Tailoring nanoscale interfaces for perovskite–perovskite–silicon triple-junction solar cells

Materials

Formamidinium iodide (FAI), formamidinium bromide (FABr), methylammonium iodide (MAI) and PDCl were sourced from GreatCell Solar Materials. Lead iodide (PbI₂), lead bromide (PbBr₂), rubidium iodide (RbI) and MeO-2PACz were purchased from Tokyo Chemical Industry. BCP was purchased from Lumtec. C₆₀ was purchased from Nano C. Tetrakis(dimethylamino)tin(IV) (TDMASn) was purchased from Strem Chemicals. Gold, copper and silver shot were purchased from ESPI Metals. Other dry chemicals were purchased from Sigma-Aldrich and all solvents were purchased from Alfa Aesar. Pre-patterned indium tin oxide (ITO) glasses with a sheet resistance of 8 Ω sq⁻¹ were purchased from Wuhan Jinge Solar Energy Technology.

Device fabrication

For the fabrication of the silicon heterojunction bottom solar cell, 6-inch N-type polished Czochralski wafers with a thickness of 150 μm and a resistivity ranging from 1 to 5 Ω cm were used. A wet-chemical process, including saw-damage removal and cleaning, was applied to the as-cut wafers. No texturing process was used for this wafer. Subsequently, an intrinsic a-Si:H passivation layer (~5 nm) was first deposited by plasma-enhanced chemical vapour deposition (PECVD) on both sides of the wafer. Then, n-type (~5 nm) and p-type (~8 nm) a-Si:H layers were sequentially deposited at the front and back sides of the wafer, respectively. After PECVD, the back contact of the silicon cells was fabricated by stacking sputtered ITO (80 nm) and then thermally evaporated Ag through a shadow mask with an opening of 1.1 × 1.1 cm² or 4.1 × 4.1 cm² on the rear side. For the front side, a 20-nm-thick ITO layer was deposited on the n front side through a shadow mask of 1.1 × 1.1 cm² or 4.1 × 4.1 cm², defining the aperture area of the silicon bottom cell and acting as a recombination layer between the silicon bottom cell and the perovskite middle cell. The silicon bottom cells were then laser-cut to a 2 × 2 cm² or 5 × 5 cm² square substrate for small-area (1 cm²) and large-area (16 cm²) tandem fabrication.

The 1.55-eV middle perovskite has the structure: MeO-2PACz/Cs_0.08Rb_0.02FA_0.9PbI₃/ C₆₀/SnO₂/Au.

The front of the silicon solar cell was first treated with ultraviolet–ozone (UVO) cleaner for 5 min. This was followed by deposition of the hole transport layer MeO-2PACz (0.5 mg ml⁻¹ in menthol) via spin-coating at 4,000 rpm for 20 s, followed by annealing at 95 °C for 10 mins. The 1.5 M Cs_0.08Rb_0.02FA_0.9PbI₃ precursor was then spin-coated at 2,000 rpm for 20 s, followed by 6,000 rpm for 30 s. N₂ was blown onto the surface in the last 20 s before the end of the spin process. The film was annealed at 105 ^oC for 10 min, producing a deep, dark, dense perovskite film. The substrates were then transferred into a thermal evaporation chamber for 20 nm C₆₀ deposition. This was followed by 20-nm SnO₂ deposition by thermal ALD in an Arradiance GEMStar reactor. TDMASn was used as the Sn precursor and was held at 60 °C in a stainless-steel container. Water was used as an oxidant and was delivered from a stainless-steel container at room temperature, and the precursor delivery manifold temperature was set to 115 °C. The TDMASn/purge1/H₂O/purge2 times were 1 s/10 s/0.2 s/15 s with corresponding nitrogen flows of 30 sccm/90 sccm/90 sccm/90 sccm to the deposition chamber at 80 °C. A 20-nm tin oxide layer was formed after 135 cycles. After that, Au was deposited via thermal evaporation for different durations. Nominal thickness reading at 0, 0.2, 0.4, 0.6, 0.8 and 1 nm (using the Inficon quartz crystal monitor with corrected tooling factor for gold material) was used to distinguish different deposition times.

