Influence of thickness and dielectric properties on implantation efficacy in plasma immersion ion implantation of insulators

Transcription

1 JOURNAL OF APPLIED PHYSICS VOLUME 95, NUMBER 7 1 APRIL 2004 Influence of thickness and dielectric properties on implantation efficacy in plasma immersion ion implantation of insulators Ricky K. Y. Fu and Paul K. Chu a) Department of Physics and Materials Science, City University of Hong Kong, Tat Chee Avenue, Kowloon, Hong Kong Xiubo Tian State Key Laboratory of Welding Production Technology and School of Materials Science and Engineering, Harbin Institute of Technology, Harbin, China Received 25 August 2003; accepted 14 January 2004 Plasma immersion ion implantation of insulators is an interesting topic both theoretically and industrially. The net energy of the incident ions is dictated by the surface potential and for conductors is equal to the voltage applied to the backside or sample stage. However, the poor electrical conductivity of insulating materials can lead not only to charging during ion bombardment but also reduced surface potential due to the capacitance effect. In the work described in this paper, we theoretically and experimentally investigate the influence of the thickness and dielectric properties of insulating materials on the implantation efficacy. The use of mesh-assisted PIII by covering the insulating materials with an electrically conducting cage to enhance the implantation efficacy is also compared experimentally. Our theoretical results suggest that a low plasma density induces less surface charges and higher surface potential. Our experimental data show good agreement with the theoretical results and mesh-assisted PIII does yield net improvement American Institute of Physics. DOI: / I. INTRODUCTION Polymeric, ceramic, and dielectric materials are widely used in the automotive, electronic, biomedical, and aerospace industry. Surface treatment of insulating materials using radio-frequency RF plasma irradiation of several tens of ev or without any sample bias has been widely studied. 1 3 However, under these conditions, a thick modified layer is hard to achieve and the process efficiency is low. Posttreatment annealing may be needed to enhance dopant diffusion but it may be undesirable in some applications as the bulk properties may be degraded. In addition, the modified surface properties such as wettability may not be retained if the treatment time is too long. Ion implantation is an effective way to obtain a thick modified layer without affecting the bulk properties. The energetic ions also give rise to ion mixing effects which may yield a more stable structure. Plasma immersion ion implantation PIII possesses advantages such as the capability of treating samples with an irregular shape, high throughput and smaller instrument footprint compared to beam-line ion implanters and has been widely applied to the metal and semiconductor industry Even though PIII of insulators is of practical interest, it is much more difficult, but by using the appropriate conditions and experimental setups, PIII of insulating materials can be conducted. There have been several theoretical studies on PIII of thin oxide films on semiconductor devices. 11,12 The results suggest that the applied voltage, plasma density, and implantation modes a Author to whom correspondence should be addressed. Tel: ; Fax: or ; electronic mail: paul.chu@cityu.edu.hk have a critical influence on the results. Experimental work carried out by Tian et al. suggests that the short-pulse mode and low plasma density are beneficial for implantation while reducing the problem associated with electrical arcing. 13 PIII has been shown to improve the surface properties of several insulating materials such as polystyrene, 14 amorphous polyolefin plastic, 15 and SiC ceramics. 16 However, the physical phenomena such as surface potential reduction, surface charge accumulation, implant depth profile, and the impact of the dielectric properties and sample thickness on the implantation efficiency have not been investigated extensively, and experimental data are particularly lacking. In this work, the effects of the thickness and dielectric properties of the insulating sample on the nitrogen PIII efficacy are studied using the dynamic sheath model. 10,17,18 The nitrogen depth profiles acquired by secondary ion mass spectrometry are used to corroborate the theoretical approach under different conditions. We also experimentally investigate mesh-assisted PIII in which the insulator is enshrouded by an electrically conducting cage biased to the implantation voltage and placed only a smaller distance from the sample surface. 19,20 Our results reveal an improvement of about 30% by using mesh-assisted PIII. II. EXPERIMENT A series of square samples (10 mm 10 mm) with different thicknesses and dielectric constants were treated in our PIII facility as shown in Fig. 1. The materials were soda-lime glass and fused quartz with an atomic density of and atoms cm 3, and dielectric constant of 4 and 7.75, respectively. The nitrogen working pressure was /2004/95(7)/3319/5/$ American Institute of Physics

