VACUUM Vol. 14, pp. 263-267.
Pergamon
Press Ltd.
Printed in Great Britain.
Mass Spectrometer
for the Study of Sputtering*
A. J. SMITH, D. J. MARSHALL, Nuclide Corporation,
L. A. CAMBEY and J. MICHAEL
P.O. Box 752, State College, Penn.,
U.S.A.
(Received 18 February 1964 ; accepted 8 June 1964)
A single focusing mass spectrometer system has been built to study sputtering of surfaces by ion Sputtered ions bombardment. Bombarding ions are formed in an oscillating electron source. are mass analyzed directly. Sputtered neutrals are ionized and detected using a pulsed sourcedetector system which allows the measurement of the ionized sputtered neutral beam intensities to be made even in the presence of much larger signals from the residual gases in the instrument. Introduction
Description
The application of mass spectrometry to the direct analysis of particles sputtered from a surface by ion bombardment is a useful technique for studying target composition and many types of surface phenomenal**. To be most useful in such work, a mass spectrometer must meet several important requirements. It is important that the system be capable of maintaining an ultra-high vacuum for two main reasons. In order that the target surface can be kept clean of background gases that remain in the system, the bombarding ion beam density should be greater than the atom flux density of these gases. To make it possible to satisfy this condition with moderate bombarding ion beam densities it is necessary that the system be capable of maintaining an ultra-high vacuum. This vacuum requirement is also necessary to reduce the magnitude of the ion peaks from the background gases which interfere with the detection of the ionized sputtered neutral peaks. Overflow of gas into the target region from the source which supplies the bombarding beam should be low enough that back diffusion of sputtered particles to the target is In addition to the standard Faraday cup ion negligible. collector, the instrument should have an electron multiplier detector in order that maximum sensitivity be provided for the detection of the small currents of sputtered species. It must be capable of analyzing both sputtered ions and sputtered neutrals because the many studies which lend themselves to ion bombardment techniques are best carried out by recording either or both types of sputtered particles. Detection of the sputtered neutrals is particularly important because their relative intensities are apparently much more representative of sample composition than are sputtered ion intensities. The detection of neutrals, however, is complicated by the fact that residual background gases are ionized simultaneously with the neutrals and, therefore, the signal-to-noise ratio is, in practice, very low for neutral species. The authors would like to report here on a mass spectrometer built for sputtering studies which meets these requirements.
The instrument, shown schematically in Fig. 1, is a 60” deflection, 6 in. radius magnetic sector field spectrometer housed in a metal vacuum enclosure and built to ultra-high vacuum standards. Gold wire-gasketed flanges join the individual housings. A removable oven canopy which encloses all of the vacuum system with the exception of the diffusion pumps and part of the pumping arms provides bakeout temperatures up to 450°C.
*This instrument was developed by Nuclide Corporation Nuclear Company for the U S. Atomic Energy Commission.
FIG. 1. Schematic
diagram
of the mass spectrometer
Two separate electron bombardment ion sources are used. The first supplies the ion beam for bombarding the target while the second one ionizes the sputtered neutrals together with any residual background gases. This second source also enables the instrument to be used as a standard gas mass spectrometer when the first source is switched off. The ion sources are shown schematically in Fig. 2. The high efficiency source which supplies the bombarding ion beam uses the oscillating electron arrangement described by Finkelsteins. Electrons emitted from the filament oscillate between the two reflection plates at either end of the ionization region by virtue of the potential distribution existing between the plates and the tubular copper anode. A magnetic field parallel to the source axis maintains a high electron density
