Nowadays, the market share of p-type silicon (p-Si) solar cells is about 70%^[1], which is still the main product in the photovoltaic market. However, the conversion efficiency is limited due to the surface recombination^[2]. Therefore, passivated contact crystalline silicon (c-Si) solar cells, such as tunnel oxide passivated contact (TOPCon) solar cells^{[3, 4]} and passivated emitter & rear totally-diffused (PERT) bifacial solar cells^[5], have been suggested due to the advantages of both high conversion efficiency and compatibility with the production line of conventional solar cells^[6−8]. The TOPCon structure is the most attractive because it exhibits immense potential for future development: The market share of TOPCon is expected to be 70% in 2033^[1]. To fabricate the passivated contact layer of polysilicon/silicon oxide stack, amorphous silicon films are deposited on top of ultrathin silicon oxide layer, followed by crystallization to form polysilicon layers. Various techniques have been employed for the deposition of amorphous silicon layers, such as low pressure chemical vapor deposition (LPCVD)^{[9, 10]}, plasma-enhanced chemical vapor deposition (PECVD)^{[11, 12]}, atmospheric pressure chemical vapor deposition (APCVD)^[13], hot-wire chemical vapor deposition (HWCVD)^{[14, 15]}, and magnetic sputtering^[16]. Generally, gas precursors are hazardous in chemical vapor deposition (CVD) process. This process requires elaborate safety equipments, which poses challenges for the mass production of solar cells. Magnetic sputtering technology is based on physical vapor deposition (PVD). Although it can degrade the interface due to the high-energy ion bombardment^[17], no hazardous gas is involved in the process. In addition, the footprint of the sputtering machines is much smaller than that of CVD equipment.

Normally, a doped polysilicon film can be obtained through two different methods: in-situ doping and ex-situ doping. The former process uses a gaseous dopant precursor such as diborane or phosphine^{[18, 19]} during the deposition of amorphous silicon film. Further, the impurities inside the deposited silicon film are thermally activated by high-temperature crystallization. On the other hand, ex-situ doping is normally realized by additional high-temperature diffusion of impurity sources, such as phosphorus oxychloride (POCl₃)^[20], boron trichloride (BCl₃)^[21], or boron tribromide (BBr₃)^[22], or spin-on dopants such as P-250^[23] and B155^[24]. Both BCl₃ and BBr₃ are hazardous, costly, and non-uniform p-type dopant. For utilizing B155 as the boron source, a time-consuming pretreatment is required, and an undesirable boron-rich layer is formed in this process^[24].

Recently, Tang et al.^[25] proposed ammonium tetraborate tetrahydrate (ATT, (NH₄)₂B₄O₅(OH)₄·2H₂O) as an environment-friendly, cost-effective, and direct boron impurity source that can achieve a uniform diffusion at a relatively low temperature. Using ATT as the direct boron source, a uniform p-type emitter with an average sheet resistance (R_sheet,m) of 80−300 Ω/sq can be obtained at a diffusion temperature less than 850 °C. In this study, ATT is employed as a boron source during the high-temperature crystallization of silicon film to fabricate a p-type polysilicon (p:poly-Si) film through ex-situ doping, with p-type amorphous silicon (p:a-Si) film deposited by magnetic sputtering. The main objective of this work is to investigate the effect of ATT on the structure and electrical properties of p:poly-Si film.

Polished phosphorous-doped n-type Czochralski silicon (n-Si) wafers with <100> orientation, a resistivity of 5 Ω·cm, a thickness of approximately 170 μm, and a size of 20 mm × 20 mm were used as the substrate for silicon film deposition, which were purchased from COMTEC Solar. The polished substrates were firstly cleaned in an ultrasonic bath of deionized (DI) water plus detergent to eliminate all the traces of dust and grease on the surface. The wafers were cleaned according to the RCA procedure, and then they were dipped into 1% (vol.) hydrofluoric acid (HF, 35%) for nearly 1 min to remove the oxides formed during the RCA cleaning. After DI water cleaning, a thin silicon oxide film was formed by immersing the wafers into hydrogen peroxide (H₂O₂) solution at ~25 °C for 10 min. The residual H₂O₂ on the wafer surface was cleaned with DI water. Subsequently, high-purity N₂ gas was flown through the substrates, and then they were degassed at 100 °C for ~1 min under N₂ atmosphere. The cleaned wafers were utilized for the deposition of amorphous silicon (a-Si) film.

