Fig. J1 Comparison of estimated SIMS detection limits (in parts per 1012) with expected solar wind abundances for elements Li-Ce
These estimates assume a background count rate of 10-2 /sec for all masses, a sputtering rate of 3 x 1012 atoms/sec, and a Si detection efficiency of 10-3. These parameters may be slightly optimistic. The calculations also assume that instrumental background does not contribute to the background. For volatile species such as oxygen and halogens, instrumental background due to re-adsorption on the sample surface during analysis is by far the largest limitation to detection. However, this is not an issue for non-volatile elements. For elements lighter than Cu, much better than 10% precision may be possible, and measurement of the isotopic abundances of major isotopes to better than 1% might be possible, especially for Mg and Si.
The sensitivity is available to measure C isotope ratios. O is discussed as a special case below. The Rb isotopic abundance may be measurable at the 1-2% level, although cm2 of collector areas would be required. If the solar wind Li/Si ratio is similar to the chondritic value, the Li isotopic composition can be precisely measured. Even if Li is depleted by the factor of 50 suggested by the photospheric Li abundance, by using the concentrator target material it should still be possible to distinguish between spallation (7Li/6Li = 2) and the terrestrial ratio of 12.5. It may not be possible to distinguish higher ratios expected on the basis of preferential 6Li destruction.
This discussion has so far not considered the issue of interferences. During SIMS analysis, molecular clusters and multiply-charged ions cause numerous interferences at multiples and fractions of the abundant species. A Si wafer was analyzed (courtesy of Charles Evans & Assoc.) over the 1-200 amu mass range using an O- beam without energy filtering and at low mass resolution, illustrating the ubiquity of interferences produced by this substrate. A large number of molecular interferences were observed, including SiO, Si2, SiO2, Si2O, SiO3, Si3, Si2O2, SiO4, etc. Molecular clusters tend to have a significantly higher mass than the single atoms of equivalent mass number. Implanted H will also tend to form hydride species. Using present SIMS instruments, most molecular species for A < 60 can be adequately resolved from atoms by analysis at higher mass resolution. In addition, suppression of the molecular interferences can be done by energy filtering; however the accompanying losses in sensitivity probably would not be acceptable for solar wind collector analysis. In many cases mass resolution of M/DM = several thousand may be sufficient to resolve molecular interferences. The most recent SIMS instruments, such as the Cameca 1270 at UCLA and the SHRIMP presently being installed at Stanford, achieve resolutions up to 6,000 with no loss in sensitivity. At this mass resolution, such an instrument also has up to fifty times higher transmission than the older Cameca 3F models. A major goal of the SIMS portion of our work is thus to determine experimental detection limits for elements with molecular ion interferences as a function of instrument mass resolving power for both Ge and Si substrates. Rb, Sr and Ba are the most important elements to test, since Si (and Ge) substrates will have interferences with the O- primary beam needed for their measurement. We presently have Rb implanted samples waiting to be analyzed on the UCLA Cameca 1270 SIMS instrument for this purpose.
Regardless of these advances, collector materials for SIMS analysis need to be chosen which will give the fewest possible interferences in the mass ranges to be studied. In general, substrates of higher atomic mass will give fewer interference peaks, since most of the interference is from ion clusters, while multiply-charged ions are relatively rare. For this reason, Ge is a better substrate than Si for SIMS analysis. In addition, a 133Cs+ primary will give much fewer interferences than a 16O- sputtering beam. Interferences from Ge, Cs, and their clusters are limited to masses 70-76, 133, and 140-152. Another potential substrate of even higher mass is Au foil, which, with a Cs beam, would have no interferences but for Au and Cs themselves. Since both elements are monoisotopic with odd mass numbers, double and even multiply-charged ions would not interfere at integer mass numbers. In Phase B a substantial effort will go into evaluating exactly which substrates can and cannot be used for each element based on analyses of ion implanted samples, including samples loaded with H.
