IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO.

1, JANUARY 2013

159

Understanding the Impact of the Doping Proles on Selective Emitter Solar Cell by Two-Dimensional Numerical Simulation
Raffaele De Rose, Mauro Zanuccoli, Paolo Magnone, M. Frei, Enrico Sangiorgi, Fellow, IEEE, and Claudio Fiegna
AbstractThe selective emitter (SE) design, featuring lower doped areas between the front contact ngers and higher doped areas underneath the front metallization, is crucial to improve the performance at the front side of a monocrystalline (c-Si) silicon solar cell. One of the most interesting and promising low-cost SE process consists of the screen printing of a phosphorus-doped paste, allowing a separate optimization of the doping proles in the metallized and nonmetallized front-side areas. By referring to this kind of process, this paper presents a simulation study with a decoupled analysis on the effect of the lowly doped and highly doped proles on the performance of an SE solar cell, by means of 2-D electro-optical numerical device simulations. Moreover, by exploiting the 2-D modeling, the effect of the alignment tolerance used in the SE diffusion process for the subsequent metallization process has been also investigated. Numerical results show that the adoption of an optimized design for the SE cell can lead to an efciency improvement above 0.4%a b s compared with the 75 /sq homogeneous emitter reference cell. Index TermsDoping prole, numerical simulation, selective emitter (SE), solar cell.

Fig. 1. Two-dimensional cross section of the SE solar cell. W su b and D su b denote the front contact pitch and the substrate thickness, respectively. W m is the width of the front contact nger, while W se represents the lateral width of the HDOP region. It is worth mentioning that the lateral size of the 2-D simulated device has been limited to half of the depicted structure.

I. INTRODUCTION

HE improvement in the front-side emitter is currently an important issue for the optimization of the silicon (Si) solar cell technological process. The emitter requires high lateral conductance to achieve a low series resistance contribution at the front side. Nevertheless, emitter recombination losses limit cell performance and become increasingly dominant with the improvement of the rear side due to the forthcoming adoption

Manuscript received January 31, 2012; revised August 2, 2012; accepted August 3, 2012. Date of publication September 28, 2012; date of current version December 19, 2012. This work was supported in part by Nanofunction Network of Excellence Project funded by the European Commission under the Seventh Framework Program and the ERG Project funded by the European Nanoelectronics Initiative Advisory Council Joint Undertaking under Grant 2010/ 270722-2. R. De Rose is with the Advanced Research Center on Electronic Systems, Department of Electrical, Electronics and Information Engineering, University of Bologna, I-47521 Cesena, Italy, and also with the Department of Electronics, Computer Science and Systems, University of Calabria, 87036 Rende, Italy (e-mail: rderose@arces.unibo.it). M. Zanuccoli, P. Magnone, E. Sangiorgi, and C. Fiegna are with the Advanced Research Center on Electronic Systems, Department of Electrical, Electronic, and Information Engineering, University of Bologna, I-47521 Cesena, Italy (e-mail: mazanuccoli@arces.unibo.it; pmagnone@arces.unibo.it; enrico.sangiorgi@unibo.it; cegna@arces.unibo.it). M. Frei is with Applied Materials, Inc., Santa Clara, CA 95054-3299 USA (e-mail: Michel_Frei@amat.com). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/JPHOTOV.2012.2214376

of a rear point contact design with passivated rear surfaces in most of the industrial Si solar cells [1]. Since transport and recombination mechanisms are strongly coupled in the semiconductor equations, the emitter design results in a strategic tradeoff between these physical phenomena for the performance of a solar cell. The development of technological solutions to contact emitters with a high sheet resistance (>80 /sq), such as the optimization of the silver (Ag) pastes [2][4] and/or the so-called seed-and-plate approach [5][7], is a challenging task in photovoltaic industry for further optimization of the emitter, but it could be limited by a narrower ring window or by the application of an additional (plating) step [8]. However, such technological solutions to contact high sheet resistance emitter are not industrially available yet. An alternative and promising approach to improve the emitter performance is the realization of a selective emitter (SE) design. SE solar cells, in contrast with standard screen-printed industrial c-Si cells based on an homogeneously diffused emitter (HE), feature different doping proles in the metallized and nonmetallized front-side areas (see Fig. 1): While a highly doped (HDOP) prole under the metal-contacted regions ensures a good ohmic metalsemiconductor contact (achieving a low contact resistance) and an adequate lateral conductivity, a lowly doped (LDOP) prole in the illuminated passivated regions between the front contact ngers leads to a reduction of Auger and surface recombination losses and, consequently, in an enhanced spectral response in the blue region of the spectrum [8][10]. However, an efciency tradeoff exists between