A flow chart of the experimental process is shown in Fig. 1. Specifically, B-doped p-type amorphous silicon (p:a-Si) film was deposited by using a radio frequency (RF) magnetron-sputtering system (JDCK-F4000D, Shenyang Judong Vacuum Technology Institute, China). Ultrahigh purity Ar gas was used to generate the necessary plasma. The Ar gas flow rate was finely tuned by a mass flow controller (MFC) and a gate valve to obtain the desired working pressure. The distance between the target and the substrate was ~11 cm. The temperature of the substrate was ~25 °C. The background pressure of the vacuum system was 8 × 10⁻⁴ Pa. B-doped p-type Si (0.01 Ω·cm) with a high purity of 99.999% (5N, Zhong Nuo Advanced Material (Beijing) Technology Co., Ltd) was employed as the Si target. The RF power (at 13.56 MHz) was set to 100 W during all the deposition processes. According to these deposition parameters, the deposition rate was around 2 nm/min. According to Padhamnath et al.^[26], thinner polysilicon layers can lead to higher efficiencies due to the reduced parasitic absorption.

Then, the p:a-Si film was annealed at a high temperature (≤1000 °C) directly (pPoly) or with 3.0% ATT-ethanol solution spin coating (ATT-pPoly) in a quartz tube furnace under continuous flow of high-purity N₂ gas. Instead of water solution, ethanol solution was used to overcome the hydrophobic property of a-Si surface. ATT-ethanol solution was spin coated at a fixed rate of 2000 rpm using a photoresist spinner (SC-1B, Beijing Jinshengweina Technology Co., Ltd., China).

Raman spectroscopy^{[27, 28]} is a non-destructive and rapid technique for the evaluation of crystallinity. Here, it was used to examine the effect of ATT doping on the local crystalline structure of p:poly-Si film. Raman spectroscopy measurements were performed on an Odyssey spectrometer (Horiba, French) with a laser wavelength of 532 nm. The surface morphology of p:poly-Si film was observed by scanning electron microscopy (SEM). The SEM results were analyzed with the nano measure software (NMS)^[29]. To ensure the analysis reliability, 140−150 particles with clear outline in each SEM image were selected, and the average grain size (δ) of the p:poly-Si film was measured from the diagonal direction.

The dopant concentration in p:poly-Si is a key limiting factor for both passivation and contact properties^[30]. The sheet resistance of the p:poly-Si film was measured by a four-point probe system (FT-340, Ningbo Rooko Instrument Co. Ltd., China). Five different measurement points were selected on each wafer, and the results were averaged to obtain the R_sheet,m. To analyze the uniformity of the p:poly-Si layer, a pure dimensionless number, namely the standard deviation coefficient (σ) of R_sheet,m, was adopted. Further, the depth distribution of boron impurity concentration in the poly-Si layer was investigated by secondary ion mass spectrometry (SIMS; TOF. SIMS 5, IONTOF GmbH, Germany)^{[31, 32]}.

Raman spectroscopy was used to analyze the structure of p:poly-Si film doped by ATT. The laser wavelength was long (532 nm), so the characteristic peak of crystalline silicon substrate could also be detected, which was not included in analyzing the crystallinity of the film structure. At the same time, a thicker silicon film (~150 nm) was prepared for Raman spectroscopy measurements to reduce the influence of the substrate.

The Raman spectra and X-ray diffraction (XRD) patterns of the as-sputtered p:a-Si and p:poly-Si films are shown in Fig. 2. The Raman spectra of the samples (blue) exhibit a broad band between 420 and 540 cm⁻¹ with a central peak near 520 cm⁻¹. To further analyze the spectra, the bands are fitted using the Gaussian distribution function to produce three symmetric peaks at ~520, 500, and 480 cm⁻¹. The full width at half maximum (FWHM) and area of each peak are listed in Fig. 2. The narrow and sharp peak at 520 cm⁻¹ (solid pink curve, FWHM ~2 cm⁻¹) is attributed to the Si−Si stretching mode of the crystalline silicon substrate. This peak is excluded in the following quantitative analysis (shown in gray font in each peak list). The peaks at ~518 (solid black curve), 500 (solid violet curve), and 480 cm⁻¹ (solid red curve) correspond to the Si−Si stretching mode of crystallized silicon^[33], grain boundary (GB)^{[34, 35]}, and amorphous silicon^[36], respectively. According to Dorz et al.^[37], the crystallinity (ϕc) can be expressed as follows:

1ϕc=AcAa+AGB+Ac×100%.