The 1.91-eV top perovskite cell had the structure: NiO_x/MeO-2PACz/Cs_0.16Rb_0.04FA_0.8Pb(I_0.45Br_0.55)₃/(PDCl)/C₆₀/SnO₂/Ag/MgF₂. The 10-nm NiO_x was deposited by sputter-coating using a 2-inch target under 60-W radiofrequency power in Ar at 2 mTorr using an AJA International sputtering system. The same concentration of MeO-2PACz was used as above for deposition, except for a longer duration of 30 s followed by annealing at a higher temperature of 100 °C. A 0.8 M, wide-bandgap perovskite precursor was then spin-coated using a single-step spin programme (3,000 rpm for 50 s), with nitrogen gas blown onto the surface during the last 25 s of the spin process. The resulting film was annealed at 105 °C for 10 min, yielding a deep red, dense film. The value of x in Cs_0.2−xRb_xFA_0.8Pb(I_0.45Br_0.55)₃ was allowed to vary between 0 and 0.12 for optimizing perovskite film quality and device performance. The results can be found in Supplementary Figs. 1 and 2. For the PDCl-treated perovskite, a solution of PDCl (~0.1 mg ml⁻¹ in isopropyl alcohol) was spin-coated onto the perovskite surface at 5,000 rpm for 20 s, followed by annealing at 105 °C for another 10 min. The same conditions were used for the deposition of C₆₀ and SnO₂ as above. Finally, the 90 nm ITO transparent electrode was deposited by sputter-coating through a metal mask with an area of 1.1 × 1.1 cm² or 4.1 × 4.1 cm² with a 35-W radiofrequency power and Ar at 1.5 mTorr using the AJA International sputtering system. To complete triple-junction cell fabrication, the silver grid was deposited by thermal evaporation to a thickness of 230 nm and 720 nm, for 1 cm² and 16 cm², respectively, through a mask. Finally, the front of the cell was deposited with 100 nm MgF₂ for antireflection.

For process optimizations, 1.91-eV perovskite single-junction opaque cells and perovskite–perovskite double-junction semitransparent test cells were also fabricated on glass substrates.

The patterned ITO-coated glass was first prepared by ultrasonic cleaning in deionized water containing 2% Hellmanex, followed by rinses in deionized water, acetone and isopropanol, each for 15 min. The cleaned ITO substrates were then subjected to UVO treatment for 15 min. After UVO treatment, the substrates were transferred to a nitrogen-filled glovebox for subsequent perovskite or perovskite–perovskite test solar cell fabrication.

The 1.91-eV perovskite single-junction opaque solar cell has the structure: glass/ITO/MeO-2PACz/Cs_0.16Rb_0.04FA_0.8Pb(I_0.45Br_0.55)₃/(PDCl)/C₆₀/BCP/Cu.

To evaluate the operational stability of Rb and MA incorporation in perovskites, three different compositions were used in fabrication test cells with the structure glass/ITO/MeO-2PACz/perovskite/C₆₀/BCP/Cu. The three compositions were:

1. (1)

   Cs_0.16Rb_0.04FA_0.8Pb(I_0.45Br_0.55)₃ (CsFARb);

2. (2)

   Cs_0.16MA_0.04FA_0.8Pb(I_0.45Br_0.55)₃ (CsFAMA);

3. (3)

   Cs_0.2FA_0.8Pb(I_0.45Br_0.55)₃ (CsFA).

The same conditions were used as for the 1.55-eV middle perovskite cell for the deposition of MeO-2PACz. The same conditions were also used as above for the deposition of 1.91-eV perovskite, PDCl treatment and C₆₀. Instead of SnO₂, 7 nm BCP was deposited via thermal evaporation. Finally, 100-nm-thick copper electrodes were deposited through a metal mask (to a defined device aperture area of 0.0706 cm²) by thermal evaporation, to finish the single-junction perovskite solar cell fabrication.

The 1.55-eV perovskite–1.91-eV perovskite semitransparent tandem test cells were fabricated using the same conditions as those used for triple-junction cells, except the Si bottom cell was replaced by ITO-patterned glass and the cell aperture area was 0.09 cm² and no metal grid was deposited. Electrical contact was made directly to the ITO.

Device encapsulation

For stability tests, 1-cm² tandem devices were laminated between two pieces of 3-mm-thick glass laminated by transparent polyolefin-type material, with polyisobutylene applied at the edges for sealing⁷. The laminating process was carried out in a Radiant YDS-1111 laminator at 110 °C for 10 min at 800 millibars of pressure. Copper tape was employed to establish electrical contact with the device electrodes, extending outward from the cover glass.

Characterizations

J–V measurements for single-junction opaque devices and perovskite–perovskite semitransparent double-junction devices were performed using a 1 lamp solar cell I–V testing system using a class AAA solar simulator under an illumination power of 100 mW cm⁻². The light was calibrated using a certified reference cell. A scan rate of 100 mV s⁻¹ was used during measuring, sweeping from near-open circuit voltage (V_OC) (1.4 V for single junctions, 2.3 V for perovskite–perovskite tandems) to near- short circuit current density (J_SC) (−0.1 V). Apertures of 0.0706 cm² and 0.09 cm² were used for single-junction opaque cells (illuminated from the glass side) and semitransparent double-junction cells (illuminated from the low-bandgap side), respectively.

