High power pulse generators for plasma based ion implantation

Plasma based ion implantation (PBII) is a large scale cost effective technique for modifying the surface properties of materials via omnidirectional ion implantation. Substrates are subjected to a train of negative high voltage pulses that extract ions from the plasma and accelerate them directly onto substrate surfaces. Principle and design of the plasma source and high voltage pulsing system are described. Distributed electron cyclotron resonance (DECR) plasma sources, which produce a peripheral ionization facing the substrate and can also be easily scaled up, appear particularly well adapted to the PBII process. The required high voltage pulsing systems are not standard industrial products and must be specially designed and built for this application. A new type of compact, low standby voltage pulse transformer has been developed. A high performance magnetic core is used as interstage step-up transformer and voltage modulation is provided by insulated gate bipolar transistor (IGBT) switches, which can transfer pulsed power with turn-on and turn-off times shorter than 1ms. The voltage transformer consists in parallel association of 96 IGBTs and primary coils, and the 96 turns secondary winding provides the 100 kV - 100 A output performances. The modulator has interlocks for over current, over voltage, and is output protected against short circuits or polarity reversal which may induce cathodic sputtering of the surrounding walls and thus contaminate the substrate. The performances of the pulse generator are presented on resistor and on substrates immersed in the DECR plasma source. Finally the possible extrapolation to higher power is discussed.

Ionic implantation consists in accelerating ions at high energy (to some hundreds of keV) to bombard the surface of a piece in order to carry out a specific superficial coat.The first applications in micro-electronics [1] have been made in local semiconductor doping (p or n) with profiles which remain unreachable by a standard diffusion technique, and then in the realization of superficial or buried SO2 insulating coats, resistance’s, etc... The ionic implantation devices were mainly composed of an ion source generating the product to implant, an electrostatic accelerator to give the desired energy to the ions, a magnetic sector to sort out them and a deflector to achieve a uniform exposition of the substrate by scanning. Other applications have been identified for a long time. They are mostly related to metallic surface processing [2] to reach particular superficial characteristics : stainless properties, hardness, coloration, adhesive capabilities or friction coefficient...

In some cases, only imposing industrial equipments could provide the processing needs despite the high costs. The recent idea to directly perform the implantation inside a plasma by PBII, initially developed by J.R. Conrad [3,4], has stimulated the development of processing of metallic or even plastic surfaces by ion implantation. However, the limitations and disadvantages of PBII lay in the lack of mass discrimination of the implanted ions, in their energy distribution which leads to a distribution of the deposition depth, and in the production of secondary electrons which generates X-rays on the sheath and imposes the calibration of the currents to monitor the correct implanted dose. Nevertheless, this method affords not only a reduction of investment and operation costs but the ability of processing complex surfaces in only one operation as well. The PBII technique then combines productivity and profitability and is very attractive for mass production in metallurgy and semi-conductor industry

2) Plasma Based Ion Implantation processing

   2.1 PBII basic mechanisms

The PBII technique consists in the creation of a plasma of the ionic species inside a grounded metallic housing. The immersed pieces are subjected to recurrent negative voltage pulses that extract ions from the plasma and accelerate them directly onto the substrate. As the sheath conformably surrounds the assembly, all surfaces are implanted at the same time. The evolution of this plasma sheath has been already extensively studied [5-7]. Three different phases can be distinguished : (1) on the time scale of the inverse plasma frequency w_pe^-1, electrons are repelled,

while ions are left back leading to an electron free ion « matrix » sheath around the substrate, (2) on the time scale of the inverse plasma frequency w_pi^-1, these ions are accelerated towards the substrate. The energy distribution depends on their initial position and the dynamics of the pulse edge, and the ion current reaches a sharp maximum before decreasing. (3) On a larger time scale, the sheath and current density evolve to the steady state regime given by the Child-Langmuir law. [8,9]

2.2 Plasma specifications

A first specification concerns the size of the plasma which must be much larger than the sheath thickness in order to avoid total depletion of the plasma between the substrate and the surrounding walls. At plasma density n=1010 cm^-3 and electron temperature of kTe = 1eV, this thickness exceeds 0.3m for an acceleration voltage of 100kV. Secondly, for a given application, the energy of implanted ions is often predetermined. Implantation of monoenergetic ions then requires additional conditions : (1) the pulse duration must be large compared to w_pi^-1 so that the fraction of ions implanted from the initial matrix is small with respect to the one of ions extracted from the plasma at the sheath edge,

(2) on the other hand, the ion transit in this sheath must be collisionless, i.e. the ionic mean free path must be longer than the sheath thickness. Under these conditions, the bombardment can be considered as quasi-monoenergetic. Finally, too much dense plasmas, which lead to short sheath thickness and high ion current densities, are not attractive due to enhanced risks of arcing in the sheath (higher electric field intensity) and increased substrate heating.

2.2 Pulse generator specifications

The specifications for the pulse generators in PBII processing directly follow from the previously described sheath dynamics : the generator has to sustain the initial peak current, the rise and fall times of the voltage must be in the order of w_pi^-1 (typ. 1ms in low density plasmas), the pulse width being quite longer (typ. 10 to 20ms) and the voltage amplitude has to be as stable as possible to avoid any energy variation during the steady-state phase. Other important requirements are firstly to avoid any reverse positive voltage, which would lead to ion bombardment of the surrounding walls and thus contaminate the substrate by a reactor wall sputtering, and secondly to insure a perfect protection of the generator against short-circuits in the load.

