Silicon heterojunction solar cells with up to 26.81% efficiency achieved by electrically optimized nanocrystalline-silicon hole contact layers

Silicon heterojunction (SHJ) solar cells have reached high power conversion efficiency owing to their effective passivating contact structures. Improvements in the optoelectronic properties of these contacts can enable higher device efficiency, thus further consolidating the commercial potential of SHJ technology. Here we increase the efficiency of back junction SHJ solar cells with improved back contacts consisting of p-type doped nanocrystalline silicon and a transparent conductive oxide with a low sheet resistance. The electrical properties of the hole-selective contact are analysed and compared with a p-type doped amorphous silicon contact. We demonstrate improvement in the charge carrier transport and a low contact resistivity (<5 mΩ cm2). Eventually, we report a series of certified power conversion efficiencies of up to 26.81% and fill factors up to 86.59% on industry-grade silicon wafers (274 cm2, M6 size). Improvements in the power conversion efficiency of silicon heterojunction solar cells would consolidate their potential for commercialization. Now, Lin et al. demonstrate 26.81% efficiency devices using a p-doped nanocrystalline silicon and low-sheet-resistance transparent conductive oxide contact layer.

Silicon heterojunction (SHJ) solar cells have reached high power conversion efficiency owing to their effective passivating contact structures. Improvements in the optoelectronic properties of these contacts can enable higher device efficiency, thus further consolidating the commercial potential of SHJ technology. Here we increase the efficiency of back junction SHJ solar cells with improved back contacts consisting of p-type doped nanocrystalline silicon and a transparent conductive oxide with a low sheet resistance. The electrical properties of the hole-selective contact are analysed and compared with a p-type doped amorphous silicon contact. We demonstrate improvement in the charge carrier transport and a low contact resistivity (<5 mΩ cm 2 ). Eventually, we report a series of certified power conversion efficiencies of up to 26.81% and fill factors up to 86.59% on industry-grade silicon wafers (274 cm 2 , M6 size).
Photovoltaic (PV) solar cells are one of the main renewable energy sources with zero operating carbon emissions; driven by ambitious carbon neutral policies worldwide, they are quickly becoming a mainstream energy supply. To a large extent, power conversion efficiency (PCE) determines whether a PV technology is competitive. Wafer-based crystalline silicon (c-Si) solar cells are the dominant technology in the global PV market. Aiming at a higher PCE, technology iteration is occurring from the passivated emitter and rear cell (PERC) to tunnel oxide passivated contact (TOPCon) and silicon heterojunction (SHJ) solar cells [1][2][3][4][5][6][7] .
SHJ technology employs an n-type (p-type) doped hydrogenated amorphous silicon (a-Si:H) layer, called n-a-Si:H (p-a-Si:H), as the electron-selective contact layer (ESC)-hole-selective contact layer (HSC). This overlays the intrinsic hydrogenated amorphous silicon (i-a-Si:H) layer, providing high-quality chemical passivation and minimizing the deficit in open circuit voltage (V OC ) [8][9][10][11][12][13] . The electrical performance of the solar cells depends strongly on the net doping of both the ESC and HSC layers. This is particularly relevant to the HSC layer, which is the emitter in SHJ solar cells based on n-type wafers. A sufficiently high doping concentration produces favourable band bending, allowing holes (minority carriers) to tunnel (selective collection of holes); efficient field-effect passivation, repelling electrons from the interface and mitigating the resulting interface recombination; and a reduced energy barrier when directly in contact with the n-type transparent conducting oxide (TCO) [14][15][16] . However, doped a-Si:H layers are always limited by unsatisfying electrical conductivity (σ < 10 -4 S cm −1 ) Article https://doi.org/10.1038/s41560-023-01255-2 technologies because of the excellent surface passivation provided by the i-a-Si:H layers. Figure 1b illustrates the FF versus V OC of different solar cell technologies, in which one can clearly see the superior FF of SHJ solar cells as compared with that of PERC and TOPCon devices.
In Fig. 1b, the lines corresponding to the Green limits 36 for different ideal factors (n) are indicated as well. The value of the ideality factor of a c-Si cell is based on the recombination mechanism: n = 2/3 for Auger recombination and n = 1 for Shockley-Read-Hall (SRH) and band-to-band recombination at low injection level. Combining high-quality c-Si wafers with the superior surface passivation we obtained, the intrinsic recombination of SHJ solar cells becomes dominant. In the case of Auger recombination at high injection level, the ideality factor value of our SHJ cells is found to be lower than one. The reduction in ideality factor causes a more square shape in the J-V curve, leading to a remarkable improvement in FF.
