Pulsed Laser Annealed Ga Hyperdoped Poly‐Si/SiOx Passivating Contacts for High‐Efficiency Monocrystalline Si Solar Cells

Polycrystalline Si (poly‐Si)‐based passivating contacts are promising candidates for high‐efficiency crystalline Si solar cells. We show that nanosecond‐scale pulsed laser melting (PLM) is an industrially viable technique to fabricate such contacts with precisely controlled dopant concentration profiles that exceed the solid solubility limit. We demonstrate that conventionally doped, hole‐selective poly‐Si/SiOx contacts that provide poor surface passivation of c‐Si can be replaced with Ga‐ or B‐doped contacts based on non‐equilibrium doping. We overcome the solid solubility limit for both dopants in poly‐Si by rapid cooling and recrystallization over a timescale of ∼25 ns. We show an active Ga dopant concentration of ∼3 × 1020 cm−3 in poly‐Si which is six times higher than its solubility limit in c‐Si, and a B dopant concentration as high as ∼1021 cm−3. We measure an implied open‐circuit voltage of 735 mV for Ga‐doped poly‐Si/SiOx contacts on Czochralski Si with a low contact resistivity of 35.5 ± 2.4 mΩ cm2. Scanning spreading resistance microscopy and Kelvin probe force microscopy show large diffusion and drift current in the p‐n junction that contributes to the low contact resistivity. Our results suggest that PLM can be extended for hyperdoping of other semiconductors with low solubility atoms to enable high‐efficiency devices.

Polycrystalline Si (poly-Si)-based passivating contacts are promising candidates for high-efficiency crystalline Si solar cells. We show that nanosecond-scale pulsed laser melting (PLM) is an industrially viable technique to fabricate such contacts with precisely controlled dopant concentration profiles that exceed the solid solubility limit. We demonstrate that conventionally doped, hole-selective poly-Si/SiO x contacts that provide poor surface passivation of c-Si can be replaced with Ga-or B-doped contacts based on non-equilibrium doping. We overcome the solid solubility limit for both dopants in poly-Si by rapid cooling and recrystallization over a timescale of ∼25 ns. We show an active Ga dopant concentration of ∼3 × 10 20 cm −3 in poly-Si which is six times higher than its solubility limit in c-Si, and a B dopant concentration as high as ∼10 21 cm −3 . We measure an implied open-circuit voltage of 735 mV for Ga-doped poly-Si/SiO x contacts on Czochralski Si with a low contact resistivity of 35.5 AE 2.4 mΩ cm 2 . Scanning spreading resistance microscopy and Kelvin probe force microscopy show large diffusion and drift current in the p-n junction that contributes to the low contact resistivity. Our results suggest that PLM can be extended for hyperdoping of other semiconductors with low solubility atoms to enable high-efficiency devices.
has a nearly three orders of magnitude higher segregation coefficient in Si versus SiO x (k Ga ∼ 20 and k B ∼ 10 −2 ), [35] which combined with the higher diffusivity prevents accumulation of Ga in SiO x and at the SiO x / c-Si interface. While the iV oc for the Ga-doped contact in Young et al. [32] was better than B-doped contacts, the higher contact resistivity limits the fill factor after various annealing methods, ranging from 850°C for 30 min to 950°C for 1 s. It was determined that not enough Ga dopants were activated under these annealing conditions. Thus, in order to activate enough Ga in poly-Si to have a low contact resistivity to the metal contacts, non-equilibrium doping above the solubility limit (hyperdoping) needs to be introduced.
The maximum solid solubility limit of Ga in monocrystalline Si has been reported as 4 × 10 19 cm −3 at 1200°C by Trumbore in 1959, [36] and the solid solubility limit is often lower (for example, 1 × 10 19 cm −3 at common cell processing temperatures of ∼900°C [37] ). However, doping concentrations >10 20 cm −3 are desired for various applications, such as shallow p + junction, [38,39] superconducting semiconductors, [40][41][42] and low contact resistivity p-type contacts for solar cells. [32,43] Conventional thermal processes cannot reach a high level of doping concentration, so to overcome the equilibrium solid solubility limit, some early studies in the 1980s based on non-equilibrium processes have shown supersaturated Ga in silicon with a substitutional doping level of 1-8 × 10 20 cm −3 [44][45][46][47][48] as well as an active doping concentration of ∼3.5 × 10 20 cm −3 . [49,50] But in all works reported, Ga was solely studied in single crystalline silicon, and no studies have been shown to investigate Ga hyperdoping in poly-Si or in poly-Si/SiO x passivating contact structures.
