A Mean-Field Approach to the Dielectric Response of Bulk Superconductors for Light Dark Matter Direct Detection

The past decade has seen the rapid maturation of dark matter (DM) interactions with condensed matter systems as a primary avenue for sub-GeV DM detection, driving enormous progress across both theoretical frameworks [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38] and experimental implementations [39, 40, 41, 42, 43, 44, 45, 46, 47] (for a recent review, see Reference [48]). Among these efforts, detection experiments based on superconducting sensors [49, 50, 51, 52, 53, 54, 55, 56, 57] (e.g., kinetic inductance detectors (KIDs) [52], transition-edge sensors (TESs) [56, 57], and superconducting nanowire single-photon detectors (SNSPDs) [51, 55]) are garnering increasing interest. The recent QROCODILE experiment [55], for instance, based on a microwire SNSPD, achieved a threshold of 0.11 eV, and placed world-leading constraints on DM-electron scattering, DM-nucleon scattering for DM masses as low as 30 keV. These results demonstrate the power of SNSPDs for probing light DM parameter space, and future upgrades, including larger arrays and even lower thresholds, promise to further extend the reach.

As DM cannot be actively manipulated like photons, probing it with superconducting sensors demands an accurate description of its interactions with the detector, particularly the many-body electronic effects within the superconductors. In the context of the linear response theory, the rate for DM-electron scattering is governed by the dynamic structure factor, which is related to the imaginary part of the inverse dielectric function. For normal metals, the Lindhard dielectric function is widely used to give a description of the electronic response within the random-phase approximation (RPA). The Lindhard function was also employed in the superconductor-based experiments to model the screening effect [50, 51, 55]. This approach relies on the assumption that the normal-state dielectric response adequately approximates the screening in the superconducting state. While this approximation is computationally convenient and has been widely used, its validity has not been rigorously tested.

In this work, we generalize the RPA framework to the superconducting ground state in the Bardeen, Cooper, and Schrieffer (BCS) theory [58]. By deriving the quasiparticle polarization function that includes the full coherence factors, the corresponding dielectric function $\epsilon_{\mathrm{BCS}}\left(\mathbf{Q},\omega\right)$ is obtained within the superconducting phase. We then perform a systematic comparison between $\epsilon_{\mathrm{BCS}}\left(\mathbf{Q},\omega\right)$ and the Lindhard function $\epsilon_{\mathrm{Lin}}\left(\mathbf{Q},\omega\right)$ for common materials relevant for SNSPD experiment, specifically aluminum and tungsten silicide (WSi), the latter being the active material in the QROCODILE detector. Natural units ($\hbar=c=1$) are used throughout the paper.

In the Nambu spinor formulation, the electron field is organized as a doublet $\psi_{\mathbf{p}}=(c_{\mathbf{p},\uparrow},c^{\dagger}_{-\mathbf{p},\downarrow})^{T}$, a representation that naturally groups electron and hole degrees of freedom. The Coulomb interaction between electrons, which plays a central role in the screening physics we investigate, is most conveniently expressed in this basis. In momentum space, the electron density operator transforms as $\sum_{\mathbf{p},\sigma}c^{\dagger}_{\mathbf{p},\sigma}c_{\mathbf{p}+\mathbf{k},\sigma}=\sum_{\mathbf{p}}\psi^{\dagger}_{\mathbf{p}-\mathbf{k}}\tau^{3}\psi_{\mathbf{p}}$, where $\tau^{3}=\operatorname{diag}(1,-1)$ is the Pauli matrix in isospin space, its diagonal elements reflecting the opposite charge of electrons ($+1$) and holes ($-1$). The full Coulomb interaction Lagrangian then takes the compact form (see the Appendix for a step-by-step derivation)

where $V^{\mathrm{C}}(-\mathbf{k})=4\pi\alpha/k^{2}$ ($k=|\mathbf{k}|$) is the momentum-space Coulomb potential, with $\alpha=e^{2}/(4\pi)$ being the electromagnetic fine-structure constant.

In the superconducting phase, the BCS mean-field Hamiltonian introduces pairing correlations that couple electron and hole sectors. The electron Lagrangian reads

where $\varepsilon_{\mathbf{p}}=|\mathbf{p}|^{2}/(2m_{e})-E_{F}$ is the single-particle energy measured from the Fermi level, and $\Delta$ is the superconducting gap, determined self-consistently from the BCS gap equation. The off-diagonal pairing field $\Delta$ renders the electron and hole states no longer eigenstates of the system, necessitating a change of basis.