As indicated in Fig. 1, the performance of SHJ solar cells has increased almost linearly both electrically and optically. The first SHJ solar cell from our group (LONGi) delivered a PCE of 25.26% (ref. 37), and we have now further boosted all the PV parameters. In this work, we show a PCE of 26.81%, with V OC of 751.4 mV (an improvement of 2.9 mV), FF of 86.07% (0.57% improvement) and J SC of 41.45 mA cm −2 (an improvement of 1.97 mA cm −2 ), an overall efficiency gain of 1.55%. Note that we achieved the highest FF of 86.59% on a different device: that is, the cell delivering an efficiency of 26.30%. Theoretical predictions yield a similar trend, as shown by the blue to red gradient solid line in Fig. 1b. Radiative, Auger and surface recombinations were taken into consideration in the calculation, while additional R S and shunt resistance (R sh ) were not included. The improvement of LONGi's SHJ solar cells follows the overall tendency apart from small deviations in practical and calculated FF, which are mainly attributed to the advance in reduction of R S . Note that the measurement approach (Methods) excludes some R S components, such as the resistance of the grid at the rear side and of the bus bars at the front side. We found that ultrahigh FF only occurs at extremely low R S and furthermore, requires high-quality passivation: that is, high V OC . This is why LONGi SHJ solar cells possess a prominent advantage in electrical performance (V OC × FF) (Fig. 1a) over PERC and TOPCon cells. Limited by insufficient passivation, the positive effect of reduced R S on FF cannot be fully unlocked for PERC and TOPCon cells.
To reveal the main contributions to the efficiency increase, two related designs with p-a-Si:H (cell 1, 25.26% PCE) and p-nc-Si:H (cell 2, 26.30%) serving as the rear emitter are investigated with the Quokka2 software 38,39 . For the sake of simplicity, hereafter we name cell 1 and cell 2 as p-a-Si:H cell and p-nc-Si:H cell, respectively. Numerical simulations and fits to experimental data of R S and power loss analysis (PLA) are shown in Fig. 2. Figure 2a shows the structure of the LONGi SHJ solar cells; the parameter variations between the p-a-Si:H cell and p-nc-Si:H cell are listed in Supplementary Table 3. The two devices are in a front and back contact architecture on an n-type c-Si (n-Si) wafer with front-sided n-type nanocrystalline silicon oxide (n-nc-SiO x :H) and a back junction (BJ). Using BJ structure alleviates the electrical requirements on the front-side TCO and metal electrodes since a large portion of the majority carriers (electrons) can be laterally collected via the n-Si wafer absorber 40 . The main difference between the two solar cells comes from the BJ stacks; the p-nc-Si:H cell features p-nc-Si:H and a tailored TCO with a sheet resistance of 40 Ω per sqaure, while the p-a-Si:H cell features p-a-Si:H and a TCO with a sheet resistance of 80 Ω per square. Due to the excellent σ and E a of p-nc-Si:H, the rear contact resistivity is reduced, which will be discussed in the next section. Figure 2b,c shows the experimental and fitted J-V curves of their related cells. The experimental curves include the real-light J-V curve (blue triangles) and the pseudo-light J-V curve tested by a Suns-V OC measurement (blue circles). The fitted J-V curves were simulated by Quokka2 (refs. 38,39). Input parameters were determined and relatively high activation energy (E a > 250 meV), which cause high contact resistivity in SHJ solar cells [17][18][19][20] .