This work explores Ga and B doping of poly-Si using an excimer laser with ns pulse duration (Figure 1a), rather than relying on the traditional furnace anneal (shown previously in Young et al. [32] ). The essence of this approach is to take advantage of the nonequilibrium nature of the anneal by utilizing the short timescale and to achieve doping concentration above the solid solubility limit. The rapid melting of the poly-Si increases the diffusion coefficient of dopants and upon the removal of the laser pulse, the temperature of the poly-Si drops rapidly, and the molten poly-Si recrystallizes back. During this process, dopants are incorporated into the Si lattice at very high concentrations due to the extremely fast regrowth rate (>1 m s −1 ), which leads to nonequilibrium doping. [51] We show the effect of laser energy density (ED) and number of pulses ( Figure 1b) on the passivation quality of the sample (see structure in Figure 1c), as well as the transport and electrical behavior. These insights are significant and lead to improvements over boron-doped p-type poly-Si/SiO x passivating contacts.  Figure 2a shows the simulated temperature-depth profiles for different laser EDs, from 700 to 1200 mJ cm −2 . For each ED, the temperature profile corresponds to the instant when the maximum melt depth is reached, which also provides the maximum temperature. Since the maximum melt depth can continue to increase even after the end of the pulse due to heat conduction, the simulated time for reaching the maximum melt depth for all the EDs investigated was ∼25 ns rather than 22 ns laser pulse length. The near-surface region of each profile shows a lower temperature gradient, followed by a higher temperature gradient further away from the surface into the wafer. These two regions correspond to the liquid phase (molten poly-Si) and crystalline phase (solid poly-Si) of Si, respectively. The difference in the two temperature gradients is due to the higher thermal conductivity of the On each 30 × 50 mm 2 sample, twenty-seven 5.1 × 5.1 mm 2 spots were processed with a KrF excimer laser with a wavelength of 248 nm. The laser beam is indicated in purple. b) Top view of the test structure in a). The numbers 1 to 7 represent seven different energy densities over the range of 300-1100 mJ cm −2 for each laser pulse. The letters A -D represent the number of pulses at each energy density ranging from 1 to 6. One spot was not annealed for reference measurement of the pre-annealed state (3, B). c) Cross-sectional image of the non-symmetrical test structure after laser processing, passivated by depositing Al 2 O 3 on both sides.

Results and Discussion
Energy Environ. Mater. 2023, 6, e12542 2 of 12 liquid phase with respect to the crystalline phase. The kink separating the two phases corresponds to the liquid-solid interface at the maximum melt depth, and its depth increases for increasing ED. Furthermore, the solid-to-liquid transition occurs in most cases at a temperature slightly higher than the melting temperature of Si, T m = 1414°C, due to the solid superheating. [52] The figure clearly shows that high-temperature gradients are reached within the poly-Si solid phase up to 150°C per 100 nm, due to the short absorption depth of ∼5 nm for the 248 nm laser wavelength and the short duration of the pulse. This feature is the basis of the method used in this work, as it allows to heat a shallow layer in poly-Si while preserving the chemical passivation of the SiO x layer at the poly-Si/SiO x interface.
In Figure 2b, we report the simulated temperature at the poly-Si/SiO x interface as a function of time. In all cases, the temperature at the poly-Si/SiO x interface increases rapidly within the first 22 ns, remains high for only a very short duration, and then decreases slowly after the end of the short laser pulse. The temporal variation in temperature at the poly-Si/SiO x interface shows a strong dependence on the laser ED. For higher laser EDs, 1100 and 1200 mJ cm −2 , the peak temperature exceeds the melting point of Si. For lower EDs, the temperature remained below T m,Si . Therefore, we can conclude that a laser ED of less than 1000 mJ cm −2 is necessary to not compromise the interfacial SiO x layer, thus preserving the passivation quality after laser processing.

Dopant Diffusion and Activation
To visualize the dopant distribution profiles, secondary ion mass spectrometry (SIMS) measurements were performed on both Ga-implanted and B-doped via plasma enhance vapor deposition (PECVD) polished c-Si samples. Figure 3 shows the Ga and B concentration depth profiles for various laser EDs with 1 or 4 pulses. For comparison, we also show the depth profiles measured in unannealed regions. The profiles show a pronounced peak within the top 10À15 nm, which is an artifact and, in the case of B, also a small dip. Other artifacts include extended profile tails at low concentrations (<5 × 10 17 cm −3 for B, and <10 18 cm −3 for Ga), and a background signal of ∼3 × 10 16 cm −3 for both dopants. Thus, SIMS can slightly overestimate the real dopant concentrations due to the background. Figure 3a,b shows common diffusion features for both dopants after PLM: a) Diffusion of Ga and B occurs into the intrinsic poly-Si region at concentrations well above the solid solubility limit in c-Si (∼4 × 10 19 cm −3 at 1200°C for Ga, and ∼ 5× 10 20 cm −3 at 1200°C for B [36] ); b) all depth profiles show a sharp decrease in concentration at depths ranging between 50 and 250 nm (as the dopants reach the maximum melt depth, the dopant diffusivity in the solid phase is orders of magnitude lower than in the liquid phase, which prevents profile broadening); and c) an increase in ED redistributes the dopants much deeper into the poly-Si due to the increase in the maximum melt depth, which is in qualitative agreement with the simulation shown in Figure 2a. In Figure 3b, as B penetrates deeper into poly-Si, the concentration in the surface region decreases as the dopant areal density is conserved during the PLM process. However, this effect is much less pronounced for Ga. Both the Ga and B profiles do not reach the tunneling SiO x layer, which is consistent with the simulation plot in   Figure 4e shows that the shape of the EELS depth profile for B matches well with that obtained from SIMS, which further confirms that the B atoms do not significantly penetrate in the c-Si bulk after PLM.