The Bogoliubov unitary transformation $U_{\mathbf{p}}$ diagonalizes the $2\times 2$ Hamiltonian matrix:

yielding the quasiparticle dispersion $E_{\mathbf{p}}=\sqrt{\varepsilon^{2}_{\mathbf{p}}+\Delta^{2}}$, which exhibits the characteristic gap $2\Delta$ in the excitation spectrum. The transformation matrix is parameterized as $U_{\mathbf{p}}=\bigl(\begin{smallmatrix}u_{\mathbf{p}}&-v_{\mathbf{p}}\\ v_{\mathbf{p}}&u_{\mathbf{p}}\end{smallmatrix}\bigr)$ with the coherence factors $\left(u_{\mathbf{p}},v_{\mathbf{p}}\right)=\left(\sqrt{\frac{1}{2}\left(1+\frac{\varepsilon_{\mathbf{p}}}{E_{\mathbf{p}}}\right)},\sqrt{\frac{1}{2}\left(1-\frac{\varepsilon_{\mathbf{p}}}{E_{\mathbf{p}}}\right)}\right)$. Physically, $v^{2}_{\mathbf{p}}$ represents the probability that the pair state $(\mathbf{p}\uparrow,-\mathbf{p}\downarrow)$ is occupied in the BCS ground state, while $u^{2}_{\mathbf{p}}=1-v^{2}_{\mathbf{p}}$ gives the probability it is empty. The quasiparticle field $\Psi_{\mathbf{p}}=U^{\dagger}_{\mathbf{p}}\psi_{\mathbf{p}}=(\gamma_{\mathbf{p},\uparrow},\gamma^{\dagger}_{-\mathbf{p},\downarrow})^{T}$ diagonalizes the BCS action and constitutes the appropriate eigenbasis for describing excitations above the superconducting condensate. The inverse transformation $\gamma_{\mathbf{p},\uparrow}=u_{\mathbf{p}}c_{\mathbf{p},\uparrow}+v_{\mathbf{p}}c^{\dagger}_{-\mathbf{p},\downarrow}$ reveals that QPs are coherent superpositions of electron and hole states, with propagator

where the infinitesimal $i0^{+}$ stems from the QP one-particle spectral function. (A detailed review of the BCS formalism is provided in the Appendix.)

For concreteness, in this work we consider a generic model where spin-$1/2$ DM particles ($\chi$) couple to the density of electrons via a Yukawa potential $V^{\chi e}\left(\mathbf{x}-\mathbf{x}^{\prime}\right)\propto e^{-m_{A^{\prime}}\left|\mathbf{x}-\mathbf{x}^{\prime}\right|}/\left|\mathbf{x}-\mathbf{x}^{\prime}\right|$, with $m_{A^{\prime}}$ being the mass of the mediator $A^{\prime}$. Similarly to the above discussion, the relevant action can be also written in terms of the QP field:

where $\tau^{3}=\operatorname{diag}(1,-1)$ is the isospin matrix, $a_{\mathbf{\mathbf{p}_{\chi},\sigma_{\chi}}}$ ($a^{\dagger}_{\mathbf{\mathbf{p}_{\chi},\sigma_{\chi}}}$) annihilates (creates) a DM particle with momentum $\mathbf{p}_{\chi}$ and spin $\sigma_{\chi}=\uparrow,\downarrow$, and $V^{\chi e}\left(\mathbf{k}\right)$ is the DM-electron interaction in momentum space. If an incident DM particle deposits energy exceeding the minimum pair-breaking threshold of $2\Delta$, it breaks a Cooper pair and excites two QPs from the BCS vacuum $\left|0;\mathrm{BCS}\right\rangle$. The four-fermion interaction in Eq. (5) directly describes the scattering process $\left|\mathbf{p}_{\chi}\right\rangle\left|0;\mathrm{BCS}\right\rangle\rightarrow\left|\mathbf{p}_{\chi}^{\prime}\right\rangle\left|\mathbf{p}_{1},\uparrow;\mathbf{p}_{2},\downarrow\right\rangle_{\mathrm{QP}}$ at tree level, where $\mathbf{p}_{1}$ and $\mathbf{p}_{2}$ represent the momenta of the two outgoing QPs. It is noted that the simple model in Eq. (5) does not change the spin of the DM particle in such a scattering, and thus the spins of the outgoing QPs are oriented in opposite directions. The tree-level $\mathcal{T}$-matrix element for $|\mathbf{p}_{\chi}\rangle|0;\mathrm{BCS}\rangle\to|\mathbf{p}^{\prime}_{\chi}\rangle|\mathbf{p}_{1},\uparrow;\mathbf{p}_{2},\downarrow\rangle_{\mathrm{QP}}$ is

where the BCS coherence factor $(u_{\mathbf{p}_{1}}v_{\mathbf{p}_{2}}+v_{\mathbf{p}_{1}}u_{\mathbf{p}_{2}})$ reflects the projection of the electron-density vertex onto the QP basis. This process and its RPA renormalization are illustrated in Fig. 1.