2156-3381/$31.00 2012 IEEE

160

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 1, JANUARY 2013

these advantages and the increased emitter resistance that affects the SE solar cells due to the lower doping concentration in the nonmetallized areas. The use of SE has become one of the most interesting lowcost concepts to boost efciency of Si solar cells. Indeed, it has already been demonstrated that a wide variety of process schemes, which are ready for industrial implementation, can be adopted to form an SE and, then, to enhance the performance of a c-Si solar cell without a relevant increase in production costs [8]. Among these methods, it is worth mentioning the selective laser doping from phosphorus silicate glass [9], the etch-back process [11], the diffusion masking [12], etc. One of the most interesting and promising low-cost SE process is based on the use of a screen-printable phosphorus-doped paste [13], [14]. An important advantage of using this technique is its ability to selectively form HDOP regions without the need for laser patterning, diffusion mask, or etch-back process. Indeed, this process consists of the screen printing of a dopant paste with the same pattern as the metallization, the ring of the paste, and, then, the subsequent printing of the front metallization over the HDOP regions (requiring a good alignment). Moreover, as demonstrated in [14], this SE process provides additional degrees of freedom to optimize the LDOP and HDOP emitters independently, achieving an efciency improvement in the range of 0.30.7%abs compared with the baseline process. By considering this kind of SE process, the aim of this paper is to perform a simulation study with a decoupled analysis of the inuence of the LDOP and HDOP proles on the performance of a c-Si SE solar cell. In this paper, we expand the analysis done in [15] by discussing the impact of LDOP prole on SE solar cell performance, by extending the comparison between the SE and HE solar cells also in terms of spectral response and by investigating in more detail the impact of the dominant recombination mechanisms in the emitter region through a careful dark loss analysis. Unlike some previous works [10], [16], [17], where the impact of the emitter doping prole was analyzed by using a 1-D analytical or numerical modeling, we perform 2-D numerical simulations using a state-of-the-art numerical TCAD simulator [18] coupled to an in-house optical simulator that provides the 2-D map of the optical generation rate inside the solar cell. The 2-D approach is mandatory to account for the shadowing losses underneath the front metallization, to distinguish the front-side passivated and metallized areas, and, in case of SE design, to differentiate the local front-side doping properties. This allows us to better model the physical mechanisms occurring inside the device (and, in particular, in the emitter region), mainly in terms of doping-dependent bulk and surface recombination and of 2-D lateral carrier transport. The comparison with a reference 75-/sq HE solar cell allows us to quantitatively evaluate the expected performance improvements. Moreover, thanks to the 2-D modeling, we are able to consider the effect on the SE cell performance of the alignment tolerance used in the SE diffusion process for the subsequent metallization process, leading to broad HDOP dopant lines. In this regard, we estimate the benet achievable by reducing the HDOP lateral width in the SE cell.

The remainder of this paper is organized as follows. The geometrical and physical details of the simulated solar cells and the simulation methodology are rst discussed in Section II. In Section III, the effect of the doping prole in the passivated LDOP regions at different front contact pitch values is analyzed. Then, the impact of the doping prole in the metal-contacted HDOP areas is investigated in Section IV. After that, Section V reports a quantitative analysis of the effect of the alignment tolerances used in SE diffusion. Finally, the main results of the work are summarized in Section VI. II. SIMULATION METHODOLOGY A two-busbar 12.5 cm 12.5 cm monocrystalline solar cell has been simulated with the following characteristics (see Fig. 1): a 1.5-cm (Nbase = 1016 cm3 ), 180-m-thick borondoped p-type base with a full-area metallized aluminum-alloyed back surface eld (Al-p+ BSF) modeled with a typical 10-mdeep prole reported in [19]; a front surface covered with a 70-nm-thick silicon nitride (SiNx ) antireective coating layer; and a silver front nger with a width Wm of 100 m. The front side of the SE solar cell is also characterized by the lateral width of the HDOP region (Wse ) and by the front contact pitch (Wsub ). By exploiting the symmetry of the device, the lateral size of the 2-D simulation domain has been limited to half of the front contact pitch. In order to achieve realistic predictions on c-Si cell performance, an ad-hoc calibration of the physical models and parameters implemented in the TCAD simulator has been performed [20]. These ne-tuned models include the bandgap narrowing (BGN) model by Schenk to account for the effective intrinsic carrier density [21], the Auger recombination model with the parameterization adopted by Altermatt in [22], the mobility model proposed by Klaassen [18], [23], [24], and the bulk ShockleyReadHall (SRH) lifetime model in boron-doped CzSi according to Glunzs parameterization [25] and in Al-p+ Cz-Si according to Altermatts parameterization [19]. Using Altermatts parameterization for the SRH lifetime in the AlBSF, we obtain a BSF saturation current density of J0 , BSF 590 fA/cm2 , similar to the typically measured values [19]. The bandgap model parameters are adjusted to achieve an intrinsic carrier density of 9.65 109 cm3 at 300 K and low doping density [26]. According to the adopted SRH lifetime model in the substrate [25], we obtain a bulk minority carrier lifetime of about 206 s for the considered substrate doping. FermiDirac statistics has been implemented to adequately model the HDOP regions, such as the emitter and the BSF. According to the applied model for BGN and to FermiDirac statistics, we adopt the parameterization proposed by Kimmerle et al. in [27] for the surface recombination velocity (SRV) at SiNx -passivated front surface, assuming an Auger-limited lifetime in the emitter bulk and a chemical phosphorus surface concentration dependence for the SRV. The SRV at the metallized surfaces is set to 106 cm/s. In addition to the electrical simulation, optical generation rate proles have been calculated with a monodimensional simulation of plane waves propagation in silicon, assuming direct

DE ROSE et al.: UNDERSTANDING THE IMPACT OF THE DOPING PROFILES ON SELECTIVE EMITTER SOLAR CELL

161

illumination with a standard AM1.5G spectrum (1000 W/m2 ) and ideal shadowing under front metal ngers and accounting for a random pyramids textured SiNx front surface. It is worth noting that the electrical simulation does not account for the surface texturing. Moreover, some of the adopted models, like the SRV model, are extracted for planar surfaces. Therefore, it is expected that the implementation of the texturing in the electrical simulations and the update of the physical models to represent textured surfaces may lead to a decrease of the performance of the simulated devices, mainly due to an increase of the surface recombination losses and to an increased emitter resistance. The calculated electrical cell parameters include the shortcircuit current density Jsc , the open-circuit voltage Vo c , the ll factor (FF), and the efciency . A postprocessing analysis has been performed on the simulated I V curves to account for the losses due to the parasitic series resistance R of the front contact according to the following expressions [28], [29]: Rc = Rm = c 1 coth 2 Lm LT 1 Lm m 3 H m Wm Wm /2 LT (1) (2) (3) (4)

Fig. 2. Analytical double error function-like phosphorus concentration prole and calculated electrically active concentration prole for the simulated HE cell. The active prole has been obtained assuming the cutoff at an activation limit of 2102 0 cm3 and neglecting shifts between the carrier and the dopant prole below the activation limit.

