Formation of iridium silicide layer by high dose iridium ion implantation into silicon

Formation of iridium silicide layer by high dose iridium ion implantation into silicon

Nuclear Instruments North-Holland and Methods in Physics Research 21 B58 (1991) 27-33 Formation of iridium silicide layer by high dose iridium i...

957KB Sizes 0 Downloads 6 Views

Nuclear Instruments North-Holland

and Methods

in Physics

Research

21

B58 (1991) 27-33

Formation of iridium silicide layer by high dose iridium ion implantation into silicon K.M. Yu a, B. Katz a, I.C. Wu a and I.G. Brown b ’ Center for Advanced Materials, Materials and Chemical Sciences Division, ‘Accelerator Received

and Fusion Research Division, Lawrence Berkeley

Laboratory,

Lawrence Berkeley Laboratory, Berkeley CA 94720, USA

Berkeley

CA 94720, USA

July 25 1990

We have investigated the formation of IrSi, layers buried in (111) silicon. The layers are formed by iridium ion implantation at room temperature using a metal vapor vacuum arc (MEWA) high current metal ion source with an averagebeam energy = 130 keV. Doses of the Ir ions ranging from 2 X lOI to 1.5 x 10”/cm2 were implanted into (111) Si. We have successfully formed a buried silicide layer in silicon by room temperature implantation and annealing. The formation of IrSi, phase is realized after annealing at temperatures as low as 500 o C. A continuous IrSi, layer of = 200 A thick buried under = 400 A of Si was achieved with samples implanted with doses not less than 3.5 x 10t6/cm2. Implanted doses above 8 X 10i6/cm2 resulted in the formation of an IrSi, layer on the surface due to excessive sputtering of Si by the Ir ions. The effects of implant dose on the phase formation mechanism, the interface morphology and the implanted atom redistribution are discussed. Radiation damage and regrowth of the Si due to the implantation process was also studied.

1. Introduction Recently, the application of ion beam techniques in the formation of metal silicides has attracted much interest. Research in this field centers around two areas, namely, ion beam mixing and ion beam synthesis. In ion beam mixing, inert gas ions are implanted into a metal-silicon structure in order to mix the interface through cascade and recoil mixing when the implantation is carried out at low temperature and through radiation enhanced diffusion when the substrate is intentionally heated during implantation [1,2]. In ion beam synthesis, the metal content needed in the metal silicide is supplied by directly implanting the silicon substrate with metal ions to a dose determined by the composition of the particular silicide studied. Silicides formed by direct metal ion implantation have several potential advantages over those formed by conventional solidstate reactions including (1) the high purity of the silicide since the metal ion beams are normally isotopitally separated, (2) the precise control of the atomic ratio and layer thickness of the silicide, (3) the precise control of forming non-equilibrium new phases and (4) the formation of buried silicide layers below the silicon surface. Buried silicide layers are particularly interesting technologically since they have great potential in the fabrication of novel devices such as metal base transistors and are also a major building block for the development of three dimensional integration of devices. 0168-583X/91/$03.50

0 1991 - Elsevier Science Publishers

In the past few years, several studies have been published describing silicide formation by high dose implantation of transition metal ions. Silicides which have been formed by this process include CoSi, [3-81, TiSi, [3,9,10], CrSi, [5,6,11], FeSi, [ll], VSi, [11,12], Nisi, [3,13] and YSi, [5]. White et al. [4-61 have fabricated epitaxial buried layers of CoSi, and CrSi, by implanting metal ions into a heated Si substrate (350-450 o C) with subsequent annealing at - 1000 o C. In fact, buried silicide formation has only been observed when the implantation was carried out at elevated substrate temperature (300 o C) [4,9,11]. This is thought to be the result of diffusion of the metal ions away from the surface so that the net implantation profile becomes deeper than that expected by theoretical calculations [4]. We have chosen to study the formation of IrSi, by Ir ion implantation because IrSi, has several interesting structural and electrical properties which make it a unique silicide system. In the last ten years, only a few reports were published on the solid-state reactions between thin film Ir on Si [14-171. Among the metal silicides, IrSi, has the highest silicon concentration (75 at.%) and therefore requires the least amount of metal atoms for its formation. When IrSi, is formed by implantation with Ir ions, this means less sputtering of the Si and shorter implantation time due to the relatively small dose of Ir ions needed. Solid-state reactions between Ir thin films and Si substrates indicate that the formation of the IrSi, phase is initiated at an annealing temperature around 1000 o C [14], higher than the for-

B.V. (North-Holland)

28

KM.