Many-body screening effect is incorporated within the RPA by summing QP bubble diagrams to all orders, as illustrated in Fig. 1. The screened amplitude becomes proportional to $V^{\chi e}(\mathbf{Q})/\epsilon_{\mathrm{BCS}}(\mathbf{Q},\omega)$, where the BCS dielectric function $\epsilon_{\mathrm{BCS}}(\mathbf{Q},\omega)=1-V^{\mathrm{C}}(\mathbf{Q})\,\Pi_{\mathrm{QP}}(\mathbf{Q},\omega)$ is defined in terms of the Coulomb potential $V^{\mathrm{C}}(\mathbf{Q})=4\pi\alpha/Q^{2}$. The QP polarization function $\Pi_{\mathrm{QP}}$ follows from evaluating the bubble diagram:

Substituting Eq. (7) into the expression for $\epsilon_{\mathrm{BCS}}$, the full BCS electronic dielectric function at zero temperature is

The zero-temperature limit is a reasonable approximation, as superconducting DM detectors operate at millikelvin temperatures. In the limit $\Delta\to 0$, $\epsilon_{\mathrm{BCS}}$ reduces to the Lindhard function for $\omega>0$.

By combining Eq. (6) and Eq. (8), we obtain the QP production rate for a DM particle incident on a superconductor with velocity $\mathbf{v}_{\chi}$:

where the integration limits are given by $Q^{\pm}=m_{\chi}v_{\chi}\left(1\pm\sqrt{1-\frac{2\omega}{m_{\chi}v^{2}_{\chi}}}\right)$, and $v_{\chi}=\left|\mathbf{v}_{\chi}\right|$ is the magnitude of the DM velocity, $m_{\chi}$ represents the DM particle mass. In momentum space, the Yukawa potential is expressed as $\left|V^{\chi e}\left(\mathbf{Q}\right)\right|^{2}=\pi\bar{\sigma}_{e}\mu^{-2}_{\chi e}\left(Q^{2}_{0}+m^{2}_{A^{\prime}}\right)^{2}/\left(Q^{2}+m^{2}_{A^{\prime}}\right)^{2}$, where $Q_{0}=\alpha m_{e}$ represents the characteristic momentum transfer in atomic scattering, $\mu_{\chi e}$ denotes the reduced mass of the DM-electron system, and $\bar{\sigma}_{e}$ serves as a reference cross section parameterizing the strength of the DM-electron coupling. $\mathrm{Im}\left[\epsilon_{\mathrm{BCS}}/\left|\epsilon_{\mathrm{BCS}}\right|^{2}\right]=\mathrm{Im}\left[-1/\epsilon_{\mathrm{BCS}}\right]$ is the energy loss function (ELF) of a superconductor. It is noted that the dielectric function is independent of the direction of the momentum $\mathbf{Q}$, i.e., $\epsilon_{\mathrm{BCS}}\left(\mathbf{Q},\omega\right)=\epsilon_{\mathrm{BCS}}\left(Q,\omega\right)$, as the isotropic symmetry is assumed for the electron system.

In experiments such as QROCODILE [55], which employs SNSPDs, the BCS electronic dielectric function $\epsilon_{\mathrm{BCS}}$ is approximated by the Lindhard function $\epsilon_{\mathrm{Lin}}$ , i.e.,

While this approximation retains the BCS coherence factor $\left(u_{\mathrm{\mathbf{p}_{1}}}v_{\mathbf{p}_{2}}+v_{\mathbf{p}_{1}}u_{\mathrm{\mathbf{p}_{2}}}\right)^{2}$ for the tree-level QP production rate, it approximates superconducting screening with the normal-state dielectric function. Given that the Lindhard function characterizes the RPA dielectric response of a homogeneous electron gas (HEG) in normal metals, its validity in the superconducting state has yet to be verified.

The accuracy of this approximation can now be examined. Taking aluminum and WSi as examples (the material used in the QROCODILE experiment [55]), the BCS dielectric function $\epsilon_{\mathrm{BCS}}$ is numerically calculated using an energy gap of $\Delta=0.17\,\mathrm{meV}$, and a Fermi energy $E_{F}=11.7\,\mathrm{eV}$ for aluminum, and $\Delta=0.26\,\mathrm{meV}$, $E_{F}=7.0\,\mathrm{eV}$ for WSi, respectively. We present the ratio $\left|\epsilon_{\mathrm{BCS}}\left(\mathbf{Q},\omega\right)\right|^{2}/\left|\epsilon_{\mathrm{Lin}}\left(\mathbf{Q},\omega\right)\right|^{2}$ in Fig. 2, which directly reflects the difference in the QP production rates calculated from the two dielectric functions.