2 3320 J. Appl. Phys., Vol. 95, No. 7, 1 April 2004 Fu, Chu, and Tian FIG. 1. Schematic diagram of the experimental setup of PIII of insulators. maintained at Torr, and the plasma was generated by a radio frequency inductive coupled plasma source RF- ICP using an input power of 1000 W. The plasma density in the vacuum chamber was monitored using a Langmuir probe and in our previous studies, 21,22 the plasma density was measured to vary less than 10% within a 600 mm diam in the vacuum chamber and the electron temperature was within 3 4 ev. During the experiments, a tetrode-based pulse modulator was used to produce short square high-voltage pulses. A Pearson model 305A voltage divider, Pearson model 3525 current monitor, and Tektronix 460A digital oscilloscope were employed to measure the applied voltage and the output current. The potential applied to the backside of the sample was 40 kv with a pulse width of 20 s and repetition rate of 200 Hz. Each sample was loaded separately into the vacuum chamber for nitrogen implantation and mesh-assisted PIII using a conducting steel grid was performed on some of the samples. III. DYNAMIC SHEATH MODEL The dynamic sheath model as applied to PIII can be derived from the Child Langmuir law. 10,11,12 To simplify the calculation, several assumptions can be made. The sheath is assumed collisionless and the electrons in the plasma have a Boltzmann distribution. The total ion current density during the high-voltage pulse is given by 2/3 1 2 s 2 0 4, 1 j i t e M V t s t where 0 is the free-space permittivity, e is the electronic charge, M is the ion mass, V (t) is the surface potential of the insulator, s 0 is the ion-matrix sheath thickness which depends on the applied potential V 0 and the ion density n i but is independent of the ion species, and s (t) is the timedependent ion sheath thickness. Integrating the total ion current density over the pulse duration yields the incident dose per pulse. The impinging ions eject secondary electrons that are repelled away from the sample to the plasma by the electric field. This leads to an increase in the measured deposited positive charges as expressed as s t FIG. 2. Surface potential of insulators versus time for different dielectric constants: pulse risetime or fall-time 2 s, V o 40 kv, n i cm 3, and d 1 mm. Q t 0 t j i t 1 dt, where the secondary electron coefficient is related to the insulator surface potential, applied potential, and materials. For simplicity, a value of 12 is used in our modeling for V 0 40 kv. 23 As the deposited charges cannot be neutralized when the voltage pulse is on, they accumulate on the insulator surface to build up a retarding electric field. Hence, the insulator surface potential cannot attain the full potential due to this surface charging as well as the capacitance effect. The final outcome of the surface potential is then related to the dielectric properties and thickness of the insulating materials such that per unit area as given by 2 V t V 0 edq t 0 r, 3 where d is the thickness of the insulator. Hence, V (t) can become more negative if the relative permittivity r is higher and the sample thickness d is thinner. IV. RESULTS AND DISCUSSION Figure 2 displays the simulated time-dependent surface potential of the insulator with a thickness of 1 mm for an ion density of cm 3 and 40 kv biased voltage during nitrogen PIII. The surface potential reaches a maximum after the pulse risetime but cannot attain the full potential of 40 kv in all cases involving various dielectric permittivities. It gradually diminishes during the pulse duration due to the charge accumulation. Our results show that the higher the value of the relative permittivity of the insulating materials, the higher the surface potential during the pulse duration. Figure 3 compares the incident ion dose between a conductor and insulator collected during a voltage pulse. The majority of the implant dose is collected during the first few microseconds and there is only a minor increase during the rest of the pulse in the case of an insulator. The dose received by an insulator is only about half of that of a conductor and

3 J. Appl. Phys., Vol. 95, No. 7, 1 April 2004 Fu, Chu, and Tian 3321 FIG. 3. Comparison of the pulse per dose for a conductor and insulator with different dielectric constants and fixed thickness. FIG. 5. Nitrogen SIMS depth profile acquired from soda lime glass with different thicknesses implanted at V o 40 kv. a longer treatment time or a higher pulsing frequency is required for an insulator in order to achieve the same dose as a conductor. However, the sample temperature is related to the pulsing frequency 24 and so care must be exercised in order to avoid damage to materials such as polymers that do not possess good thermal conductivity. Figure 4 shows the theoretical prediction on the maximum surface potential of the insulator for different ion densities when a 40 kv pulse is applied to the sample platen, which holds the insulator sample of with various dielectric constants and a thickness of 1 mm. The maximum surface potential drops rapidly when the ion density is increased slightly. The lower the dielectric constant of the insulator, the more substantial is the surface potential drop. The results are expected from Eq. 3. Hence, a lower ion density is recommended for treating insulating materials and our observation is consistent with that of Ref. 13. Experimentally, the plasma sheath expands dynamically and continuously when a voltage pulse is applied to the substrate, whereas the sheath thickness and expansion velocity are dependent on the applied potential and nearby plasma density. According to Eq. 1, the time-dependent current density is reduced as the plasma sheath continuously expands. Thus, the collected current not shown here on the exposed area is reduced accordingly. However, the surface potential of the insulator diminishes even when the ion flux on the sample surface decreases during the pulse. Hence, a thorough knowledge of all the plasma parameters is necessary in order to conduct a highly accurate simulation on the surface potential and implantation depth. In this work, since we are addressing the general phenomenon, such exact knowledge is not needed. A different implantation energy or modified layer thickness is known to change the properties of the modified materials. 