Investigation of Boron Thermal Diffusion from Atmospheric Pressure Chemical Vapor Deposited Boron Silicate Glass for N-Type Solar Cell Process Application

An atmospheric pressure chemical vapor deposition (AP-CVD) system has been newly developed for boron silicate glass (BSG) film deposition dedicating to solar cell manufacturing. Using the system, thermal boron diffusion from the BSG film is investigated and confirmed in terms of process stability for surface property before BSG deposition and BSG thickness. No degradation in carrier lifetime is also confirmed. A boron diffusion simulator has been newly developed and demonstrated for optimization of this process. Then, the boron thermal diffusion from AP-CVD BSG is considered to be the suitable method for N-type silicon solar cell manufacturing.


Introduction
In development of advanced solar cell, improvement of efficiency and reliability with maintaining low manufacturing cost is one of crucial issues.To achieve the high efficiency and reliability, the N-type silicon solar cell is the most likely candidate although the P-type silicon solar cell is now industrial standard [1,2].Indeed, the P-type silicon cell has a reliability problem which is efficiency degradation due to light illumination, which is called light induced degradation (LID) caused by formation of boron and interstitial oxygen pairs [3][4][5].On the other hand, the N-type silicon cell which uses the phosphorus doped silicon wafer has no such LID and higher immunity to metallic contaminations comparing to the P-type silicon cell [6].In order to manufacture the Ntype silicon cell, a P + diffusion layer which is usually a boron doped layer has to be formed.Thus, many boron diffusion processes have been proposed and used to fabricate the Ntype silicon solar cell.Open-tube liquid source diffusion of BBr 3 [7,8] must be a common industrial standard process but the number of process steps increases because of diffusion protection film formation on the opposite side.Ion implantation is the most advanced process and has precise controllability of the dopants [9].However, crystal defects are induced by the implantation and thermal annealing after the implantation required for recovering of the crystal defects.In addition, low throughput of the implantation is also one of the problems.On the other hand, boron thermal diffusion from a deposited boron silicate glass (BSG) film is a simple and well matured method [10][11][12].Particularly, the BSG film deposition using atmospheric pressure chemical vapor deposition (AP-CVD) is a cost-effective process because of the simple machine composition.In this paper, characteristic stability for process variation of the boron thermal diffusion from the BSG film deposited by newly developed AP-CVD system has been investigated.For optimization of the process, boron diffusion simulator has been developed and demonstrated for prediction of the solar cell performance.

Experimental Procedure
Czochralski-(CZ-) grown n-type (100)-oriented silicon wafers were used as starting materials.The dopant of silicon wafer is phosphorus with 6-9 ohm cm in resistivity.The wafer thickness is 675 m for the boron diffusion experiments.We compared the boron diffusion profiles between 675 m and 180 m thick wafers which are usually used in solar cell manufacturing but no difference was observed in the boron diffusion profiles.For the other experiments such as lifetime measurements, the 180 m thick wafers were used.Prior to a BSG film deposition, the wafers were cleaned by diluted hydrofluoric acid (DHF) to remove native oxide on the wafers.To study the variation of diffusion process, some wafers were dipped in HCl/H 2 O 2 solution after the DHF cleaning to make chemical oxide on the wafer surface.The wafers with a 5 nm CVD oxide film on the surface were also prepared.The BSG film was deposited by the AP-CVD system with SiH

AP-CVD BSG Deposition System
The newly developed AP-CVD BSG deposition system consists of multigas dispersion heads (DH1, DH2, and DH3), SiC trays, resistive heaters, and predeposition lamp heaters as shown schematically in Figure 1 [13].To deposit the BSG film, thermal decomposition of SiH 4 , O 2 , and B 2 H 6 gases is used.These gases are delivered from the dispersion heads to the wafer surface.The wafers are automatically loaded and unloaded from the wafer carriers to the SiC trays.The wafers are heated by the resistive heaters.The resistive heaters are covered with quartz glass underneath the SiC trays.The SiC trays move from the wafer loading portion to the unloading portion and carry the wafers through the oxide deposition zones under the dispersion heads.The predeposition heaters prevent wafer bending especially for thin wafers.We set the important factors of system development for the solar cell manufacturing which are (i) boron concentration uniformity within a wafer and between wafers, (ii) suppression of metallic contamination, and (iii) better Cost of Ownership (CoO).Owing to the dispersion heads designed by a process simulation, good sheet resistance uniformity within 10% after the thermal diffusion is realized as shown in Figure 2. Because the wafers contact only with the trays, the SiC trays prevent metallic contamination during the process.Previous AP-CVD tools commonly use Inconel mesh belts to transfer wafers underneath the deposition head zone which means that there is possibility of metallic contaminations such as Nickel, Iron, and Chrome in the wafers.The metallic contamination in the wafers makes the problems of solar cell performance degradation.Less footprint and simple machine composition with no vacuum system is feasible for better CoO.In addition, because of the multihead concept, a stacked film such as undoped silicate glass (USG)/BSG is easily deposited by one path.The oxide is deposited only on the surface of wafers, which is suitable for the codiffusion process [14][15][16] and the manufacturing cost reduction.In this process, deposited BSG film thickness and boron concentration have to be controlled.The deposited film thickness can be controlled by the tray speed as well as the deposition temperature and the SiH 4 /O 2 gas flow rate.The BSG boron concentration is controlled by the gas flow ratio of B 2 H 6 to SiH 4 .The relation between the gas flow ratio and the BSG boron concentration which was measured by SIMS is shown in Figure 3.

