Gas dynamics of the ICP-MS interface: impact pressure probe measurements of gas flow profiles Terry N. Olney, Wei Chen and D. J. Douglas* Department of Chemistry, University of British Columbia, 2036 Main Mall, Vancouver, BC, Canada V6T 1Z1 Received 6th August 1998, Accepted 13th November 1998 A versatile ICP-vacuum interface was constructed to investigate the gas dynamics and gas flow profiles of the ICP-MS interface. This new interface combines a low interface pressure,0.24 Torr, with the ability to accommodate a variety of skimmer designs at sampler to skimmer spacings up to 80 mm.Plasma beams were extracted by placing a 0.9 or 2.0 mm diameter skimming orifice at distances of 6.7–47 mm behind the sampling orifice. Gas flow profiles of each resulting plasma beam were measured using an impact pressure probe placed at distances of 35–105 mm behind the skimmer. The centreline flux and width of the gas beam were compared with those calculated for an ideally skimmed beam.The results for the 0.9 mm diameter skimmer orifice showed that placing the skimmer at 17.0 and 27.3 mm downstream of the sampler formed the highest intensity and narrowest beams. In contrast, by placing the skimmer closer to the sampler, as in common interface designs, the beam profile is less intense on the centreline and much wider. Using a larger 2.0 mm diameter skimming orifice at 16.5–26.8 mm downstream of the sampler produced a more intense beam than any arrangement using the 0.9 mm skimming orifice.However with the 2.0 mm diameter skimmer, centreline intensities were still about half those of an ideally skimmed beam. These eVects are consistent with the formation of a shock Introduction wave or some other disturbance at the skimmer tip. In the The ion sampling interface that is used on nearly all inductively most extreme case the supersonic expansion stops, a shock coupled plasma mass spectrometer (ICP-MS) systems was first wave forms across the skimmer tip and there is a secondary described by Douglas and French1 for use with a microwave expansion through the skimmer.It is also possible that shock induced plasma. The design is based on molecular beam waves can form inside the skimmer, partially scattering and techniques developed by Campargue.2–6 In this interface, the disrupting an otherwise ideal beam. In this case an attenuated plasma expands from atmospheric pressure into a region at a beam can be formed.pressure of several Torr. A molecular beam skimmer is placed The ion extraction process at the skimmer base is ineYcient. in the free jet expansion and the centreline flow passes through In part, this has been attributed to space charge eVects in the skimmer into a region at a pressure of about 10-4 Torr. which mutual charge repulsion causes the ion beam to ‘blow At this pressure electric fields from the first ion optic elements up’.8b,c,13 However, if a molecular beam does not pass cleanly extract ions from the rarefied plasma. The theory of the gas through the skimmer tip, there will be an additional contridynamics of this interface has been described in detail.7–9 bution to the beam ‘blow-up’.The local density of gas within Formation of a molecular beam was considered desirable the skimmer will be higher and this will lead to increased because the gas would then flow through the skimmer undis- scattering of ions in the extraction process.Thus, if a true turbed. In this case there is a minimum number of collisions beam can be formed it may help improve the eYciency of the of ions and neutrals in the expanding gas and hence the least ion extraction process. Further, Tanner et al.14 have described chance for reaction or recombination. The gas flow down- a three aperture gas dynamic interface that relies on formation stream of the skimmer behaves as though it comes from a of a molecular beam through the skimmer.In this interface, point source very close to the sampling orifice and the density downstream of the skimmer, the beam hits a flat plate where in the