Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Low ohmic contacts to the front side of pn-junction Si solar cells require high doping in the “emitter” (usually the n-type) part of the cell. Unfortunately, the high doping goes hand-in-hand with increased Auger recombination and degraded quantum efficiency for short wavelength radiation. The “selective” emitter (SE) concept uses laterally different emitter doping: (i) high doping under the front side metallization for low contact resistance between contact metal and semiconductor interface, (ii) lower doping between the contact fingers for a better short wavelength response due to less Auger recombination 1 as well as improved emitter passivation 2, 3. Unfortunately, most SE concepts developed so far in research imply a high process complexity due to the required masking steps for selective diffusion 4 or emitter etch back 5. In particular, such existing concepts are hard to implement into industrial production of solar cells. Several different SE approaches for industrial production are currently under development 6, 7 or already being transferred to the industry 8.
The present paper reports an SE formation by an elegant add-on process using laser tailoring of furnace diffused emitters. This process does not only work in a research environment but is also compatible with industrial mass production. The add-on laser doping increases the doping concentration as well as the emitter depth underneath the contact fingers, thus yielding good ohmic contacts. The increased emitter depth allows one using screen printed contacts, which cannot be applied on shallow diffused emitters. Hence, laser tailoring is well suited for industrial solar cells; we prove an increase of 0.5% (absolute) cell efficiency.
Figure 1 depicts the process sequence for silicon solar cells with add-on laser tailoring for SE formation. We use 170 µm thick, p-type Czochralski grown wafers of 12.5 cm × 12.5 cm in size. Step 1 textures the wafer surface by a standard random pyramid etching process in order to reduce the optical reflectivity of the solar cell. Step 2 creates the n-type phosphorus doped emitter together with a phosphosilicate glass (PSG) layer on the wafer surface by conventional POCl3 furnace diffusion. The PSG mainly consists of SiO2:P2O5 with the ratio of SiO2 and P2O5 depending on processing time, gas composition, gas flow, and temperature 9. In Step 3, our patented laser process 10 locally melts the wafer surface, using a pulsed Nd:YAG laser with 532 nm wavelength, 20 kHz pulse repetition rate, and 65 ns pulse duration. The shape of our laser beam consists of a Gaussian profile in the short axis with a full width half maximum FWHM = 8 µm and a 225 µm wide tophat profile in the long axis. For the laser processing, we use a pulse overlap of 35%. Simulations show an almost complete absorption of the laser irradiation within a 50 nm thick surface layer due to the strong increase of the absorptance of silicon at higher temperatures 11. Hence, the melting of up to 1000 nm deep areas at the wafer surface mostly takes place by heat diffusion. The high phosphorus diffusion constant D = 5.1 × 10−4 cm2/s in liquid silicon 12 enables the fast incorporation of phosphorus atoms from the PSG-layer, up to 800 nm deep into the molten silicon within a few hundred nanoseconds. Subsequently to the laser pulse, the molten silicon cools, re-crystallizes epitaxially, and forms a highly phosphorus doped selective n-type emitter without incorporation of any grain boundaries and dislocations 13. Step 4 removes the PSG layer with hydrofluoric acid, and a SiNx anti-reflection coating (ARC) is deposited by plasma enhanced chemical vapor deposition on the front side. In the last processing step, step 5 screen printing, firing of the front, and back contacts as well as laser edge isolation complete the solar cell.