R = Rc + R m
2 PM = PM 0 IM P R

where Rc is the metalsemiconductor contact resistance, c is the specic front contact resistivity, LT = (c /Rsq ) is the transport length that is a function of both c and the emitter sheet resistance Rsq [30], Rm is the metal nger resistance, m = 6 106 cm is the front metal resistivity, Lm = 3 cm is the nger length, Hm = 12 m is the nger thickness, IM P is the current at the maximum power point, PM 0 is the maximum output power without the resistance losses, and PM is the effective maximum output power. The effect of the full-area back contact is not accounted because it is a vertical contact (i.e., in which the current ows vertically) and, then, its entire area is active [30]. In (1), the contact resistance is calculated by considering the metal nger composed of two parallel metals featuring a width of Wm /2, which account for the current coming from the two sides of the nger. Since the parallel of two identical resistances is considered, we introduce a factor 1/2 in (1). Concerning the specic front contact resistivity, it is worth considering that it depends on the surface concentration, according to Schroder and Meier [30]. However, it has been demonstrated in [31] and [32] that the contact resistivity of screenprinted silver contacts depends not only on the surface doping concentration, but also on several other process-dependent factors related to the contact formation. In particular, c is mainly limited by the number and the size of the silver crystals grown into the emitter, i.e., by the silver crystal covered surface area and by the current paths from the silver crystals to the bulk of the nger [31]. In addition, it has been reported that the surface texture, leading to a pyramidal shape of the silver crystal, inuences the contact resistance on emitters with high sheet resistance [32].

The values reported by Schroder and Meier in [30] are referred to planar contacts formed by evaporated metals on silicon and, hence, are signicantly lower than the typically measured values for screen-printed silver contacts on industrial emitters [33]. Vinod in [34] suggests c below 1 mcm2 for phosphorus surface concentrations NS 21019 cm3 in case of screenprinted Ag contacts after sintering and annealing. In our work, we adopt a surface concentration-dependent model for the specic front contact resistivity, according to [34] c = c 0 exp C1 b NS (5)

where c 0 is the specic contact resistivity for an innite active surface doping concentration, C1 is a constant (71010 cm3 /2 eV1 for Si), NS is the surface doping concentration and b is the barrier height between the metal and the semiconductor (0.78 eV at Ag contact/n+ -Si interface). In particular, whereas the simulated emitter proles at the metal/n+ Si interface feature surface concentrations NS 11020 cm3 both in HE and SE cells, we obtain c < 1 mcm2 , accordingly to Vinods data [34], which does not strongly affect our simulation results. All the simulated emitter phosphorus doping proles are modeled by an analytical double error function (two error functionlike curves with different peak position and depth), in order to reproduce the typical kink-and-tail shape [35]. Furthermore, we account for the cutoff at an activation limit of 21020 cm3 [36] to calculate the electrically active concentration proles from the generated doping proles (see Fig. 2), assuming full dopant ionization below the activation limit and neglecting shifts between the carrier and the dopant proles [27]. In the following, Cp eak represents the surface peak doping concentration of the chemical prole (which determines the SRV), Jdepth is the junction depth, and Rsq is the sheet resistance extracted from the electrically active proles applying the Klaassen mobility model [23], [24].

162

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 1, JANUARY 2013

TABLE I LDOP AND HE PROFILES

Fig. 3. (Left) Short-circuit current density Jsc and (right) open-circuit voltage V o c versus front contact pitch W su b for the SE cell at different simulated LDOP proles and for the 75- /sq HE cell.

III. ANALYSIS OF THE EFFECT OF THE LOWLY DOPED PROFILE In this section, the effect of the LDOP prole in the passivated front-side areas has been evaluated following the procedure suggested in [37]: For a given HDOP prole (50 /sq with Cp eak = 51020 cm3 and Jdepth = 0.6 m), several LDOP proles (see Table I) have been simulated by assuming a HDOP lateral width Wse = 250 m (150 m wider than the nger width) as a function of the front contact pitch Wsub (ranging from 1.4 to 2.4 mm with a step of 200 m) in order to highlight the tradeoff between the emitter resistance, the illuminated area, and the recombination losses in the LDOP regions. The different LDOP proles are obtained by changing the surface peak doping concentration from 11020 to 41019 cm3 for a xed Jdepth = 0.25 m, corresponding to a sheet resistance range between 90 and 188 /sq (see Table I). As a reference, a 75-/sq HE solar cell featuring a doping prole with Cp eak = 4 1020 cm3 and Jdepth = 0.35 m (see Table I) has been also simulated (see Fig. 2). Jsc and Vo c curves versus front contact pitch for some of the simulated LDOP proles and for the HE cell are reported in Fig. 3. For a given front contact pitch value, both Jsc and Vo c of the SE solar cells are larger with respect to the HE cell due to the lower doping concentration and, hence, lower Auger and surface recombination losses in the illuminated nonmetallized areas. In fact, the highest Jsc and Vo c values are obtained by the SE solar cell featuring the LDOP prole with the lowest doping concentration and, consequently, the highest sheet resistance (i.e., 188 /sq) within the considered range. Moreover, in all the simulated cells, the increase of the front contact pitch Wsub , leading to larger illuminated areas, results in an increasing trend of both Jsc and Vo c . An opposite trend can be observed in Fig. 4, where the FF curves versus Wsub are reported. Simulation results highlight the strong dependence of the FF on the emitter resistance. Indeed, the increase of the front contact pitch, leading to larger emitter resistances, results in a steep decrease of the FF, particularly for the SE cells featuring a high sheet resistance LDOP