Contrary to B, in the case of Ga, the diffusion profiles are much shallower and exhibit a less sharp dopant drop off at the maximum melt depth. Additionally, the number of laser pulses does not seem to have a large effect on the doping behavior. This can be attributed to the difference in equilibrium segregation coefficients for Ga and B defined as k eq = C S /C L , where C S and C L are the concentration of the dopant in the solid and the liquid phase, respectively. B has a k eq = 0.8, which is two orders of magnitude higher than Ga, k eq = 0.008. [36,53] Therefore, during the recrystallization process, B is more likely to incorporate in the crystalline poly-Si, whereas Ga tends to be pushed towards the surface with the molten front: this explains the gradual decrease in the Ga doping concentration profile moving deeper into the film in Figure 3a.
The O concentration depth profiles for the Ga and B doped samples are shown at the top in Figure 3a,b, which can be used to identify the SiO x layer. The dopant profiles for Ga and B behave very differently in the SiO x layer: B appears to pile up in the SiO x layer, while Ga does not show any pileup. Indeed, this B concentration peak at the oxide is likely due to an oxygen matrix effect in SIMS, [54] as the witness unannealed region in Figure 3b also shows a higher concentration of B in the SiO x layer. Thus, we can conclude that no accumulation of Ga or B in the SiO x layer occurs due to PLM.
To gain insights into the electrical behavior of the Ga and B doped regions, Van der Pauw -Hall (VdP-Hall) measurements were performed in each laser annealed spot separately with the four probes directly contacting the four corners of the laser annealed region. Figure 5 shows the sheet resistance, and the hole Hall areal density for Ga (Figure 5a,b) and B (Figure 5c,d) for different laser EDs and numbers of pulses. We see for both Ga and B that the sheet resistance decreases while the hole Hall areal density increases with laser ED, which indicates a higher degree of doping. Notably, regardless of the laser ED and number of pulses, a hole areal density between 7 × 10 15 cm −2 -13 × 10 15 cm −2 was measured for B, whereas for Ga, the hole areal density is about one order of magnitude lower, ranging between 0.9 × 10 15 cm −2 -3 × 10 15 cm −2 . Figure 6 reports the maximum active concentrations measured as above in different samples for Ga and B as a function of the corresponding maximum chemical concentrations (total atomic concentration). To calculate the maximum active doping concentration, a conversion factor called the Hall scattering factor, r H , needs to be considered to obtain the true active carrier concentration, according to the relationship p ¼ r H p H where p is the hole concentration, and p H is the Hall hole concentration. The Hall scattering factor is an empirically determined quantity that depends on various scattering mechanisms and detailed band structure. [55,56]   Here, we assume a Hall scattering factor of r H ≈0.62 based on a previous work by Romano et al., who determined r H by fitting the Hall areal density and substitutional doping concentration for B and Ga in c-Si. [44] Taking r H into consideration, we multiplied the hole Hall areal density in Figure 5 by r H to obtain the true hole concentration for Ga and B. To compare the active dopant concentration to the total dopant concentration, we extract the maximum active dopant concentration, N max (yaxis in Figure 6) using the SIMS profiles in Figure 3 following the procedure in Impellizzeri et al. [57] We assume that in regions with dopant concentration below N max , 100% of the dopants are active (see inset in Figure 6). In the region closer to the surface, the dopant concentration in the SIMS profile is higher than N max . We further assume that dopants at concentrations above N max are inactive. We then integrate the SIMS profile with N max set as the upper limit (see the shaded region in the inset of Figure 6) such that the integrated SIMS profile (cm −2 ) is equal to the measured active drift areal density obtained from the Hall measurements. [57] The maximum chemical concentrations in Figure 6 were calculated as the mean of the SIMS profiles in the top 10-25 nm after excluding the surface artifacts through extrapolation.
Boron at low chemical concentrations, up to ∼5 × 10 20 cm −3 , approaches nearly 100% activation, but the fraction of active dopants decreases to ∼50% at higher chemical concentration and levels off at ∼10 21 cm −3 . This upper limit for B dopant activation is twice the solubility limit of B in Si. Conventional furnace annealed B-doped low pressure chemical vapor deposition (LPCVD) poly-Si in Hollemann et al. [58] reported ∼75% active B atoms, but at a lower doping concentration of ∼5 × 10 19 cm −3 . Our result is also much higher than the ∼1.3% activation level for B reported by Nemeth et al. [59] using conventional annealing in a furnace. In comparison, we show with PLM ∼100% of B atoms are activated in poly-Si at a much higher doping concentration of up to ∼5 × 10 20 cm −3 .