O isotopic analysis by SIMS is not limited by sensitivity but by the
presence of instrumental background due to adsorbed constituents from the
residual gases, primarily CO, of the instrument vacuum system. For the
common Cameca 3F instrument we measured apparent O concentrations of the
order of thousands of atomic ppm (ppma) on high purity Au and Pt, where
for comparison the concentrator target is expected to have about 600 ppma O
in the outer 100 nm. With the state-of-the-art Cameca 1270 instrument at
UCLA we measured much lower apparent concentrations on the same materials,
down to 1 ppma for a single crystal Au sample, presumably reflecting the
much better vacuum of the 1270. However, in achieving the lower detection
limits very high primary Cs ion beam currents were used in order to rapidly
clean the sample surface. With this approach, the sputter rates were such
that the entire solar wind O profile would be sputtered away in seconds.
Such rapid data acquisition is possible in principle with a multicollector
instrument, but it is far better to see the profile develop on a time scale
compatible with human comprehension. What is needed for O isotopic analysis
of collector materials is an instrument with a means of removing adsorbed
gas layers independently of the SIMS primary ion beam. A variety of
techniques are possible but two are mentioned here. One is simply to heat
the sample before and during analysis. A temperature of 100-200o C
may be sufficient to reduce adsorption by several orders of magnitude. A
second promising technique is to use an infrared laser tuned to the primary
vibrational frequency of CO. In collaboration with Charles Evans and Assoc.
we propose to test these schemes on one of their existing SIMS instruments
measuring the factors by which O surface contamination is decreased. For
actual collector material analysis, we propose that an advanced instrument
be built which will have ultra-high vacuum, a surface desorption system,
and at least 3 isotope multicollector detector capabilities.
Mass fractionation corrections are necessary for SIMS isotopic analyses.
Transient fractionation is a special case of fractionation which occurs
until a sputtering steady state is achieved, only affects at most the outer
~10 nm and can be at least partially accounted for by calibration experiments.
The bulk of the solar wind is implanted to depths of ~50 nm, and very little
solar wind should be affected by transient fractionation in the outer 5-10
nm for substrates such as Si and Ge. Even when steady state is achieved,
there is still a mass fractionation which is both element and substrate
specific, so that calibrations must be made on the same element with a
standard of similar composition to the substrate to be analyzed. For O,
implantation is the best way to produce standards which can be calibrated
precisely by other analytical techniques, such as dynamic mass spectrometry.
A major Phase C effort will be made to demonstrate that a suitable oxygen
isotope standard can be produced. Implanting a sufficient quantity of
oxygen for conventional gas mass spectrometric analyses may require several
cm2 of material, since the dosage must be kept low enough to avoid
radiation damage and resulting diffusive losses. However, many ion
implantation facilities, e.g. the one we have used at Hughes, Malibu,
scan the ion beam to produce implants with a high degree of uniformity
over areas of many cm2. Thus, although separate 18O and
16O implants are required, it should be possible to produce high
dose ion implants with a uniform 18O/16O ratio, which will then
be calibrated by GSMS. Because SIMS can analyze submillimeter spots,
the isotopic uniformity of the implant can readily be checked.
One of the recent developments in commercial SIMS instruments promises
a significant advantage for solar wind analysis. Typically SIMS analyses
consist of a single spot of up to ~100 microns diameter where the primary
beam excavates target material for analysis. Since the solar wind
fluence is very low, analyses of many such spots would be required to
acquire a sufficient number of counts for the desired statistical accuracy.
For 17O
this is true even in a concentrated sample of the solar wind. For example,
with a concentration factor of ~40--near the center of the concentrator
target--with a SIMS primary beam size of 100 µm, the combined
statistics from ~50 spots would be needed to achieve the desired two-sigma
uncertainty of 0.1%. A dynamic transfer optics (DTO) package
enables the analysis of a much larger area (~500 µm square)
by rastering the primary beam and tailoring the optics so the area from
which ions are extracted coincides with the beam location at any given
time. This technique could drastically cut the number of SIMS analyses
required for the solar wind and improve the quality of analyses. Rather
than 50 analysis spots, the 0.1% precision could be achieved with only
2-3 analysis spots. The result could be further constrained by additional
analysis spots. The DTO feature is available for testing on several
nearby SIMS instruments, including UCLA and Livermore. The value of DTO
for volatile elements such as O is coupled to the workability of an
independent means of surface desorption as discussed above because the
outer margins of the rastered primary beam are presumably the source of
most of the instrumental background.