Fig. 4. (Left) FF and (right) efciency versus front contact pitch W su b for the SE cell at different simulated LDOP proles and for the 75- /sq HE cell.

prole. Therefore, for a given front contact pitch, among the investigated SE cells, the highest FF is obtained for the LDOP prole with the lowest sheet resistance (i.e., 90 /sq). The efciency tradeoff between the emitter resistance, the illuminated area, and the recombination losses can be better appreciated in Fig. 4, where the efciency versus Wsub curves shows a typical bell shape. On one hand, for lower Wsub values, the efciency increases with the front contact pitch due to the increase of both Jsc and Vo c . On the other hand, for higher Wsub , the emitter resistance losses dominate and, hence, the efciency decreases. Moreover, it can be observed in Fig. 4 that the position of the efciency peak changes depending on the sheet resistance of the LDOP prole for the simulated SE cells: In particular, the efciency peak moves toward lower Wsub for the cells featuring a higher emitter sheet resistance. The maximum efciency (18.93%) is obtained by the SE cell featuring a 100-/sq LDOP prole at Wsub = 2 mm, while the 75-/sq HE cell shows a peak efciency of 18.52% at Wsub

DE ROSE et al.: UNDERSTANDING THE IMPACT OF THE DOPING PROFILES ON SELECTIVE EMITTER SOLAR CELL

163

TABLE II SIMULATION RESULTS

TABLE III HDOP PROFILES

= 2 mm. In Table II, we report the comparison between the HE and SE cells featuring the maximum efciency. The SE cell shows an increase in terms of Jsc and Vo c of 0.63 mA/cm2 and 4.6 mV, respectively, an FF decrease of 0.17%abs , and an efciency gain of 0.41%abs . In terms of dark saturation current density J0 , tot , the SE cell shows a lower value (860 fA/cm2 ) with respect to the HE cell (1010 fA/cm2 ), according to the higher performance. The decrease of J0 , tot shown by SE cell is a consequence of the reduction of the recombination losses in the emitter: As a matter of fact, the emitter saturation current density J0 , e decreases from 284 fA/cm2 for HE to 135 fA/cm2 for SE (see Table II). Through a careful analysis of the loss mechanisms in the emitter region, we can ascribe the J0 ,e decrease to a moderate reduction of Auger recombination losses (J0 , e(Auger) decreases from 63 fA/cm2 for HE to 36 fA/cm2 for SE) and to a strong reduction of the surface recombination losses (J0 , e(Surf ) decreases from 197 fA/cm2 for HE to 81 fA/cm2 for SE). IV. ANALYSIS OF THE EFFECT OF THE HIGHLY DOPED PROFILE In this section, the impact of the HDOP prole in the metallized front-side areas on SE cell performance has been investigated according to [15]. Various HDOP proles have been examined at three different phosphorus dose concentrations (D) (6 1015 , 3 1015 , and 1.5 1015 cm2 ), by varying the Cp eak from 5 1020 to 11020 cm3 and, consequently, the junction depth of the doping prole (see Table III). Accordingly, the sheet resistance of the simulated HDOP proles ranges from 12 to 64 /sq. It is worth noting that only the HDOP proles whose junction is deeper than that of the previous LDOP proles have been considered in Table III. The effect of the different HDOP proles in the SE cell has been evaluated by using the optimal 100 /sq LDOP prole at Wsub = 2.0 mm, derived from the analysis reported in the previous section and by assuming the same HDOP lateral width (Wse = 250 m). The trend of Vo c versus Cp eak for the three different doses is reported in Fig. 5. We observe that, for a given dose, a lower Cp eak and, hence, a deeper junction is slightly benecial for the Vo c because of the following combined effects: 1) a lower peak doping leads to a lower SRV in the illuminated HDOP area outside the front metal nger and 2) a deeper junction is benecial for the reduction of the emitter saturation current density in the metallized regions J0 , e(m et) , as already demonstrated in [10], due to the shielding effect of the minority carriers in the emitter,

Fig. 5. (Left) Open-circuit voltage V o c and (right) short-circuit current density Jsc versus peak doping concentration C p e a k at three different dose concentrations (6 101 5 , 3 101 5 , and 1.5 101 5 cm2 ) for the simulated HDOP proles, by assuming a HDOP lateral width W se of 250 m.

