Plasma-immersion ion implantation

Plasma-immersion ion implantation

Nuclear Instruments and Methods in Physics Research B 139 (1998) 37±42 Plasma-immersion ion implantation Rainer W. Thomae 1,2 Institut f ur Angewa...

138KB Sizes 0 Downloads 36 Views

Nuclear Instruments and Methods in Physics Research B 139 (1998) 37±42

Plasma-immersion ion implantation Rainer W. Thomae

1,2

Institut f ur Angewandte Physik der Johann Wolfgang Goethe-Universit at, Robert-Mayer-Strasse 2-4, D-60054 Frankfurt am Main, Germany

Abstract Plasma immersion ion implantation (PIII) is a technique for surface modi®cation. In contrast to conventional ion implantation techniques, the target is surrounded by the plasma and then pulse biased to high negative voltages. In this paper, the principle of plasma and high voltage generation will be shortly reviewed. Further on, the paper concentrates on experiments that have been carried out at Frankfurt university. We report on measurements of surface sputtering occuring during oxygen and argon implantation into silicon targets. Surface sputtering accounts for dose limitation of implanted ions. By combining vapor-deposition of neutral atoms with implantation and di€usion layers of stoichiometric silicon and oxygen ratio with a thickness of more than 300 nm have been obtained. Other experiments in which vapor-deposition only is applied will demonstrate the potential of this simple setup. Ó 1998 Elsevier Science B.V. PACS: 52.75-d Keywords: Plasma immersion ion implantation; Di€usion; Sputtering; Vapour-deposition; Dose limitation

1. Introduction Plasma-immersion ion implantation (PIII) [1] has been established during the last years as an additional surface modi®cation technique. In more than 30 laboratories experiments have been carried out in which the improvement of metal surfaces against wear and corrosion and the variation of the electrical conductivity of semiconductors are studied [2]. In contrast to conventional ion implantation, in a PIII-device the target is surrounded by

1

Corresponding author. Tel.: +49 69 7982 3492; fax: +49 69 7982 8510; e-mail: [email protected] 2 This work is supported by BMBF (06OF359).

the plasma and then pulse biased to high negative voltages. There exist di€erent methods for the generation of the plasma and the production of the high voltage pulses. The advantage of PIII is the capability of implanting targets with complicated 3-dimensional geometries, which avoids scanning systems or target manipulators. This is due to the fact that the electrical ®eld between plasma sheath and target is ending perpendicular everywhere with the target surface. However, on planar targets, a distinct angle distribution of the ions, which starts from a spherical formed plasma sheath, has been measured [3]. Furthermore, since a neutral gas pressure of the order of 10ÿ4 ±10ÿ3 hPa is necessary to maintain the plasma, the mean free path length of the

0168-583X/98/$19.00 Ó 1998 Elsevier Science B.V. All rights reserved. PII S 0 1 6 8 - 5 8 3 X ( 9 7 ) 0 0 9 5 2 - X

38

R.W. Thomae / Nucl. Instr. and Meth. in Phys. Res. B 139 (1998) 37±42

accelerated ions is of the order of the sheath thickness and the probability of collisions is enhanced and the ions may have a continuous energy distribution. Both e€ects decrease the penetration depth and increase the surface sputter probability, which results in a lower applicable dose [4]. For that reason most of the PIII-experiments combine implantation with vapor-deposition and di€usion processes, which can be simultaneously handled in the plasma. For example, vapor-deposition can be achieved just by placing an additional sputter electrode close to the target to be implanted and supplying a negative DC-voltage. Di€usion processes are enhanced due to the ion bombardment and by the possibility to adjust the target temperature by the power load, i.e. implantation energy, pulse length and repetition rate.

Due to the target power load it is not possible to run PIII-experiments in continuous wave-mode. Furthermore, during high voltage application to the target the plasma withdraws from it, i.e. the plasma sheath increases. Therefore, the voltage has to be pulsed with a pulse length short enough to ensure plasma stability. There are two convenient ways to generate these pulses, which are shown in Fig. 1. In Fig. 1(a), a capacitor is positively charged via a resistor by means of a DC power supply and simultaneously loaded by negative charges from the plasma. At a certain voltage the positive connector of the capacitor is pulled down to approximately 0 V. The other connector, previously on plasma potential is thus on the corresponding negative potential. By this voltage the