3. Resonance Ionization Mass Spectrometry (RIMS)
Resonance ionization mass spectrometry is a relatively new technique
which has only been used for materials analysis in the last decade. The
basic idea is to use laser beams to pump atoms of the desired element to
an excited state. From there they can be easily ionized by one or more
additional photons. Since the intermediate resonant state is unique to
the given element, only atoms of the selected element will be ionized,
eliminating interfering isobars. A second advantage of RIMS is its high
ionization efficiency. RIMS has the possibility of ionizing within the
ballpark of 100% efficiency. In practice, ion extraction geometry and
timing considerations associated with beam pulsing (used to reduce
secondary ion background) and incomplete transmission by the mass
analyzer result in detection efficiencies of 1-20%. However, this is a
great improvement over SIMS where at best 0.1% is achieved. These two
features--elimination of interferences and high ionization efficiency--make
RIMS extremely attractive for low-level detection such as is needed for
solar wind analysis.
Isotopic analysis of a given element is complicated by even-odd isotopic
selectivity (e.g., Spiegel, 1994, Wunderlich et al., 1992).
Nonlinear mass discriminations of up to 50% can occur (e.g., Spiegel
et al., 1991; Wunderlich et al., 1992). At present, gross
isotopic differences can be detected with corrections for both linear and
non-linear mass fractionations by the use of standards. Further work is
being done by a number of groups to study the problem of isotopic selectivity
(e.g., Whitten and Ramsey, 1990; Wunderlich et al., 1992), while some
isotopic studies are already being carried out by RIMS (e.g., Spiegel
et al., 1992; Ma et al., 1995; Nicolussi et al., submitted).
If this problem can be overcome, RIMS would be
capable of providing solar wind isotope ratios for nearly all of the
elements lighter than and including Zn; however, our emphasis at present
is on the use of RIMS for elemental analysis.
Because of the uniqueness of the energy levels of a given element, each
element must be analyzed by a different ionization scheme. The
feasibility of ionizing a specific element depends on 1) the energy level
of the first excited state, 2) the quantal character of accessible
lower-lying states, and 3) the magnitude of the ionization potential
(Hurst et al., 1978). A few elements, namely He, Ne, Ar, and F,
do not have low enough first excited states, or the transition from the
ground state to the lowest excited state is forbidden by quantum
mechanical selection rules. For a number of other elements the
ionization potential is high enough to require three steps. This is
generally true of elements with ionization levels above ~10 eV since
tunable dye lasers, most commonly used for RIMS experiments, operate
over most of the range of wavelengths corresponding to ~1.6-5.7 eV with
frequency doubling (e.g., Hurst et al., 1978). There are numerous
schemes for achieving ionization, some of which can use the same
wavelength for more than one step, reducing the number of laser beams
required.
Atoms must be removed from the sample surface to be ionized by the laser.
Two techniques have been used: (a) sputtering and (b) laser ablation.
Both have excellent depth resolution optimizing the ability to distinguish
surface contamination from implanted solar wind.
With an ion beam as a sputter source using the SARISA IV
instrument at Argonne National Laboratory, we have made
a first-order estimate as to which elements can presently be
analyzed by RIMS. These estimates are based on measured
backgrounds and sensitivity from a three-color RIMS analysis
of Zr in Si (Hansen et al., 1996). In Fig. J3 we compare the estimated
detection limits with the calculated solar wind concentrations from
a two-year exposure averaged over the top 100 nm. The RIMS
detection limits (signal/noise = 3) were calculated using 5 x 105
pulses, a Zr sensitivity of ~3 ppm/count/pulse (useful yield =
1.5%), and the measured backgrounds at the respective masses.
Allowing for future instrument improvements, assuming this
sensitivity for all elements should be conservative. The backgrounds
were measured with lasers set for Zr while sputtering a Si substrate;
however, these backgrounds should be reliable for other elements,
assuming the same substrate, because the background was
dominated (>90%) by non-photoion sources.
Also included in Fig. J3 are two estimates of detection limits
that assume improvements to the present instrument. The
solid line predicts what the detection limits would be if a
high-current ion source (100ľA) was installed in the present
instrument and if the resolution of the present mass spectrometer
of SARISA IV was increased by a factor of ten. Both these
improvements are possible with current technology, and so,
the solid line is our best estimate of the detection limits
for today's state-of-the-art RIMS instrument. In fact, if noise
does not rise linearly with increased current, detection limits
may even be below the solid line. The dashed line predicts
the detection limits that could be reasonably anticipated in the future
when laser repetition rates increase to the point that larger numbers
of averages (108 pulses) are achievable in a reasonable time frame
(a few hours). This is also a conservative estimate since laser technology
is improving continually and no other improvements from the present
state-of-the-art instrument have been incorporated into this estimate.