Effect of Native Oxide Underneath BSG for Boron Diffusion.
With consideration of the stable boron diffusion layer formation, queue time control between precleaning and BSG deposition is one of the key manufacturing parameters.When the queue time is long, the native oxide must be formed on the silicon surface.Thus, there is a possibility to alternate  the boron diffusion profiles due to existence of the native oxide.We have investigated the boron profiles after the thermal diffusion by SIMS when there is no oxide, chemical oxide, or CVD oxide between the BSG film and the silicon surface.The precleaning by DHF just before the BSG deposition was used to obtain the "no oxide" sample.To make chemical oxide on the surface, HCl/H 2 O 2 precleaning was used and the thickness of the chemical oxide was estimated to be 1-2 nm.The sample with a 5 nm CVD oxide film on the surface was also prepared.After these surface treatments, a 50 nm BSG film was deposited.The boron diffusion profiles for these samples annealed for 20 min at 950 ∘ C are shown in Figure 4.
As shown in the figure, the profiles are almost identical.It can be concluded that such chemical or CVD oxide film does not affect boron diffusion from BSG.Thus, this process has enough stability for the queue time control between the precleaning and the BSG deposition which is a superior point for the production control.control.The BSG boron concentration of both samples is 8.41 × 10 21 cm −3 .The boron profiles after the diffusion are shown in Figure 5.There is no difference in the boron profiles.The BSG thickness of 40 nm is confirmed to be enough to make the infinite diffusion source of boron in this diffusion condition.Furthermore, no thickness dependence in the diffusion profile is another superior point for the production control.

BSG Boron Concentration Dependency.
In this process, the P + layer sheet resistance can be controlled by changing the BSG boron concentration and/or thermal diffusion conditions such as temperature and time.Figure 6 shows the P + sheet resistance as a function of the BSG boron concentration with the different diffusion temperatures.The diffusion time was fixed at 20 min.As shown in this figure, the sheet resistance decreases with the BSG boron concentration up to certain BSG boron concentrations and then slightly increases.From this minimum sheet resistance point, BRL is starting to form at the surface.With the BSG boron concentration increasing, the surface boron concentration also increases.When the surface concentration reaches the solid solubility limit, BRL is formed on the surface of silicon.Because BRL consists of silicon borides, the crystal structure changes.Thus, BRL existence on the surface affects boron diffusion as described in [17].To manipulate and predict the boron diffusion profile, a simulator to calculate the boron profile after the diffusion is indispensable because the diffusion mechanism is quite complicated [18,19].Then, we have made the boron diffusion simulator by solving diffusion equation BSG boron concentration (cm −3 ) Figure 6: Sheet resistance dependence on BSG boron concentration as a parameter of diffusion temperatures.The sheet resistance can be controlled by the BSG boron concentration and diffusion conditions such as temperature and time.BRL is formed when the BSG concentration is higher than the concentration at the minimum sheet resistance.
numerically and the boron diffusion model of this simulation will be explained in the next section.