beam decreases as 1/x2, where x is the distance from a shock wave is formed. In this shock wave the directed gas the sampler. flow stops and the gas stagnates at a temperature near that of Recent results10–12 suggest that in common ICP-MS the ICP. However, the density in the shock wave is orders of interfaces a molecular beam is not being formed by the magnitude lower than in the atmospheric pressure source.Ions skimmer. Niu and Houk10 measured electron densities in the can then be extracted through a small orifice in the plate under interface with a Langmuir probe. They found that in the near collisionless conditions. The total ion current is reduced region between the sampler and skimmer the electron density but, because there are no collisions in the ion extraction decreased as 1/x2, as expected for a free jet.However, down- region, scattering is eliminated and a brighter ion source can stream of the skimmer the centreline electron density decreased be formed. Ying and Douglas15 used this approach for highmuch more rapidly than that of an ideally skimmed beam. resolution quadrupole ICP-MS where the brightest possible Also, laser induced fluorescence measurements11 have shown ion source is required. Tanner et al.14 found that measured a broader distribution of ions and neutrals than expected for ion intensities from this source were less than expected from a skimmed beam.Preliminary impact pressure probe measure- an ideal beam forming a shock wave on the plate. This again ments have shown lower gas densities and a wider gas flow can be attributed to formation of an attenuated or perturbed beam with a conventional sampler–skimmer arrangement. profile than expected for a beam behind the skimmer.12 J. Anal. At. Spectrom., 1999, 14, 9–17 9In the ICP interface the plasma expansion from atmospheric The probe was moved across the gas flow at diVerent distances from the sampler and skimmer.Calibration of the probe with pressure through the sampler forms a free jet with supersonic flow, terminating at the Mach disk.8 The distance (xM) from a free jet showed that the measured pressure was linearly proportional to the gas flux into the probe within a few the sampler to the Mach disk is given by per cent.The gas flow profiles were measured for an interface xM Do =0.67SP0 P1 (1) arrangement with a first stage pressure, P1, of 3.3 Torr and a sampler–skimmer spacing of 6.4 mm. This interface is similar where Do is the sampling orifice diameter and P0 and P1 are to that used on commercial ICP-MS systems and is referred the source and interface background pressures, respectively to here as a ‘conventional’ interface. Furthermore, we con- (see Table 1 for a summary of symbol definitions).In most structed a new interface in which the pump speed on the current ICP-MS interfaces the pump speed on the region region behind the sampler has been increased to about 110 l s-1 behind the sampler is about 5–15 l s-1. This gives a back- in order to lower the interface pressure, P1, to 0.24 Torr. This ground pressure, P1, of about 2–5 Torr, which, in the absence causes the Mach disk to move to 37.7Do (approximately of a skimmer, gives rise to a Mach disk at a distance 8–13Do 42 mm with a 1.12 mm diameter sampler) downstream of the downstream of the sampler (where Do is often about 1 mm).sampler. Hence, the skimmer can be placed much further For beam formation the skimmer must be placed in the downstream from the sampler while remaining inside the free directed flow upstream of the Mach disk. The relatively small jet. Measurements of gas flow profiles under these conditions sampler–skimmer separation (6–12 mm) in standard ICP allowed us to investigate the eVects of moving the skimmer interfaces means that the gas density is relatively high at the further away from the sampler into lower gas density regions skimmer tip and this may give rise to a disturbance in front of the free jet. In addition, a skimmer with a larger diameter of the skimmer.Furthermore, the