For Ga in Figure 6, there is no clear trend in the fraction of active Ga atoms in poly-Si. For a maximum chemical concentration confined in the relatively narrow range over 7-12 × 10 20 cm −3 , the corresponding active concentrations is over the range of 1.2-2. 5 × 10 20 cm −3 (11%-24% activation). We ascribe this difference in dopant activation to lower solubility and low segregation coefficient (∼10 −2 ) in solid Si/molten Si of Ga compared with B (∼1). As the liquid-solid melt front moves towards the surface during recrystallization (see discussion for Figure 3a), Ga is pushed out from the growing solid into the remaining melt. This leads to the observed steeper Ga concentration depth profiles compared with the "box-like" depth profiles for B. This reduces the amount of Ga dopants in the poly-Si portion closer to the SiO x /wafer interface. Nevertheless, 19% dopant activation level (averaged over the doped portion of poly-Si) in Figure 6 is almost six times higher than the Ga solubility limit (at 1200°C). These values indicate that hyperdoping of Ga in poly-Si was successfully achieved via ion implantation and subsequent PLM with active dopant concentrations similar to those previously reported in c-Si. [49,50] Figure 7 shows the hole drift mobilities as a function of the maximum hole concentration for Ga and B, calculated from the data in Figure 5 using the expression μ D ¼ 1=R s eN H r H , where μ D is the hole drift mobility, R s is the sheet resistance, N H is the Hall areal density, and e is the electron charge. The Hall scattering factor, r H , has been assumed equal to 0.62 according to Romano et al. [44] Figure 7 also shows hole drift mobility for Ga-and Bdoped c-Si from previous studies. [60,61] These values differ from each other due to the lattice strain induced by the dopants, which in turn influences the hole effective mass. The same study [61] also extracted the strain-corrected mobility, i.e., the mobility for unstrained p-type doped Si, which is the same for both dopants. It is interesting to note that the B mobility data in poly-Si obtained in this work lie between the values reported in the literature for B in c-Si, and the trend depicted by the strain-corrected curve. We speculate that in the case of B the deviation from the c-Si curve towards the unstrained Si curve could be attributed to strain relaxation in poly-Si during the pulsed laser processing. On the contrary, the hole mobility for Ga-doped poly-Si is much lower than the trend in the c-Si and the strain-corrected curves. We attribute this to a lower activated fraction of Ga dopants and additional scattering channels that limit the carrier mobility, such as Ga clusters in the grain boundaries [62,63] and neutral impurity scattering. [64,65] Finally, comparing the B mobility data with Nemeth et al., [59] which represents the state-of-the-art furnace-annealed PECVD poly-Si, the laser processed sample shows ∼four times higher mobility, again highlighting the advantage of using PLM.   Figure 8a arises from differences in orientation that cause the backscattered electrons to interact with varying degrees with the annular detector. In Figure 8b, the Kikuchi patterns generated by backscattered electrons are indexed relative to the silicon crystal structure and each pixel is assigned a color related to the crystallographic direction that is normal to the sample surface. The black pixels represent regions where no solution was found based on the diffraction pattern at those pixels. This can occur when multiple small grains overlap within the volume where the diffraction pattern is generated. However, this does not seem to be the case for this sample. The diffraction pattern for the black regions does not contain any Kikuchi patterns (see Figure S1) which suggests that these regions could be amorphous. In general, EBSD shows that the laser annealed poly-Si has an average grain size of ∼100 nm.

Electron Microscopy of the Contact Structures
To confirm that the sample structure (poly-Si/SiO x /c-Si) remained intact after laser processing, we performed cross-sectional transmission electron microscopy (TEM) on the Ga sample post laser processing, and the oxide interface was not impacted and stayed intact (see Figure S2).
