A silicon MBE-compatible low-energy ion implanter

Nuclear Instruments and Methods in Physics Research B55 (1991) 314-317 North-Holland

314

A silicon MBE-compatible

low-energy

ion implanter

J.S. Gordon VS W Scientific Instruments,

A. Bousetta, Department

Ml6

OJT, England

J.A. van den Berg and D.G. Armour

of Electronic and Electrical Engineering,

R. Kubiak Department

Manchester

and E.H.C.

University of Salford

Salford M5 4 WT, England

Parker

of Physics, University of Wanuick,

Coventry CV4 4AL, England

The design of an ion implanter for the doping of shallow layers grown by silicon molecular beam epitaxy to a concentration 10” cm 3 or higher is described. This requires a capability to deliver beams into the UHV growth chamber with ion energies in the range 500 eV down to 50 eV and lower, fluxes in the range 1 to 50 PA cm-* and uniformity over a 100 mm diameter wafer. The instrument has been designed to operate simultaneously with the MBE growth process, without having to compromise the latter. The practical problems which arise from attaching to an existing silicon MBE installation in which the position, electrical potential and environment of the target are predetermined have also influenced the final design.

1. Introduction

The provision of consistent and controllable in-situ doping of molecular beam epitaxy (MBE) grown semi-

MBE

conductor layers remains a major challenge. This is particularly so in silicon MBE, where there are physical and technical problems with the usual dopant species. Doping methods compatible with the proven ability of

chamber Ion

source

II

,

I Gate PSU

Fig. 1. Schematic diagram of the low-energy ion implanter. P indicates a high-vacuum pump, V a vacuum valve and C an ion current monitor. The dump and gate power-supply units allow zero-contamination mass switching. and the scan power supply rasters the beam over the target. 61’&-58~X/91/$03.50

6 I991 - Elsevier Science Publishers B.V. (North-Holland)

J.S. Gordon et al. / A silicon’ MBE-compatible

MBE to grow high-quality layers with thicknesses in the nanometer range and atomically sharp interfaces are required. The most flexible doping method is to incorporate a mass-analyzed very-low-energy ion implanter with the MBE system [l-4]. This allows essentially any dopant species to be introduced without growth chamber contamination problems, and with accurate dosimetry. In this note we describe a very-low-energy ion implanter now being designed and built to attach to an existing silicon MBE system (VG Semicon V90) as part of the UK SERC Low Dimensional Structures and Devices Programme. The equipment is being used to grow and dope silicon and silicon-germanium low-dimensional structures. A high-temperature effusion cell for boron has been fitted to the MBE system which is giving good p-type co-evaporation doping performance, particularly at substrate temperatures below 500°C. Compatible ntype doping is required at the growth temperatures below 600°C which are needed to realise the full potential of Si and SiGe low-dimensional structures. The prime task of the low-energy ion implanter is therefore to enable doping with phosphorus, arsenic or antimony, with a similar level of control to that being achieved with co-evaporation boron doping. In the longer term, it is planned to investigate the use of ion-assisted growth to enable epitaxial-growth temperatures to be reduced towards room temperature. The features of the low-energy implanter are illustrated schematically in fig. 1. Ions are extracted from a universal source, accelerated to a transport potential of 10 kV, mass analysed and directed into the MBE growth chamber, where they are decelerated to the required energy prior to striking the target. Electrostatic deflectors are provided for beam gating and rastering, and there are current monitors at five locations along

Table 1 Design specifications

of the silicon MBE-compatible

Source Acceleration voltage Mass-resolving magnet Neutral-elimination magnet Magnet switching, 0 to B,,,,, Gas load into MBE chamber Ion energy Neutral content of beam Beam size on target Beam-current density Uniform coverage Raster

frequency

x Y

low-energy

315

ion implanter

the beam line and in the growth chamber. Specifications for the instrument are given in table 1.