2. Basics of PIII The general setup of a PIII experiment consists of an implantation chamber to which the plasma generator and the high voltage feedthrough are mounted. The volumes of di€erent chambers vary between several liters and some cubic meters. Some of the experiments make use of magnetic con®nement of the plasma by means of permanent magnets or solenoidal coils. For the most part, the plasma is generated by means of RF-power (13.56 MHz) which is coupled to the plasma by means of matched antennas or coils, which are either placed inside the chamber or outside with a suitable glass window. Typical values for the plasma density and neutral gas pressure are 108 ±1011 cmÿ3 and 10ÿ3 ±10ÿ4 hPa, respectively. Some experiments use ®lament discharge for plasma generation. One or more ®laments emit electrons which are accelerated towards the plasma by means of the so-called arc voltage which is of the order of 100 V. This kind of generation allows a high plasma density (108 ±1012 cmÿ3 ) but needs also a high working pressure (10ÿ3 ±10ÿ1 hPa). Furthermore, cathode sputtering may enhance plasma impurities. Dense plasma (1011 cmÿ3 ) at low pressure (10ÿ3 hPa) can be obtained by using a magnetron at 2.45 GHz.

Fig. 1. In (a) a capacitor is positively charged via a resistor by means of a DC power supply and simultaneously loaded by negative charges from the plasma. At a certain voltage the positive connector of the capacitor is pulled down to approximately 0 V. The other connector, previously on plasma potential is thus on the corresponding negative potential. In scheme (b) the DC power supply is connected to the electrode by a high voltage switch, which generates short pulses.

R.W. Thomae / Nucl. Instr. and Meth. in Phys. Res. B 139 (1998) 37±42

plasma ions are accelerated towards the target and discharge the capacitor. The shortening can be achieved by means of a tube. This allows a control of the pulse length and repetition rate. Depending on the setup, the implantation voltage can be as high as 150 kV. Typical values for the pulse length and repetition rate are 100 ls and 100 Hz, respectively, i.e. corresponding to a duty cycle of 1%. In some experiments high voltage transistor switches are used which are connected in series with the DC power supply as shown in Fig. 1(b). These switches are limited in high voltage capability to values below 65 kV. The averaged implantation current varies between 1 and 1000 mA for di€erent experiments and is related to the target surface dimensions. In addition, it should be mentioned that it is not possible to suppress the secondary electron current that is produced during implantation, which reduces the eciency.

3. PIII at the Frankfurt University The plasma is generated by a Leybold 13.56 MHz RF ion source which is operated without extraction system. The coupling system consists of a cylindrical copper coil which surrounds the quartz glass source body and a tuning ring in order to optimize the coupling. The implantation chamber is rectangular with a volume of 0.06 m3 . It contains a target support and a support for an extra sputter target. The additional equipment consists of a turbomolecular pump with a capacity of 2000 l/s, a 13.56 MHz RF transmitter and two high voltage feedthroughs for the high voltage pulsing system and the connection of the sputter target to a DC power supply. The pulsing system consists of a 65 kV, 10 mA power supply and a transistor switch with repetition rates between 1 and 1000 Hz, a ®xed pulse length of 250 ls with a rise time less than 1 ls and a decay time of 50 ls. Source operation parameters for all experiments are: RF power 150 W, working gas pressure 3.5 ´ 10ÿ4 hPa, gas ¯ow 2 l/h, sputter voltage 200±1000 V, repetition rate 5±20 Hz, and implantation voltage 20±30 kV. For a detailed description of the experimental setup see [4,5].

39

4. Experimental results and discussion Experiments in which the implantation of oxygen ions into a silicon target was studied have shown that the maximum dose which could be achieved is of the order of 2 ´ 1017 ions/cm2 for an implantation voltage of 30 kV [5]. Assuming as a rough estimate a sputter coecient SY ˆ 0.3 and a projected range Rp ˆ 50 nm, which has been obtained from TRIM simulation (Transport of Ions in Matter, version 94.04), a dose of 8 ´ 1017 cmÿ2 is calculated following the theory described in [6]. In this theory, it is derived that after a certain time an equilibrium of implantation and sputtering is achieved at the target surface which leads to a limited dose. Therefore we carried out measurements to determine the sputter coecient for di€erent ion±target combinations. For that reason silicon targets are vapor-deposited with a copper layer of some nm thickness by means of a copper sputter target in an argon plasma. Thereafter the copper sputter target is replaced by one made out of silicon and a silicon layer is brought on top of the copper layer. In Fig. 2 the vapor-deposition rate as a function of the sputter voltage is plotted. We found a linear dependence. Due to the larger sputter coecient of argon ions on copper compared to silicon, the deposition rate is two times larger. After vapor-deposition the targets are connected to the high voltage pulsing system and implant-

Fig. 2. Vapor-deposition rate as a function of sputter voltage for silicon and copper sputter targets in an argon plasma. The plasma parameters are: neutral gas pressure 3.5 ´ 10ÿ4 hPa, RF power 150 W.