close proximity of the Mach orifice could be placed in the region of lower gas density in disk to the sampler leaves little room to explore the eVect of the free jet and still produce a suYciently low gas flow such moving the skimmer to diVerent positions in the free jet.that high vacuum (about 10-4 Torr) could be maintained In this work we measured gas flow profiles downstream of behind the skimmer with a modest pump speed. For comparia skimmer using an impact pressure probe. The probe consisted son, flow profiles were measured with the skimmer replaced of a small 0.254 mm diameter orifice in a flat plate at the end by an orifice in a flat plate.Under these conditions no beam of a 6.35 mm od tube that was connected to a pressure gauge. is expected. The gas flow profiles were characterised by (i) their absolute Table 1 Definitions of symbols centreline flux compared with that expected for an ideal beam and (ii) their measured full width at half maximum compared Symbol Units Description with the geometric width of an ideal beam. It was necessary to correct the measured impact pressure for scattering between x mm Distance downstream from the source the skimmer and gas probe to obtain a true measure of the xM mm Distance of the Mach disk downstream from the source beam intensity produced by the skimmer. The results show xs mm Sampler–skimmer separation that the skimming of the free jet using a small diameter xp mm Sampler–impact probe separation on the skimmer (0.9 mm) produces a highly perturbed beam for all centreline sampler–skimmer separations (xs).This eVect is particularly xsk–p mm Skimmer–impact probe separation on the pronounced for separations commonly used in conventional centreline interface designs (xs12 mm).In contrast, the larger diameter Do mm Sampler orifice diameter Ds mm Skimmer orifice diameter skimmer (2.0 mm) produces beams much closer to ideal for P0 Torr Source pressure 16.5 mmxs<xM mm. However, even with the 2.0 mm diam- P1 Torr Interface pressure eter skimmer orifice the centreline intensity of the beam from Pi Torr Impact probe pressure the skimmer is about half that of an ideally skimmed beam.PMax Torr Impact probe pressure on the centreline G1 s-1 Gas flow out of the probe Gx s-1 Gas flow through an orifice a distance x from Experimental the source k erg K-1 Boltzmann constant The ICP and extraction interface used in this work, shown in Kn — Knudsen number Fig. 1, were constructed at the University of British Columbia. n cm-3 Number density The ICP was formed using a centre tapped load coil, quartz n0 cm-3 Gas number density in the source torch (SCP Science, Montreal, PQ, Canada Cat.No. 020– nx cm-3 Gas number density at a distance x from the 050–0710) and Meinhard nebulizer (Meinhard Associates, source v: cm s-1 Average thermal speed Santa Ana, CA, USA), and was operated at 27.7 MHz and vx cm s-1 Flow velocity at a distance x from the source 1000 W. The plasma, auxiliary and nebulizer argon gas flow A cm2 Orifice area rates were 15, 1 and 0 l min-1, respectively.Gas flow profiles m g Molecular or atomic mass were initially recorded using nebulizer flow rates of 0 and Trs — Skimmer transmission 1.0 l min-1. No significant diVerences in the flow profiles were T0 K Source temperature found. Therefore, a nebulizer gas flow rate of 0 l min-1 was Tp K Probe temperature FWHM mm Full width measured at half the maximum used for all the gas flow profile measurements. The plasma peak height expanded into vacuum through a sampler and skimmer.The FWHMG mm Full width of a geometric projection of a sampler, which had a 1.12 mm diameter aperture, was mounted point source at the sample through the in a water cooled plate. Skimmers with orifice diameters of skimmer 0.9 and 2.0 mm, termed the ‘small’ and ‘large’ diameter r mm Distance from source to impact probe skimmers, respectively, were used. The height and the internal r0 mm Free jet source orifice radius w Polar angle between the source and pressure and external half angles