It is observed that $\left|\epsilon_{\mathrm{BCS}}\right|^{2}$ and $\left|\epsilon_{\mathrm{Lin}}\right|^{2}$ agree closely over most of the parameter space of interest, except in two specific regimes: (1) above the red dotted curve $\omega=\left(\frac{Q}{2}+p_{F}\right)\frac{Q}{m_{e}}\simeq v_{F}Q$ (with $p_{F}$ being the Fermi momentum), where the imaginary part of the Lindhard function vanishes (see the Appendix) while that of the electronic BCS dielectric function remains finite; and (2) near the edge of the energy gap $\omega=2\Delta$, only above which the QP production is possible.

We also plot the kinematic regimes relevant for incident DM with a typical halo velocity of $v_{\chi}=10^{-3}c$. The momentum transfer $Q$ and energy deposition $\omega$ are bounded by $\omega\leq Qv_{\chi}-Q^{2}/\left(2m_{\chi}\right)$, shown by the contours for two benchmark DM masses, in black solid ($m_{\chi}=1\,\mbox{MeV}$) and dashed ($m_{\chi}=0.1\,\mbox{MeV}$) lines, respectively. The two curves coincide in the low-$Q$ region, converging to the asymptotic limit $\omega=v_{\chi}Q$.

Our numerical results confirm that the Lindhard dielectric function provides an accurate description of the electronic screening effect in QP production induced by halo DM in bulk superconductors. For experiments based on superconducting nanowires or thin films with a typical layer size of $10\,\mbox{nm}$, the geometric finite-size effects of the superconductor can be considered negligible for momentum transfer $Q\gtrsim 0.2\,\mbox{keV}$ (or equivalently, $1/Q\lesssim 1\,\mbox{nm}$). Consequently, the results presented here are applicable to DM particles with masses $m_{\chi}\apprge 0.1\,\mbox{MeV}$, where the bulk approximation remains valid. In the case of thin layer size of $\mathcal{O}\left(1\,\mbox{nm}\right)$, Refs [59, 55] computed the response function by numerically solving Maxwell’s equations with the appropriate boundary conditions for free electrons. Given that the free-electron-based Lindhard function effectively describes the screening effect in bulk superconductors, it is natural to expect this free-electron-based numerical method can be extended to superconducting thin films.

In summary, within the RPA framework, we have extended the Lindhard function of a normal-metal HEG to the superconducting ground state, thereby constructing the dielectric function of a BCS superconductor. Our analysis delineates a clear regime of validity for the Lindhard dielectric function: for energy transfers above $5\Delta$, the normal-state Lindhard function provides an accurate description of the in-medium response of superconducting targets.

Yet further extension is possible. Throughout this study we have described the target at the BCS mean-field level, which is expected to be adequate for the QP-excitation processes induced by DM scattering at a finite $Q$. In the long-wavelength limit, however, the collective dynamics of the condensate phase becomes prominent and calls for a description beyond mean-field theory. One must then reinstate the phase degree of freedom by restoring the phase of the order parameter $\Delta\rightarrow\Delta e^{i\theta}$, which gives rise to a phase field-electron coupling term $\propto\Delta\theta\psi^{\dagger}_{e}\tau^{2}\psi_{e}$, with $\tau^{2}=\left(\begin{array}[]{cc}0&-i\\ i&0\end{array}\right)$. Consequently, in addition to the QP bubble $\Pi_{\mathrm{QP}}$, the response of the target acquires a further contribution $\Pi_{\theta}$ describing the phase exchange. Dimensional analysis of the corresponding diagram in Fig 3 suggests that $\Pi_{\theta}$ is parametrically suppressed with respect to $\Pi_{\mathrm{QP}}$ by $\Delta^{2}/\left(v_{F}Q\right)^{2}\sim\left(\xi^{-1}_{0}/Q\right)^{2}$. The inverse coherence length $\xi^{-1}_{0}=\pi\Delta/v_{F}$ thus emerges as a natural crossover scale separating the mean-field, QP regime ($Q\gg 1/\xi_{0}$) from the collective, hydrodynamic regime of the superconducting state ($Q\ll 1/\xi_{0}$). While the mean-field approach is sufficient for the direct detection of halo DM with energy transfers above $5\Delta$, where the relative error $\Delta^{2}/\left(v_{F}Q\right)^{2}$ is at most $1/25$ (see Fig. 2), an accurate description of many-body physics in superconducting targets at the crossover scale becomes essential, for the regime approaching the pair-breaking threshold $\omega\rightarrow 2\Delta$. A detailed analysis of this issue is left for future work.