40

R.W. Thomae / Nucl. Instr. and Meth. in Phys. Res. B 139 (1998) 37±42

Table 1 The sputter rates in 1015 cmÿ2 minÿ1 for argon and oxygen ions on copper or silicon layers for implantation voltages of 20 and 30 kV, respectively

Ar ) Cu Ar ) Si O ) Si

20 kV

30 kV

10.0 2.3

14.0 4.8 0.6

The pulse repetition rate amounts to 20 Hz. Note that these are results for our particular ion energy distribution.

ed in an oxygen or argon plasma for di€erent intervals. Subsequently the thickness of the remaining layer is again measured by the RBS technique (Rutherford Backscattering Spectroscopy). The determination of the sputter coecient of argon ions on copper is of importance for experiments which are described below. In Table 1 the sputter rate in 1015 /(cm2 min) for argon and oxygen ions on copper or silicon layers for implantation voltages of 20 and 30 kV at a pulse repetition frequency of 20 Hz is given. In Table 2 the averaged values of the sputter yield, i.e. the number of sputtered target atoms per incident ion, are calculated from Table 1. Note that for this calculation the number of ions on the target have been determined from the averaged current from the high voltage power supply assuming a secondary electron emission coecient of 1. This coecient is uncertain because the ions hit the target and target support which consists of di€erent materials. Nevertheless, for a sputter yield for oxygen ions of 0.6 we calculate a dose limit of 4 ´ 1017

cmÿ2 . The discrepancy by a factor of 2 compared to the experimental results is probably related to the uncertainness in the secondary electron emission coecient. Furthermore, it should be mentioned that these results are averaged values for our particular ion energy distribution. Next we carried out experiments in which deposition and implantation were combined. The basic idea is to compensate the target etching due to sputtering by depositing neutral atoms at the surface during implantation. With the knowledge of the deposition and sputter rates it is possible to ®nd an equilibrium that serves for a surface layer of constant thickness which is thin enough to be penetrated by the energetic ions. We started with the implantation of a mixture of argon with 10% oxygen. In Fig. 3 the RBS-spectrum of a silicon target is shown which has been implanted for 825 min (20 kV, 20 Hz) in such a plasma with simultaneous deposition of copper ()250 V on sputter target). The simulation of the spectrum (Fig. 4) by means of the RUMP-code [7] shows a very thin copper layer of 20 nm followed by a 350 nm thick layer of stoichiometric ratio of oxygen and silicon atoms. This thickness cannot be achieved by implantation only. The oxygen di€uses due to the high dose implantation from the surface into the

Table 2 The averaged values of the sputter yield, i.e. the number of sputtered target atoms per incident ion, calculated for Table 1

Ar ) Cu Ar ) Si O ) Si

20 kV

30 kV

9.4 2.4

9.0 3.0 0.4

For this calculation the number of ions on the target have been determined from the averaged current from the high voltage power supply assuming a secondary electron emission coecient of 1.

Fig. 3. RBS-spectrum and RUMP simulation of a silicon target implanted with 20 kV at a repetition rate of 20 Hz in an argon plasma with small fraction of oxygen ions for 825 min. Simultaneously, neutral copper atoms generated by an extra sputter target (250 V) are deposited on the target surface.