We also performed PLM on randomly textured c-Si samples, which are otherwise nominally identical to the polished c-Si samples discussed above. The EDs used in these experiments were over the range of 200-500 mJ cm −2 . Figure 9a shows the plan-view scanning electron microscopy (SEM) image of the laser-processed region. It is evident that the vertices of the pyramids are deformed. However, the simulated maximum melt depth at a higher ED of 700 mJ cm −2 on a planar surface is <25 nm (see Figure 2a). Nevertheless, Figure 9b shows that ∼500-600 nm of the c-Si pyramid melted and recrystallized. The melting behavior at the vertices of the pyramids indicates that the pyramid tips have been subjected to a temperature above the melting point of Si. We attribute this difference to the heat dissipation at the pyramid tips versus the faces or valleys. For a given ED, the heat flux to a textured surface is lower because of a higher surface area by a factor of ∼√3 compared with a planar surface. However, at the pyramid tips, there are four melt fronts corresponding to each pyramid face that converge, and with no heat dissipation into the bulk c-Si, this causes melting over a much larger length scale than a planar surface. Once the tips melt, the liquid surface tension tends to round off the surface and modifies the shape of the vertices. This deformation of the pyramid tips also resulted in poor passivation, which was confirmed by photoluminescence imaging ( Figure S3). SIMS and VdP-Hall measurements show that the lower ED (200-500 mJ cm −2 ) used in PLM of textured samples was not as effective in spreading and activating the Ga dopants (see Figures 3a and  5b). Therefore, we conclude that textured samples require careful tuning of the laser conditions to preserve passivation, and this technique is more suitable for the planarized back surface. Figure 10a shows the photoluminescence (PL) image of a Ga-doped polished sample after PLM and a 2nd passivation for different ED and number of pulses. Most of the laser-processed 5.1 × 5.1 mm 2 regions (see Figure 1b) appear bright in this image except for those that are close to the edges. To qualitatively analyze the passivation quality of each region, we first determine the change in the iV oc , ΔiV oc , by using the PL intensity of each laser processed spot before PLM (I PL before ) and after PLM (I PL after ) in the expression in Equation 1. [66] ΔiV oc ¼ kT q ln I PLafter I PLbefore

Passivation Quality of Ga-Doped Contacts
In this expression, k is the Boltzmann's constant, T is the temperature of the sample, q is the unit charge of an electron. Using   [60] and Romano et al. [61] for hole mobility of B and Ga in c-Si, respectively. The straincorrected hole mobility in c-Si was also reported in Romano et al. [61] The blue circles and red squares indicate the Ga and B mobilities after PLM, respectively, and the red triangles indicate the B mobilities after conventional furnace annealing in Nemeth et al. [59] Energy Environ. Mater. 2023, 6, e12542 6 of 12 photoconductance decay measurement, we determined the iV oc of a larger sample that was processed in a similar manner as the one in Figure 10a, except that it did not undergo PLM. For each laser processed spot in Figure 10a, the absolute iV oc is obtained by adding ΔiV oc to the reference sample's iV oc . We see that the PL brightness increases with increasing laser ED but adding additional laser pulses had little effect on the passivation. The field regions around the laser processed spots appear darker indicating that the passivation quality of the laser annealed regions is better than the surrounding field regions. The highest iV oc value of 721 mV was achieved with a pre-PLM passivation anneal, for a laser ED of 900 mJ cm −2 and 4 pulses. The iV oc values listed here are an average of the passivation quality of both the rough and PLM-processed sides of the wafer. The passivation quality of the back rough side of these single-side-polished samples was not measured, and likely lowered the measured iV oc . Therefore, the iV oc values listed in Figure 10a are conservative and the actual passivation of the PLM surface is likely much higher. The tube furnace processed witness samples have an iV oc of 716 and 700 mV for Ga-and B-doped samples, respectively.
To extend PLM processing to a larger area, we used overlapping laser square spots across the c-Si wafer. We used an ED of 900 mJ cm −2 and 2 pulses at each position, which resulted in four PLM pulses at the edges, and eight PLM pulses at the corners of the squares. Figure 10b shows the PL image of a 30 × 50 mm 2 sample after PLM rastering. A part of this sample was not laser processed for comparison. The iV oc of the laser-processed region of the sample was 735 mV, with a J 0 value of 4.1 fA/cm 2 , which is even higher than the iV oc for the small laser processed spot shown in Figure 10a. Remarkably, the PL intensity is uniform across the sample without any deterioration in passivation of the edges or corners of the individual laser spots due to the overlapping laser positions. The PL intensity for the non-laser processed region is darker than the laser processed region, which is expected due to the lack of dopant activation. This result demonstrates our ability to PLM large area samples and lays the groundwork for incorporating it into a back-junction front/back poly-Si/SiO x passivating contacts devicesuch as high-performance p-type TOPCon solar cells.

Electrical Characterization of Ga-Doped Contacts
The metal-to-poly-Si contact resistivity for poly-Si: Ga/SiO x contacts on n-type wafers was determined to be 0.9 mΩ•cm 2 using the transfer length method (TLM) [67] on the sample shown in Figure 10b. To extract the poly-Si-to-c-Si contact resistivity, we evaporated 1 μm thick dots of Ti/Ag of varying sizes (35 μm, 70 μm, 0.125 mm, 0.5 mm, and 1 mm; see inset of Figure 11a) onto each laser processed spot for the sample shown in Figure 10a. Isolated diode structures were produced by etching the unmetallized poly-Si:Ga/SiO x contacts into the c-Si wafer using an SF 6 plasma (see inset of Figure 11b). An ohmic contact was also formed on the back of the sample by etching off the poly-Si/ SiO x and contacting the bulk c-Si with a Ga-In eutectic alloy. Currentvoltage (I-V) measurements were performed for the diode structures by contacting the circular electrode at the front and the Ga-In eutectic at the back. The contact resistivity was obtained from Equation (2) [68] by plotting dV=dln I ð Þ versus I for various sized metal contacts (see Figure 11a).