Table I compares eight add-on SE solar cells having an emitter sheet resistance ρs = 110 Ω/sq with eight conventionally processed reference cells with ρs = 60 Ω/sq. The higher doping of the conventional processed reference cells is required for low contact resistivities ρc = 3 mΩ cm2, measured by the transmission line method 14. For SE formation, we use a pulse energy density Ep = 2.5 J/cm2. The laser irradiation decreases the contact resistivity of the un-irradiated ρs = 110 Ω/sq emitter from ρc > 400 mΩ cm2 down to ρc = 4 mΩ cm2. The back side of all solar cells features a screen printed aluminum-alloyed back surface field. The average efficiency η of add-on SE solar cells increases by Δη = 0.5% compared to the reference cells. The best efficiency is ηmax = 18.1%. The increase Δη = 0.5% results from a higher short circuit current density Jsc = 37.1 mA/cm2 and an improved open circuit voltage Voc = 629 mV due to less Auger recombination and thus also better blue response. The lower emitter doping of the SE solar cells increases the lateral resistivity, thus decreasing the fill factor FF. Suns-Voc measurements offer the so-called pseudo I/V-characteristics without series resistance losses 15. The pseudo fill factor PFF = 82.0% for standard as well as for SE cells proofs that the loss in fill factor results from an increased series resistance and not from laser induced defects. In future, the FF reduction will be avoided by a redesigned contact grid.
Table I. Mean values of the cell characteristics of eight selective emitter (SE) and eight standard solar cells. The shallow emitter of the SE cells increases the cell performance by Δη = 0.5% absolute as a result of higher open circuit voltage Voc and short circuit current density Jsc. The SE cell uses the same front grid layout as the standard solar cells. The front grid is optimized for the standard emitter thus decreasing the fill factor FF of the SE cells as a result of higher emitter sheet resistance ρs.
Figure 2 shows the internal quantum efficiency (IQE) of a SE compared to a standard cell. The lower doped 110 Ω/sq emitter of the SE cell increases the IQE at wavelengths λ < 600 nm due to better blue response. The increase at shorter wavelengths results from less Auger recombination in the lower doped emitter between the contact fingers.
Figure 3 presents phosphorus doping profiles of add-on laser irradiated furnace diffused emitters and a non-irradiated reference, all measured with secondary ion mass spectroscopy (SIMS). Increasing the laser pulse energy density Ep results in longer and deeper melting of the silicon surface. Thus, the phosphorus atoms from the PSG-layer together with the atoms already incorporated during furnace diffusion have more time to diffuse deep into the melt, increasing the emitter depth. The integrated areal concentration of phosphorus atoms increases from CS = 1.4 × 1016 cm−2 for the furnace diffused emitter up to CS = 1.7 × 1016 cm−2 and CS = 2.2 × 1016 cm−2 at pulse energy densities Ep = 2.2 J/cm2 and Ep = 3.2 J/cm2, respectively. The limited amount of doping atoms in the PSG-layer decreases the maximum phosphorus concentration Cmax at the wafer surface, changing the doping profile from error function-type to Gaussian-type. The inset in Figure 3 shows the sheet resistance ρs of furnace diffused emitters processed with different laser pulse energy densities Ep. For Ep < 1.4 J/cm2, the sheet resistance ρs remains un-changed. At higher pulse energy densities Ep, the silicon surface starts to melt, enabling the in-diffusion of additional phosphorus atoms from the PSG-layer and the redistribution of already incorporated doping atoms, thus decreasing the emitter sheet resistance ρs down to ρs = 11 Ω/sq. The ρs-drop is caused by two different mechanisms: the in-diffusion of additional phosphorus from the PSG and the activation of already incorporated atoms.
Despite the significantly higher surface concentration Cmax of the purely furnace diffused emitters compared to laser tailored emitters, the contact resistance ρc of the un-tailored emitter is significantly higher than for the laser processed cells. Therefore, a low contact resistance depends more on the doping profile instead on a high doping concentration at the wafer surface. During the contact firing, the etching of the screen printing paste into the emitter 16 changes the doping concentration at the metal/silicon interface and thus the contact resistivity.