R.W. Thomae / Nucl. Instr. and Meth. in Phys. Res. B 139 (1998) 37±42

41

bulk material. The growth rate for the silicon-oxygen layer amounts to 0.4 nm/min for this example. Other experiments under similar conditions resulted in growth rates between 0.2 and 0.8 nm/min. The peak at channel 710 in Fig. 4 has been identi®ed as copper by means of RBS-measurements with di€erent incident angles. Further experiments have to be carried out to understand why the sputter rate has to be distinctly larger than the vapordeposition rate to create layers of silicon and oxygen. Experiments, in which both rates were in the

same order of magnitude, resulted in no layer at all. Next we discuss experiments with a pure oxygen plasma without implantation. In Fig. 5 the RBS-spectrum of a silicon wafer which has been vapor-deposited for 3930 min by a copper sputter target is shown. The sputter voltage amounts to )400 V for this example. The simulation results in a layer of 3400 ´ 1015 atoms/cm2 with a composition of 57.5% oxygen, 40.5% copper, and 2% of silicon. Due to the anity of the oxygen it is not possible to create a pure copper layer on the silicon substrate, but it shows that in principle vapordeposition of compounds is possible. The silicon part is believed to be sputtered from the chamber wall where it has been deposited in preceding experiments. A compound layer is also demonstrated in Fig. 6, in which the RBS-spectrum of a probe is shown, which has been processed in an oxygen plasma. For this experiment on one side of the copper sputter target, a silicon wafer was mounted. This serves during sputtering for both, copper and silicon atoms. To allow a de®nite interpretation again a silicon wafer with a copper layer of 200 nm on top has been vapor-deposited for 1725 min with a sputter voltage of )800 V. For this example a layer of 22% silicon, 15% copper,

Fig. 5. RBS-spectrum and RUMP-simulation of a silicon target which has been vapor-deposited in an oxygen plasma with a copper sputter target for 3930 min. The sputter voltage amounts to )400 V. The plasma parameters are: neutral gas pressure 3.5 ´ 10ÿ4 hPa, RF power 150 W.

Fig. 6. RBS-spectrum and RUMP-simulation of a silicon target which has been vapor-deposited in an oxygen plasma with a sputter target consisting of copper and silicon for 1725 min. The sputter voltage amounts to )800 V. The plasma parameters are: neutral gas pressure 3.5 ´ 10ÿ4 hPa, RF power 150 W.

Fig. 4. Depth distribution of Fig. 3. A stoichiometric layer of Si and O of 350 nm thickness was obtained, covered by a thin layer of copper at the surface.

42

R.W. Thomae / Nucl. Instr. and Meth. in Phys. Res. B 139 (1998) 37±42

and 63% oxygen had been built up. The thickness of the compound layer amounts to 1900 ´ 1015 atoms/cm2 . Assuming compounds of CuO and SiO2 the ratio of the three elements is close to a stoichiometric ratio. The microstructure of the layers has to be identi®ed by means of additional diagnostics.

The RBS measurements have been carried out at the Institut f ur Kernphysik der JWG-Universitat. We would like to thank especially Dr. H. Baumann and F. Link for continuous support with this equipment.

5. Conclusion

References

The ®rst PIII-experiments have been set up 10 years ago. Nowadays in more than 30 laboratories of universities, research institutes, and industry PIII is applied to the surface modi®cation of metals and semiconductors. The plasma generators and high voltage pulsing systems are working reliably and research work is concentrated on target physics. Problems still under investigation are the homogeneity of the implants, impurities, ion energy distribution and target sputtering. Nevertheless, the possibility of combining implantation with other techniques, e.g. vapor-deposition, opens a wide ®eld of applications with a quite simple device.

[1] J. Conrad, J. Radtke, R. Dodd, F. Worzala, N. Tran, J. Appl. Phys. 62 (1987) 4591. [2] Proc. PBII'96, Third Int. Workshop on Plasma Based Ion Implantation, 15±18 September 1996, Research Center Rossendorf, to be published in Surface and Coatings Technology. [3] Ch. Gabor, M. Galonska, R. Thomae, to be published. [4] H. Bender, J. Brutscher, W. Ensinger, R. G unzel, J. Halder, H. Klein, B. Rauschenbach, J. Sch afer, B. Seiler, R. Thomae, Nucl. Instr. and Meth. B 113 (1996) 266. [5] R. Thomae, B. Seiler, H. Bender, J. Brutscher, A. Jakob, H. Klein, A. Maaser, J. M uller, M. Sarstedt, M. Weber, in: S. Co€a, G. Ferla, F. Priolo, E. Rimini (Eds.), Proceedings of the IIT 94, Elsevier, Amsterdam, 1995, p. 989. [6] H. Ryssel, I. Ruge, Ionenimplantation, B.G. Teubner, Stuttgart, 1978. [7] L.R. Doolittle, Nucl. Instr. and Meth. B 9 (1985) 344.

Acknowledgements