In contrast to defect-rich amorphous silicon, hydrogenated nanocrystalline silicon (nc-Si:H) dramatically improves film crystallinity, which straightforwardly favours the improvement of carrier mobility and effective doping concentration. However, depositing a sufficiently thin layer of highly crystalline nc-Si:H on amorphous surfaces at low temperature is challenging. Depositing p-type doped nc-Si:H (p-nc-Si:H) is even more difficult, as boron doping restricts grain growth. Sophisticated techniques from deposition engineering, including pretreatment methods, adjusting deposition parameters (pressure, temperature, power, frequency and silane dilution) and posttreatment, help to address these issues [21][22][23][24][25] . For example, researchers discovered that CO 2 treatment facilitates fast nucleation for the growth of nanocrystalline silicon film 26 . High crystalline volume fraction (X C )-up to 50%-has been reported by several groups through finetuning gas flow rates 21,25 . Umishio et al. 27 found a clear relation between the nanostructure evolution of the p-nc-Si:H layers and their electrical properties and resulting performance. They concluded that surface coalescence of the p-nc-Si:H nanocrystals, rather than doping concentration, dominantly determines the E a of the film 27 . In other words, hole selectivity and hole transport through the TCO-p-nc-Si:H contact and hence, the solar cell's V OC and fill factor (FF) are crucially influenced by the electrical contact properties between TCO and the crystalline phase at the surface of the p-nc-Si:H layer 19 . Based on this understanding, the contact resistivity of p-nc-Si:H-based HSCs has been reduced to about 100 mΩ cm 2 , yielding a series resistance (R S ) of 0.65-1.3 Ω cm 2 and a 23-25% PCE (Supplementary Table 1) [27][28][29][30][31] . Nevertheless, realizing the advantageous electrical properties of SHJ solar cells over their TOPCon counterparts still poses a challenge, and the opportunity to unlock the full potential of SHJ technology remains open.
In this contribution, we report the successful introduction of nanocrystallization processes for fabricating cutting-edge HSCs, which-when paired with correspondingly tailored TCO-result in improved PCEs and FFs on wafer-scale single-junction SHJ solar cells. We demonstrate a 26.30% SHJ solar cell with an FF of 86.59%; to the best of our knowledge, this FF outperforms any other silicon solar cell. By reducing the shading ratio from 2.8 to 2.0% and modifying the window layers at the front to minimize the parasitic absorption, we further boost the PCE to 26.74% by increasing the short-circuit current density (J SC ) to 41.16 mA cm −2 . Finally, by introducing an additional reflective MgF 2 /Ag stack at the rear side and regulating the transmittance of TCO, we achieve a PCE of 26.81% with a J SC of 41.45 mA cm −2 . Investigation of the power and series resistance losses reveals outstanding performance of these HSCs, with reduced contact resistivity (ρ c < 5 mΩ cm 2 ) and improved passivation (recombination current density J 0 = 0.5 fA cm −2 ). Structural and electrical characterizations of the p-nc-Si:H indicate high X C (63%), ultralow E a (<115 meV) and excellent σ (>3 S cm −1 ), which is four orders of magnitude higher than that of the traditional p-a-Si:H film 32,33 . HSCs endowed with these electrical improvements easily trigger band-to-band tunnelling (BTBT) transport behaviour and induce a sharp band bending, enhancing hole extraction efficiency.

Efficiency increase analysis of SHJ solar cells
In mass production, the competition between SHJ and TOPCon technologies is fierce. As can be seen from Fig. 1a, SHJ solar cells feature greater electrical performance measured by V OC × FF, while TOPCon and PERC hold relatively superior J SC . The inferior J SC of SHJ solar cells can be attributed to the strong parasitic absorption inherent in the functional layers at the front side; PERC and TOPCon usually yield higher J SC (>41 mA cm −2 ) due to the use of conventionally diffused front junctions and optically transparent antireflective coatings 34 . Benefiting from the unique design of rear-sided passivating contact with an SiO x /poly-Si(n + ) stack, TOPCon wins out over PERC with intrinsically improved V OC (ref. 35). SHJ produces the highest V OC among the c-Si solar cell Article https://doi.org/10.1038/s41560-023-01255-2 by the measured sheet resistance of the TCO, the line resistance of the finger, the contact resistivity and recombination of the heterojunction. A full description of these parameters is in Supplementary Table 4. For comparison, the corresponding intrinsic light J-V curves computed from the theoretical limit in electrical performance are also shown (grey lines). Observe that the deviation between the Suns-V OC curve and the light J-V curve (arrows enclosed by black circles) differs between the p-a-Si:H cell and the p-nc-Si:H cell. To explain this, note that we can define a pseudo-fill factor (pFF) from a Suns-V OC curve ignoring all resistance. Both the p-a-Si:H cell and p-nc-Si:H cell have similar pFF values, which means that the difference in FF at the maximum power point (MPP) originates from a difference in the series resistance R S . Therefore, a smaller gap between the light and pseudolight J-V curves in the p-nc-Si:H cell indicates a smaller R S .
