Rapid in-line diffusion of P emitters

In this section, different P emitters are diffused with the new in-line RTP system located at the Laboratory and Service Centre (LSC) in Gelsenkirchen. They are characterized by sheet resistance and profile measurements. The design of the in-line RTP system has been presented in section 1.3.2.

The diffusion process

The diffusion zone of the in-line RTP furnace features an approximately 1 m long segment exhibiting a plateau of constant temperature. The time the wafer dwells in this segment can be varied easily by varying the speed of the walking string drive . Since the maximum drive speed is 1.8 m/min, the minimum dwell time, or equivalently the diffusion time , becomes 30 to 35 s. If one wants to reduce further, one has to switch off someof the

Tab. 3.5:Wafer dwell (diffusion) time in the diffusion zone of the in-line RTP furnace as a function of drive speed. The diffusion zone exhibiting constant temperature is approximately 1 m long. In the case of process F, the length of the diffusion zone was reduced to 0.5 m by switching off two of its segments.

Process

A B C D E F

[mm/min] 200 400 600 1000 1600 1700

[s] 300 150 100 60 35 18

four segments forming the diffusion zone. Table 3.5 shows as a function of for the diffusion processes utilized in this work. All diffusion processes used in this work have in common that wafer heating by the tungsten halogen lamps was performed solely in the diffusion zone. The heating of the remaining zones (including the RTP unit) was switched off completely.

They were heated only indirectly by light and convection of heat coming from the diffusion zone. In addition, the mercury UV lamps were not used.

Due to the low thermal mass of the walking string drive and due to wafer heating by visible light from the THLs, the wafer can be heated up rapidly. Also, because of the water-cooled cooling zones it also cools down quickly. Hence, the emitter diffusion takes place in the actual diffusion zone mainly. Exemplarily, Fig. 3.12 shows the temperature time cycle of diffusion process D. The temperature was acquired by a thermocouple (TC) mounted on a 10 10 cm² wafer which was sent through the furnace. The in-line RTD processes are regulated by the temperature of thermocouples attached to the quartz channel. In the case of process D, their temperature in the diffusion zone is set to 900 C. Apparently, in the diffusion zone, the actual wafer temperature is 40 to 50 C higher than the temperature of the surrounding quartz channel.

This is due to the prevailing non-equilibrium conditions. Compared to real cold wall processing as present in a single-wafer RTP system this temperature difference is comparatively small.

Nevertheless, one has to keep in mind that the process temperature adjusted by setting the temperature of the TCs located within the quartz chamber deviates significantly from the actual wafer temperature, i.e. from the diffusion temperature. According to Fig. 3.12 the diffusion plateau lasts approximately 60 s. This is exactly the value calculated from the extension of the diffusion zone and the applied string speed. Taking the derivative of the temperature-time cycle we obtain a heating rate of up to 40 C/s and a cooling rate of 25 C/s at the maximum.

Not shown here, in the case of RTD process F, heating and cooling rates of 50 and 30 C/s are achieved, respectively. Such high heating and cooling rates can only be achieved due to the low thermal mass of the walking string drive, wafer heating by irradiation and the well cooled cooling zones. For emitter diffusion, the spin-on source P508 was applied single-sided.

Emitter sheet resistance

has been studied as a function of drive speed (i.e. ) and temperature of the quartz channel within the diffusion zone as measured by TCs. The result is plotted in Fig. 3.13. From this figure, necessary to obtain the desired 40 to 50 /sq emitter has

20 40 60 80 100 120 140 160 180 200 220 100

200 300 400 500 600 700 800 900 1000

RTP unit

Te m pe ra tu re [° C ]

Time [s]

t_diff = 60 s T_quartz= 900 °C T_wafer

cooling zone diffusion zone

Fig. 3.12: Wafer temperature as acquired by a TC mounted on a 10 10 cm² wafer, which was sent through the furnace at a speed of 1000 mm/min (process D from Tab. 3.5). The process was controlled by the TCs measuring the temperature of the quartz channel which was set to 900 C.

