Identification and Mitigation of Transient Phenomena That Complicate the Characterization of Halide Perovskite Photodetectors

Halide perovskites have shown promise to advance the field of light detection in next-generation photodetectors, offering performance and functionality beyond what is currently possible with traditional inorganic semiconductors. Despite a relatively high density of defects in perovskite thin films, long carrier diffusion lengths and lifetimes suggest that many defects are benign. However, perovskite photodetectors show detection behavior that varies with time, creating inconsistent device performance and difficulties in accurate characterization. Here, we link the changing behavior to mobile defects that migrate through perovskites, leading to detector currents that drift on the time scale of seconds. These effects not only complicate reproducible device performance but also introduce characterization challenges. We demonstrate that such transient phenomena generate measurement artifacts that mean the value of specific detectivity measured can vary by up to 2 orders of magnitude even in the same device. The presence of defects can lead to photoconductive gain in photodetectors, and we show batch-to-batch processing variations in perovskite devices gives varying degrees of charge carrier injection and photocurrent amplification under low light intensities. We utilize the passivating effect of aging to reduce the impact of defects, minimizing current drifts and eliminating the gain. This work highlights the potential issues arising from mobile defects, which lead to inconsistent photodetector operation, and identifies the potential for defects to tune photodetection behavior in perovskite photodetectors.


Note 1: Photoresponse Intensity Dependence and Low Light Illumination
To characterise the photodetectors, we performed an incident light intensity dependence of photoresponse, which provides a range of figures of merit of a photodetector (Figure S3) and importantly, how these change in different illumination regimes.The linear dynamic range can be obtained from the range of incident powers between which the response is linear.The responsivity can be obtained from the gradient, and any variations in sensitivity as light intensity changes can be determined.Finally, if the detection measurement set-up is sensitive enough, the intensity can be reduced to find the point where the response is indistinguishable from the device noise, the noise equivalent power (NEP).From the NEP the detectivity can be determined, using the following equation: Where A is the detector area and Δf is the measurement bandwidth.This gives a wide picture of the detection performance from a photodetector, and how these change with intensity.Then the temporal response can be determined separately to give the speed of the detector.
Intensity dependence measurements are sometimes performed over a small intensity range, and then linearly extrapolated to the independently measured noise level, assuming the responsivity of the photodetector remains constant.However, this is not always the case and responsivity has been shown to vary with light intensity, leading to potential errors in LDR, NEP and D* values.For example, deviations from linear light intensity dependence measurements are commonplace in detectors exhibiting gain in their response, as introduced by Fang et al. 3 Depending on the mechanistic origin of the gain, the impact of light intensity is different, showing both super-and sub-linear behaviour.
5][6][7] When the trap states at the interfaces are filled with charge carriers, band-bending can occur and charge injection through the transport layer enables photoconductive gain, increasing responsivity.However, as light intensity is reduced, and the filling of these traps no longer occurs, this gain mechanism is lost, and a reduction in responsivity is seen.Some devices only show gain at reduced light intensities, demonstrating super-linear deviations at low illumination levels. 8,9These systems require carrier densities to be on the order of defect densities, increasing the fraction of carriers in trap states.The trapping of one carrier (either electrons or holes) increases the lifetime of the other character, by reducing the likelihood of recombination.The increase in lifetimes of carriers relative to the carrier transit time enables re-circulation of carriers and gain.At lower light intensities, the deepest and longest lived traps are filled, enhancing gain further.Similarly, in systems without gain, but high trap densities, sublinear deviations can also be seen at low intensities where trapassisted non-radiative losses reduce the charge carrier collection efficiency.As a result, measurements were performed down to the lower limits of our experimental set-up, in order to accurately report the figures of merit, and also investigate any trap related phenomena that may be present.

Note 2: Reverse Biasing
All measurements were performed with and without reverse bias, to assess any changes in photodetector performance and any additional dynamic behaviour that may arise.Photodiodes can function without an external bias due to the in-built electric field, known as photovoltaic or self-powered mode.However, in applications they are often operated under reverse bias, known as photoconductive mode, due to enhanced perfromance. 1 Reverse bias can increase the response speed of a photodiode through two effects: increasing the drift velocity of carriers through the device, and lowering the diode capacitance by widening the depletion region.
Moreover, improvements in charge dissociation and collection with an externally applied bias results in improved linearity of response and a wider LDR. 2 However, reverse bias also increases dark current, through increased charge injection, which increases the shot noise from the device.Also, excessive reverse bias can degrade the device, and so care must be taken to balance any improvements in performance.
The amount of reverse bias was selected after IV curve analysis of the devices (Figure S2), ensuring the bias was well before the breakdown of the device.Commonly, this was selected as -1 V.

Figure S1 :
Figure S1: Drift of the dark current under short-circuit conditions, with the corresponding photocurrent shown in the inset (405 nm, 1.3 mW cm -2 ).

Figure S3 :
Figure S3: Initial measurement of the dark JV curves to select a reverse bias that is below the breakdown voltage.Reverse biases of -1 V were subsequently used as standard.

Figure S4 :
Figure S4: Measurement protocol for intensity dependence of photoresponse, referred to as the 'dynamic' measurement procedure.Taking a measurement of Ilight and Idark at each illumination intensity avoids any characterisation issues that arise when the current drifts with time.

Figure S7 :
Figure S7: Current density variation Jfinal -Jinitial vs. applied reverse bias, for the transient current curves in Figure 4a

Figure S9 :
Figure S9: The response time limits of the measurement set-up, measured with an ultrafast Si photodiode (Thorlabs FDS010).a) and b) show the rise and fall limits exceeds the