Yu et al. / Formation of IrSi, layers in Si

mation temperature of any other metal silicides. Because of its high formation temperature, IrSi, formation does not follow the diffusion controlled kinetics as in most cases since at such a high temperature diffusion of the Ir and Si is very fast in the layer. Instead, IrSi, formation was governed by nucleation controlled kinetics 1141 where the growth of the silicide did not follow a layer-by-layer process. According to Ishiwara et al. [18], the lattice mismatch between IrSi, (hexagonal with a = 4.354 A and c = 6.628 A) and (111) Si is 1.878, so that epitaxial growth of IrSi, on (111) Si is very likely. Recently Chu et al. [17] investigated the epitaxial growth of IrSi, on (111) and (100) Si and observed that large grains of - 30-40 urn in diameter of epitaxial IrSi, can be formed on (111) Si. Formation of an epitaxial buried IrSi, layer is therefore very possible with high dose ion implantation. In addition to the interesting structural properties, the Schottky barrier formed between IrSi, and n-Si exhibit barrier height = 0.94 eV, the highest in any silicide/silico’n contact structure. In this work we investigated the formation of IrSi, in (111) Si by direct Ir ion implantation into Si substrate at room temperature. The silicides were characterized by Rutherford backscattering spectrometry (RBS) and glancing angle X-ray diffraction. The interface morphology and the microstructures were studied by transmission electron microscopy (TEM). The regrowth of the Si layer due to radiation damage was analyzed by ion channeling method.

2. Experimental The implantation was performed by MEWA (metal vapor vacuum arc) high current ion source [19]. Briefly, in this ion source the highly ionized metal plasma that is created in the metal vapor vacuum arc discharge provides the “feedstock” from which the ion beam is extracted. For this work, the beam extraction voltage was 50 kV. The measured flux distributions were 6%, 41%, 52% and 1% for 1 + , 2 + , 3 + and 4 + particles, respectively. The mean charge state of the iridium ions produced were therefore = 2.6. Since no energy analysis was carried out, the mean implantation energy for this work was = 130 kV. The wafer was cooled by thermal conduction of the support structure. The beam pulse repetition rate was limited to 5 pulse/s with a pulse duration of 250 ps to avoid overheating of the target. No intentional heating of the wafer was applied and hence the implantation was considered to be carried out at room temperature. However, the actual temperature of the substrate during implantation was not measured. The ion current density was - 20-30 pA/cm2. Implantation doses varied from 2 X lOI6 to 1.5 X 10i7/cm2.

200

CHANNEL

#

450

Fig. 1. The RBS spectra from the as-implanted samples for various doses. The doses indicated are the retained doses measured by RBS. Note that the Ir signal moves toward the surface as the dose increases. Annealing was carried out in the temperature range of 400-900°C for time duration of 20 min to 5 h in flowing N, with the samples covered by a blank Si wafer. Another set of samples has undergone rapid thermal annealing (RTA) in Ar ambient at 820°C for 20 s. Rutherford backscattering spectrometry (RBS) with 1.8 MeV 4He+ ion beams was used to study the ion dose, depth profiles, atomic composition and interface sharpness of the structures. The backscattering angle 0 was 165O and the samples were tilted with the beam to a surface normal angle = 60-65 o to improve the depth resolution to I 50 A. Channeling/RBS in the (111) axis were carried out with the ion beam at normal incidence to the sample in order to investigate the damage recovery of the samples. X-ray diffraction (XRD) in the Seeman-Bohlin geometry was carried out using Cu K, X-ray, for high sensitivity phase identification. Phases formed in a layer 2 50 A thick can be identified by this technique. The microstructures and the interface morphology were studied by transmission electron microscopy (TEM) of both plan-view and cross-sectional specimens. 3. Results and discussion Fig. 1 shows a series of RBS spectra for Ir implanted samples with different doses. The total Ir doses in these

K.M. Yu et al. / Formation GOOtaj

\

Surface Si vs. lrnplanted

of IrSi, 40

thickness Dose

29

layers in Si Ir

f b)

Peak Concentration vs. Implanted Dose

5-J

\

“t \

5 030: E \

zioom .I/)

