Ultralow dark current in near-infrared perovskite photodiodes by reducing charge injection and interfacial charge generation

Metal halide perovskite photodiodes (PPDs) offer high responsivity and broad spectral sensitivity, making them attractive for low-cost visible and near-infrared sensing. A significant challenge in achieving high detectivity in PPDs is lowering the dark current density (JD) and noise current (in). This is commonly accomplished using charge-blocking layers to reduce charge injection. By analyzing the temperature dependence of JD for lead-tin based PPDs with different bandgaps and electron-blocking layers (EBL), we demonstrate that while EBLs eliminate electron injection, they facilitate undesired thermal charge generation at the EBL-perovskite interface. The interfacial energy offset between the EBL and the perovskite determines the magnitude and activation energy of JD. By increasing this offset we realized a PPD with ultralow JD and in of 5 × 10−8 mA cm−2 and 2 × 10−14 A Hz−1/2, respectively, and wavelength sensitivity up to 1050 nm, establishing a new design principle to maximize detectivity in perovskite photodiodes.

In panel b, the spectra obtained with two device configurations with opaque (top Ag) and semi-transparent (top ITO) electrodes are shown. This shows that the minor signal between 0.8 and 1.0 eV is enhanced by optical interference due to photons being reflected by the silver layer. This interference is less when an optically transparent ITO top contact is used. The intensity of sub-bandgap EQE is enhanced by this effect. . d, Detectivity calculated from dark current density using D* = SR / (2qJ D ) 1/2 and from measured noise current using D* = SR A 1/2 /i n for photodiode with PTAA:poly-TPD EBL.

Supplementary Tables
Supplementary Table 1 1 Specific detectivity derived from noise spectral density using D * = SR A 1/2 i n −1 is marked with † . Specific detectivity values calculated using D* = SR (2qJ D ) −1/2 (assuming dominant shot noise) are reported with * . 2 The reported times are measured using a standard square wave and/or a short single peak pulse of light (represented in brackets). 3 The reported times have been estimated based on mobilities of the materials used in the device and the thickness of each layer.

Supplementary Note 1: Drift-diffusion simulations
The drift-diffusion simulations were performed using the SCAPS software version 3.  ITO and perovskite. The energy levels were as reported in the band diagram of Figure 1 (main text).
To simulate the thermally-activated charge generation at the interface between the blocking layers and the perovskite, we treated the blocking layers as contacts under the assumption of no transport loss beyond their interface, as described by Wang et al. 10 . The simulated stack consisted then of PTAA/ perovskite/ C 60 , with contacts parameters as reported in Supplementary Table 2. Perovskite properties were the same as previously listed in Supplementary Table 1, while the energetic information is reported in Figure 1.

Supplementary Note 2: Calculation temperature activation energies from sensitive EQE measurements
One possible source of J D is thermally driven charge carrier generation within the bulk of the perovskite. This can occur with the excitation of carriers across the bandgap (i.e., from perovskite VB to CB), or between trap states that lie within the bandgap. To determine whether either of these mechanisms makes a significant contribution to J D in the PPDs measured here, we adapt an approach commonly applied to calculate the short-circuit current density (J sc ) in solar cells that was also employed in our recently published work on the origins of OPD dark current 11,12 . More specifically, this widely used approach [13][14][15] proposes that J sc is proportional to the overlap integral between the incident photon flux density ϕ(E) of the AM1.5G spectral radiance (determined from the spectral irradiance I(E) via ϕ(E) = I(E)/E and the external quantum efficiency spectrum EQE PV (E) of charges collected per incident photon: The thermally generated dark current in the radiative limit rad 0 J can then be calculated by replacing the AM1.5G solar spectrum with the spectral photon flux from thermal black-body spectrum ϕ BB (E) at the device temperature: where c is the speed of light in a vacuum and k B is Boltzmann's constant. Dividing by the energy of each photon E and multiplying by a factor of c gives the spectral photon flux ϕ(E) (in units of J −1 m −2 s −1 ) emitted into a hemisphere from a planar unit surface, again in the interval from E to E + dE, Multiplying by a factor of two to account for the photon flux from both sides of the planar device (and converting to units of eV -1 cm -2 s -1 ) gives the spectral photon flux over at a certain temperature.
Since for a black-body emission and absorption are equivalent, the thermally generated dark current density can thus be determined from Equation S1. This approach assumes that EQE is temperature invariant; which may not be perfectly true.
Determining whether charge carrier excitation between sub-bandgap trap states contributes to observed J D requires extrapolation of the EQE spectrum to low energy regions that would correspond to plausible trap state distributions. Experimentally, we can determine the EQE down to ~0.63 eV, leaving the spectral shape below this energy, or when the limits of experimental sensitivity are reached, undetermined. Furthermore, because the black-body photon flux spectrum at room temperature and below is concentrated at low energies (e.g., maximum flux is at ~0.2 eV at 300 K), ignoring photon energies below 0.63 eV when determining EQE PV will underestimate J D calc17 .