of the small diameter skimmer were probe 19.1 mm, 25.8° and 30.3°, respectively, and those of the large l(xs) mm Mean free path at the skimmer tip diameter skimmer were 37.1 mm, 30.3° and 34.3°, respectively.zp mm Distance from the centreline to the pressure The skimmer was mounted in a water cooled flange which probe was positioned on spacer rings such that the distance between 10 J. Anal. At. Spectrom., 1999, 14, 9–17rate given by G1= 1 4 nv:A (4) where n is the number density in the probe and v: is the average thermal speed of the gas in the probe, given by v: =A8kTp pm B1/2 (5) where Tp is the probe temperature.At the steady state, G1=Gx so that 1 4 nv:A=nxvxA (6) and so Fig. 1 Schematic diagram of the interface and impact pressure probe. n=nxvxA4 v: B (7) Ip, moveable impact pressure probe; Sa, sampler; Sk, skimmer; Sm, skimmer mount; Sp, spacers. Typically, the probe reached a steady state pressure within 1 min of being placed in the gas flow.The number density in the probe, and hence the pressure, are proportional to the gas flux in the beam (nxvx). Because vx is essentially constant for the sampler and skimmer (xs) was varied from 6.7 to 47 mm x>5Do,8 the probe pressure is proportional to the gas density with the small diameter skimmer and from 11.5 to 36.8 mm in the beam. Substituting vx from Eqn. (3) and v: from Eqn. (5) with the large diameter skimmer. In addition, a flat plate with gives a 2.0 mm diameter orifice was used in place of the skimmer to model a system which is known not to produce a molecular beam when placed inside of a free jet. The plate was placed between 26.7 and 47.0 mm from the sampler.n nx = vx v: 4 =A5kT0/m 8kTp 16pm B1/2 =A10p T0 TpB1/2 (8) The region between the sampler and skimmer was pumped by a mechanical booster or roots pump (Leybold WSU 501, 168 l s-1; backing rotary pump, Leybold D40B, 13.3 l s-1; Taking T0=Tp gives n/nx=5.605 and taking T0=5000 K and Leybold Vacuum Products, Export, PA, USA) to a pressure Tp=295 K gives n/nx=23.07.of 0.24 Torr, measured with a capacitance manometer Each gas flow profile was obtained from pressure (Baratron Model 122AA, MKS. Instruments, Andover, MA, measurements made at several points as the probe was scanned USA). The eVective pump speed at the interface was 110 l s-1, radially across the gas beam at a fixed centreline distance limited by the conductance of the line to the pump.The region behind the sampler (xp) and skimmer (xsk–p). The gas flow downstream of the skimmer was evacuated with a 345 l s-1 profiles were recorded at several sampler–probe distances, xp, turbomolecular pump (Leybold, Turbovac 361). The backfor each skimmer position xs (small diameter skimmer, xs= ground pressure, which was dependent on the skimmer diam- 6.7, 12.0, 17.0, 27.3, 37.3, 47.0 mm; large diameter skimmer; eter and the sampler–skimmer separation, was measured with xs=11.5, 16.5, 26.8 and 36.8 mm).In addition, gas flow an ionisation gauge (Model 0571, Varian Vacuum Products, profiles were obtained using a flat plate (2.0 mm diameter Lexington, MA, USA). Background pressures varied between orifice) at distances from the sampler of 26.7, 37.0 and 5.0×10-5 and 1.4×10-3 Torr. 47.0 mm. An impact pressure probe tip is shown in the inset of Fig. 1. The flow out of the probe is modelled as eVusive [Eqn.(4)]. The probes had a 0.254 mm diameter orifice in a 0.051 mm This requires that the mean free path, l, within the probe be thick plate mounted on the end of a 6.35 mm od stainless steel substantially greater than the orifice diameter. In measure- tube. The tube was connected to a 0.00–1.00 Torr capacitance ments of gas flow profiles from the interface, the highest probe manometer (Baratron Model 120A, MKS Instruments; manupressure encountered was 30 mTorr. For an assumed, but facturer’s stated accuracy±0.12% of reading). To change the reasonable, collision cross-section for Ar of 50 A° 2, the smallest sampler–probe distance, tubes of diVerent length were used.mean free path is 0.20 