i.e., the holes, from the metalsemiconductor interface. As a matter of fact, for a dose of 31015 cm2 , ranging the Cp eak from 51020 to 11020 cm3 , the J0 , e(Surf ) decreases from 80 to 70 fA/cm2 due to the lower SRV in the exposed HDOP region, while the J0 , e(m et) , weighted on the metallized area, decreases from 333 to 157 fA/cm2 due to the better shielding effect of the deeper junction. In order to further highlight this effect, we performed a simulation of a cell with Wse = 0 (i.e., no HDOP region) featuring the 100-/sq emitter prole (L100 in Table I) with a shallow junction of 0.25 m. As expected, this cell exhibits a low value of emitter dark saturation current (J0 ,e 132 fA/cm2 ) because of reduced Auger and surface recombination losses. However, the same cell shows a higher value of J0 , e(m et) = 588 fA/cm2 due to the shallower junction that causes an increased recombination at the metal/semiconductor interface, according to [10]. Nevertheless, simulation results reported in Fig. 5 only reveals a weak dependence of the Vo c on the HDOP prole: the maximum variation of the Vo c is limited to 0.7 mV (0.11%). On the contrary, a stronger impact of the HDOP prole on the Jsc is highlighted in Fig. 5, where a

164

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 1, JANUARY 2013

Fig. 6. (Left) Efciency and (right) FF versus peak doping concentration C p e a k at three different dose concentrations (6101 5 , 3101 5 , and 1.5101 5 cm2 ) for the simulated HDOP proles, by assuming a HDOP lateral width W se of 250 m.

Fig. 7. Efciency versus peak doping concentration C p e a k at three different dose concentrations (6101 5 , 3101 5 , and 1.5101 5 cm2 ) for the simulated HDOP proles, by assuming a HDOP lateral width W se of (left) 400 m and (right) 150 m.

maximum change of 0.76 mA/cm2 (2.1%) has been observed. For a given dose, a higher Cp eak leads to a larger Jsc (see Fig. 5), despite higher Auger and surface recombination losses in the exposed HDOP area, because of the shallower junction. The dominant impact of the HDOP prole on Jsc is conrmed by the efciency graph (see Fig. 6), where a similar trend can be found. Fig. 6 shows that an improper HDOP diffusion may result in an efciency loss up to 0.35%abs . It is worth noting that the lower the dose, the lower the impact of the HDOP prole. This is due to the fact that at low doses, by varying the Cp eak , the change in the junction depth of the corresponding HDOP proles is lower than that observed at higher values of dose, and hence, the effect on Jsc is similar for the HDOP proles at such doses. The analysis also shows in Fig. 6(a) slight effect of the HDOP prole on the FF with a maximum variation of 0.15%abs , due to the fact that the emitter resistance of the SE cell is dominated by the LDOP regions. A higher value of FF is observed by increasing the dose and, for a given dose, by decreasing the peak concentration, due to the reduction of the HDOP sheet resistance (see Table III). However, for Cp eak = 11020 cm3 , the higher contact resistance leads to a slight decrease of the FF. V. INVESTIGATION ON THE ALIGNMENT TOLERANCE The analysis performed in Section IV, which was carried out by considering a large lateral width of the HDOP region (Wse ) equal to 250 m (150 m wider than the nger width), clearly shows a strong impact of different HDOP proles on the Jsc and, then, on the efciency of the SE solar cell, particularly at low sheet resistances (hence at high dose concentrations) of the HDOP prole. In order to understand the impact on the SE cell performance of the alignment tolerance used in the SE diffusion process for the subsequent metallization process, the previous analysis has been repeated for other two values of Wse , 400 and 150 m. The relative efciency curves are reported in Fig. 7.

This further investigation reveals a strong inuence of the HDOP lateral width on the Jsc and, consequently, on the efciency of the SE cell, especially, at low sheet resistance HDOP proles. In fact, as long as the alignment tolerance (and, then, the exposed HDOP area) decreases, the efciency of the SE cell becomes less dependent on the dose and the peak doping. This is because the exposed HDOP region (outside the front metal nger) is a source of large variation on Jsc , due to a Jdepth dependent effect (as discussed in Section IV). Indeed, while for a Wse = 400 m a large efciency variation of 0.70%abs is observed (see Fig. 7), a smaller Wse = 150 m results in a much lower variation (0.12%abs ). This means that an improved alignment process and, consequently, a smaller alignment tolerance can reduce the dependence of the SE performance on the HDOP prole and, hence, makes less critical the SE diffusion process, allowing a wider process window and, consequently, a better control of the diffusion process. Obviously, a reduced Wse also enhances the cell performance. This results in an increased gain efciency of 0.47%abs for the SE cell with respect to the 75-/sq HE cell (see Fig. 7). It is particularly important to note (focusing on the lower dose curves in Figs. 6 and 7) that a lower dopant dose and, hence, a higher sheet resistance lead to a reduction of the benet achievable from a decreased alignment tolerance. This means that the alignment tolerance becomes a less critical issue when the HDOP regions feature a higher sheet resistance. The SE (Wsub = 2 mm, 100-/sq LDOP, and 50-/sq HDOP) and HE (Wsub = 2 mm, 75 /sq) solar cells have been also compared in Fig. 8 in terms of collection efciency of the photogenerated electronhole pairs, which is calculated as the ratio between the short-circuit current density Jsc and the photogenerated current density Jph , within the wavelength range 3001200 nm. As expected, the SE cell shows a better spectral response in the blue region of the spectrum, resulting in a higher short-circuit current density and open-circuit voltage (as already discussed previously). In addition, the simulation results