825 850 875 900 925

20 40 60 80 100

R

[

/s q]

T

diffusion zone [°C]

v_drive [mm/min]

200 400 600 1000 1600

45 Ω/sq

Fig. 3.13: as a function of drive speed (i.e. ) and temperature of the quartz channel . The P SOD P508 was used. Please note that the actual wafer temperature is 40 to 50 C higher than due to the prevailing non-equilibrium conditions.

been deduced. As shown, must not be confused with the actual wafer temperature. In the case of in-line RTD the actual wafer temperature is almost impossible to measure. For this reason, we decided to characterize an in-line RTD process solely by the drive speed and the sheet resistance necessary to obtain a certain sheet resistance and not to mention the respective temperature of the quartz channel. Nevertheless, we can say that the wafer temperature is a monotonous function of the quartz temperature. Hence, from Fig. 3.13 we can conclude that the faster the diffusion process, the higher the required temperature to obtain a 40 to 50 /sq emitter.

Emitter profiles

Comparative Stripping Hall and SIMS measurements were performed on polished samples for a 5 min (process A) and a 1 min (process D) diffused emitter. In order to get a similar , the diffusion time was higher in the case of the faster diffusion. As can be seen in the left graph of Fig. 3.14, the profiles show the following characteristics:

The P508 yields a high concentration (up to 10²¹ cm^-3) of inactive P in the surface near region. This can be deduced from the fact that the P concentration exceeds the electron concentration.

Both profiles exhibit a plateau of constant electron concentration followed by a kink and a tail (which is typical for CFD emitters also [178]).

The electron and the P concentration lie above each other for concentrations below the plateau concentration.

The surface near concentration of the active as well as the surface near concentration of the inactive P seems to increase with diffusion temperature.

Further samples were diffused in the in-line RTD system with a diffusion plateau time of approximately 18 s (process F) and by CFD at 900 C for 15 minutes using POCl3. The first one features an of 49 /sq and the latter one of 45 /sq. The right graph in Fig. 3.14 shows a comparison of the respective electron profiles measured by Stripping Hall. The CFD emitter exhibits the deepest profile of all emitters. Evaluated at = 10¹⁶ cm^-3, the junction depth! is approximately 350 nm in the case of the CFD emitter and only approximately 180 nm in the case of the 18 s RTD emitter, though is almost the same. Clearly, for similar , ! increases for diffusion processes using the longer diffusion time. This effect is not outweighed by the higher diffusion temperature necessary in the case of the faster diffusions. But, the higher temperature applied in the fast processes leads to a higher near surface electron concentration.

This is because the solubility of P in Si increases with temperature. In conclusion:

In general, RTD emitters feature shallower junction depths and higher near surface electron concentrations than CFD emitters of similar sheet resistance. This is caused by the higher diffusion temperature in combination with a generally shorter diffusion time applied in RTD.

0 100 200 300

Fig. 3.14: Left) Comparison of electron and P profiles of in-line RTD emitter using the SOD P508.

Right) Electron profiles of different RTD emitters and of a CFD emitter of similar sheet resistances. The CFD emitter was diffused using POCl₃.

We have established an approximately 18 s in-line diffused 50 /sq emitter with a junction depth of only 180 nm. In section 4.3.4 we demonstrate that this extremely shallow emitter can in fact be contacted by industrially feasible screen-printing of Ag front contacts. In section 4.6.7 the in-line RTD processes are used to fabricate solar cells from EFG silicon. For our single-track prototype furnace, the associated throghput of 125 125 mm² wafers is in the range of 600 wafers/h, taking a wafer spacing of 45 mm and a transportation speed of 1700 mm/min.

Assuming that a production type version would feature at least four tracks, the 18 s RTD process will enable a throughput of 2400 wafers/h with a just 1 m long diffusion zone and an overall furnace length well below 4 m.