To explore how the record FF of 86.59% was achieved, we fit the J-V curves in Fig. 2b,c to determine power loss and series resistance, seen in Fig. 2d. The losses are divided into three regions: front ESC, bulk silicon (bulk) and rear HSC. Figure 2d also shows the series resistance determined from the power loss at MPP. From the p-a-Si:H cell to p-nc-Si:H cell, the total electrical power loss is reduced from 0.41 to 0.13 mW cm −2 at the rear HSC alone; this is nearly equal to the full power loss reduction, 0.3 mW cm −2 (from 1.01 to 0.71 mW cm −2 ). This indicates that the improvement in electrical performance comes primarily from the rear HSC (as expected). In particular, the reduction of power loss at the rear HSC is attributed to the use of p-nc-Si:H and the updated TCO, which facilitates excellent passivation and contact performance. The series resistance analysis shows the same trend as the PLA. The total R S is reduced from 381 mΩ cm 2 for the p-a-Si:H cell to 206 mΩ cm 2 for the p-nc-Si:H cell ( Supplementary Fig. 2a); this difference mostly appears at the rear HSC, at which R S is reduced from more than 130 mΩ cm 2 to less than 20 mΩ cm 2 (Fig. 2d). The slightly increased power loss in the bulk seen in the p-nc-Si:H cell could be caused by a small fluctuation in wafer quality. A slightly reduced R S of the front ESC (from 47 to 41 mΩ cm 2 ) lies in the reduction of the sheet resistance of front TCO from ~150 to ~50 Ω per square. Note that to ensure the credibility of the R S data, we measured the total R S at MPP with four different methods 41 , obtaining R S = 353-381 mΩ cm 2 (p-a-Si:H cell) and 175-206 mΩ cm 2 (p-nc-Si:H cell), as shown in Supplementary Fig. 2b.

Characterization of p-nc-Si:H
As discussed above, the implementation of p-nc-Si:H together with matched TCO leads to a dramatic reduction in the contact resistivity at the rear side, resulting in an efficiency increase to 26.30%. As this contact, the overall resistance depends mainly on the bulk resistance of p-nc-Si:H itself and on the contact resistivity at the p-nc-Si:H-TCO interface. Therefore, gaining the optimal p-nc-Si:H layer is of critical importance to achieve high-efficiency SHJ solar cells. We investigated p-nc-Si:H layers grown from different recipes; corresponding deposition conditions and characterization results are elaborated in Fig. 3 and Supplementary Table 5. We studied the structural properties of the p-a-Si:H and p-nc-Si:H layers by both Raman spectroscopy and transmission electron microscopy (TEM). The Raman measurement was carried out with a 325-nm laser and performed on the p-a-Si:H or p-nc-Si:H layer deposited on a planar i-a-Si:H-glass substrate. The deposition process was identical to that of the fabrication procedure for solar cells. The Raman spectra were fitted with multiple (or single) Gaussian functions, as shown in Fig. 3a. For p-nc-Si:H, three Gaussian peaks are identified with the centres at 482, 507 and 518 cm −1 , representing separately the transverse optical phonon mode of amorphous silicon and two optical vibrational modes of nanocrystalline silicon 42,43 . In contrast, only one peak at 478 cm −1 is observed for p-a-Si:H, indicating a fully amorphous structure. X C is determined from the integrated intensities (I) of the Gaussian peaks via equation (1) as follows 44,45 :  where β = 0.8 represents the ratio of the back-scattering cross-sections 46 .