Figure 4 demonstrates the wide process window of the add-on laser doped SE for screen printed contacts due to the modification of the doping profiles. We calculate the contact resistivity for the doping profiles of Figure 3 using the Wentzel–Kramers–Brillouin approximation 17 and a Schottky barrier Φb = 0.78 eV 18, which is typical for Ag on n-type Si. The deeper the screen printed paste etches into the emitter, the lower is the doping concentration at the silicon–metal interface, thus increasing the contact resistivity ρc. The ohmic contact of screen printed fingers is formed by small Ag crystallites covering only a small part of the total contact area at the silicon surface 19. Ballif et al. 20 showed that typical macroscopic contact resistivities ρc < 10 mΩ cm2 of screen printed contacts imply very low microscopic contact resistivities ρmic < 1 × 10−3 mΩ cm2 for the Ag crystallites. Here, we define the maximum allowable etch depth zmax for screen printed contacts to be reached if the microscopic contact resistivity ρmic increases above ρmic = 1 × 10−3 mΩ cm2. The inset of Figure 4 shows zmax to increase linearly with the applied laser pulse energy density Ep. The low maximum etch depth zmax = 45 nm of the un-irradiated shallow furnace diffused emitter is smaller than typical etch depths zpaste ≈ 60 nm of screen printing pastes 21, explaining the measured high macroscopic contact resistivities ρc > 400 mΩ cm2 without add-on laser doping. The significantly deeper doping profile of the laser irradiated samples increases zmax to values high above the etch depth, thus widening the process window for the firing of screen printed contacts and avoiding metal spiking through the pn-junction 22.
Figure 5 illustrates the efficiency potential of SE compared to standard cells by reducing the finger width wf. An advanced grid simulation method 23 calculates the efficiency limit for different finger widths wf considering the optimum number of fingers. The lower efficiency limit curves for standard as well as SE cells are calculated assuming screen printed contacts with an aspect ratio a = 0.125 (finger height/finger width) and a resistivity ρ = 3.6 × 10−6 Ω cm for the metallization. The simulation of the upper limit curves implies fine line metallization, for example, metal aerosol jet printing, with an aspect ratio a = 0.5 and a resistivity ρ = 1.9 × 10−6 Ω cm 24. The area between the lower and upper limit represents varied aspect ratios 0.125 ≤ a ≤ 0.5 and resistivity 1.9 × 10−6 Ω cm ≤ ρ ≤ 3.6 × 10−6 Ω cm. The two experimental data points show the efficiency η of our SE and standard cells from Table I. As expected, the experimental result of the standard cells with an optimized front grid correlates well with the simulation. Using the same grid for the SE, cells show a small gap between measurement and simulation due to the not yet optimized number of fingers. The numerical front grid optimization demonstrates that it is possible to combine our SE with fine line metallization in order to obtain a further increase of the cell efficiency η by Δη = 0.7% up to η = 18.8%.
The wafer processing time tp = 1 s is a prerequisite for industrial solar cell production. Therefore, the laser doping of 15.6 cm × 15.6 cm large wafers requires a pulse re-petition rate frep = 30 kHz. A SE cell with optimized screen printed contacts uses nf ≈ 70 front fingers. To reach the processing time tp = 1 s, the area under each contact finger has to be processed simultaneously. The processed area A for each laser pulse is A = lpwpnf, with pulse length lp = 225 µm and pulse width wp = 8 µm. Hence, an industrial laser doped SE needs a laser power:
considering optical transformation losses η0 = 30% of the focusing optics and Ep = 2.5 J/cm2. Frequency doubled disc laser systems have the potential to achieve the necessary output power P = 135 W together with the required beam quality.
In conclusion, we developed a laser add-on SE concept with only one additional processing step which locally increases the doping underneath the screen printed contacts. The emitter profile modification from a shallow error-function type to Gaussian type results in good ohmic contacts between screen printed fingers and emitter for a wide range of firing parameters. Additionally, the significantly deeper doping profiles of laser add-on SE prevents metal spiking through the pn-junction. This additional laser diffusion step increases the cell efficiency η by Δη = 0.5%. Combining our SE with fine line metallization has the potential to boost the cell efficiency further up to η = 18.8%. The simulated absolute efficiency gain Δη = 0.7% with one additional laser irradiation step makes this process promising for industrial application.
We are grateful to J. Straub for wafer processing and G. Bilger for SIMS measurements. The authors gratefully acknowledge funding by the German Federal Ministry for Environment, Nature Conservation, and Nuclear Safety (BMU) under project no. 327519.