‘*

\*

,/ _______---__--~~_----I'Si, /

;J

\ \

4

-:oc

\

/

\

E D

\

2 200:

'

00

Sputter

yield

2

=

2.54

:

\

‘\

\

8

*\ \

\ \ \ \ \ \ . I,,mI.IIsIIII ,.I 8 ‘10 Total

D%se

(X10”

/cm’)

0

'3

L

\ *' \

Y ; IOL i : ,k -/ ;( 00

/

2/ /

c/ /

k’ g-’ 2

Tota:

Dose

8 (“X 1 016 /cm2

10

12

)

Fig. 2. (a) A plot of the surface Si layer thickness measured by RBS as a function of the total implanted Ir. (b) A plot of the Ir peak concentration in Si as a function of implanted Ir dose.

samples as measured by RBS are 3.5, 6.3 and 8.5 X 10’6/cm2 while the intended doses were = 3.5, 7.0 and Therefore, the maximum 10 X 10’6/cm2, respectively. retention dose for 130 keV Ir implanted into Si is = 8.5 x 10’6/cm2. Note that the thickness of surface Si layer decreases and the signal moves to the surface as the implanted dose increase due to sputtering of Si by the Ir ions. For low dose implants (< 3 X 10’6/cm2), the Ir profile in Si shows a Gaussian shape with range straggling = 190 w while the calculated value according to the LSS theory for 130 keV implant is = 100 A. The broadened Gaussian distribution comes from the energy spread as a result of the multiple charge states of the Ir ions in this work. For the high dose samples (> 7 X 10’6/cm2), the implant profile becomes flat-top with a low energy tail.

200

290 360 CHANNEL

#

The surface Si layer thickness as measured by RBS is plotted against the total dose in fig. 2(a). When extrapolating the curve to high dose, we observe that - 1017/cm2 of implant dose is needed in order to sputter away ail the Si on the surface. This is consistent with the RBS results in fig. 1 where we observe that all the surface Si is sputtered away in the sample with an intended dose of 10’7/cm2 (retained dose of 8.5 X 1016/cm2). The projected range can also be estimated to be = 530 _& by extrapolating this curve to zero dose. This is in good agreement with the calculated value of 550 A for 130 keV Ir ions implanted into Si from the LSS theory. In fig. 2(b), the peak concentration of the Ir atoms are plotted against the total measured dose. A linear dependence up to 30 at.% of Ir in Si is observable in this figure. For the stoichiometric IrSi,, Ir content in

460

Fig. 3. (a) The RBS spectra from the low dose sample, dose = 3.5 X 1016/cm2, as-implanted and annealed at 700’ C for 1.5 h. (b) XRD spectrum for the annealed sample in (a) indicating the presence of IrSi, and polycrystalline Si phases.

30

K.M. Yu et al. / Formation of IrSi,

the silicide is 25 at.%. From fig. 2(b), we can estimate that a total dose of - 6.3 X 10’6/cm2 is needed so that the peak Ir concentration reaches 25 at.% which we call the stoichiometric dose level (Nt),,. XRD analysis of the samples with retained Ir doses varying from 1.5-8.5 x 10i6/cm2, reveals that the IrSi, phase appears for all doses following annealing at temperatures above 500°C while the as-implanted samples exhibit an amorphous phase. This is consistent with the findings of Petersson et al. [14] who reported that the formation of IrSi, is a heterogeneous process nucleated at distinct points in the films at very high temperature ( - 1000 o C) for solid state reactions between Ir thin film and Si substrate. For implanted samples, it is expected that many IrSi, nucleation sites are already available. Therefore, a very low annealing temperature - 500 o C is needed for the formation of this phase after implantation. The RBS spectra from the samples with low dose, 3.5 X 1016/cm2, are shown in fig. 3(a) both for as-implanted and after annealing at 700” C for 1.5 h. The spectrum for the as-implanted sample indicates that the original Ir profile is Gaussian with Ir peak concentra= 400 A of Si. The tion = 15 at.% buried underneath spectrum for the sample after annealing shows that the Ir atoms diffuse toward the peak of the profile from both sides of the distribution. The buried layer formed after annealing has a nominal composition of IrSi,, with thickness of = 200 A. It is also observed from the