J–V measurements for 1-cm² triple-junction devices were performed using an LED solar simulator (6,060 A, 350–1,200 nm, AAA class, Qingdao Solar Science Instrument Hightech) under an illumination power of 100 mW cm⁻². A scan rate of 100 mV s⁻¹ was used during measurement, sweeping from 3.2 V to near J_SC (−0.1 V). An aperture of 1.0 cm² was used. J–V measurements for 16 cm² triple-junction devices were performed by the Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Science, using a dual light source AAA steady-state solar simulator (YSS-T155A-2M) under an illumination power of 100 mW cm⁻² with an aperture of 16.0 cm².

EQE measurements for single-junction solar cells were carried out using the QuantX-300 Spectral Response (Newport) system with monochromatic light from a xenon arc lamp.

EQE measurements for triple-junction tandem solar cells were carried out in AC mode using Enli Technology (model QE-R) Taiwan system. The EQE response was calibrated using a certified reference cell for 300–1,100 nm. For measuring EQE of the silicon bottom cell, a blue LED (450 nm) and infrared LED (730 nm) were used to saturate the top and the middle cells. For measuring EQE of the middle perovskite cell, a blue LED (450 nm) and near-infrared LED (940 nm) were used to saturate the top and the bottom cells. For measuring EQE of the top perovskite cell, an infrared LED (730 nm) and near-infrared LED (940 nm) were used to saturate the middle and the bottom cells.

Transient photocurrents of solar cells were measured using a Keysight MSO9254A oscilloscope. The 520 nm wavelength excitation light was provided by a Thorlabs NPL52B pulsed laser with a 5-ns pulse width, repetition rate of 1 MHz and pulse energy of 1.2 nJ. The diameter of the beam was approximately 2 mm.

Temperature-dependent open circuit voltages of solar cells were measured using a Keysight 2636B source meter, illuminated by a Thorlabs OSL2 Fiber-Coupled Illuminator with intensity equivalent to 1 Sun. The temperature was controlled using a cryogenic cryogen-free variable temperature cryostat, with a Lakeshore 350 temperature controller.

The transmittance and reflectance of samples were measured using a Perkin Elmer Lambda1050 UV/Vis/NIR spectrophotometer.

Absorbances of samples were measured using an FS 5 (Edinburgh Instruments).

Thermal admittance spectroscopy and Mott–Schottky were conducted using a Keysight E4990A impedance analyser, operating from 20 Hz to 10 MHz with the ‘enhanced measurement speed’ option.

Time-of-flight secondary ion mass spectrometry (TOF-SIMS) was performed using an IONTOF TOF-SIMS 5 system, operating in positive polarity mode with Bi³⁺ primary ions at an energy of 30 keV and Cs⁺ sputtering ions at 1 keV, in the MC^s+ operational mode

X-ray diffraction patterns were recorded using a Bruker ECO D8 diffractometer with a CuKα (λ = 1.5418 Å) radiation.

XPS and UPS were performed using an ESCALAB250Xi (Thermo Scientific). For XPS analysis, we employed X-ray emission using an anode with MgKα line (12 kV–200 W) from an ultrahigh vacuum non-monochromatic source. Following an initial survey scan to assess chemical states, we performed high-resolution scans at a pass energy of 10 eV. The excitation energy used was 1,253.6 eV. The Φ was calculated according to the formula Φ = hν (21.22 eV) − E_{cutoff-measured}.

Top-view and cross-sectional SEM images were obtained using a field-emission microscope (NanoSEM 230).

TEM including specimen preparation

To characterize gold Au deposited (for different durations) on the ALD SnO₂ layer, a carbon-coated TEM grid (EMSCF200-CU-UL, Proscitech) was used, which was coated with a 20-nm SnO₂ layer by ALD, followed by thermal evaporation of Au for different times.

The samples were introduced into a JEOL 2100F FEG-TEM, which was fitted with a Gatan Ultrascan camera for imaging. For the TEM imaging process, we applied an electron dose rate of approximately 2 e/Ų per second.

Photoluminescence characterizations

Steady-state photoluminescence spectra of perovskite films were measured using an FS 5 (Edinburgh Instruments) with an excitation wavelength of 450 nm.