DE ROSE et al.: UNDERSTANDING THE IMPACT OF THE DOPING PROFILES ON SELECTIVE EMITTER SOLAR CELL

165

cell becomes more efcient than the SE cell, due to the reduced emitter resistance and, then, to the higher FF. VI. CONCLUSION In this paper, we have investigated the dependence of the performance of an SE solar cell on the doping proles both in the nonmetallized LDOP areas and in the metallized HDOP areas. First, the effect of the doping prole in the LDOP areas at different front contact pitch values has been highlighted. The analysis shows that the efciency tradeoff, resulting from the emitter resistance, the illuminated area, and the recombination losses, leads to an optimal 100-/sq LDOP prole at a front contact pitch of 2 mm. Afterward, the impact of the HDOP prole underneath the front metallization has been analyzed. This study demonstrates that the HDOP prole strongly inuences the short-circuit current density and, consequently, the efciency of the SE cell, particularly at low sheet resistances of the HDOP prole; therefore, an improper diffusion may lead to an efciency loss up to 0.35%abs . This strong dependence can be mainly attributed to the large alignment tolerances used in the SE diffusion process for the subsequent metallization process, resulting in broad HDOP areas. Thus, a consequent analysis of this aspect has also been carried out, showing that the reduction of the alignment tolerance could be highly benecial for the SE solar cell, since a smaller lateral width of the HDOP areas (Wse ) allows us to reduce the inuence of the HDOP prole on the SE performance and, hence, to make less critical the SE diffusion process. Indeed, while for a lateral width of the HDOP region Wse = 400 m a large efciency variation of 0.70%abs is observed by varying the HDOP dose, a smaller Wse = 150 m results in a much lower variation (0.12%abs ). On the other hand, the same analysis reveals that the alignment tolerance becomes a less critical issue when the HDOP regions feature a higher sheet resistance. REFERENCES
[1] T. Dullweber, S. Gatz, H. Hannebauer, T. Falcon, R. Hesse, J. Schmidt, and R. Brendel, Towards 20% efcient large-area screen-printed rearpassivated silicon solar cells, Prog. Photovoltaics: Res. Appl., vol. 20, pp. 630638, 2012. [2] M. M. Hilali, A. Rohatgi, and S. Asher, Development of screen-printed silicon solar cells with high ll factors on 100 /sq emitters, IEEE Trans. Electron Devices, vol. 51, no. 6, pp. 948955, Jun. 2004. [3] M. M. Hilali, A. Rohatgi, C. Khadilkar, S. Kim, T. Pham, J. Salami, A. Shaikh, and S. Sridharan, Understanding and development of Ag pastes for silicon solar cells with high sheet-resistance emitters, in Proc. 19th Eur. Photovoltaic Solar Energy Conf., 2004, pp. 13001303. [4] A. Ebong, A. Rohatgi, W. Zhang, and P. Bokalo, Understanding the role of glass frit in the front Ag paste for high sheet resistance emitters, in Proc. 22nd Eur. Photovoltaic Solar Energy Conf., 2007, pp. 17341737. [5] A. Ebong, D. S. Kim, A. Rohatgi, and W. Zhang, Understanding the mechanism of light induced plating of silver on screen-printed contacts for high sheet resistance emitters with low surface phosphorous concentration, in Proc. 33rd IEEE Photovoltaic Spec. Conf., May 2008, pp. 15. [6] A. Ebong, B. Rounsaville, I. B. Cooper, K. Tate, A. Rohatgi, S. Glunz, M. Horteis, A. Mette, and M. Gundermann, High efciency silicon solar cells with ink jetted seed and plated grid on high sheet resistance emitter, in Proc. 35th IEEE Photovoltaic Spec. Conf., Jun. 2010, pp. 13631367. [7] M. H orteis and S. W. Glunz, Fine line printed silicon solar cells exceeding 20% efciency, Prog. Photovoltaics: Res. Appl., vol. 16, no. 7, pp. 555 560, 2008.

Fig. 8. Collection efciency of the photogenerated electronhole pairs, which is calculated as the ratio between the short-circuit current density Jsc and the photogenerated current density Jp h , within the wavelength range 3001200 nm for the simulated HE (W su b = 2 mm, 75 /sq) and SE (W su b = 2 mm, 100- /sq LDOP, 50- /sq HDOP) cells.

TABLE IV AUGER AND SURFACE RECOMBINATION LOSSES

in terms of collection efciency (see Fig. 8) show an improved blue spectral response at Wse = 150 m, conrming that the reduction of the alignment tolerance is benecial for the SE cell. The benet of the SE cell in terms of blue spectral response with respect to the HE cell is mainly due to the lower doping concentration in the LDOP emitter regions, leading to reduced Auger and surface recombination losses as well as to the shallower LDOP junction which improves the collection of minority carriers generated close to the front surface mainly contributed by photons at lower wavelengths (in the range 300600 nm). In Table IV, we individually evaluate the effect of Auger and surface recombination losses in our simulated cells, by disabling alternately the two mechanisms in the simulation. When Auger recombination mechanism is not taken into account, the efciency gain of SE over HE slightly reduces from 0.41%abs to 0.37%abs . On the other hand, by disabling the front surface recombination, the efciency gain of SE over HE reduces to only 0.06%abs . Therefore, numerical simulations show that the decrease of the surface recombination losses is the main advantage of the SE design compared with the standard HE solar cell, as already highlighted by the loss analysis results reported in Section III. Furthermore, in Table IV, it has been demonstrated that by neglecting both the recombination mechanisms, the HE