cm, which is about eight times greater For an impact probe placed in a free jet at a distance x from than the orifice diameter. Hence the probe remained under the source, the directed gas streams into the probe at a flow eVusive flow conditions in this work. rate Gx given by To verify the calibration of the pressure probes, the ICP Gx=nxvxA (2) and interface were replaced with the flow system of Fig. 2. The pressure of Ar behind a 0.313 mm diameter orifice in a where A is the area of the probe orifice and nx and vx are the 0.051 mm thick plate was controlled by a needle valve. The number density and gas velocity, respectively, in front of the pressure was measured with a 0–100 Torr capacitance man- orifice. The gas velocity of Ar expanding from the source is ometer (Baratron Model 122AA, MKS Instruments; manufac- given by the terminal speed of the expansion: turer’s stated accuracy±0.5% of reading).Argon expanding through the orifice produced a free jet with well defined vx=A5kT0 m B1/2 (3) centreline and angular density distributions. The density on the centreline, nx, at a distance x from the orifice is given by where k is Boltzmann’s constant, T0 is the source temperature (approximately 5000 K in the ICP) and m is the mass of argon.nx n0 =0.161 ADo x B2 (9) Gas can leave the probe and pressure gauge only by flowing back out through the orifice. Provided that the probe pressure is suYciently low, gas leaves the probe in eVusive flow with a where n0 is the density behind the orifice with diameter Do. J. Anal. At. Spectrom., 1999, 14, 9–17 11Fig. 2 Flow apparatus used to calibrate the impact pressure probe. Fig. 4 Free jet profile. Filled circles show the profile of a free jet Ip, impact pressure probe; Nv, needlevalve; So, source orifice; Sp, formed from 101.4 Torr of Ar expanding through a 0.313 mm source spacers. orifice and measured using an impact probe with a 0.254 mm orifice at a distance of 13.0 mm from the source.The dotted line represents the free jet density profile calculated from Eqn. (10). OV the centreline the density is given by16 n(r,w) n0 =B cos2 Apw 2CBAr r0B-2 (10) increases over that of Eqn. (4). Hence, for a given flux into the probe, a lower steady state pressure is reached. Data sets recorded under the same conditions on two diVerent days where w is the polar angle between the centre axis and the (shown on Fig. 3 as open circles and filled squares and as point of interest, r0 is the orifice radius, r is the distance of open inverted triangles and open squares) agreed within 2%. the probe from the source (r2=xP2+zp2, where zp is the The solid line in Fig. 3 is the calculated impact pressure, which distance from the probe to the centreline) and B and C are has a slope of 3.09×10-21 Torr cm2 s.This is within 6% of constants, 0.643 and 1.365, respectively, for Ar. the linear regression of all the combined data sets shown in Fig. 3, which gives a slope of 2.92×10-21 Torr cm2 s. This Results agreement is considered more than adequate for this work. Fig. 4 shows the free jet gas flow profile obtained from Probe calibration impact pressure measurements made at several points as the Fig. 3 shows the measured impact pressure versus the impact probe was scanned radially across the gas beam at a corresponding calculated flux on the centreline of the free jet. fixed centreline distance (xp=13.0 mm) behind the free jet This calibration curve is linear for gas fluxes less than orifice. This profile is compared in Fig. 4 with the density field 1×1019 cm-2 s-1 and the measured free jet flux density agrees of Eqn. (10). In Fig. 4 the density field of Eqn. (10) was with the calculated free jet flux density [Eqns.