The expected characteristics of the pulse generator have then been defined as follows :

Maximum output voltage	about 100 kV
Maximum output current	100 A
Rise and fall times at max. load	<1ms
Pulse duration	from 3ms to about 20ms
Maximum pulse frequency	50 Hz
Maximum average power	5 kW

3) Plasma Reactor

The uniform distributed plasma results in the latest outgrowth of DECR [10,11] and has proved to be a flexible concept. The reactor (see fig. 4) is a 60 cm diameter, 70 cm height cylinder in which the inside wall is covered with an array of 24 tubular magnets, 2.45 GHz microwave power feeds and wave propagators. At 1 mTorr pressure and 1.3kW input power, The N₂ plasma has a density of 2x10¹⁰ ions/cm³, and electron temperature is 1.2 eV. The N⁺/N₂⁺ ratio of 7/3 is favorable for PBII applications. The figure 1 shows a top view of the plasma chamber, in which we can distinguish the peripheral tubular configuration, and the aluminium sphere (120mm diameter) in an N₂ plasma at a pressure of few mTorr. Its typical pink coloration is uniform in the whole reactor's volume and its composition has been analyzed by quadrupole mass spectrometry.

4) The 100kV - 100 Amps pulse generator

The Generator is based on a low leakage high voltage pulse transformer. The voltage modulation is provided by 96 1200V high power IGBTs which supply 96 on turn primary coils in parallel, coupled to a 96 turns secondary coil with an high performance amorphous magnetic core (figure 2). Between the pulses, a supply resets the magnetic circuit to allow the maximum flux variation in the smallest volume.

Fig. 2 : Inside of a Pulse Generator

Fig. 3 : Pulse generator schematic

On figure 3, one can see the copper windings surrounding the magnetic core, the power transistors, the high frequency capacitors on the bottom PCB and the electrolytic energy tank at the very bottom of the picture. On the top PCB, one can see the ultra fast diodes which protect against voltage reversal, the output and the voltage divider are located on top of this PCB which is not visible on this picture.
This configuration insures an efficient energy transfer from the internal mid voltage (1000V) energy tank to the high voltage output in the smallest volume. The whole assembly is immersed in a dielectric oil. The output connection presents a coaxial symmetry to insure a perfect matching with the load and the lowest electromagnetic radiation, as well as a high safety.
The generator has a cylindrical shape (50cm diameter, 55cm height) see figure 4. The Generator is controlled by a micro-processor, and allow the user to select with an external interface panel (or a PC via RS 232) all the working conditions (voltage, pulse width, frequency rate,...)

5) Results on resistive and plasma loads

The performances of the generator have first been tested on a water load (fig. 5a&b) in normal conditions (bold lines) and with a deliberate short-circuit (thin lines) resulting from an arcing condition (fig. 7) at 75 kV. The generator sustained as expected these hard conditions ans showed perfect dynamic behaviors.

Connected to the plasma reactor, the results are in accordance with the theoretical behavior (fig 6a&b) and proved the performance of the assembly in terms of efficiency and ability to process complex structures on one run.

Fig. 6a-6b : Voltage and current pulse characteristics on 300cm2 stainless steel

6) Conclusion

In terms of efficiency, the PBII technique offers future prospects : the mean current in a classical implantation system reaches some fractions of mA, and in this new application it is about 2 decades higher. This capability associated to the simple configuration meets the industrial demands in the micro-electronic field, in the surface treatment industry or others. On one other hand, the pulse generator relates compactness and safety, and can easily be monitored. The simplicity of the method equally requires low maintenance in comparison to classical systems based on high voltage supplies and vacuum tube switches.

Finally it exhibits an excellent reliability. The architecture of the DECR plasma source can be scaled up to several cubic meters [12,13] and the performances of the generator can easily be extrapolated to 200A and/or 150kV and the repetition rate be stepped up to some kHz ; the mean power transferred to the plasma could then exceed some kW as long as the substrate holder and reactor cooling system can remove the total heat.

6) References

[1] P.K. Chu, S. Qin, C. Chan, N.W. Cheung and L.A. Larson
Mater. Sci. Eng. R17,(1996) 207

[2] R. Hutchings
Mater. Sci. Eng. A184,(1994) 87

[3] J.R. Conrad
« Method and apparatus for plasma source ion implantation », U.S ; Patent No 4.764.394, (August 1988)

[4] J.R. Conrad, J.L. Radtke, R.A. Dodd, F.J. Worzala and N.C. Tran
J. Appl. Phys. 62,(1987) 4591

[5] J.R. Conrad
J. Appl. Phys. 62,(1987) 777

[6] I.J. Donnelly and P.A. Watterson
J. Phys. D22,(1989) 90

[7] M.A. Lieberman
J. Appl. Phys. 66 (7),(1989) 2926

[8] G.A. Collins and J. Tendys
J. Vac. Sci. Technol. B(12), Mar/Apr (1994) 875

[9] M. Shamin, J.T. Scheuer and J.R. Conrad
J. Appl. Phys. 69, (1991) 2904

[10] T. Lagarde, Y. Arnal and J. Pelletier
Plasma Sources Sci. Techn. 6, (1997) 53

[11] F. Le Coeur, Y. Arnal, R. R. Burke, O. Lesaint and J. Pelletier
Surface & coating Techn. 93, (1997) 265

[12] J. Pelletier, Y Arnal, R. Debrie, L. Pomathiod and J.C. Rifflet
Rev. Sci. Intrum. 55, (1984) 1636

[13] L. Pomathiod, R. Debrie, Y Arnal and J. Pelletier
Phys. Lett. A 106, (1984) 301