Here A_c, A_GB, and A_a are the area under the Gaussian peaks centered at ~520, 500, and 480 cm⁻¹, respectively. The value of ϕc should be considered as a lower limit for the actual crystalline volume fraction.

The Raman spectrum of the sputtered p:a-Si layer (Fig. 2(a)) shows two peaks centered at 520.44 and 480.90 cm⁻¹, respectively. The peak at 520.44 cm⁻¹ comes from the n-Si substrate, while the peak at 480.90 cm⁻¹ can be attributed to the p-type amorphous silicon film^[36]. After high-temperature annealing at 950 °C, however, the broad peak centered at 480.90 cm⁻¹ disappears almost, and two new broad peaks appear in the Raman spectra of both pPoly and ATT-pPoly films (gray and violet solid curve in Figs. 2 (b) and 2(c)). The peak at 518.98 or 518.00 cm⁻¹ can be assigned to the polycrystalline silicon, and the peak at 497.06 or 491.23 cm⁻¹ corresponds to the grain boundaries (GBs)^[35].

Comparing Fig. 2(b) with Fig. 2(c), it is found that the Raman peaks of both polysilicon and grain boundary are redshifted after the addition of ATT: the peak of polysilicon shifts from 518.98 to 518.00 cm⁻¹, and the peak of grain boundary shifts from 497.06 to 491.23cm⁻¹. The redshift of ATT-pPoly may be ascribed to the tensile stress^[38] due to the high-temperature annealing process, which arises from the substitutional doping of boron atom with a smaller size than silicon. The broadening of the ATT-pPoly spectrum may result from the Fano effect^[39], which occurs when the Fermi energy resides in the valence band due to the heavy boron doping, leading to destructive and constructive interference that affects the spectral shape.

Fig. 2(d) shows the XRD patterns of pPoly and ATT-pPoly films. It can be seen that the structure of p:poly-Si films is not altered by the incorporation of ATT: the preferential orientation of both pPoly and ATT-pPoly films is the (111) direction. Besides, the grain size (δ) can be effectively improved by incorporating ATT in the film (Fig. 3). An example of the grain size δ distribution obtained through SEM and NMS is shown in Fig. 3(a). The average grain size δ obtained using the Gaussian fitting function (16.22 nm) is almost the same as that obtained by the log-normal fitting function (16.32 nm). However, the latter is more consistent with the grain size histogram obtained using NMS. Therefore, all δ values in Fig. 3(b) have been obtained using the log-normal fitting function. The average grain size δ of pPoly ranges from 16−23 nm and is nearly independent of the value of the film thickness d. The average grain size δ of ATT-pPoly film has an inverted U-shape. It obviously increases with the addition of ATT, ranging from 21−47 nm, which corresponds to an average increase of ~70%. The improved grain size of the polysilicon film is beneficial for silicon solar cells^[24].

The electrical properties of the silicon films were tested with Hall measurements (HL 5500, Nanometrics, American), including the resistivity (ρ), carrier concentration (p^*), and carrier mobility (μ). The results are shown in Fig. 4. According to Joshi et al.^[40], the electrical properties are mainly determined by the doping concentration and grain size. Compared with the pPoly film, it is clear from Fig. 4(a) that the resistivity ρ of ATT-pPoly is decreased by more than 80%. The carrier concentration p^* of ATT-pPoly is higher (~10 times higher on average than the pPoly film) which is shown in Fig. 4(b). This improvement is ascribed to the increased boron dopant concentration and the enhanced grain size by the usage of ATT. The lower carrier mobility μ of ATT-pPoly film can be attributed to the Coulomb scattering effect due to the higher doping concentration.

In addition, it can be seen from Fig. 4 that the film thickness d is a significant parameter that affects the electrical properties (e.g. the resistivity ρ) of the silicon films. According to the results of Sahraoui et al.^[41], the conduction mechanism of silicon films can be well described by a thermally activated process: r=r0e−(Ea/kBT), where r, r₀, E_a, k_B and T are electrical conductivity, pre-exponential conductivity factor, activation energy, the Boltzmann constant and the temperature in Kelvin, respectively. The thermal activation energy E_a for electrical conductivity is lowered by the increasing crystallinity ϕc^{[41, 42]}. The calculated ϕc from Eq. (1) shows that it decreases with the increase in the thickness d. Specifically, the crystallinity ϕc decreases from 60.50% and 54.80% at 80 nm to 47.76% and 42.40% at 150 nm for pPoly and ATT-pPoly films, respectively. This suggests that the activation energy E_a of the silicon films increases with the increase in d, which may result from the fewer electrically activated impurity. The lower carrier concentration induce the weaker Coulomb scattering effect. So, the carrier mobility μ would increase with the increasing thickness d (Fig. 4(c)). It is important to note that the calculated crystallinity ϕc of ATT-pPoly can only be used to qualitatively analyze the effect of the thickness d on the electrical properties. This is because the crystallinity ϕc of ATT-pPoly is smaller than the actual value due to Fano effect^[39], which can lead to a broad low-energy tail that contributes to the GB peak.