For the p-nc-Si:H sample, an X C of 63% is obtained. The cross-sectional TEM images of the HSCs based on p-a-Si:H and p-nc-Si:H are shown in Fig. 3b,c. The samples, namely TCO-p-nc-Si: H-i-a-Si:H-n-Si and TCO-p-a-Si:H-i-a-Si:H-n-Si, were taken by focused ion beam from as-prepared SHJ solar cells. In the TEM images, each layer in the stack is identified by its thickness, which was indivi dually measured and confirmed by ellipsometry. In both samples, the thickness of the i-a-Si:H layer is almost same (about 6 nm), while the p-nc-Si:H layer (about 21 nm) is much thicker than that of p-a-Si:H (about 5 nm). A thicker p-nc-Si:H layer is required for incubation and for the subsequent phase transition from amorphous to crystalline. The higher X C of the p-nc-Si:H layer is evidenced by the observation that nanocrystalline domains embedded in an amorphous phase coalesce laterally to one another. No crystalline feature is observed in the i-a-Si:H and p-a-Si:H layers. To identify nanocrystalline domains, fast Fourier transforms (FFTs) were performed on the TEM images (insets in Fig. 3b,c). In both samples, the reciprocal spots derived from the c-Si substrate are highlighted by red circles; these include high-order spots. In the p-nc-Si:H sample, additional reciprocal spots derived from nanocrystalline domains can be seen. We performed inverse FFT calculations on the FFT images, masking selected reciprocal spots, to identify the corresponding nanocrystalline domains in the TEM images. The nanocrystalline domains with different crystalline orientation in the   where the crystalline phase should be localized within a thin layer to balance out detrimental parasitic absorption and resistive transport. Here, the strategy for highly crystalline p-nc-Si:H involves fast nucleation at the initial stage, which was facilitated by CO 2 plasma treatment on its underlying i-a-Si:H layer 47,48 , and selective removal of the amorphous fraction by the plasma with high hydrogen dilution. This led to the accumulation of dense crystallites into the a-Si:H matrix 28 . Apart from the X C , electrical characteristics such as σ and E a can reflect the quality of doped films. As shown in Fig. 3d, the p-nc-Si:H layer on glass exhibits a high σ > 3 S cm −1 and a low E a < 16 meV. By contrast, the p-a-Si:H layer on glass shows values of σ ≈ 10 -4 S cm −1 and E a ≈ 350 meV, respectively. The higher σ and lower E a values are thus obtained from the p-nc-Si:H layer. We computed E a from the equation where k B is the Boltzmann constant and T is the temperature in kelvin 49 .
The values of E a for p-type semiconductors describe the energy difference between the Fermi level and the valence band maximum.
A lower E a in the p-nc-Si:H layer, therefore, indicates higher effective doping and work function. Note that the test films are deposited on two types of substrates (SiO 2 -coated textured silicon wafer and planar glass) using the same deposition processes. Correspondingly, the deposited p-nc-Si:H layer on glass is about 1.7 times thicker than the layer grown on the pyramidal textured surface of c-Si wafer. Considering  49 . e,f, Equilibrium band diagrams of HSCs based on p-a-Si:H (e) and p-nc-Si:H (f) related to the cross-sectional structures in b and c. E C , E V and E F denote conduction band energy, valence band energy and Fermi level, respectively. Insets: enlarged view of the black wire frames; there, ΔE equals the difference between E F and E V at the i-a-Si:H-n-Si interface. The collection path of holes across the heterojunction is depicted as a more complicated curve (red arrows) for p-a-Si:H to illustrate a more challenging transport mechanism at the relative interfaces, with respect to the p-nc-Si:H counterpart. Current is generated once the holes meet and recombine with the electrons (along blue arrows) at the interface of the TCO-hole transport layer. , the much smaller E a value of the p-nc-Si:H layer on glass than that on the c-Si wafer is caused by variations in the thickness-dependent crystallization fraction. It should also be noted that the above tests are on samples without light soaking; the E a plots of the samples under light soaking are shown in Supplementary Fig. 4. With light soaking, the E a and σ values of both films are slightly improved, except for the σ of p-a-Si:H; this increases by a factor of two to three. Although light soaking is used in final devices, it occurs at the front side of the solar cell. Thus, we assume that the properties of HSCs at the rear side are negligibly influenced, and the p-a-Si:H-and p-nc-Si:H-layer properties without light soaking should approximate conditions seen by real devices.
To evaluate the real device performance more accurately, we chose E a values of 110.7 meV (346.8 meV) for p-nc-Si:H (p-a-Si:H) tested on textured silicon wafers as input parameters for simulated band diagrams. Here, to fit the tested I-V curves in Supplementary Fig. 5, the simulated I-V curves in Supplementary Fig. 6 were carefully regulated by adjusting the simulation parameters so that the accuracy of the technology computer-aided design (TCAD) simulated band diagrams shown in Fig. 3e,f is ensured. The p-layer change leads to adjustments from three aspects: (1) conductivity of the p layer itself, (2) band bending at the c-Si surface and (3) carrier transport at the TCO-p-layer interface. All the three aspects have favourable influences on minimizing resistive losses in the HSC and consequently, yielding maximal FF in the devices. The relevant interpretation is elaborated as follows.
