Enhanced bulk photovoltaic effect in two-dimensional ferroelectric CuInP2S6

The photocurrent generation in photovoltaics relies essentially on the interface of p-n junction or Schottky barrier with the photoelectric efficiency constrained by the Shockley-Queisser limit. The recent progress has shown a promising route to surpass this limit via the bulk photovoltaic effect for crystals without inversion symmetry. Here we report the bulk photovoltaic effect in two-dimensional ferroelectric CuInP2S6 with enhanced photocurrent density by two orders of magnitude higher than conventional bulk ferroelectric perovskite oxides. The bulk photovoltaic effect is inherently associated to the room-temperature polar ordering in two-dimensional CuInP2S6. We also demonstrate a crossover from two-dimensional to three-dimensional bulk photovoltaic effect with the observation of a dramatic decrease in photocurrent density when the thickness of the two-dimensional material exceeds the free path length at around 40 nm. This work spotlights the potential application of ultrathin two-dimensional ferroelectric materials for the third-generation photovoltaic cells.

strong depolarization effect in CIPS. As shown in Figure S6, high bias up to -0.7 V was applied in the I-V curve measurement of device #6. The I-V curve deviates from the linearity and shrinks when the DC voltage is above -0.3 V. As a result of the depolarization effect, the nominal Voc (-0.55 V) from the I-V curve should be smaller than that from the initial state of the device. By linear fit to the I-V curve, we find a much larger Voc at -1.04 V (see Figure S6). To avoid this problem, when measuring the I-V curve at bright state of our devices the reading voltage is therefore limited within the range of ±0.1 V (equivalent to an electric field strength at ~10 5 V/cm). The Voc is then determined through a linear fit to the I-V curve with extended DC bias range (see

Supplementary note 2: The power conversion efficiency estimation of CIPS device.
We calculated the power conversion efficiency (PCE) by the formula where Jsc is the short-circuit current density, Voc is the open-circuit voltage in the light irradiation, Pin is the power of the incident light, is the absorption coefficient of 2D CIPS, and FF is the fill factor, which represents the ratio between actual power and the ideal power obtained from the product Jsc×Voc. Considering the thickness of the CIPS used in our study, the ultrathin CIPS is transparent in a broad range of the light spectrum.
Most intensity of the irradiation light is not absorbed by the device. To quantify the absorbance of ultrathin CIPS at 405 nm irradiation, we measure the transmission spectrum of CIPS on PDMS substrate as shown in Fig. S15. We find more than 97% of the light intensity transmits through the 10 nm thick CIPS. The absorption coefficient α at 405 nm is found to be lower than 3% for 10 nm CIPS. According to the data shown in Fig. 2a (device #2 with Jsc ~0.8 mA/cm 2 , Voc ~-1.65 V, and FF ~25 %) under onsample light power density at 0.3 W/cm 2 , the PCE of device #2 is estimated to be 3.67 %, which is more than half of the value at the S-Q limit for 2D CIPS with 2.85 eV band gap. We compare this efficiency with the conventional ferroelectrics based photovoltaic cells. For example, the PCE of 50 nm thin film ferroelectric BaTiO3 is 2.1×10 -6 . Some more data can be found in the last column of (b) Raman spectrum of the bottom graphene in device #3. Typical G mode and 2D mode peaks are observed. The intensity ratio of I2D to IG is 3/2, which indicates that the graphene used is bilayer.            electron and hole will be driven The created electron and hole can reach the two electrodes when the distance is less than 2l0, where l0 is the carrier diffusion length.
Here we assume the electron and hole have the same diffusion length. When d > 2l0, the created electron and hole cannot reach the electrodes simultaneously, yielding a sudden drops of photocurrent.

Fig. S14
Photovoltaic transport character of devices in Fig. 5b for different thicknesses (in the unit of nm) and different excitation power (in the unit of μW). The intercept photocurrent at zero bias (for example in (a), Isc = Ids (Vds =0) is 1.2 nA) divided by the excitation laser spot size yields the photo response density used in the main text.

Fig. S15
Transmission spectra of 10 nm (a) and >200 nm (b) CIPS thin films measured at 405 nm. In our experiment, the 405 nm laser is adopted as the irradiation source. The transmission (T) is defined as T = CIPS / sub , where CIPS and sub are the transmitted light intensity of CIPS on PMDS and bare PDMS substrate. We estimate the transmission coefficient for 10 nm CIPS to be about 97%, which is used in the main text for the estimation of PCE.