Detection limits published in the literature appear to be consistent with our predictions.
Ma et al.
(1991) report a two-color detection limit of 50 ppt for Ru, while Pappas
et al. (1989) gave a 9 ppt detection limit for In. The two-color Ru
detection limit is consistent with the detection limit in our three-color
set-up, probably because photoions do not contribute to the background
in thie mass range. The In detection limit is significantly lower, as
expected, since a much larger ion gun was used (50 µA instead
of 2 µA). If one accounts for the increase in sensitivity due to
the larger ion current, the In results also appear to agree with our
estimates in Fig. J3. The detection limit for 56Fe of
20ppt (Pellin et al., 1988), where a three-color ionization
scheme was used, is nearly the same as our prediction in Fig. J3,
indicative that non-photoion background is minimal.
Comparison of Figs. J1 and
J3 is somewhat unfair, since the SIMS detection
limits (Fig. J1) are theoretical based on the latest commercial instrument,
while the RIMS detection limits (Fig. J3) are experimental, obtained from
what is no longer considered a state-of-the-art instrument. (Note also
that Fig. J3 does not contain the abundant, more easily analyzed, elements
below mass 50 and does contain the low abundance nuclei in the mass 100-120
mass range). Fig. J3 represents the SARISA IV instrument operating in its
simplest configuration, and although improvements of 10-1000 in detection
limit are required for solar wind analysis, such factors are possible.
The principal reasons why RIMS detection limits were not lower are:
(a) a relatively low analysis rate, and especially (b) higher general
background levels than SIMS. There are a variety of options for major
improvements in both factors, as discussed below.
(a) RIMS is a relatively efficient atom counter when compared to other
beam techniques due to its high ionization efficiency. From our recent study
(Hansen et al., 1996) where Zr was measured in Si,
assuming a sputtering current of 3 ľA, a Si sputtering yield of 1.5,
and a pulse duration of 800 ns, 4500 Zr atoms were sputtered from
the target during 105 pulses (about an hour with present lasers).
A total of 75 counts were detected in the Zr mass region, 68 above
background, giving a detection efficiency of 1.5%. The efficiency
is much higher than SIMS, but the total number of counts is low.
Since the efficiency is within a factor of 3-5 of the best that could
be expected, improvements in the detection limit are unlikely by increases in
detection efficiency are unlikely. However, negligible sample is consumed for
the analysis conditions just described. Thus, our measured detection
limits are, in part, rate- or time-limited rather than sample-limited.
Two options available for increasing the number of counts detected
are to either increase the amount of material desorbed per pulse or to
improve the duty cycle. As
demonstrated by Pappas et al (1989), detection limits can be
improved by a higher primary ion fluence because the signal is proportional
to the primary ion current.
Signal counting rates are also heavily influenced by laser duty cycle,
and it is likely that laser duty cycles will increase substantially in
the next few years. From these two improvements much higher counting
rates, and thus improved detection limits, can be obtained as shown by the
solid and dashed lines in Fig. J3. Any improvements beyond this will be
dependent upon reducing background levels.
(b) Background sources can be grouped into roughly four categories.
These are 1) background signal due to secondary ions, 2) background
signal dependent on laser light, 3) detector dark current, and 4)
instrumental contamination. We
will examine each of these in turn, but #3 is much less important.
1) Secondary ions as a source of background come from two sources.
Secondary ions directly produced during the sputtering
process is the present major source of noise (70% or the total). These are
suppressed in SARISA by maintaining a high electric potential on the
target during sputtering, then dropping the potential during extraction
of the photoions. Secondary ions then have excess energy and do not
pass the spherical energy analyzers. The rejection ratio in this method
depends both on the voltage of the electrical pulse and its fall time. This
past year an order of magnitude improvement in SARISA's rejection ratio
was obtained by upgrading the target pulse capability from the 300 V / 100 ns
fall time previously available to a 3 kV / 20 ns pulse created with new solid state
high voltage switches (Behlke).