For TRPL decay measurements, a LabRAM HR Evolution system was used with a time-correlated single photon counting system (DeltaPro-DD, Horiba). Using a 485-nm diode laser (DD-510L, Horiba) as the excitation source, a laser with a pulse duration of 110 picoseconds, a reception rate of 312.5 kilohertz and a fluence of approximately 5–6 microjoules per square centimetre per pulse was used. The PL signal was captured at a wavelength of 660 nm. Both the incident and reflected light were directed through a ×50 objective lens (Leica PL FLUOTAR L 50/0.55), resulting in a spot size of approximately 2 μm. The samples were maintained in a nitrogen environment to prevent degradation during the measurement process. To determine the PL lifetime from the TRPL decay curves, a bi-exponential model was applied:

where A₁ and A₂ are weightings of the τ₁-fast decay component recombination via defect trapping and of the τ₂-slow decay component associated with radiative recombination¹⁵ was used in decay analysis software to fit the experimental results. The average lifetime, τ_avg was calculated using the following equation:

For PL imaging, a custom PL imaging system, featuring 430-nm royal-blue LED chips and 451/106-nm bandpass filters, was employed. The cells were secured in a nitrogen-filled, temperature-controlled custom jig during the imaging process and exposed to an intensity equivalent to 1 Sun. To capture the PL image, a Peltier-cooled (at −70 °C) Si CCD camera from Princeton Instruments (model Pixis 1024) along with a 700-nm long-pass filter was used, with an exposure time of 0.03 s. The PL image was then processed using Fiji software, which was also used to add a colour bar and calibration bar to the image.

Stability testing

For MPPT of the solar cells to evaluate operational stability of triple-junction tandems, encapsulated devices were placed inside an environmental chamber for continuous Xe illumination (100 mW cm⁻²). The temperature and relative humidity were kept at 25 ± 5°C and 60 ± 20%, respectively. The MPPT algorithm is based on the well-established perturb-and-observe methodology, integrated into a LabVIEW program for efficient implementation. The algorithm begins by deriving an initial estimation of the MPP through a rapid initial J–V measurement. In the regular operation of the algorithm, the applied voltage is perturbed using a double step of both +10 mV and −10 mV around the voltage corresponding to the maximum power point, denoted as V_MPP. Subsequently, the output power of the solar cell is measured at these three distinct voltage levels. The algorithm then selects the new V_MPP on the basis of the voltage configuration that yields the highest power output. One critical aspect of the algorithm’s execution is the duration of each voltage step. It is imperative to ensure that this duration is sufficiently long to allow for transients within the system to equilibrate before computing the power at the newly set voltage level. This careful consideration ensures the accuracy and effectiveness of the MPPT process.

For IEC 61215 standard thermal cycling testing, the encapsulated triple-junction tandem device underwent a thermal cycling regime and was measured ex situ regularly. The temperature cycle was between −40 °C and 85 °C, and for 204 times, in this work. During the cyclic testing, the device was held at both −40 °C and 85 °C for a duration of 10 min each. The temperature transitions between these points were executed at a controlled ramp rate of 45 °C per hour.

Simulation

Simulation of the electrostatic surface potential of PD⁺ was carried out by the DFT/B3LYP method with a basis set of 6-31G(d)(p) for determining the dipole moment. All the calculations were performed using the Gaussian 16 program package.

A commercial software package, Silvaco technology computer-aided design, was used to model the energy band structure of the SnO₂/(Au)/NiO_x stack under thermal equilibrium.

To calculate the optical effect of the Au nanoparticles, the Python-based software RayFlare²⁸ was used, which uses a modified version of the solver S⁴ (ref. ²⁹). The nanoparticles were represented as Au pillars in an NiO_x background material. It is assumed that the nominal thickness of Au (d_Au,nominal) deposited can be used to calculate the total volume of Au deposited per unit area, so that the height of the Au pillars (h_pillar) can be calculated from h_pillar = d_Au,nominal/C, where C is the area coverage fraction of the Au. The coverage fraction was determined from scanning TEM images of Au deposited at different nominal thicknesses. Since RCWA calculations assume a periodic unit cell, the random structure of the Au nanoparticles was simulated by generating a random unit cell, containing eight non-overlapping Au pillars randomly placed within the unit cell. The radius of the discs and size of the unit cell were chosen to give the correct coverage fraction with the pillar height as calculated above. The cell structure used is shown in Fig. 3a. Twenty random unit cells were generated for each coverage fraction, with the reflectance, transmittance and absorptance per layer calculated for each unique unit cell assuming normally incident unpolarized light. We then took the average of these results. The maximum possible short circuit current of the two perovskite junctions was calculated as:

where Φ_AM1.5G(λ) is the photon flux in the AM1.5G solar spectrum as a function of wavelength, q is the elementary charge and A_i(λ) is the fraction of incident photons absorbed in the relevant perovskite layer, from the average of the 20 randomly generated unit cells. For the results without antireflection coating, the simulations were performed using RCWA only. To simulate cells with textured polydimethylsiloxane antireflection coating, RayFlare’s ray tracer was used, assuming a regular inverted pyramid structure with an opening angle of 52°.

Simulation codes for results in Supplementary Figs. 18–21 can be found in ref. ³⁰.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.