(8) and (9)] to multiplied by an additional factor cosw to account for the within 2%. For higher fluxes, which give probe pressures decrease in the apparent area of the probe (solid angle) viewed greater than 30 mTorr, the measured pressures are slightly less from the source orifice as the probe moves oV-axis. When the than expected from extrapolation of the linear low pressure probe is moved oV-axis there is a decrease in the apparent data.This may be caused by a slight deviation from purely area because the probe orifice remains perpendicular to the eVusive flow from the probe. As the flow moves from purely centreline. The agreement between the measured profile and eVusive to transition flow (mean free path comparable to the calculated profile is good. The measured FWHM is 0.93 orifice diameter), the flow out of the probe for a given pressure of the FWHM calculated from Eqn. (10). The small diVerence may be due to edge eVects of the probe.As the impact probe is moved oV-axis the finite thickness of the orifice plate can cause scattering of the incoming gas flow so that the flux into the probe is no longer simply described by Eqn. (2). Measurements of the free jet profile were also carried out using tube probes. These were stainless steel tubes (id 2.3 mm, od 3.0 mm) of various lengths (to give diVerent sampler–probe separations). The measured gas flow profiles and centreline impact pressures of the interface were qualitatively similar to the orifice impact probe results.However, after calibration with a free jet it was found that the tube probe response was not linear with the calculated centreline flux of the free jet. Also, the gas flow profile of the free jet measured with the tube probe was much narrower than the profiles either recorded using the orifice impact probe or calculated from Eqn. (10). The gas flow into and out of a tube is complex and cannot be described simply by Eqns.(2) and (4).17 Conventional interface Fig. 3 Measured impact pressure versus calculated free jet flux density. Shown are six individual data sets for sampler orifice–probe separa- Radial gas flow profiles were recorded downstream of the tions of 13.0–58.8 mm. Linear regression (not shown) of a composite small diameter skimmer in an interface typical of many of all the data produced a slope of 2.92×10-21 Torr cm2 s. The solid line represents the impact pressure calculated from Eqns.(8) and (9). ICP-MS instruments. This interface, which employed a 12 J. Anal. At. Spectrom., 1999, 14, 9–17Table 2 Results from gas flow profiles produced using a conventional interface (0.9 mm diameter orifice) xs/mm xp/mm FWHM/FWHMG Centreline/ideal pressure 6.4 41.0 2.8 0.062 51.0 2.8 0.053 71.0 2.8 0.041 112.0 2.8 0.030 sampler, the centreline gas flux of an ideally skimmed beam is the same as the intensity of the free jet measured at the same distance xp from the source.At xp=41.0, 51.0, 71.0 and 112.0 mm, the centreline impact pressures, Pi, measured using the conventional interface were 0.062, 0.053, 0.041 and 0.030 times, respectively, that expected from an ideally skimmed beam, Pideal (Table 2). The ratio of the measured centreline pressure to the pressure of an ideally skimmed beam shows a systematic decrease as xp increases. This decrease is Fig. 5 Gas flow profiles produced from a conventional ICP-MS attributed to scattering of the gas beam between the skimmer interface.Do=1.12 mm; DS=0.9 mm; xs=6.4 mm; P1=3.3 Torr. The and probe, which attenuates the gas flow and causes the probe horizontal lines indicate the FWHM for each curve. pressure to decrease. Scattering is expected to attenuate the beam by an additional factor exp (-nsxsk–p), where n is the 1.12 mm diameter sampler separated by 6.4 mm from a 0.9 mm number density of the background gas and s is the collision diameter skimmer, was evacuated to a pressure of 3.3 Torr cross-section.To determine the intensity of the unscattered using a rotary pump (Leybold S25B, 8.5 l s-1). The resulting beam at the skimmer tip a plot of log (Pi/Pideal) versus xp was flow profiles, recorded at four distances behind the sampler extrapolated back to xp=xs to estimate the ratio that would (xp), are shown in Fig. 5. The full width measured at half of be observed at the skimmer tip without scattering.For the the maximum height (FWHM) of each