In Equation 2, R T is the total resistance, which can be obtained from the slope of each curve, and n is the ideality factor. [68] R T is a combination of diode resistance R diode , spreading resistance R s , and the resistance  of the wafer and the bottom contact R 0 . R s was determined according to Denhoff [69] and R 0 was measured by bypassing the diode region of the sample and directly contacting the bulk c-Si. The five different values of R diode ¼ R T ÀR s ÀR 0 are plotted as a function of 1/S, where S is the area of the circular electrodes (see Figure 11b) to obtain the average contact resistivity. For an ED of 900 mJ cm −2 and 4 pulses (laser spot "5, C" in Figure 1b), the contact resistivity is 35.5 AE 2.4 mΩ cm −2 . B-doped samples were analyzed similarly, resulting in a lower contact resistivity of 13.6 AE 5.2 mΩcm −2 . We only report the best contact resistivity values due to the inability to measure this quantity on laser spots with lower dopant activation. The low contact resistivity is consistent with the Hall data for B (see Figure 5c,d) where the sheet resistance is low with a high hole areal density. Interestingly, both B and Ga SIMS profiles in Figure 3 show a lowly doped poly-Si region (below the SIMS background of 10 17 cm −3 ) near the tunneling SiO x . Yet, despite this lowly doped poly-Si region, we obtained low contact resistivity for both Ga and B.
To validate the resistivity measurements for the Ga sample that was laser processed with an ED of 900 mJ cm −2 and 4 pulses, we performed cross-sectional scanning spreading resistance microscopy (SSRM) to probe the local spreading resistance as a function of depth from the metal electrode into the Ga-doped poly-Si film (schematic in Figure 12a). Due to the small atomic force microscopy (AFM) tip-sample contact area, the resistance measured by SSRM is dominated by the resistivity of local sample volume right beneath the probe in nm-scale. Figure 12b shows a spreading resistance map over a 1 × 1 μm 2 region. In this image, the metal appears dark (low resistance) and the c-Si wafer, which has a much higher resistivity, appears lighter (high resistance). This sharp change in resistance occurs within the poly-Si. In the 1D line scan shown in Figure 12c, we can see that within the Ga-doped poly-Si contact, there is a high resistance region with a low doping concentration that is closer to the c-Si wafer that spans from about 100-150 nm. Thus, SSRM results confirm that the resistance profile is consistent with the SIMS depth profiles in Figure 3a.
Typically, low contact resistivity is achieved in poly-Si/SiO x contacts when the poly-Si is heavily doped, and dopants diffuse through the SiO x and form a sharp decreasing gradient tail inside the c-Si. However, our results show that despite a lowly doped poly-Si region in-between the heavily doped poly-Si and the SiO x , a relatively low contact resistivity of <40 mΩ•cm 2 is obtained. To better understand the transport in these PLM-processed samples, we performed Kelvin probe force microscopy (KPFM) measurements on the structure shown in the inset of Figure 13a to map the electric field across the p-n junction. The height profile across the cross-section of the sample is shown in Figure 13a, which can be used to distinguish between the metal and poly-Si region due to the surface morphology difference of metal and Si on the cross-section of the device. Figure 13b represents the electric field, which is the first derivative of the measured potential. The maximum of the electric field curves indicates the transition of the p-type to n-type semiconductor, and the overall shape of the electric field curves is directly correlated to the carrier concentration profiles around the p-n junction. Three different reverse bias voltages were applied to measure the electric field changes induced by the bias  voltages in order to minimize the effect of surface charges trapped on the cross-sectional surface. All three curves show a peak at ∼450 nm, which is ∼250 nm from the metal/poly-Si interface. Given that the poly-Si has a thickness of ∼250 nm, the transition from p-to n-type Si occurs around the interface between poly-Si/SiO x contact and the n-type c-Si wafer. Interestingly, for all three bias voltages, the electric field decays much more slowly into the poly-Si region than into the c-Si region, which suggests that most of the depletion region lies within poly-Si, with a small portion extending into c-Si. The slope of the electric field also decays slower from the poly-Si/c-Si interface to the blue line, which indicates that the poly-Si closer to the oxide is lowly doped. The slopes of the field increase after the blue line to the interface between metal and poly-Si, which shows that the poly-Si near the metal surface is more heavily doped. These results are again in agreement with the SSRM data in Figure 13c and the SIMS profiles in Figure 3a.
In our electrical measurement, we are detecting the diffusion-limited recombination current into the p + poly-Si/SiO x contact, similar to the classical diode dark current. However, in this measurement we cannot establish relative contributions to the overall current from electrons and holes. At this point, we can conclude that the hyperdoped contact passes the recombination current at low apparent contact resistivity. Hole transport properties can be established separately in a solar cell device or on p-type wafer, which is planned for future work.

