 3. Frequency tuning, quality factor, and form factor Frequency tuning is achieved by rotating eight 3 mm-thick alumina rods with relative permittivity $\varepsilon_r\approx 9.8$, one in each cell. All rods are mounted on a common shaft driven by a single piezoelectric actuator. The resonator was tuned from 5.83 GHz to 5.94 GHz, with an average tuning step of $\Delta f\approx 17.4$ kHz and a total span of $\Delta f_{\rm tot}\approx 110$ MHz [2507.04344]. The unloaded and loaded quality factors are defined as $$ Q_0=\frac{\omega_0\,U}{P_{\rm diss}}, \qquad Q_L=\frac{\omega_0\,U}{P_{\rm diss}+P_{\rm ext}}=\frac{Q_0}{1+\beta}, $$ where $\beta$ is the coupling coefficient. The measured cryogenic value was $Q_0\approx 4.5\times 10^4$, and with $\beta\approx 2$ during the run, the typical loaded quality factor was $Q_L\approx 1.5\times 10^4$ [2507.04344]. For mode $n$ in an external field $\mathbf B_{\rm ext}$, the form factor is $$ C_n= \frac{\bigl|\displaystyle\int_V \mathbf E_n(\mathbf r)\!\cdot\!\mathbf B_{\rm ext}\,dV\bigr|^2} {V\,\displaystyle\int_V \varepsilon |\mathbf E_n|^2\,dV}. $$ COMSOL simulations showed $C$ rising from approximately 0.50 at 5.83 GHz to approximately 0.70 at 5.94 GHz. Mechanical misalignments degraded $C$ by up to 5%, but by optimizing rod diameters this was reduced below 1%. The experiment adopted an average $C\approx 0.6$ in its sensitivity estimates [2507.04344]. These quantities determine the cavity’s usefulness as a haloscope resonator. The combination of a large volume, a loaded quality factor of order $10^4$, and a form factor near 0.6 places the device in a regime where high-frequency operation remains experimentally competitive rather than purely demonstrative. ## 4. Cryogenic environment and quantum-limited amplification The cavity and JPA were cooled by a dilution refrigerator to below 40 mK. The superconducting magnet bore ran at 4 K and produced up to 8 T at the cavity center. This establishes the low thermal-noise and high-field environment required for a resonant axion search at microwave frequencies [2507.04344]. The front-end amplifier was a flux-driven JPA implemented as a $\lambda/4$ coplanar-waveguide resonator terminated by a SQUID. It provided three-wave-mixing amplification: a pump tone at $f_p$ with power $P_p$ yields a signal at $f_s$ and an idler at $f_i=f_p-f_s$. During operation, a gain of approximately $G_{\rm JPA}\approx 20$ dB was maintained via a look-up table and rapid fine-tuning of $P_p$ on timescales $\lesssim 3$ s to within $\pm 0.5$ dB, with in situ monitoring by a VNA [2507.04344]. Because JPAs are highly field sensitive, the amplifier was shielded from residual magnetic fields of at most 100 mT by a three-layer $\mu$-metal/aluminum/niobium assembly. This shielding is an enabling subsystem rather than an incidental detail: without it, stable near-quantum-limited gain in the vicinity of a high-field solenoid would be difficult to maintain [2507.04344]. Noise characterization used the Y-factor method. With $$ P_{\rm on}=k_B\,G_{\rm on}(T_{\rm cav}+T_{\rm on})\Delta\nu, \qquad P_{\rm off}=k_B\,G_{\rm off}(T_{\rm cav}+T_{\rm off})\Delta\nu, $$ one solves for $T_{\rm sys}=T_{\rm cav}+T_{\rm add}$. Over all tuning steps, the system noise temperature varied from 380 mK to 500 mK, with a step-to-step spread of about 10 mK; uncertainties in $G_{\rm JPA}$ and $T_{\rm off}$ were kept below 1% and 2%, respectively [2507.04344]. ## 5. Signal model, sideband summing, and data analysis The expected axion-conversion power is maximized on resonance, $\nu_a=\nu_c$. In the normalization quoted for the experiment, it is 14.2 yW for the reference parameters $g_\gamma=0.97$, $\rho_a=0.45\,\mathrm{GeV/cm^3}$, $\nu_a=5.9\,\mathrm{GHz}$, $B_{\rm rms}=6.97\,\mathrm T$, $V=3.1\,\mathrm L$, $C=0.6$, and $Q_L=1.5\times10^4$. The Dicke-radiometer signal-to-noise ratio after integration time $t$ is $$ \mathrm{SNR} = \frac{P_{a\gamma\gamma}}{k_B\,T_{\rm sys}} \sqrt{\frac{t}{\Delta\nu_a}}, $$ where $\Delta\nu_a\approx \nu_a/Q_a$ is the axion linewidth [2507.04344]. A distinctive feature of the analysis is sideband summing. Because the JPA idler tone provides a correlated copy of an axion-induced signal, the normalized excess power in signal and idler bins, $\delta_k$ and $\delta_{-k}$, can be combined as $$ \delta_{ss,k} = \frac{w_k\,\delta_k + w_{-k}\,\delta_{-k}} {\sqrt{\,w_k^2 + w_{-k}^2 + 2\,\rho\,w_k w_{-k}\,}}, $$ where $w_{\pm k}$ are SNR-weighted and $\rho$ is the signal-idler correlation coefficient. This yielded an approximately 4% SNR gain over conventional single-sideband analysis [2507.04344]. The scan covered 5.83–5.94 GHz in 17.4 kHz steps, each step spanning a 1 MHz IF window with 62.5 Hz bins. At each step, 85,200 spectra in the lower half or 88,000 spectra in the upper half were recorded and then averaged into five groups. The total DAQ live-time efficiency was approximately 92%. Preprocessing excluded 25 kHz around the pump idler, removed the baseline with a Savitzky–Golay filter of 4th order and 1921-bin window, approximately 120 kHz, and normalized the data to zero mean and unit variance [2507.04344]. After sideband summing, spectra were aligned in the RF domain, weighted by the expected SNR for a KSVZ signal, and combined into a grand spectrum by further summing over 25 kHz, taken as the axion linewidth. The final spectrum had $\sigma\approx 0.85$ because of filter-induced correlations and was then rescaled to unit variance. Injection tests with synthetic axions showed an overall SNR efficiency of $\varepsilon_{\rm SNR}\approx 84\%$. A 90%-CL detection threshold of 3.47, assuming a nominal SNR of 5, produced 85 initial candidates; follow-up rescans ruled them all out [2507.04344]. ## 6. Exclusion result and broader significance No statistically significant excess was observed. The experiment excludes axion-photon couplings down to $$ g_{a\gamma\gamma}<1.2\times 10^{-14}\,\mathrm{GeV}^{-1} $$ at 90% confidence level over the interval $5.82