experimental gas flow data of the conventional interface, this is shown in Fig. 7. In profile is indicated on Fig. 5. In Fig. 6, the experimental this case at xs=xp=6.4 mm the ratio is 0.0836. That is, the FWHMs are compared with the width of a geometric projecskimmer transmits a beam with an intensity 8.36% that of an tion of a point source, through the skimmer, to the probe ideal skimmer. We refer to this as the ‘skimmer transmission,’ position. This width, FWHMG, is given by Trs.The same procedure was used to correct for scattering and to calculate the skimmer transmission for both skimmers FWHMG=Ds Axp xsB (11) and the flat plate at all positions, xs, of the new interface. where Ds is the skimmer orifice diameter. The measured Small diameter skimmer FWHMs were 2.8 times the widths based on the geometric The small diameter skimmer used in the conventional interface projection. The laser-induced fluorescence measurements of was then mounted on the apparatus of Fig. 1 with an interface Duersch et al.11 showed beam widths about twice those pressure of 0.24 Torr. Gas flow profiles were recorded at expected for a skimmed beam. several skimmer–probe separations (xsk–p#35, 45, 65 and 105 mm) for four sampler–skimmer positions (xs=6.7, 17.0, Skimmer transmission 27.3 and 37.3 mm). Additional profiles were recorded at xsk–p= Under conditions of ideal skimming, the centreline intensity 39.0, 49.5, 66.5 and 109.5 mm for xs=12.0 mm and at xsk–p= of the beam is unaVected by the presence of the skimmer.The 34.5 mm for xs=47.0 mm. Typical gas flow profiles recorded gas flows undisturbed through the skimmer and no Mach disk at xs=17.0 mm are shown in Fig. 8 and the FWHM/FWHMG is formed on the centreline. Thus, at a distance xp from the and centreline Pi/Pideal ratios for all of the above experiments are summarised in Table 3. Fig. 7 The measured centreline pressure with the conventional Fig. 6 The FWHM versus impact probe distance behind the sampler interface divided by the pressure for an ideally skimmed beam for diVerent probe positions xp. The straight line is an exponential fit to from the data in Fig. 5. Also shown are the widths (FWHMG) of a geometric projection from the sampler, through the skimmer, to the the data. The intercept at xp=6.4 mm gives the skimmer transmission Trs=0.083. impact probe position. J. Anal. At.Spectrom., 1999, 14, 9–17 13were about 1.6–1.7 times the widths of a geometric projection and the centreline pressures were 0.194–0.128 times the pressures of an ideally skimmed beam. The skimmer transmission increased to Trs=0.23. A comparison of the gas flow profiles obtained at xp=52 mm for various values of xs showed that the FWHM decreased from 38.3 to 6.7 to 4.3 mm as xs was increased from 6.7 to 12.0 to 17.0 mm, respectively. In addition, the centreline probe pressure was observed to increase from 3.4 to 7.1 to 13.6 mTorr as xs was increased from 6.7 to 12.0 to 17.0 mm, respectively.Thus, as the skimmer is pulled back from 6.7 to 17.0 mm behind the sampler, the gas flow profile becomes 8.8 times narrower and the centreline intensity becomes 4.0 times greater. At xs=27.3 mm the gas flow profile FWHM were even narrower, typically 1.3–1.7 times the geometric projected widths. In addition, the measured centreline pressures were about 0.19–0.087 times that of an ideally skimmed beam.The skimmer transmission was Trs=0.30. At a common xp of Fig. 8 Gas flow profiles using the small diameter skimmer. Do= 61.5 mm the FWHM for xs=27.3 mm was narrower than that 1.12 mm; DS=0.9 mm; xs=17.0 mm; P1=0.24 Torr. The horizontal for 17.0 mm, as expected, and the centreline intensities were lines indicate the FWHM for each curve. comparable. Increasing the sampler–skimmer separation to xs=37.3 mm Table 3 Results from gas flow profiles produced using the small produced gas flow profiles with FWHM of 1.4–1.7 times that skimmer (0.9 mm diameter orifice) of the geometric projected width and measured centreline pressures of 0.084–0.047 times the