A comparison of the electrical properties of boron doped p:poly-Si films between this work and other studies is presented in Table 1. Both the carrier concentration and mobility of p^* and μ obtained with ATT in this work are better than the reported values due to the heavier doping and improved grain size: the carrier concentration p^* is higher than 1 × 10²⁰ cm⁻³, while the carrier mobility μ retains a high value (>30 cm²/(V·s)). Obviously, ATT is a good boron source to obtain high-performance p:poly-Si film.

Fig. 5 shows the R_sheet,m and σ of films (with a thickness of ~80 nm) annealed at different temperatures for a duration of 45 min. Fig. 5(a) shows the R_sheet,m of the sample prepared at a pressure of 0.2 Pa as a function of annealing temperature. The R_sheet,m of pPoly (black solid line with circular markers) varies from 1148 to 25 800 Ω/sq as the annealing temperature changes from 1000 to 850 °C. The R_sheet,m of ATT-pPoly (blue solid line with circular markers) decreases sharply. It varies from 109 to 16 106 Ω/sq when the annealing temperature decreases. The relative variation of R_sheet,m (pink dashed line with circular markers) between pPoly and ATT-pPoly is nearly −30% at 850 °C, which implies that the R_sheet,m of ATT-pPoly film decreases by at least 30%. Besides, it can be further decreased (by more than 90%) by increasing the annealing temperature. Fig. 5(b) shows the R_sheet,m of the samples prepared at a pressure of 0.5 Pa as a function of the annealing temperature. The R_sheet,m of pPoly film (dark yellow solid line with circular markers) varies from 412 to 5474 Ω/sq as the annealing temperature changes from 1000 to 865 °C. The R_sheet,m of ATT-pPoly film (orange solid line with circular markers) decreases sharply. It varies from 106 to 2693 Ω/sq when the annealing temperature decreases. The relative variation is at least −50% (magenta dashed line with circular markers). Besides, it can be further decreased (by nearly −90%) by increasing the annealing temperature. Yan et al.^[16] doped the p:poly-Si layer by co-sputtering method, and the R_sheet,m was ~250 Ω/sq after annealing at 950 °C. In another study^[11], the p:poly-Si layer was doped with BBr₃, and the R_sheet,m was 200−1.5 × 10⁴ Ω/sq, which depended on the annealing temperature (870−980 °C) and interfacial tunneling layer. In the present study, the R_sheet,m of the sample annealed at 950 °C is ~150 Ω/sq. Consequently, a lower R_sheet,m can be obtained with ATT. Obviously, irrespective of the sputtering pressure (0.2 or 0.5 Pa), the incorporation of ATT as a boron source can effectively reduce the R_sheet,m of the p:poly-Si film. In addition, the uniformity becomes better with ATT spin coating, as shown in Figs. 5(c) and 5(d): the σ of ATT-pPoly is less than 5.0% at annealing temperature >850 °C for 45 min, and it further decreases with the increase in annealing temperature. This is further confirmed in the following experiments.

Fig. 6 shows the R_sheet,m and σ of the p:poly-Si films (with a thickness of ~80 nm) annealed at 900 °C for different annealing durations. Fig. 6(a) shows the R_sheet,m of the sample prepared at a pressure of 0.2 Pa as a function of the annealing duration. The R_sheet,m of pPoly film (black solid line with circular markers) varies from 2136 to 6374 Ω/sq when the annealing duration decreases from 120 to 45 min. The R_sheet,m of ATT-pPoly film (blue solid line with circular markers) sharply decreases (by at least 30%) with the incorporation of ATT. It varies from 250 to 4428 Ω/sq as the annealing duration decreases. The relative variation is less than −80% when the annealing duration is longer than 60 min (pink dashed line with circular markers). Fig. 6(b) shows the R_sheet,m of the samples prepared at a pressure of 0.5 Pa as a function of the annealing duration. The R_sheet,m of pPoly film (dark yellow solid line with black circular markers) varies from 1489 to 5436 Ω/sq when the annealing duration decreases from 120 to 45 min. The R_sheet,m of ATT-pPoly film (orange solid line with circular markers) decreases sharply, by at least 60%. It varies from 131 to 1920 Ω/sq when the annealing duration decreases. The relative variation is less than −90% when the annealing duration is longer than 60 min (magenta solid line with circular markers). Furthermore, the film uniformity is better with ATT spin coating, as shown in Figs. 6(c) and 6(d): the σ of ATT-pPoly is less than 5.0% when it is annealed at 900 °C for more than 45 min, and it further decreases with the increase in annealing duration. In addition, according to the results of Meyer et al.^[50], the sheet resistance of polycrystalline film for silicon solar cells is generally ~200−300 Ω/sq. This can be realized for ATT-pPoly at a relatively low annealing temperature of 900 °C as the duration is 60 min which is impossible for pPoly, shown in Fig. 6(b).