First, the p-nc-Si:H layer presents a much higher dark conductivity than the p-a-Si:H layer, which is beneficial for reducing resistive loss in the p layer itself. Second, compared with the p-a-Si:H case, the p-nc-Si:H layer has favourable band bending at the c-Si surface region. This is reflected by the smaller gap between the valence band maximum and the Fermi level (ΔE) (insets in Fig. 3e,f). The reason for this is the lower E a , hence higher work function, of the p-nc-Si:H sample. The accumulation of holes at the surface of n-Si for the p-nc-Si:H sample is also much higher than that of the p-a-Si:H case, leading to a larger difference between hole and electron concentrations and thus, an improvement in electrical passivation. Moreover, this enhanced band bending at the p layer-i layer-n-Si stack provides a lower and sharper energy barrier, which favours a collection of holes through tunnelling at the HSC.
Accordingly, lower J 0 and ρ c in the p layer-i layer-n-Si stack is obtained. Third, holes in the p-layer valence band recombine with electrons in the TCO conduction band. As elaborated by Procel et al. 50,51 , the ρ c variations at the interface are linked to the dominating carrier transport mechanisms and can be influenced by the energy alignment of the p layer-TCO contact. The p-nc-Si:H-TCO interface features a lower hole transport barrier than the p-a-Si:H-TCO case, as indicated in Fig. 3e,f. According to Fig. 3f, the energy alignment at the p-nc-Si: H-TCO interface could facilitate carrier transport through dominating BTBT, which is widely considered to be the most efficient carrier transport mechanism. By contrast, at the p-a-Si:H-TCO interface, trap-assisted tunnelling, which is usually a less efficient tunnelling mechanism than BTBT, is likely to dominate; this is confirmed by our simulated results in Fig. 3e. Therefore, a higher ρ c at the p-a-Si:H-TCO interface is to be expected. It should be mentioned that the band gap of p-nc-Si:H (~1.7-2 eV) is simplified to be consistent with that of i-a-Si:H because it has little effect on the simulation result (J-V) if the E a of the doped layer is small enough [50][51][52] .

Evaluation of holes selectivity
The replacement of the p-a-Si:H layer with a p-nc-Si:H layer improves the passivation quality and reduces ρ c values at both the p layer-i-a-Si:H-n-Si contact and the TCO-p layer interface. These positive effects contribute to delivering a larger hole selectivity of the HSCs. Figure 4a shows the distributions of simulated solar cell efficiency and hole selectivity as functions of ρ c and J 0 . We extract ρ c and J 0 of HSCs based on p-nc-Si:H directly from the measurements shown in Fig. 4b,c, while those of p-a-Si:H are evaluated according to the fitting parameters in Fig. 2 and Supplementary Fig. 1 Article https://doi.org/10.1038/s41560-023-01255-2 Three methods-the Cox and Strack method (CSM) 57 , the expanded Cox and Strack method (ECSM) 58,59 and the transfer length method (TLM) 60 -were utilized to extract the ρ c of HSCs based on p-nc-Si:H. Emphasis was placed on the ECSM, which can directly extract the ρ c values of the TCO-p-nc-Si:H-i-a-Si:H-n-Si stack, as shown in Fig. 4b. The other two methods, which evaluate the ρ c through measuring symmetric samples on a p-type silicon wafer (p-Si), are shown in Supplementary Fig. 7. The coloured lines in Fig. 4b are the dark current-voltage (dark I-V) curves of the ECSM disks with different diameters (the same as the CSM disks shown in Supplementary  Fig. 7a but with the n-Si wafer as the substrate). The inset in Fig. 4b shows total series resistance (R t ) from different disks plotted against inverse area (1 S −1 ). From the linear fit (red dashed line), we can extract a ρ c of 3.6 mΩ cm 2 . For comparison, the results extracted from CSM and TLM are 4.2 and 6.5 mΩ cm 2 , respectively.
A symmetric structure, shown in the inset of Fig. 4c, was used to characterize J 0 for the HSC based on p-nc-Si:H. The transient mode of a Sinton measurement was used for testing minority carrier lifetime, and the results are indicated by red circles. The intrinsic recombination mode presented by Black and Macdonald 61 was used to fit the data, tuning the parameters of SRH recombination (τ SRH; reflecting the quality of silicon wafer) and J 0 (representing the quality of surface passivation). The fitted total J 0 is 1.0 fA cm −2 . Considering the symmetry of the structure, it should be 0.5 fA cm −2 for the J 0 of the p-nc-Si:H-i-a-Si:H-n-Si HSC.