The second source background ions is scattered ions
that reach the detector after several bounces via a path other than the
normal ion flight path. With the recent reduction in secondary ion
background in SARISA, this source of noise is now about 30% of
the total noise. Since these ions reach the detector by some path
other than through the mass spectrometer, it should be possible
to nearly eliminate it by constructing shielding between the detector
and the sample target.
2) Undesired photionization as a source of background is especially difficult to
eliminate since it has the same time signature as signal. It usually
arises from non-resonant ionization of higher concentration, isobaric
species in the desorbing flux (often it is molecular in nature, e.g.,
Si2 interference with Fe). As discussed above, the use
of a large number of low energy photons can dramatically reduce the
background, especially from non-resonant molecular ions. The
three-color resonant ionization scheme used to measure Zr and to
calculate the detection limits in Fig. J3
contributes < 10% to the total background.
3) Dark counts in the detector constitute a third souce of background.
After installing new microchannel plates in the
detector this past year, this noise source is now <1% of the secondary
ion noise. Even lower-noise detectors can be obtained.
4) Instrumental contamination comes from two sources.
One important source of contamination comes from what is usually
called "sample memory effects." Deposition of the element of interest
on the interior walls of the chamber during previous use is the start
of this process, e.g. by analysis of a high concentration target. Then,
during subsequent analysis of the sample this impurity is transported
onto the sample surface or into the laser beam paths. The material
that composes the Faraday cup used to measure primary beam
characteristics and the target material used to align and calibrate
the instrument are the primary memory sources. The SARISA instrumental
design has several patented features to
reduce this contamination. Recent experiments on Ca in Si using
the SARISA instrument have demonstrated that the memory effect caused
by aligning and calibrating the instrument using a pure Ca target can be
eliminated by covering over the sputter deposited Ca by extended
sputtering (1 hour) of a Si target before loading the sample into the
instrument. (Calaway et al. to be published). Similar work with Zr in Si
has resulted in lower measured values than reported in Hansen et al. (1996).
In addition, we propose to machine Faraday cups of different materials
which can be inserted and removed in the same manner as a sample to
reduce this contamination.
The second source of surface contamination occurs prior to sample
loading. The recent calcium study and others at ANL suggest
that contamination of samples during transport and loading into the
SARISA IV instrument must be carefully controlled. To minimize
this, we plan to add a small class 10 handling area around our sample
loading cell.
In summary, background source #4 can be minimized by proper and well
defined contamination avoidance procedures; background source #3 can be
easily minimized with higher quality detectors;
source #2 can essentially be cured by developing multi-step
ionization scheme and using enough lasers to implement them. Therefore,
the near term design challenge is to reduce background source #1.
Laser ablation, as an alternative to ion beam sputtering, has attractive
features. Laser ablation offers control on the atom removal rate.
However, as with ion sputtering, laser desorption can produce non-resonantly
ionized species in greater numbers than the secondary ions from ion beams
if the laser intensity is not carefully controlled. With experience it
should be possible to obtain an adequate ablation rate with minimal amounts
of non-resonant ionization background. Undesired resonantly-excited neutral
atoms are also produced, which can contribute to the background when ionized
by the lasers intended to ionize only the element of interest. Since 1995 we have had
a half-time post-doctoral appointee, resident at Argonne, working specifically on improving
detection limits for both the ion-initiated instrument (SARISA) and the
newer CHARISMA laser desorption instrument for the eventual analyses of
solar wind samples.
Fig. J2 RIMS Ionization Schemes; vertical axis represents energy
Fig. J2 gives a schematic illustration of some of the different paths
to ionization. Scheme 1 shows single photon non-resonant ionization.
Any element with an ionization potential less than the photon energy
(represented by the length of the arrow) will be non-resonantly ionized.
Scheme 2 shows resonant ionization in which two different photons of the
same color are used to boost the atom first to an intermediate excited
state and from there into the ionization continuum. Since photons of this
color are too low in energy to ionize atoms directly, and the resonance
level is unique to the desired element, only atoms of the desired element
are ionized. Two different wavelengths can also be used in the same way,
as shown in scheme 3. Three-step resonant ionizations are shown in schemes
4 and 5. In scheme 4 one of the first two colors is used for the third
step, while in scheme 5 all three photon inputs are from different lasers.
Fig. J3 Comparison of SARISA RIMS detection limits
with expected solar wind abundances.