166

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 1, JANUARY 2013

[8] G. Hahn, Status of selective emitter technology, in Proc. 25th Eur. Photovoltaic Solar Energy Conf., 2010, pp. 10911096. [9] T. R oder, P. Grabitz, S. Eisele, C. Wagner, J. R. Kohler, and J. H. Werner, 0.4% absolute efciency gain of industrial solar cells by laser doped selective emitter, in Proc. 34th IEEE Photovoltaic Spec. Conf., Jun. 2009, pp. 871873. [10] U. J ager, S. Mack, A. Kimmerle, A. Wolf, and R. Preu, Inuence of doping prole of highly doped regions for selective emitter solar cells, in Proc. 35th IEEE Photovoltaic Spec. Conf., Jun. 2010, pp. 31853189. [11] B. Raabe, F. Book, A. Dastgheib-Shirazi, and G. Hahn, The development of etch-back processes for industrial silicon solar cells, in Proc. 25th Eur. Photovoltaic Solar Energy Conf., 2010, pp. 11741178. [12] F. Book, B. Raabe, and G. Hahn, Two diffusion step selective emitter: Comparison of mask opening by laser or etching paste, in Proc. 23rd Eur. Photovoltaic Solar Energy Conf., 2008, pp. 15461549. [13] M. Edwards, J. Bocking, J. E. Cotter, and N. Bennett, Screen-print selective diffusions for high-efciency industrial silicon solar cells, Prog. Photovoltaics: Res. Appl., vol. 16, no. 7, pp. 3145, 2008. [14] P. Kumar, K. Wijckoon, D. Tanner, K. Rapolu, D. Zhe, H. Huang, B. Zhang, S. Wang, A. Bhanap, and D. Chen, Industrial selective emitter processes for Czochralski-silicon solar cells, in Proc. 26th Eur. Photovoltaic Solar Energy Conf., 2011, pp. 13341339. [15] R. De Rose, M. Zanuccoli, P. Magnone, D. Tonini, M. Galiazzo, G. Cellere, M. Frei, H.-W. Guo, C. Fiegna, and E. Sangiorgi, 2-D numerical analysis of the impact of the highly-doped prole on selective emitter solar cell performance, in Proc. 37th IEEE Photovoltaic Spec. Conf., Jun. 2011, pp. 25562559. [16] N. Stem and M. Cid, Physical limitations for homogeneous and highly doped n-type emitter monocrystalline silicon solar cells, Solid-State Electron., vol. 48, no. 2, pp. 197205, 2004. [17] A. Cuevas and D. A. Russell, Co-optimisation of the emitter region and the metal grid of silicon solar cells, Prog. Photovoltaics: Res. Appl., vol. 8, no. 6, pp. 603616, 2000. [18] Synopsis, Sentaurus device user guide, Version C-2009.06, 2009. [19] P. P. Altermatt, S. Steingrube, Y. Yang, C. Sprodowski, T. Dezhdar, S. Koc, B. Veith, S. Herrman, R. Bock, K. Bothe, J. Schmidt, and R. Brendel, Higly predictive modelling of entire Si solar cells for industrial applications, in Proc. 24th Eur. Photovoltaic Solar Energy Conf., 2009, pp. 901906. [20] R. De Rose, M. Zanuccoli, P. Magnone, E. Sangiorgi, and C. Fiegna, Open issues for the numerical simulation of silicon solar cell, in Proc. 12th Int. Conf. Ultimate Integr. Silicon, 2011, pp. 14. [21] A. Schenk, Finite-temperature full random-phase approximation model of band gap narrowing for silicon device simulation, J. Appl. Phys., vol. 84, no. 7, pp. 36843695, 1998. [22] P. P. Altermatt, Models for numerical device simulations of crystalline silicon solar cellsa review, J. Comput. Electron., vol. 10, no. 3, pp. 314 330, 2011. [23] D. B. M. Klaassen, A unied mobility model for device simulationI: Model equations and concentration dependence, Solid-State Electron., vol. 35, no. 7, pp. 953959, 1992. [24] D. B. M. Klaassen, A unied mobility model for device simulationI: Temperature dependence of carrier mobility and lifetime, Solid-State Electron., vol. 35, no. 7, pp. 961967, 1992. [25] S. W. Glunz, S. Rein, J. Y. Lee, and W. Warta, Minority carrier lifetime degradation in boron-doped Czochralski silicon, J. Appl. Phys., vol. 90, no. 5, pp. 23972404, 2001. [26] P. P. Altermatt, A. Schenk, F. Geelhaar, and G. Heiser, Reassessment of the intrinsic carrier density in crystalline silicon in view of band-gap narrowing, J. Appl. Phys., vol. 93, no. 3, pp. 15981604, 2003. [27] A. Kimmerle, A. Wolf, U. Belledin, and D. Biro, Modelling carrier recombination in highly phosphorus-doped industrial emitters, in Proc. Int. Conf. Crystalline Silicon Photovoltaics, 2011, vol. 8, pp. 275281. [28] P. Magnone, R. De Rose, M. Zanuccoli, D. Tonini, M. Galiazzo, G. Cellere, H.-W. Guo, M. Frei, E. Sangiorgi, and C. Fiegna, Understanding the impact of double screen-printing on silicon solar cells by 2-D numerical simulations, in Proc. 37th IEEE Photovoltaic Spec. Conf., Jun. 2011, pp. 21772180. [29] A. Goetzberger, J. Knobloch, and B. Voss, Crystalline Silicon Solar Cells. Hoboken, NJ: Wiley, 1998. [30] D. K. Schroder and D. L. Meier, Solar cell contact resistanceA review, IEEE Trans. Electron Devices, vol. 31, no. 5, pp. 637647, May 1984. [31] G. Schubert, J. Horzel, and S. Ohl, Investigations on the mechanism behind the benecial effect of a forming gas anneal on solar cells with silver thick lm contacts, in Proc. 21st Eur. Photovoltaic Solar Energy Conf., 2006, pp. 14601466.