The achievements in electrical performance of the p-nc-Si:H cell (26.30% PCE) define a state-of-the-art SHJ process on which advanced optical designs can be adopted to further promote PCE. As shown in Fig. 5a, with the same BJ processing as the p-nc-Si:H cell, reducing the shading ratio from 2.8 to 2.0% after using the laser transfer process and modifying the window layers of n-nc-SiO x :H, i-a-Si:H and TCO at the front (such as reducing the thickness of n-nc-SiO x :H and i-a-Si:H, increasing the oxygen content of n-nc-SiO x :H and decreasing the carrier density of TCO) to minimize the parasitic absorption, we achieved a higher PCE of 26.74% with the increase of J SC to 41.16 mA cm −2 . The slight reduction of FF is mainly due to the change of front metal fingers. In addition, V OC increases by 0.9 mV, possibly enhanced by the higher J SC . The external quantum efficiency spectrum and main optical loss of each layer are shown in Fig. 5b. The absorption by n-Si fits well with the Institute for Solar Energy Research in Hamelin (ISFH)-certified one, and the slight deviation at short wavelength may arise from the refractive indices used in the simulation. The front grid electrode causes a decrease of 0.97 mA cm −2 in J SC , accounting for a shading fraction of 2.0%. In the ultraviolet range, the front TCO, n-nc-SiO x :H and i-a-Si:H dominate the loss due to parasitic absorption. In the near-infrared range, the major source of loss is the escape of light from the front, which is attributed to reflection by the test chunk due to the limited wafer thickness. With introducing an additional reflective MgF 2 /Ag stack at the rear side and regulating the transmittance of TCO, the J SC was further improved to 41.45 mA cm −2 due to the optical gain, while FF decreased to 86.07% due to the decrease of TCO conductivity. Finally, the PCE was improved to 26.81%. The certified J-V curve of the device is reported in Supplementary Fig. 8.

Conclusion
Through introducing nanocrystallization technology in the doped layer at carrier-selective contacts for both polarities, we achieve a record efficiency of 26.81% and on a different cell, an extremely high FF (86.59%) on M6-sized BJ SHJ solar cells. Two types of SHJ solar cells equipped with p-type transporting layers of amorphous silicon and nanocrystalline silicon are comprehensively investigated; we study their power loss, contact resistivity, transport mechanism and so on. Structural and electrical characterizations of the boron-doped nanocrystalline silicon films indicate that a higher X C (>63%), an increase in conductivity of four orders of magnitude (in comparison with p-type amorphous silicon) and an ultralow E a (<115 meV) are the main causes of its excellent electrical performance. Because of the low E a of p-nc-Si:H, it is easy to enable BTBT at the n-type TCO-p-nc-Si:H interface and achieve sharp band bending at the i-a-Si:H-n-Si interface, facilitating efficient transport and collection of holes across the whole junction. Our study shows that the implementation of p-nc-Si:H together with a modified TCO greatly reduces the contact resistivity of HSC, from >100 to <5 mΩ cm 2 . The total series resistance of the solar cell is reduced from the original 0.37 to 0.2 Ω cm 2 , yielding a record FF for single-junction silicon solar cell.

Solar cell fabrication
In this work, solar cells were fabricated by the commercial SHJ research and development line on LONGi M2 (the 25.26% efficiency SHJ solar cell) or on an M6 Czochralski n-Si wafer with a resistivity of 1.2-1.5 Ω cm and a thickness of 130 μm in (100) orientation. The n-Si wafer was first cleaned and textured by a wet chemical process. Before subsequent deposition, the thickness of the n-Si wafer was confirmed by weight measurement. Radiofrequency plasma-enhanced chemical vapour deposition was used to prepare i-a-Si:H, and then, very high-frequency plasma-enhanced chemical vapour deposition (40.68 MHz) systems   Fig. 2), and the other is referring to a simple recombination model (as shown in Supplementary Fig. 1). For the Quokka2 simulation, the unit cell was modelled in two dimensions to calculate the power loss in the transversal transport of carriers between two fingers. The input parameters were primarily obtained from measurements 'Characterizations'. The line resistance of the finger and the contact resistivity of the heterojunction were considered as series resistance in an external circuit. The optical path-length factor (Z) was set as 4n 2 , and the transmittance was adjusted to match the simulated J SC to that of actual cells. The τ SRH of the silicon wafers was set by adjusting the values of σ n , σ p and the defect density of the SRH defect (N t ). τ SRH , J 01 (surface recombination) and R s were further tuned to obtain agreement between the light J-V and the Suns-V OC curves. Richter's Auger mode was chosen, and the value of radiation recombination was changed to 0.4 × B rad with a photon recycling probability of 0.6 (ref. 65).