Table 1 A summary of the XRD patterns obtained in the sample implanted with Ir to a total dose of 3.5 X 1016/cm2 and annealed at 700°C for 1.5 h Experimental d (A) Intensity 4.001 3.437 3.240 2.554 2.220 1.936 1.883 1.847 1.671 1.531 1.447 1.407 1.343 1.257 1.207 1.089

m s s m m s w s w w w w m w w w

Si

IrSi, a b d (A)

hkl

Intensity

3.77 3.28

100 m 101 s

2.491 2.177 1.907

102 m 110 m 103 m

1.815 1.657 1.517 1.434 1.395 1.318 1.250 1.198 1.085

201 004 104 203 211 114 105 213 205

d (A)

hkl

3.136

111 s

1.810

003 m

layers in Si

RBS spectrum that in the annealed sample = 2 at.% of Ir is present in the surface Si layer. This Ir distribution in the surface Si may be the result the singly charged Ir particles (6%) produced by the MEVVA ion source during the implantation process. The XRD analysis of the same sample shown in fig. 3(b) reveals patterns of IrSi, and polysilicon phases. A summary of the XRD results is tabulated in table 1 together with the diffraction patterns for the IrSi, and Si standards. Table 1 clearly shows the presence the IrSi, phase in the sample. The IrSi, phase observed is hexagonal in structure with lattice parameters a = 4.44 A and c = 6.684 A which are slightly larger than those reported in the literature 1141. With these measured lattice parameters for the IrSi, phase, the lattice mismatch on (111) Si is < 1.8% [17] and therefore favors epitaxy. A transmission electron diffraction study of the surface Si of the annealed sample also shows the ring pattern of polycrystalline Si. The polysilicon diffraction pattern in fig. 3(b), however, has peaks shifted to lower Bragg angle is compared to the Si standard in the XRD data file. This shift in the Bragg peaks can be related to change in the lattice parameter of the Si layer through the relationship: 2(Aa/a) = -cot 0(A28), where a is the lattice parameter and 0 is the Bragg angle. The increase in the Si lattice parameter obtained this way is As/a = 0.04. Since the RBS result shows a uniform Ir distribution of - 2 at.% in the surface Si layer in this sample, it is very likely that the incorporation of the Ir atoms in the surface Si layer during regrowth of this layer results in the expansion of the lattice parameter. From the RBS, XRD and TEM results, it is expected that a continuous buried IrSi, layer will be formed at a retained Ir dose of 2 3.5 x 10’6/cm2 (with as-implanted Ir peak concentration = 15 at.%). The RBS spectra from the samples with retained dose = 6.3 x 10n’/cm2 = (Nt),,, as implanted and fur-

Intensity

Medium

Dose

Ir

lmplonted

S1

s w m m m m m m m

’ JCPDS standard pattern #19-59X. ’ v, m, and w represent strong, medium, intensities, respectively.

and weak diffraction

Fig. 4. The RBS spectra from the samples with medium dose, 6.3 X 1016/cm2, as-implanted and annealed at 750 o C for 2 h.

K. M. Yu et al. / Formation of IrSi

annealed at 750 o C for 2 h are shown in fig. 4. The spectrum for the annealed sample shows sharpening of the Ir profile forming a layer = 400 A thick with nominal composition of IrSi, buried under = 300 A of Si. Similar structure is observed for the sample annealed with RTA at 820 o C for 20 s. The RBS results show that the IrSi,/Si interfaces in these annealed samples are very abrupt to within 50 A resolution (in the glancing incident angle geometry). XRD on the furnace annealed sample shows diffraction peaks of the IrSi, and polysilicon phases indicating that the surface Si layer is polycrystalline. A cross-sectional TEM micrograph of the sample after RTA is shown in fig. 5. The TEM micrograph shows a continuous IrSi, buried layer with sharp Si/IrSi,/Si interfaces consistent with the RBS data. Fig. 6 shows the dark field TEM micrograph on the plan-view specimen of the same sample with the corresponding electron diffraction pattern. The micrograph in fig. 6 shows rod-shaped IrSi, grains of - 5000 A-1 urn in length formed in Si. The diffraction pattern reveals that these grains are indeed IrSi, grains. Fig. 7(a) shows the RBS spectra from the samples with retained dose = 8.5 X 10i6/cm2 as-implanted and annealed at 700 o C for 2 h and 900 o C for 20 min. The as-implanted sample has an Ir-Si layer = 400 A thick on the surface with [Ir]: [Si] ratio - 1 : 2.34. The Ir signals from fig. 7(a) are shown in fig. 7(b). From fig. 7(b) we observe that the sample annealed at 700 o C for 2 h has layered structure of (170 A) IrSi,_,,/(250 A) IrSi, on Si. A uniform layer of stoichiometric IrSi, on

3layers in Si

31

nace

Fig. 5. Cross-sectional

Fig. 6. Plan-view TEM dark field image of the same sample as in fig. 5. The corresponding electron diffraction pattern is also shown.