for the Oak
263
Ridge
National
Laboratory
which
is operated
by Union
Carbide
264 in the ionization
by the authors
A. J.
SMITH,
D. J. MARSHALL,L. A. CAMBEYand E. J. MICHAEL
region. The detailed source geometry chosen is similar to that described by Almen and
SAMPLE
GAS ,
FILAMENT
TERMINALS _
Nielsenh. The magnetic
field of about 200 gauss needed to operate the source is supplied by a solenoid of _rb in. dia. anodized aluminium wire wound onto a form concentric with the source housing. Standard magnet wire would not do for this application since it is necessary to leave the magnet in place during bakeout of the instrument. Sample gas from a gas handling system is let into the Finkelstein source through a Granville-Phillips variable leak valve. A 2 1. reservoir is used to store the sample gas. With argon, a reservoir pressure of 80 torr provides an essentially constant flow rate for about one hour before the gas pressure decreases by one per cent. The second source is a Nier electron bombardment types which is modified to accept particles sputtered from the target. A slit cut into the rear of the ionization box allows the sputtered particles to pass through the electron beam, where some of the neutrals are ionized. Sputtered ions pass The target is mounted in a through essentially undisturbed. holder which is located to the rear of the Nier source, directly behind the ionization region. A micrometer operated lever allows the target to be rotated about an axis perpendicular to the plane of the ion beams from outside the vacuum. The angle between the normal to the target and the bombarding beam is adjustable from 0’ to 90”. The two sources are housed separately and are isolated from each other by a .250 in. .030
FINKEL SOU
I+
:E
-ACCELERATING ELECTRODE
-COLLIMATING
I
rlONlZlNG
ELECTRON
SLIT
t
BEAM
iR -.-.--.--TO ANALYZER
SC RCE
FIG. 2. Diagram showing the source geometry of the mass spectrometer. Ions and neutrals sputtered from the target pass through the standard gas source where residual background gases and some of the sputtered neutrals
are ionized in the electron beam. The resulting ion beam is then mass analyzed in the magnetic sector field.
With the ion beam accelerating voltage supplies connected as shown in Fig. 2, the bombarding ion beam is accelerated to ground after it leaves the Finkelstein source and then is decelerated between the grounded collimating slit and the target. Since thesupplies are connected in series, the maximum accelerating voltage on the Finkelstein source is + 10 kV, while on the Nier source it is just + 5 kV. The maximum possible energy of a singly charged bombarding ion is 5 keV. This particular arrangement could have been avoided by electrically connecting the bombarding beam accelerating electrode to the target but the added trouble of insulating this electrode and shielding the beam from the grounded housings prohibited the change. One important advantage of the present means of connection is that a higher ion extraction potential is available with the same supplies. The ion current obtainable from the source is much greater at 10 kV than at 5 kV. Focusing effects, as noticed by variation in the bombarding beam current, are observed for particular combinations of the accelerating potential applied to each This effect is partially overcome, however, by source. readjusting the total emission current in the Finkelstein source to keep the target ion current constant. Two collectors, a retractable Faraday cup and a 20-stage electron multiplier, provide a large dynamic ion current measurement range. A vibrating reed electrometer and strip chart recorder record the output from either detector. To observe sputtered neutrals without interference from the background gases a synchronous source-detector system
265
Mass Spectrometer for the Study of Sputtering With this system sputtered neutral peaks was devised. orders of magnitude smaller than the residual background peaks could be measured with no interference from the latter. The synchronous source-detector system is extremely simple, both in construction and operation. It consists of two electriThe first cal choppers and a wide-band a.c. amplifier. chopper pulses the bombarding ion beam by switching the Finkelstein source on and off. The Mier source is left on continuously so the ion current leaving that source has a d.c. component due to the ionized background gases, and an a.c. component due to the sputtered neutrals. A retarding potential on the target with respect to the Nier source suppresses sputtered positive ions. The resulting ion beam is analyzed and then detected with the electron multiplier in normal fashion. The output current from the multiplier is fed into The signal from the amplifier, which the a.c. amplifier. consists only of the ionized sputtered neutrals, is mechanically rectified with the second chopper and then recorded. The signal currents obtained with the synchronous sourcedetector system are shown schematically in Fig. 3.
o-J
I
ANALYZED
BEAM
+
!_1n
n
}
IONIZED N&i:: DC
TIME
E!ACKGRO”ND
CURRENT CURRENT
-
FIG. 3. Sketch of current waveforms in different sections of the instrument when synchronous source-detector system is in use
Performance
A nominal mass range of l-300 amu is obtainable with the 6600 Gauss electromagnet and the variable 5000 V Nier source accelerating potential supply. Mass scanning is done magnetically by sweeping the magnet current manually or automatically. The overall regulation of the instrument measured with the beam halfway in the collector slit is better than l/30,000, including beam intensity fluctuations. When used as a standard gas analyzer with .005 in. source slits and .012 in. collector slit the resolving power of the instrument is somewhat greater than 250 as can be seen in the mercury spectrum in Fig. 4. The sensitivity under these conditions is approximately IO-4A of collected ion current per torr of gas pressure in the source region, when source magnets are used. With the multiplier, ion currents as small as 1 x lo-1sA can be detected. When the multiplier was installed in the instrument its gain was stable and of the order of 107. However, when the spectrum was scanned for sputtered neutrals, the conversion dynode became poisoned from intermittent exposures to the
264
2b2 2bl 2bo
I49
Al3
FIG. 4. Magnetic spectrum scan of mercury isotopes 198 to 204 with instrument operated as a residual gas analyzer. Accelerating voltage 2lOOV. source slits width of ,005 in. collector slit width of ,012 in.