2. Physical

design

constraints

There are a number of constraints on the nature of the beam at the target for it to be compatible with simultaneous silicon MBE. These concern the following: _ choice of species, - beam purity and accompanying gas load, _ ion energy range and incidence angle, - flux and area1 coverage. The use of the well-established Freeman ion source [5] ensures compatibility with all the silicon dopant elements, plus a wide range of other species, including silicon itself, that may prove valuable for ion-assisted growth, ion etching, or ion-beam deposition within the MBE processing environment. The low emittance of the source is also important for magnetic mass analysis and for well-controlled ion deceleration. The need for a very pure source feed material is considerably relaxed by comparison with UHV evaporators or non-mass-analyzed ion-beam attachments. The calculated typical resolving power of 150 (5% valley between adjacent masses) provides sufficient mass resolution to give pure monoisotopic beams of the silicon dopant species. Sputtering of beam-line components by the ion beam could introduce very undesirable metallic impurities into the growing film. High-density, high-purity graphite components are used where 10 keV ions will strike surfaces, and careful arrangement of baffles along the beam line minimizes the possibility of sputtered material proceeding towards the MBE growth chamber. The final aperture before the growth chamber is constructed

implanter

Freeman universal type, 45 o angled 10 kV 45 ‘=‘,350 mm radius, 0.65 T maximum field 90 O, 300 mm radius, 0.75 T maximum field (5s ~5x10~8mbarls-’ 10 keV down to 50 eV or lower < 0.1% 25x7.0 mm2 up to 50 uA cm-’ before rastering (final target value 100 PA cm-2) central 100 mm diameter of 150 mm diameter wafers to better than 10% uniformity (final target value 2%) 3000 Hz 12 Hz

IV. SOURCES

& BEAM TRANSPORT

316

J.S. Gordon et al. / A silicon MBE-compatible

from a silicon wafer to ensure that any sputtered material in this critical region is benign. The gas load into the growth chamber is arranged to be no greater than that expected from a typical effusion cell. This is achieved using five stages of differential pumping to reduce the pressure from lop4 to lo-’ mbar in the implanter source chamber to lo-* mbar prior to the MBE growth chamber. Whilst the first mass-resolving magnet removes the neutral flux from the source from the ion beam, it is vital to take further steps to eliminate the high energy neutrals formed when ions in the beam charge exchange with residual gas species. These neutrals have the full acceleration energy and are of course unaffected by the deceleration field. Tbe second electromagnet, therefore, as well as performing beam steering, removes unwanted neutrals from the ion beam. The opportunity for further fast neutral creation beyond the magnet is considerably reduced in the ultrahigh-vacuum (UHV) conditions that then prevail (lo-’ mbar and better). Experience on a UHV low-energy ion-beam implantation and deposition system at the University of Salford has shown that neutral trapping at this stage effectively eliminates damage to silicon substrates from high-energy neutrals

161.

Ion energies in the sub-500 eV range are needed to ensure that the dopant can be confined only to the growing surface region. Choice of ion energy for MBEcompatible implantation involves more than simply selecting on the basis of the projected range, however. It is also necessary to take into consideration radiation damage, sticking coefficients and the efficiency of incorporation of the dopant ions into active sites for a given dopant/ substrate combination. The bombardment of silicon at MBE growth temperatures with ions capable of producing significant atom displacements leads to the formation of highly stable complex defects which cannot be annealed out below about 1000 K. High temperature anneals after an implant are usually out of the question, as the resulting diffusion would destroy the nanometer scale doping profiles created during the growth process. Since it has been shown that several silicon atoms are displaced from lattice sites per incident ion during bombardment by an argon beam of only 60 eV [6], it is clearly important to try to reach the lowest possible energies to search for the point where there is still good incorporation efficiency, but stable damage structures are not formed. The ability to achieve a uniform beam at low energies is dependent upon having a low-emittance beam undergoing a well-controlled deceleration. Experience on the University of Salford low-energy implanter has shown that beam quality can be maintained with less than 10 V difference between source and target potential [7]. In the instrument described here, however, the retardation optics must not alter significantly the MBE

ion implanter

growth chamber. The solution adopted has been to use the cold shields of the chamber itself as the final lens element. Initial ion-optic simulations have shown acceptable performance down to 50 eV. It is an aim of the project to explore the low-energy performance limit. The retardation geometry features normal beam incidence onto the target, as this has been found to give the best control at low energies. Normal incidence is also predicted to give better doping profiles than non-normal incidence [7]. The possible detrimental effects of channeling on doping profiles are currently under investigation [8], but it is not expected that the choice of normal beam incidence will be a significant disadvantage. Practically any incidence angle will reveal a channelling effect at these low energies because of the large critical angles for channelling. The required beam flux at the target can be derived simply from the required doping level and known MBE growth rates. Taking the volume doping level N, [cmp3] to be given by

where N, is the atom density of the film, J, the molecular beam flux (atoms cme2 s-i), fi and f,, the fractions of the incident ion and molecular beam fluxes which are incorporated into the growing film, I the beam current density (A cmm2), n the ionic charge state, e the electronic charge, a the beam area and A the area over which the beam is rastered. Taking the case of silicon growing at one monolayer per second (J, = lOi cm-* s-i), assuming f/f, is unity and a/A is l/32, the ion-beam current must be 10 lr,A cm-* to reach doping levels in the 102’ cmm3 range. Based on the performance of the system at Salford University, a maximum current density of 50 PA cmp2 is expected, relatively independent of the selected energy. Even higher doping levels should be achievable by reducing the MBE growth rate (or halting it temporarily) or by reducing raster area.