To further study the impurity distribution in the film, SIMS measurements were carried out, and the results are shown in Fig. 7. It is evident that the boron concentration of ATT-pPoly film (blue circles) is nearly 10 times higher than that of pPoly film (black circles). This further proves that ATT can be used to obtain a heavily doped p:poly-Si layer with a concentration of ~3 × 10²⁰ cm⁻³ at 950 °C for 45 min, about 2 times higher than that obtained by Park et al. using spin coating^[24]. In Ref. [51], the boron concentration of p:poly-Si film annealed at 875−1050 °C with borosilicate glass (BSG) layer deposited by APCVD was less than 1 × 10²⁰ cm⁻³. Using the laser doping method, Wuu et al.^[52] found that the boron dopant concentration of p:poly-Si layer was ~8 × 10¹⁹ cm⁻³. All these results further verify that ATT can realize a much heavier doped polycrystalline silicon film.

Besides, the following conclusions can be obtained from the SIMS results: (1) A large amount of boron is accumulated at the surface (~0 nm) of p:poly-Si layer in both pPoly and ATT-pPoly films. This may be attributed to the oxidation of p:a-Si film under a long-term exposure to the air or to the segregation phenomenon on the surface during the annealing process. (2) The boron concentration of the pPoly sample decreases sharply with the depth near the surface (≤20 nm), while that of the ATT-pPoly sample also decreases, but slowly. (3) The boron concentration in the p:poly-Si layer (≥20 nm) continuously decreases in the pPoly sample, but it remains almost unchanged in the ATT-pPoly sample. (4) Notably, the "p:poly-Si/n-Si" interface (pink solid line in Fig. 7) is not clearly defined, possibly due to the high etching rate in the SIMS test process^[53].

In this study, heavily doped polycrystalline silicon films (ATT-pPoly films) were successfully fabricated by sequential processes of p:a-Si magnetic sputtering, ATT solution spin coating, and high-temperature annealing. The ATT source solution concentration was fixed at 3.0%. To understand the effects of ATT on the structure and properties of the p:poly-Si film, pPoly films were fabricated in the same batch with ATT-pPoly as a reference through direct annealing of the p:a-Si films. Both ATT-pPoly and pPoly films were composed of a mixture of GB and crystalline regions. An obvious redshift in the Raman spectrum was observed in both pPoly and ATT-pPoly films. It was further downshifted for ATT-pPoly film, accompanied with a broader line shape. The preferred orientation of ATT-pPoly film was the (111) direction, similar to the pPoly film. Further, the incorporation of ATT was beneficial to the grain size of the polycrystalline silicon film. The grain size increased by ~70% on average. Besides, the electrical properties of the polysilicon film could be improved by ATT: (1) The resistivity ρ was reduced by more than 80%. (2) The carrier concentration p^* was increased by ~10 times on average. (3) The carrier mobility μ was reduced due to the higher Coulomb scattering. However, the carrier mobility μ obtained in this work by incorporating ATT was better than the previously reported values. The R_sheet,m sharply decreased by more than 30% by the addition of ATT, irrespective of the sputtering pressure, and it could be further decreased by ~90% through the increase in annealing temperature or duration. With ATT, a favorable uniform p-type polysilicon layer could be obtained: the σ of ATT-pPoly film, which was annealed at a temperature higher than 950 °C for 45 min or at 900 °C for a time longer than 45 min, was less than 5.0%. The measurement results of SIMS showed that the boron concentration distribution of the ATT-pPoly film annealed at 950 °C for 45 min was almost constant, and the concentration was ~3 × 10²⁰ cm⁻³, except near the surface. The concentration was higher than that reported in previous studies.