Conclusion
In this work, we demonstrated a novel poly-Si:Ga/SiO x passivating contact through non-equilibrium doping enabled by PLM. We explored a range of laser energy densities from 300 to 1100 mJ cm −2 and laser pulses (1)(2)(3)(4)(5)(6) to optimize the melt depth and the resulting Ga and B depth profiles in poly-Si. The SIMS depth profiles indicate that higher ED and number of pulses promote dopant penetration into the depth of poly-Si and allow hyperdoping within the poly-Si. VdP-Hall measurements reveal active doping concentrations as high as ∼2.5 Â 10 20 cm −3 for Ga and ∼ 10 21 cm −3 for B. The active dopant fraction for B in poly-Si approached ∼100% up to a chemical concentration of 5 x 10 20 cm −3 after which the activated fraction decreased to ∼50% for a chemical concentration of ∼2 Â 10 21 cm −3 . This high fraction of activated B dopants in poly-Si beyond the solubility limit combined with  c-Si-like hole mobilities has been reported for the first time. For Ga, we report hyperdoping 6Â above the solubility limit but with active dopant fractions between 11%-24%, and lower mobilities likely due to Ga clusters in the grain boundaries and neutral impurity scattering. These results unambiguously demonstrate that PLM can be used for hyperdoping of poly-Si with Ga and B. Despite the melt/freeze cycle of the poly-Si during PLM processing, the passivation of the SiO x /c-Si interface remained high or improved with annealing. For the Ga-doped poly-Si/SiO x contacts on polished n-type Cz c-Si wafers, we demonstrated a high iV oc of 735 mV. Electron microscopy and photoluminescence measurements show that alkaline-textured Si(100) surfaces with random pyramids require additional work with PLM due to melting and deformation of the pyramid tips as the ∼500 nm melt region penetrates well beyond the SiO x layer and degrades the chemical passivation. Additionally, we extended PLM to a larger 3 × 4 cm 2 area by patching together multiple 5.1 × 5.1 mm 2 spots and showed good passivation quality with no degradation from the overlapping laser spots at the edges and corners. For Ga-doped poly-Si/SiO x contacts, there was a lowly doped poly-Si region of ∼100 nm between the hyperdoped surface region and the SiO x /c-Si interface. Nevertheless, we obtained a contact resistivity of 35.5 AE 2.4 mΩcm 2 , using the expanded Cox and Strack method. We also tested the metal-to-semiconductor contact resistivity using TLM and obtained a low value of 0.9 mΩ cm −2 . We think that the hyperdoped contact passes the recombination current at low apparent contact resistivity, but further investigation is needed. Overall, this result shows that by using PLM a sufficiently low contact resistivity can be obtained for Ga-doped poly-Si contacts compared with the ∼10 Ω cm 2 resistivity obtained by conventional furnace annealing. [32] Therefore, PLM is a promising technique for the fabrication of Gadoped poly-Si/SiO x passivating contacts that can address the lower iV oc obtained for conventionally annealed B-doped poly-Si/SiO x passivating contacts.

Experimental Section/Methods
Fabrication of test structures: Both double-side textured (180 μm) and singleside polished (400 μm) phosphorus-doped, 3-5 ΩÁcm resistivity, Czochralski (Cz) Si wafers (Woongjin) were cleaned with piranha and the standard RCA method. [70,71] A ∼1.5 nm low-temperature SiO x layer was then formed on both sides of the wafer, followed by the growth of ∼200 nm of intrinsic poly-Si on textured samples and ∼250 nm of intrinsic poly-Si on polished samples via lowpressure chemical vapor deposition. Subsequently, both textured and polished Si wafers were laser-scribed into 30 Â 50 mm pieces. These wafers were then split into two groups. For the first group, we deposited ∼20 nm of B-doped hydrogenated amorphous Si (a-Si:H) on one side via PECVD using a SiH 4 /H 2 /B 2 H 6 capacitively-coupled, radio-frequency (rf) plasma operating at 13.56 MHz and a pressure of 1 Torr. The c-Si wafers were placed on the grounded electrode, which was kept at 300°C. The plasma power was 8 W, and the gas flow rates were set to 2 and 100 standard cm 3 /min (sccm) for SiH 4 and H 2 , respectively, with 1 sccm of B 2 H 6 (2.6% in H 2 ). For the second group of samples, 69 Ga ions were implanted at 10 keV with a nominal dose of 6 Â 10 15 cm −2 . However, the implanted Ga dose measured by Rutherford backscattering spectrometry (RBS) was found to be 4.5 AE 0.2 Â 10 15 cm −2 (see Figure S4) which is ∼25% lower than the nominal implant dose. We attributed this loss of Ga to the sputtering of the highly doped poly-Si surface layer during ion implantation. To limit the projected depth to <50 nm, the implant energy and doses were estimated using simulations with the Stopping Range of Ions in Matter (SRIM-2013) software. [72] Both Ga-doped and B-doped samples were further split into two groups: one group was used for PLM studies and the second group was annealed in a tube furnace. Prior to pulsed laser melting, ∼15 nm of Al 2 O 3 was deposited on the Ga-implanted samples using atomic layer deposition (ALD) from trimethylaluminum and water. This was followed by annealing in forming gas (1:9 H 2 :N 2 ) at 400°C for 60 min. After annealing, the Al 2 O 3 was removed in a 1% aqueous HF