Carrier Lifetime after Boron Diffusion.
Another important issue to apply the process to solar cell manufacturing is the effect on the carrier lifetime.It is reported that the opentube liquid source BBr 3 diffusion induces the crystal defects and the carrier lifetime degradation due to the formation of BRL at the surface [7,20].On the contrary, it is also reported that the formation of BRL improves the lifetime owing to metallic impurity gettering of BRL [21].To investigate the effect on the lifetime using this process and existence of BRL on the surface, the BSG films were deposited on both sides of 156 mm square 180 m thick silicon wafers with two different boron concentrations, 3.72 × 10 21 cm −3 and 6.72 × 10 21 cm −3 , which are without and with BRL after the diffusion, respectively.In addition, both side USG deposited Si wafers are also prepared as references.The carrier lifetimes after diffusion at different temperatures for 20 min are measured by microwave photoconductive decay (-PCD).Before the measurements, the P + diffusion layers were etched off by a mixed solution of HNO 3 and HF.After RCA precleaning, the surfaces of wafers were chemically passivated by iodine to reduce the effect of surface recombination in the lifetime measurements.The lifetime was measured on area of whole wafer and averaged values were used.The results are shown in Figure 7 including the lifetime of a starting Si wafer without annealing.In the case of no existence of BRL on the surface (3.72 × 10 21 cm −3 ), the lifetimes decrease with the diffusion temperature.The lifetimes are, however, almost the same as those of the USG deposited samples.This means that the decrease of lifetimes is not due to the boron diffusion.When BRL is formed on the  The lifetimes of low BSG boron concentration decrease with diffusion temperature but almost the same as that of USG deposited samples.The lifetimes of samples with BRL are higher than that of the initial silicon wafer.surface (6.72 × 10 21 cm −3 ), the lifetimes are more than 1 ms which is higher than that for the initial Si wafer.From this result, the gettering effect by BRL is confirmed in this process.It is also concluded that the thermal boron diffusion from AP-CVD BSG does not affect the carrier lifetime.This process was also applied to N-type silicon solar cell fabrication instead of BBr 3 diffusion.We got 19.35% in efficiency using this process with comparing to 19.13% for the BBr 3 diffusion.This result also supports that the boron diffusion from AP-CVD BSG does not induce any crystal defect.

Boron Diffusion Profile Prediction Simulator
The boron thermal diffusion simulator called DADiS (D&S AMAYA Diffusion Simulator) has been newly developed to predict the diffusion profile using process parameters which are the BSG boron concentration, the thermal diffusion temperature profile including stand-by temperature, ramping up and down rate, and diffusion temperature and time.In this simulator, the diffusion equation, where  is the boron concentration in silicon,  is the diffusivity of boron in silicon,  is the depth from the surface, and  is the diffusion time, is solved numerically with the constant surface boron concentration because the BSG film is the finite boron diffusion source.The important factor  to predict the profile is the accurate diffusivity model.The diffusivity model in this simulator is used as where   is the intrinsic diffusivity showing BSG boron concentration dependence and   is the intrinsic carrier concentration at the diffusion temperature [22].The intrinsic diffusivity has temperature dependence with the activation energy   as where  is the Boltzmann constant,  is the diffusion temperature, and   ref is the intrinsic diffusivity at the reference temperature  ref .The surface boron concentration equations as a function of the BSG boron concentration were also modeled.The boron profile in BRL is assumed to be a complimentary error function with fitting parameters.Detail explanation of the model is found in [23].Examples of comparison between measured and DADiS simulated boron profiles are shown in Figure 8. the diffusion temperature, and 20 to 60 min in the diffusion time which are covered commonly used process conditions.Another important prediction by the simulator is the sheet resistance after the diffusion because the sheet resistance is commonly used for monitoring of the boron diffusion in the solar cell manufacturing.When the boron diffusion profile is known, the sheet resistance   is given as where  is the elementary electric charge,  is the mobility of holes, and  boron is the number of boron atoms in the diffusion layer [24].It is well known that the mobility has impurity concentration dependence and is empirically given as where  1 and  2 are the fitting parameters [25].Using the boron profile calculated by DADiS, (4), and (5),   can be predicted when the BSG boron concentration and the diffusion conditions are known.Figure 10 shows the comparison between measured and simulated   by DADiS [26].The simulated   agrees well with the measured   for a wide range of 30 to 2000 Ω/sq.and the error in resistance is within 20%.The boron profiles and sheet resistance after the thermal diffusion are accurately predicted by DADiS with conditions of the BSG boron concentration and thermal diffusion temperature profiles.Owing to the accuracy of DADiS, the optimization of diffusion process can be achieved with less experiment [27].