16 At low implantation energy, nuclear stopping dominates and more damage can be created in the near surface. In the more serious case, amorphization occurs and it may not bode well for the treatment of crystalline materials and annealing may be necessary to recover the crystal damage. Figures 5 and 6 display the nitrogen depth profiles acquired by secondary ion mass spectrometry SIMS from samples with different thicknesses and dielectric constants. FIG. 4. Maximum surface potential of insulating samples versus ion density for different dielectric constant for V o 40 kv and d 1 mm. FIG. 6. Nitrogen depth profile acquired from soda lime glass and quartz with the dielectric constant of 7.75 and 4.0 and thicknesses of 1.2 mm and 1.0 mm, respectively.

4 3322 J. Appl. Phys., Vol. 95, No. 7, 1 April 2004 Fu, Chu, and Tian FIG. 7. Nitrogen depth profile acquired from soda lime glass implanted using the mesh-assisted approach at V o 40 kv. As usual, the nitrogen distribution is broader than that typically observed in a beamline implanted sample. As shown in Fig. 5, a thicker modified layer is observed in the thinner sample and this is due to a smaller surface potential reduction. In addition, as depicted in Fig. 6, the sample with higher dielectric constant shows a deeper nitrogen profile. As the thickness and dielectric constant can result in the variation of the surface potential on insulator to several kvs, the implant projected range can be shifted by several nanometers. The lower implantation dose obtained from the thicker and higher dielectric constant samples is to be expected as the lower surface potential can result in smaller collected ion current using the same plasma parameters. So far, we have demonstrated the implantation efficacy for different substrate thickness and dielectric properties. The implantation efficiency can be enhanced by coating a thin conducting layer on the insulator surface 25 or enshrouding the insulator with an electrically conducting grid. Figure 7 exhibits the nitrogen depth profile acquired from an insulating sample covered by a conducting grid. The grid is in contact with the sample holder and placed very close to the insulating sample surface, so that ions accelerated through the grid holes impact the surface without suffering much energy loss while traveling from the grid to the sample surface. In our experiments, the glass sample with a thickness of 0.15 mm could be implanted at 40 kv without experiencing any electrical arcing. The implantation process is stabilized by the grid that also repels secondary electrons back to the sample surface to improve the implantation efficiency. Using TRIM simulation, the curves can be fitted by the superposition of several profiles with different implantation energies and the results are then summarized in Fig. 8. Theoretically, the maximum surface potential drops with increasing insulator thickness due to the decrease of the capacitance of the insulator. Two plasma densities: cm 3 and cm 3 are used in our simulation and our experimental results align more closely with those simulated using an ion density of cm 3. However, it should be noted that both our experiments and theoretical simulation have a certain degree of uncertainty. For instance, the measured ion FIG. 8. Predicted maximum surface potential on insulator versus sample thickness for different dielectric constants and comparison with experimental results. density can be off by several tens of percent 26,27 and the secondary electron yield is only an assumed value and in fact changes with the sample voltage for example, continuously during the voltage rise time. Nonetheless, the dynamic sheath model can yield some insight on the process and is quite powerful when combined with experimental verification. Moreover, our mesh-assisted PIII experiments demonstrate the effectiveness of the approach and the improvement in the ion energy is about 30%. In spite of the encouraging results, it should be noted that mesh-assisted PIII is not a mature technique and a number of issues including shadowing effects, plasma dynamics between the grid and sample surface, as well as other phenomena must be investigated further. To alleviate the charging effects on insulator surface, a bipolar pulse can in principle be used so that electrons can be attracted to the surface during the positive cycle of the pulse. However, the hardware implementation, that is, the power modulator, is not trivial. Nonetheless, plasma discharge and the dynamic sheath expansion can take into account bipolar pulse and the subsequent anode effects. V. CONCLUSION The dynamic sheath model is employed to predict the surface potential on insulating samples with different sample thicknesses and dielectric constants. The results disclose that a higher surface potential can be achieved if the insulator is thin or has a high dielectric constant. A low plasma density is beneficial for treating insulating materials in PIII since less surface charging takes place and a higher implantation energy can be obtained. Our experimental results are in good agreement with the modeled results. Our experiments using mesh-assisted PIII shows improvement in the implantation energy.