[32] G. Schubert, J. Horzel, R. Kopecek, F. Huster, and P. Fath, Silver thick lm contact formation on lowly doped phosphorus emitters, in Proc. 20th Eur. Photovoltaic Solar Energy Conf., 2005, pp. 934937. [33] S. Kontermann, Characterization and modeling of contacting crystalline silicon solar cells, Ph.D. dissertation, Dept. Phys., Univ. Konstanz, Konstanz, Germany, 2009. [34] P. N. Vinod, SEM and specic contact resistance analysis of screenprinted Ag contacts formed by re-through process on the shallow emitters of silicon solar cells, J. Mater. Sci., vol. 20, pp. 10261032, 2009. [35] D. Fenning, M. Bertoni, and T. Buonassisi, Predictive modeling of the optimal phosphorus diffusion prole in silicon solar cells, in Proc. 24th Eur. Photovoltaic Solar Energy Conf., 2009, pp. 18931896. [36] S. Solmi, A. Parisini, and R. Angelucci, Dopant and carrier concentration in Si in equilibrium with monoclinic SiP precipitates, Phys. Rev. B, vol. 53, no. 12, pp. 78367841, 1996. [37] M. Zanuccoli, P. F. Bresciani, M. Frei, H.-W. Guo, H. Fang, M. Agrawal, C. Fiegna, and E. Sangiorgi, 2-D numerical simulation and modeling of monocrystalline selective emitter solar cells, in Proc. 35th IEEE Photovoltaic. Spec. Conf., Jun. 2010, pp. 22622265. Raffaele De Rose received the B.S. and M.S. degrees in electronic engineering from the University of Calabria, Rende, Italy, in 2007 and 2009, respectively. He is currently working toward the Ph.D. degree with the University of Calabria within the Ph.D. School Archimedes in Science, Communication, and Technologies. He is currently with the Advanced Research Center on Electronic Systems for Information and Communication Technologies E. De Castro (ARCES), University of Bologna, Italy. In 2011, he did an internship of six months with the Interuniversity MicroElectronics Center (IMEC), Leuven, Belgium, where he worked on 3-D numerical simulation of advanced silicon-based solar cells. His research interests include the modeling and the electro-optical numerical simulation of semiconductor devices and silicon solar cells.

Mauro Zanuccoli received the M.S. degree in electronic engineering from the University of Bologna, Italy, in 2000 and the Ph.D. degree in information technology from the University of Bolgna, in 2012. In 2009, he joined the Advanced Research Center on Electronic Systems for Information and Communication Technologies E. De Castro (ARCES), University of Bologna. During the Ph.D program, he wasa Visiting Scientist with the Applied Materials Inc., Santa Clara, CA, the Grenoble-INP, France, and the Institute of Physics and Technology of the RAS (FTIAN), Moscow, Russia. His research interests include the electro-optical numerical simulation and modeling of photovoltaic silicon solar cells.

Paolo Magnone received the B.S. and M.S. degrees in electronic engineering from the University of Calabria, Rende, Italy, in 2003 and 2005, respectively, and the Ph.D. degree in electronic engineering from the University of Reggio Calabria, Italy, in 2009. From 2006 to 2008, he was with the Interuniversity MicroElectronics Center (IMEC), Leuven, Belgium, within the Advanced PROcess Technologies for Horizontal Integration project (Marie Curie Actions), where he worked on parameter extraction and matching analysis of FinFET devices. He was a Postdoctoral Researcher with the University of Calabria from 2009 to 2010. He is currently with the Advanced Research Center on Electronic Systems for Information and Communication Technologies E. De Castro (ARCES), University of Bologna, Italy. His research interests include the electrical characterization, electro-thermal simulation, and modeling of advanced metal-oxide semiconductor eld-effect transistor devices and silicon solar cells.

DE ROSE et al.: UNDERSTANDING THE IMPACT OF THE DOPING PROFILES ON SELECTIVE EMITTER SOLAR CELL

167

M. Frei, photograph and biography not available at the time of publication.

Enrico Sangiorgi (F05) received the Laurea degree from the University of Bologna, Italy, in 1979. He has been a Visiting Scientist with Stanford University, Stanford, CA, and Bell Laboratories, Murray Hill, NJ. In 1993, he was appointed Full Professor of Electronics with the University of Udine, Italy. In 2002, he joined the University of Bologna, where he is now Director of the Electrical Engineering Department Guglielmo Marconi. His research covers device modeling. Dr. Sangiorgi is a Distinguished Lecturer of the IEEE; he has been Chairman of the Electron Devices Society (EDS) and the TRANSACTIONS ON COMPUTER-AIDED DESIGN Technical Committee and a member of the Cledo Brunetti Award and Education Award Committees of the EDS. He has been Editor of the IEEE ELECTRON DEVICE LETTERS and on the editorial boards of the IEEE TRANSACTIONS ON ELECTRON DEVICES, and the IEEE JOURNAL OF PHOTOVOLTAICS.

Claudio Fiegna received the Laurea and Ph.D. degrees in electronics engineering from the University of Bologna, Italy. He is currently a Full Professor of electronics with the University of Bologna, where he is also the Director of the Advanced Research Center on Electronic Systems for Information and Communication Technologies E. De Castro (ARCES). His research interests include numerical device simulation and its application to the analysis of advanced metal-oxide semiconductor device structures, and silicon-based solar cells. Dr. Fiegna serves on the committee for the Educational Award of the IEEE Electron Devices Society.