Other parameters in the simulation are listed in Supplementary Table 4. It should be noted that since band gap narrowing is not considered in this model, the corresponding V OC may deviate somewhat from the experimental value. For the simulation by the simple recombination model method, we first assume a uniform quasi-Fermi level and then fit the Suns-V OC and light J-V curves by adding the formulas of intrinsic recombination, SRH recombination, surface recombination, series resistance and so on. The specific fitting parameters, consequent results and corresponding descriptions are shown in Supplementary Discussion 1 and Supplementary Fig. 1.

Electrical simulation
The band diagrams in Fig. 3e,f and the dark I-V curves in Supplementary  Fig. 6 were calculated by using Sentaurus TCAD based on drift-diffusion models. The simulation structure of the testing device is illustrated in Supplementary Fig. 7a but with n-Si as the substrate: that is, a stacked film of Ag-TCO-p-nc-Si:H (or p-a-Si:H)-i-a-Si:H on bulk n-Si with a disk shape on the front surface and a full-area ohmic contact on the rear surface of the n-Si substrate. Physical models, including surface recombination, Auger recombination in substrate and thermionic emission at proper interfaces, are taken into consideration. Moreover, p-a-Si:H and p-nc-Si:H layers feature spatially uniformly distributed traps in the exponential and Gaussian energy distributions 50 . Trap-assisted tunnelling and BTBT models are also considered at the TCO-p-nc-Si:H (or p-a-Si:H) interface. The specific film parameters are shown in Supplementary Table 7. The simulated results are consistent with the experiments (Supplementary Figs. 5 and 6 and Supplementary Discussion 2).

Optical simulation
The optical simulation in Fig. 5b was performed using the SunSolve software provided by PV Lighthouse. In the simulation, we use refractive indices from the literature 66-70 except for the TCO film, which we measured by ellipsometry. In SunSolve, ray tracing and thin-film optics were adopted with a Monte Carlo algorithm for sampling and averaging the absorption in each layer. Here, 1 million rays with zero zenith angle and air mass 1.5G were randomly generated for better confidence of the statistics; the 95% confidence interval of cell absorption is less than 0.01 mA cm −2 .

Characterizations
The light J-V curves of the solar cells were tested and certified by ISFH ( Supplementary Fig. 8); gold-coated brass chuck was used to mount the solar cells, and the resistance of the grid at the rear side and of bus bars at the front side was neglected. The pseudo light J-V curves were extracted from a Suns-V OC measurement. The Suns-V OC module of a Sinton WCT-120 instrument was used to collect the changes in voltage of the device by reducing the light intensity of the flashlight; these were computationally transformed them into Suns-V OC curves. Raman spectra were obtained with a Horiba LabRAM Odyssey Raman spectrometer with a 325-nm excitation laser. Test films were deposited on cleaned glass substrates using the same deposition processes as the solar cells. The morphology of heterojunctions, consisting of TCO-coated doped amorphous and nanocrystalline silicon films on c-Si bulk, was observed by TEM, and the test samples were prepared by the focused ion beam method. Activation energies were calculated from measurements of electrical conductivity as a function of test temperature. Electrical conductivities were calculated from I-V curves measured by a semiconductor analyser (Keithley 4200A-SCS). It should be noted that the 50-nm isolation layer of SiO x film was deposited by plasma-enhanced chemical vapour deposition before the deposition of p-nc-Si:H or p-a-Si:H on the pyramid-textured silicon wafers. Contact resistivity tests were carried out using the CSM, ECSM and TLM; the details are shown in Fig. 4b and Supplementary Fig. 7. Sinton's minority carrier lifetime test was used to characterize the passivation quality on a symmetric structure of p-nc-Si:H-i-a-Si:H-n-Si-i-a-Si:H-p-nc-Si:H (Fig. 4c). The transient mode of a Sinton measurement was used for testing the minority carrier lifetime. Pyramid-textured n-Si wafers with a thickness of about 130 μm and resistivity of about 1.59 Ω cm were used as the substrates.

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

Data availability
All data generated or analysed during this study are included in the published article and its Supplementary Information. Source data are provided with this paper.