3. Technical and practical considerations The physical arrangement of the system is shown in fig. 2. The overall geometry is fixed by the position and orientation of the target in the MBE growth chamber, and the normal beam incidence requirement. The electron beam evaporators used for silicon growth require the target to be horizontal, and facing downward. The normally incident ion beam must therefore enter the chamber vertically from below. This is achieved using a 90” deflecting magnet placed under the growth chamber. This section of the beam line must be disconnected from the remainder of the system and from the growth chamber, and moved clear whenever it is necessary to

J.S. Gordon et al. / A silicon MBE-compatible MBE

growth

base

flange

ion implanter

317

chamber Ion

source /

Electromagnets

i-

3.1

m

Fig. 2. Elevation viewof the low-energyion implanter.

lower the growth chamber flange to recharge the MBE evaporators. The ion source is conventional except in that the arc chamber is attached at 45” so that a beam can be extracted at that angle down towards the first electromagnet. The target wafer and cooling shields of the MBE system cannot be taken to high voltage, therefore it has been necessary to design an isolated high-voltage flight tube to transport the ions from the source extraction lens to the retardation stage in the growth chamber. This runs within the (grounded) vacuum envelope, supported by ceramic insulators. Pumping conductance into the inner tube is provided with high-transmission mesh. A special mechanism withdraws the inner flight tube to allow gate valves at either end of the beam line to close. The choice of a 10 kV transport potential is a compromise which gives acceptable transport efficiency in a system in which totally effective space-charge compensation cannot be relied upon in the ultrahigh-vacuum stages, with good deceleration performance, operational convenience and acceptable levels of sputtering and heating in the beam line. Using the two electromagnets with Hall probe feedback, and an electrostatic beam dump and fast beam gate, it is possible to switch between two species being extracted from the source with zero possibility of target contamination, and accurate dose control. A two-axis beam raster provides one complete coverage of 100 mm diameter in 40 ms, a time short compared to the time for one monolayer to be deposited. The practical performance of the retardation optics when the beam is deflected well off-axis is the subject of investigation.

4. Conclusions The design of a low-energy MBE-compatible ion implanter differs considerably from conventional

medium- and high-energy implanters. The combination with silicon MBE should allow the growth of low-dimensional structures with p- and n-type doping, and also longer-term investigations into ion-assisted growth processes. Practical investigations are necessary into the achievable low-energy limit and the behaviour of the beam when it is deflected well off-axis for large wafer coverage,.

Acknowledgements The authors acknowledge the mechanical design skills of Simon Head of VSW. The project has been funded as part of the SERC Low Dimensional Structures and Devices Programme.

References [l] Y. Ota, J. Appl. Phys. 51 (1980) 1102. [2] J.C. Bean and E.A. Sadowski, J. Vat. Sci. Technol. 20 (1982) 137. [3] D.C. Houghton, M.W. Denhoff, T.E. Jackman, M.L. Swanson and N. Par&h, J. Electrochem. Sot. 135 (1988) 3109. [4] J.-P. Noel, J.E. Greene, N.L. Rowe11and D.C. Houghton, Appl. Phys. Lett. 56 (1990) 265. [5] J.H. Freeman, Nucl. Instr. and Meth. 22 (1963) 306. [6] A.H. AI-Bay&, K.G. Orrman-Rossiter, R. Badheka and D.G. Armour, accepted for publication in Surface Science. [7] D.G. Armour, MRS Symp. Proc. 100 (1988) 127. [8] A. Bousetta, J.A. van den Berg, D.G. Armour and P.C. Zalm, these Proceedings (8th Int. Conf. on Ion Implantation Technology, Guildford, UK, 1990) Nucl. Instr. and Meth. B55 (1991) 565. IV. SOURCES & BEAM TRANSPORT