High-efficiency detection of single microwave photons is a challenging task due to their extremely small energy. However, once realized, such a detector could find applications in quantum information processing as a way to actively correct qubit energy relaxation errors or to improve readout for cavity-based dark matter axion searches- as photon counting does not add a half photon of noise like phase preserving amplifiers.
A simple scheme to detect flying photons consists of a qubit absorber coupled to an input waveguide and a readout resonator. Once the qubit changes state by absorbing a flying photon, the dispersively coupled readout resonator shifts in frequency which can be measured with a homodyne detection chain. The bandwidth of the detector must be designed such that the photon is trapped long enough to perform high-fidelity readout, leading to inherently narrowband detectors. Keeping bandwidth large but compensating with strong coupling to the readout resonator leads to excessive backaction which causes the photon to reflect off the detector rather than being absorbed. In order to get around these constraints, we are currently testing a detector  based on an ensemble of qubits coupled in such a way that bright states, which have linewidths larger than the inherent qubit linewidth, are formed along with dark states, with linewidths approaching zero. By absorbing a photon into the bright state and transferring it to the long-lived dark state, we circumvent the bandwidth/interaction time tradeoff.
 B. Royer, et al., PRL 120, 203602 (2018)