large ion currents resulting from the argon gas ionized in the Nier source. A subsequent drop-off in gain of a factor of 100 was observed for cumulative exposure times of 5-10 min. at ion currents on the order of IO-sA. After repeated exposures over several days time it was observed that if left alone, this problem would clear itself up but high temperature bakeouts speeded the process considerably. The resultant permanent effect of repeated exposure to these large currents was an eventual reduction of the clean multiplier gain to 106. Typical background pressures obtainable in the instrument are in the low 10-9 torr range, the principal gas species being CO and COz. Originally, two “ tin valves “6 were used to isolate the source diffusion pumps from the source housings. The tin reservoirs of these valves were made of inconel, and the plungers were fabricated from cold rolled steel. Besides being unreliable in operation, these valves also trapped large quantities of gas each time they were operated. This gave a persistent CO, CO2 and Hz0 background spectrum even after intense bakeout procedures. This virtual leak at times supplied enough CO2 to keep the source region pressure in the mid 10-s torr range. The tin valves eventually were replaced with allmetal Granville-Phillips valves when large stress-corrosion cracks formed in the walls of the tin reservoirs. If the bombarding ion beam is left off, the instrument functions as a residual gas analyzer and only peaks due to the residual gas in the vacuum system appear in the mass spectrum. When the bombarding ion beam is turned on, peaks due to sputtered ions and sputtered neutrals which are subsequently ionized in the Nier source then appear. The peaks due to the sputtered ions are easily observable since they are usually equal to or greater than the intensity of the residual background peaks. This is not true of the ionized sputtered neutrals, however, even though the sputtered neutrals are much more abundant than the sputtered ions. It has been our general observation for copper and inconel targets that the normal background peaks are two or perhaps three orders of magnitude greater than the sputtered neutral peaks. The reason for this is not fully understood, but it is believed to be at least in part due to a lower ionization probability for the neutrals because of their large initial kinetic energies. Bombarding ion currents of I-3,uA at a current density of 3-lO,uA/cmz could be obtained for 1 x 10-4 torr argon gas pressure in the Finkelstein source with an electron emission of 25mA Under these conditions, the target region pressure
266
A. J.
SMITH, D.
J. MARSHALL, L. A. CAMBEY and E. J. MICHAEL
rises to 1 - 3 x lo-6 torr because of overflow of the unionized argon gas into this region. The instrument was tested using several different polycrystalline OFHC copper and beryllium-copper targets. In some of the runs the inconel target holder was unintentionally exposed to the bombarding beam with the result that some additional data were obtained for this material. The data from these tests have already been reported elsewhere7 so that only the general features will be discussed here as related to instrument operation. The targets were cleaned first by mechanical polishing to remove oxides and then degreased with methanol and rinsed with distilled water. It was found that the secondary Cu+ ion yield varied greatly from one target to the next, although the targets were obtained from the same source and received similar preparatory treatments before sputtering. The maximum initial sputtered Cuf ion currents ranged from 1 x IO-t2A to 5 x lo-1sA for the group of targets used. Typically, currents of lo-13A to lo-i4A were obtained from a new target. It was observed that the sputtered ion current intensity always decreased gradually by about 25 per cent from its maximum initial value to a steady value after the first 5-10 min of sputtering.