TEM micrograph of the sample with the same dose as those in fig. 4 but after RTA. interfaces between the buried silicide layer and the surface and substrate Si are very sharp.

Note that both

the

32

K.M.

Yu et al. / Formation

of IrSi,

layers in Si

7000 High

b)

d

Dose

Ir

lmplonted

SI

SOOT 20 min.

f I Y E 3 8

300

200

0

370

CHANNEL

460

370

CHANNEL

#

4;

#

Fig. 7. (a) The RBS spectra from the samples with high dose, 8.5 X10’6/cm2, as-implanted and annealed at 700°C for 2 h. and 900°C for 20 min. (b) The Ir signals of (a) in a different scale. Note that the redistribution of the Ir signals as the samples are annealed.

Si is achieved after annealing at 900 o C for 20 min. The samples annealed at lower temperatures (500-700 o C) only show a sharpening in the Ir-Si layer/Si interface but the [Ir] : [Si] ratio in the layer remains = 1 : 2.34. However, XRD shows only diffraction pattern of IrSi, phase for the samples annealed at temperatures above 500 o c. From the RBS and XRD results, we conclude that for samples with doses higher than (Nt),,, annealing at temperatures in the range of 500-700 o C results in the formation of IrSi, phase with nominal composition of IrSi,_,. The excess Ir atoms in the layer diffuse into the Si substrate when the samples are annealed at higher temperatures (> 700 o C) forming a thicker layer of stoichiometric IrSi,. RTA at 820” C for 20 s on this sample also results in a sharp layer of stoichiometric IrSi, on Si.

5000

(4

High

Oose

85OT

40

min.

A (111) aligned RBS spectrum for the high dose sample after annealing at 850 o C for 40 min is shown in fig. 8(a) together with the spectrum of the sample taken in the random orientation. The Ir signals in these spectra show no reduction in the yield when the sample is aligned in the channeling orientation indicating small grain polycrystalline IrSi, in the layer. This is consistent with our TEM results which show rod-shaped grains of IrSi, in the sample. Fig. S(b) shows the Si signals from the RBS spectra in the (111) channeling alignment for the high dose samples after annealing at 500 o C for 1 h, 850°C for 40 min, and after RTA (the (111) spectrum for the as-implanted sample was similar to that after the 500 o C anneal). A damage layer of = 1000 A Si, possibly amorphous, below the IrSi, layer is observed in the sample which had been annealed at 500 o C for 1 h. This sample after RTA shows epitaxial regrowth of this Si

+



Aligned

High

Oose

Samples

--ill> + Random AlIgned

d 5 ;: Y r

+

200

300 CHANNEL #

400

480

170

5OO'C

1 hr.

CHANNEL

#

Fig. 8. (a) The RBS spectra from the high dose sample (Nt = 8.5 X lOI at/cm*) after 850 o C 40 min anneal in the random aligned direction. (b) The Si signals from the RBS spectra min and RTA aligned with the (111) axis. The channeling

320

and (111)

of the high dose sample after annealing at 500 o C for 1 h, 850 o C for 40 spectra were taken with the ion beam at normal incidence to the sample.

K. M. Yu et al. / Formation

layer from the substrate leaving < 200 A of damaged Si below the IrSi, layer as indicated by the reduced backscattering yield of the Si signal below the surface. Complete regrowth of the damaged layer is achieved after annealing at 850 o C for 40 min.