TCAD Simulation for Prediction of Solar Cell Characteristics
Using DADiS and well-known solar cell simulator, PC1D [28], the solar cell characteristics such as Jsc, Voc, and efficiency can be easily simulated as a function of the BSG boron concentration or diffusion conditions.The boron diffusion profiles calculated by DADiS can be used as input files of  PC1D to simulate the solar cell characteristics.This method, which is so called technology computer aided design (TCAD) for the solar cell, reduces the development time and cost.For example, the optimized condition of the BSG boron concentration must exist for the solar cell efficiency.The simulation results using DADiS and PC1D are shown in Figure 11.We can find out the optimized condition as shown in the figure.It is indicated that the TCAD simulation can accelerate solar cell development speed and reduce developing cost greatly.

Conclusion
The AP-CVD system to deposit the BSG films has been newly developed for boron diffusion layer formation in the Ntype silicon solar cell manufacturing.This system has good performance for manufacturing with less CoO.Using this system, the process stability of thermal boron diffusion is investigated.It is found that the boron diffusion profiles are almost identical even though there is 5 nm CVD oxide between the BSG film and silicon or BSG thickness is changed from 40 nm to 80 nm, which indicates stable productivity by the system.The boron diffusion profiles can be manipulated by the BSG boron concentration or thermal diffusion conditions such as diffusion temperature and time.It is also confirmed that this process does not affect carrier lifetime.Because the boron diffusion mechanism is complicated, the boron diffusion profile prediction simulator, DADiS, has been developed and demonstrated for the optimization of BSG boron concentration with PC1D.It is concluded that the P + diffusion layer formation by thermal boron diffusion from AP-CVD BSG is the suitable method for the N-type silicon solar cell manufacturing.

2 InternationalFigure 1 :
Figure 1: A schematic of newly developed AP-CVD system.Deposited wafers are carried by SiC trays underneath the multideposition gas dispersion heads (DH1-3).The wafers are also heated by resistive heater coils and lamp heaters at the predeposition zone.

Figure 2 :
Figure 2: Sheet resistance uniformity within a wafer after diffusion.BSG boron concentration is 6.72 × 10 21 cm −3 .Diffusion was performed at 950 ∘ C for 20 min.The BSG film was deposited on a 156 × 156 mm 2 pseudo square silicon wafer.

Figure 3 :
Figure 3: BSG boron concentration measured by SIMS as a function of flow ratio of B 2 H 6 to SiH 4 .The inset graph shows the boron concentration profile in BSG measured by SIMS as an example.The BSG boron concentration can be varied by changing the gas flow ratio of B 2 H 6 to SiH 4 .

Figure 4 :
Figure4: Boron profiles for the sample with DHF preclean (blue line), HCl/H 2 O 2 preclean (red line), and CVD USG 5 nm on the silicon surface (green line).In the case of HF preclean, there must be no oxide between BSG and silicon.In the case of HCl/H 2 O 2 preclean, around 1-2 nm chemical oxide exists between BSG and silicon.Even though the chemical or 5 nm CVD oxide exists, the profiles are identical to no oxide case.

Figure 7 :
Figure 7: Carrier lifetimes as a function of diffusion temperature.The lifetimes of low BSG boron concentration decrease with diffusion temperature but almost the same as that of USG deposited samples.The lifetimes of samples with BRL are higher than that of the initial silicon wafer.

Figure 8 :
Figure 8: Examples of comparison between measured and simulated boron profiles after diffusion by DADiS.The simulated profiles including surface BRL area agree well with the measured profiles for wide range of the BSG boron concentrations and diffusion temperatures.

Figure 9 Figure 9 :
Figure 9: (a) Simulation error in depth at the same boron concentration.(b) is a list of examined BSG boron concentration, diffusion temperature, and time.Even though wide range of the conditions was used, the error is almost within 15%.It is confirmed that the simulator DADiS is accurate and useful for the boron diffusion analysis.

Figure 10 :
Figure 10: Comparison between measured and DADiS simulated sheet resistance.The error of simulation in the sheet resistance is within 20%.

Figure 11 :
Figure 11: Simulated (a) efficiency, (b) short circuit current, and open circuit voltage using PC1D with DADiS as a function of BSG boron concentration.The evaluated BSG boron concentration range is lower than 4 × 10 21 cm −3 and no BRL is formed.The maximum point in efficiency is observed at around 3 × 10 21 cm −3 .
To extract the diffusion activation energy,   , boron profiles of different diffusion temperatures are used.After the extraction of   , the   ref dependence on BSG boron concentration is empirically modeled based on the boron profiles of different BSG boron concentration samples.