FIG. 5. Magnetic spectrum scan of sputtered ions obtained from OFHC copper target and inconel target holder with 3 ke V Ar+ ions. Bombarding ion current 0.5 PA, Nier source accelerating voltage 3 kV
A typical spectrum scan is shown in Fig. 5. The low mass side of each peak is well defined while the up mass side is smeared out so that there is a great deal of overlapping among adjacent peaks. Since the peaks obtained when the instrument is operated as a residual gas analyzer are very narrow and well-resolved, the broadened peaks indicate that the sputtered ions possess an energy spread. A retarding potential on the target showed that this indeed was the case and that over 1 per cent of the ions had initial energies in excess of 350 eV. From spectrum scans taken at different Nier source accelerating potentials it was observed that the degree of overlapping among the peaks increased or decreased as the source. potential was decreased or increased, while the energy spread of the ions remained constant. This is what one would expect to observe if the spread in energy was not occurring in the Nier source. Further, the ions making up
the high mass side of the peak could only have been displaced towards higher mass numbers if they had energies in excess of the energy they gained by acceleration in the Nier source. This is further evidence that the Nier source is not responsible for the energy spread and that the ions must be ejected from the target with this energy. The spread in energy of the bombarding ions was not determined, but for singly charged ions the maximum spread could only be 150 eV, the source electron accelerating potential. Electron accelerating potentials differing by as much as 50 V were tried, with no noticeable change in the energy spread of the sputtered ions. Preliminary checks made on the effect of target temperature and angle of incidence were not conclusive. Background contamination has a large effect on the yield of sputtered ions; this is especially noticed after heating the target. A slight discoloration of the copper targets is noticed after heating to 600°C while the background pressure is in the low 10-7 torr range. Peaks due to CuOz+ sputtered ions equal to about 10 per cent of the 63Cu+ peak height appear after this treatment. A search was made for Cu20+ as well as fragments of these ions, but none were found. The intensity of the ion currents of sputtered neutrals is so low that even with the synchronous source-detector and Nier source magnet the total electron current emission from the filament in the Nier source had to be increased to 30 mA to produce a 63Cu peak ion current of 2 x lo-isA. At this ionizing current the background peaks are very large in comparison to the peaks of interest, but they are effectively suppressed with the synchronous source-detector system. In contrast to the broad peaks characteristic of the sputtered ions, the peaks due to the sputtered neutrals are much narrower and well-resolved, somewhat similar to the background peaks, but not quite as narrow. On the basis of the peak width it is estimated that the collected sputtered neutrals have an energy spread of 1 eV or less. However, the initial energy and energy spread of sputtered neutrals is much greater than thiss. Since the ionization efficiency of the Nier source varies as (E)-‘, where E is the energy of a sputtered neutral atom, the more energetic particles are strongly discriminated against.
Recommendations
From the experience gained with this system several recommendations can be made to improve its overall performance. The detection sensitivity for sputtered neutrals can be increased by increasing the ionization efficiency in the second source, increasing the bombarding ion current intensity, and improving the operation of the synchronous sourcedetector system. Replacing the Nier type source with a Weiss9 ionizer should make a significant difference in ionization efficiency for the sputtered neutrals. Presently, only a fraction of the ion beam leaving the Finkelstein source is available for bombarding the target since it is not focused and only collimated to confine it to the target area. A lens system here would be an obvious improve-
Mass
Spectrometer
obtainable.
instrument technician, Dale Wilson, now at the University of Maryland, for his help in assembling and checking out the instrument.
; thus it is essential that a double focusing instrument be used when working with them.
References
greater should be
Acknowledgements
The authors gratefully acknowledge the help of A. L. Southern, of the Oak Ridge National Laboratory, who wrote the original instrument specifications. Thanks are also due to the various individuals of Nuclide Corporation for their assistance. In particular we wish to thank T. J. Eskew for his many electronic circuit designs, L. F. Herzog for his suggestions and criticisms, R. J. Yinger for his work with the “ tin ” valves, and B. R. F. Kendall for his assistance with some of the vacuum system problems. Thanks are also due to the
1 R.E. Honig, J. Appl. Phys., 29, (1958), 549. R. E. Honig, Advances in Mass Spectrometry, 2, 25, Pergamon Press, New York (1963). 2 R. C. Bradley, J. Appl. Phys., 30, (1959), 1. R. C. Bradley, A. Arking and D. S. Beers, J. Chem. Phys., 33, (1960) 764. R. C. Bradley and E. Ruedl, J. Appl. Phys., 33, (1962), 880. 3 A. Theodore Finkelstein, Rev. Sci. Inst., 11, (1940), 94. 4 0. Almen and K. 0. Nielsen, Nucl. Inst., 1, (1957), 302. 5 A. 0. C. Nier, Rev. Sci. Instr., 18, (1947), 398. 6 C. M. Haaland. Rev. Sci. Instr.. 30. (1959). 947. 7 A. J. Smith, L.‘A. Cambey and D.‘i. Marshall, J. Appl. Phys., 34, (1963X \-~ ~~,,2489. ~~ 8 R. V. Stuart and G. K. Wehner, Trans. 9 Nat. Vat. Symp., Pergamon Press, London, 1962. 9 Rainer Weiss, Rev. Sci. Instr., 32, (1961), 397.