5 J. Appl. Phys., Vol. 95, No. 7, 1 April 2004 Fu, Chu, and Tian 3323 ACKNOWLEDGMENTS The work was jointly supported by Hong Kong Research Grants Council CERG No. CityU 1137/03E CityU designation and National Natural Science Foundation of China No C. Riccardi, R. Barni, E. Selli, G. Mazzone, M. R. Massafra, B. Marcandalli, and G. Poletti, Appl. Surf. Sci. 211, D. Ferrante, S. Iannace, and T. Monetta, J. Mater. Sci. 34, C. J. Jahagirdar and Y. Srivastava, J. Appl. Polym. Sci. 82, P. K. Chu, R. K. Y. Fu, X. C. Zeng, and D. T. K. Kwok, J. Appl. Phys. 90, X. B. Tian, R. K. Y. Fu, L. W. Wang, and P. K. Chu, Mater. Sci. Eng., A 316, M. Ueda, C. M. Leipienski, E. C. Rangel, N. C. Cruz, and F. G. Dias, Surf. Coat. Technol. 156, R. K. Y. Fu, P. Peng, X. C. Zeng, D. T. K. Kwok, and P. K. Chu, IEEE Trans. Plasma Sci. 31, W. Moeller, S. Parascandola, T. Telbizova, R. Gunzel, and E. Richter, Surf. Coat. Technol. 136, X. B. Tian, R. K. Y. Fu, and P. K. Chu, J. Vac. Sci. Technol. A 20, P. K. Chu, S. Qin, C. Chan, N. W. Cheung, and L. A. Larson, Mater. Sci. Eng., R. R17, S. Qin, J. D. Bernstein, Z. F. Zhao, W. Liu, and C. Chan, J. Vac. Sci. Technol. B 13, W. En, B. P. Linder, and N. W. Cheung, J. Vac. Sci. Technol. B 14, X. B. Tian, R. K. Y. Fu, J. Y. Chen, P. K. Chu, and I. G. Brown, Nucl. Instrum. Methods Phys. Res. B 187, A. Lacoste, F. Le Coeur, Y. Arnal, J. Pelletier, and C. Grattepain, Surf. Coat. Technol. 135, M. Tonosaki, H. Okita, Y. Takei, A. Chayahara, Y. Horino, and N. Tsubouchi, Surf. Coat. Technol. 136, R. K. Y. Fu, K. L. Fu, X. B. Tian, and P. K. Chu, J. Vac. Sci. Technol. A 22, M. A. Lieberman and A. J. Lichtenberg, Principles of Plasma Discharges and Materials Processing Wiley, New York, 1994, p J. T. Scheuer, M. Shamim, and J. R. Conrad, J. Appl. Phys. 67, J. N. Matossian, R. W. Schumacher, and D. M. Pepper, U.S. Patent No , Hughes Aircraft Company, Los Angeles, CA, R. K. Y. Fu, X. B. Tian, and P. K. Chu, Rev. Sci. Instrum. 74, D. L. Tang, R. K. Y. Fu, X. B. Tian, and P. K. Chu, Rev. Sci. Instrum. 74, D. L. Tang and P. K. Chu, J. Appl. Phys. 93, M. M. Shamim, J. T. Scheuer, R. P. Fetherston, and J. R. Conrad, J. Appl. Phys. 70, X. B. Tian and P. K. Chu, J. Phys. D 34, T. W. H. Oates, D. R. McKenzie, and M. M. M. Bilek, Surf. Coat. Technol. 156, G. J. H. Brussaard, M. van der Steen, M. Carrere, M. C. M. van de Sanden, and D. C. Schram, Surf. Coat. Technol. 98, N. Hershkowitz, M. H. Cho, and J. Pruski, Plasma Sources Sci. Technol. 1,