solution. Both Ga-and B-doped samples underwent PLM with a KrF excimer laser (λ = 248 nm) and a pulse duration of 22 ns (based on full width at half maximum). The laser has a spot size of 5.1 Â 5.1 mm 2 with a uniformity within 2%. Figure 1a shows a schematic of Ga-and B-doped test structures that were processed with the excimer laser. The samples were translated laterally using a motorized X-Y stage to process multiple locations with varying energy densities and numbers of pulses. The matrix of the varying laser conditions on a 30 Â 50 mm 2 test sample is illustrated in Figure 1b, which features seven columns with different laser EDs and four rows corresponding to different numbers of laser pulses. After laser processing, the final test structure is shown in Figure 1c. For comparison purposes, the second set of ion-implanted samples was annealed in a tube furnace at 850°C for 30 min in N 2 ambient to crystallize the a-Si:H into poly-Si, and to diffuse and activate dopants. The samples were repassivated by Al 2 O 3 deposited by ALD and annealed in forming gas. Simulation of laser processed temperature-depth profiles: Simulations based on heat flow calculations were performed by inputting physical and optical data for Si from literature into the LIMP-Laser Induced Melting Prediction, the Harvard simulation software package. [73] We assume that the ∼200 nm layer of poly-Si has properties similar to c-Si. In this model, there is a 20 nm a-Si:H layer on top of poly-Si. We assume a flat surface with a reflectivity R = 0.67. The maximum melt depth profiles were modeled with laser EDs between 700-1100 mJ/cm 2 . Characterization of the test structures: The Ga and B depth profiles in the test structures were analyzed using SIMS with a 3 keV O 2 + beam, collecting 69 Ga + , 11 B + , 16 O + , and 30 Si + secondary ions. The Ga concentration was calibrated by measuring the as-implanted Ga dose in selected samples by RBS, and the calibration of the B concentration was performed by measuring a certified standard. For both elements, the concentrations have a relative error of AE10%. The depth scales were calibrated with an accuracy of AE1% by measuring the crater depths with a profilometer (Tencor P17) and assuming a constant sputtering rate. The sheet resistance, carrier areal density and mobility of the dopants were determined by VdP-Hall measurements, using a 0.65 T permanent magnet, with relative errors of AE5%. The Ga doped samples were further analyzed using the following techniques. PL was used to map the passivation quality for different laser processing conditions. [74] The PL intensity for different laser processed spots was used to calculate the iV oc by comparing it to a larger control sample that was annealed in the tube furnace. [66] The iV oc for this control sample was determined using quasi-steady state photoconductance (QSSPC) decay measurements performed using a Sinton WCT-120 instrument at 1-Sun illumination. [75][76][77][78] The contact resistivity was determined using the expanded Cox and Strack method described by Wang et al. [68] Cross-section SEM was performed using Hitachi S-4800 FE-SEM. TEM specimens were prepared by the focused ion beam (FIB) liftout technique using a Nova 200 Nanolab Dual-Beam FIB. [79] Bright-field and phase-contrast TEM images were acquired using a FEI Tecnai F20 TEM operated at 200 kV. EELS spectrum imaging was performed using a Gatan Enfinium spectrometer with dispersion set to 0.25 eV/channel an acquisition time of 0.2 s at each pixel and an approximate pixel size of 2 nm within the map regions. EBSD mapping was performed on a FEI Nova 200 Nanolab dual-beam focused ion beam (FIB) equipped with an Oxford Nordlys detector. Data were collected using an accelerating voltage of 15 kV and beam current of 2.2 nA. EBSD data were processed using Oxford's AZtecICE software package.
The poly-Si region was characterized by cross-section SSRM [80] (Veeco Instruments Dimension 5000 scanning probe microscope with a Bruker SSRM module) to determine the local resistivity across the metal, poly-Si, and c-Si. The microscope was housed in an Ar glovebox (<0.1 ppm O 2 , <0.1 ppm H 2 O) to prevent oxidation of the sample. A conductive wear-resistant diamond-coated probe (Bruker DDESP-V2 nanoelectrical) with a radius of ∼25 nm, was used for the SSRM measurements. The actual probe/sample contact size is usually smaller than the probe radius, giving better spatial resolution than the probe radius. The SSRM measurement was performed at an applied direct current bias of 3 V. All scans were conducted incrementally in 1 Â 1 μm 2 scan areas. Additionally, crosssectional KPFM was employed for potential imaging of the metallized test structure. The KPFM setup includes the first resonant oscillation of the PtIr-coated cantilever (50-70 kHz), which was used for noncontact atomic force microscopy (AFM, Veeco D5000 and Nanoscope V) topographic imaging, and the second resonant frequency (300À500 kHz) for potential imaging, which yields an enhanced Energy Environ. Mater. 2023, 6, e12542 energy resolution of ∼10 mV. Three reverse bias voltages of −0.5, −1, and −1.5 V were applied to the device to measure the change in the potential as a function of distance from the metal surface.