4. Summary and conclusions We have studied the formation of IrSi, layers by direct Ir ion implantation into Si. A buried IrSi, layer in Si has been fabricated by implantation of Ir ions into Si at room temperature at an average energy of 130 keV. The critical minimum dose for the formation of a buried continuous layer of IrSi, in Si is 2 3.5 x 10i6/cm2. This critical dose is much lower than those observed for other buried silicides studied mainly because of the fact that IrSi, has the lowest metal concentration in any metal-silicide. The IrSi, phase is formed after annealing at temperature 2 500°C for doses as low as 1.5 X 10i6/cm2. For the samples with doses below the stoichiometric dose, (Nt),,, upon annealing the Ir atoms diffuse to the center of the implant from the tails of the profile. For the samples with doses higher than (Nt),,, annealing at temperature below 700” C results in IrSi, phase with higher Ir content (IrSi,_,). Redistribution of the Ir atoms into the substrate occurs during annealing at temperatures above 700” C forming a thicker layer of stoichiometric IrSi,. Complete regrowth of the damaged Si below the IrSi, layer due to implantation is observed after annealing at 800” C for 40 min. However, no epitaxial layer of IrSi, is formed in this study. It is likely that a buried epitaxial IrSi, in Si can be formed by Ir implantation in a heated Si substrate where in situ annealing is performed.

Acknowledgement This work was supported by the Director, Office of Energy Research, Office of Basic Energy Sciences and Materials Science Division of the U.S. Department of Energy under Contract No. DE-AC03-76SF00098.

of IrSi,

layers in Si

33

References [I] H. Okabayashi, Nucl. Ins&. and Meth. B39 (1989) 246. [2] A. Hamdi and M.A. Nicolet, Thin Solid Films 119 (1984) 357. [3] F.H. Sanchez, F. Namavar, J.I. Budnick, A. Fasihudin and H.C. Hayden, Mater. Res. Sot. Symp. Proc. 51 (1986) 439. [41 A. White, K.T. Short, R.C. Dynes, J.P. Gamo and J.M. Gibson, Appl. Phys. Lett. 50 (1987) 95. [51 A. White, K.T. Short, R.C. Dynes, J.M. Gibson and R. Hull, Mater. Res. Sot. Symp. Proc. 100 (1988) 3. WI A. White, K.T. Short, R.C. Dynes, R. Hull and S.M. Vandenberg, Nucl. Instr. and Meth. 339 (1989) 253. [71 A.H. Van Ommen, J.J.M. Ottenheim, A.M.L. Theunissen and A.G. Mouwen, Appl. Phys. Lett. 53 (1988) 669. PI K. Kohlhof, S. Mantl, B. Sttitzker and W. Jager, Nucl. Instr. and Meth. B39 (1989) 276. G.J. Clark, F.K. Leguoues, F.M. d’Heurle [91 P. Madakson, and J.E.E. Baglin, Mater. Res. Sot. Fall Meeting 1987, Symp. H, Boston, MA, Nov. 30-Dec. 5, 1987. B.M. Arora, DOI V.P. Salvi, S.V. Vidwan, A.A. Rangwala, Kuldeep and A.K. Jain, Nucl. Instr. and Meth. B28 (1987) 242. F.H. Sanchez, J.I. Budnick, A.H. Fasihudin [111 F. Namavar, and H.C. Hayden, Mater. Res. Sot. Spring Meeting 1987, Symp. C, Anaheim, CA, April 21-23, 1987. WI S. Vidwans, V.P. Salvi, A.A. Rangwala, B.M. Arora, A.K. Jain and Kuldeep, Nucl. Instr. and Meth. B39 (1989) 280. P31 J.K.N. Lindner and E.H. teKaat, J. Mater. Res. 3 (1988) 1238. F. D’Heurle, T.S. [I41 S. Petersson, J. Baglin, W. Hammer, Kuan, I. Ohdomari, J. de Sousa Pires and P. Tove, J. Appl. Phys. 50 (1979) 3357. T.S. Kuan and K.N. Tu, J. Appl. Phys. 50 u51 I. Ohdomari, (1979) 7020. 1161 M. Wittmer, P. Oelhafen and K.N. Tu, Phys. Rev. B35 (1987) 9073. u71 J.J. Chu, L.J. Chen and K.N. Tu, J. Appl. Phys. 63 (1988) 1163. Jpn. J. Appl. [W H. Ishiwara, S. Saitoh and K. Hikosaka, Phys. 20 (1981) 843. P91 LG. Brown, J.E. Galvin, B.F. Gavin and R.A. MacGill, Rev. Sci. Instr. 57 (1986) 1069.