intensities of an ideally xs/mm xp/mm FWHM/FWHMG Centreline/ideal intensity skimmed beam.The skimmer transmission decreased to Trs= 6.7 41.3 5.33 0.051 0.093. When compared with skimmer positions of xs=17.0 52.1 5.47 0.043 and 27.3 mm, at a common distance downstream of the 71.5 4.37 0.039 sampler, the beam profiles were narrower, as expected, but the 111.5 4.00 0.019 centreline intensities dropped to half of the intensity of the 12.0 51.5 1.79 0.090 profiles of the smaller sampler–skimmer separations of 17.0 61.5 1.99 0.085 and 27.3 mm. 78.5 1.83 0.084 121.5 1.85 0.071 When the skimmer was placed downstream of the Mach 17.0 52.1 1.56 0.194 disk, at xs=47.0 mm (recall xM=42 mm), the FWHM again 61.5 1.72 0.177 became twice the width of a geometric projection.Also, the 81.5 1.71 0.155 centreline intensity suddenly dropped to 0.02 times that of an 121.5 1.74 0.128 ideally skimmed beam in the absence of the Mach disk. At a 27.3 61.5 1.33 0.190 common xp of 81.5 mm, when the skimmer was placed down- 71.5 1.44 0.187 91.5 1.52 0.157 stream of the Mach disk, the centreline intensity dropped to 131.5 1.75 0.087 0.25 times the intensity when xs=36.8 mm and 0.12 times the 37.3 71.5 1.39 0.084 intensity when xs=17.0 mm. 81.5 1.65 0.061 101.5 1.55 0.056 141.5 1.99 0.047 Large diameter skimmer 47.0 81.5 1.92 0.020 The small skimmer was replaced with the large diameter skimmer having a 2.0 mm orifice and the interface pressure was maintained at 0.24 Torr. Gas flow profiles were recorded The gas flow profiles recorded for xs=6.7 mm were broader than the profiles obtained using the conventional interface at xsk–p#35, 45, 65 and 105 mm for skimmer position xs= 16.5, 26.8 and 36.8 mm and at xsk–p=41.0, 50.0, 67.0 and arrangement. The FWHM were about 4–5.5 times larger than the geometric projected widths.The centreline pressures for 110.0 mm for xs=11.5 mm. Typical gas flow profiles recorded at xs=16.5 mm are shown in Fig. 9 and the results are xs=6.7 mm at xp=41.3, 52.1, 71.5 and 111.5 mm were 0.051, 0.043, 0.039 and 0.019 times, respectively, the pressures of an summarised in Table 4. At a sampler–skimmer separation of 11.5 mm the gas flow ideally skimmed beam. The skimmer transmission was 0.0825, very similar to that of the conventional interface at xs= profiles had FWHM which were typically only 1.1–1.2 times the width obtained from a geometric projection.The measured 6.4 mm (T=0.0836). The measured FWHM and centreline intensities were about 1.7 and 0.8 times, respectively, those of centreline pressures were 0.278–0.161 times the pressures of an ideally skimmed beam. The skimmer transmission was 0.39. the conventional interface. As the skimmer was moved back to a position of xs= These data can be compared with the profiles produced by the small skimmer at xs=12.0 mm where the measured FWHM 12.0 mm, the FWHM became much narrower than at xs= 6.7 mm, but were still about 1.8–2.0 times the widths based were about 1.8–2.0 times the geometric projection and the skimmer transmission was 0.102.on a geometric projection. The centreline pressures for xs= 12.0 mm were 0.090–0.071 times the intensity of an ideally The gas flow profiles measured for xs=16.5 mm had FWHM which were 1.0–1.1 times the widths of a geometric projection skimmed beam.The skimmer transmission was 0.102. At a common xp of about 52 mm, comparison of the gas flow and had measured centreline pressures which were about 0.365–0.233 times the intensity calculated for an ideally profiles at xs=12.0 mm and 6.7 mm shows that the FWHM becomes 5.5 times narrower and the centreline intensity skimmed beam. The skimmer transmission increased to 0.536.These centreline pressures were about 1.4–1.5 times the press- becomes 2.1 times greater as the skimmer is moved further from the sampler. ures obtained at xs=11.5 mm for the same values of xp. For xs=17.0 mm the centreline impact pressure at xp=51.5 mm At a sampler–skimmer spacing of 17.0 mm the FWHM 14 J. Anal. At. Spectrom., 1999, 14, 9–17Table 5 Results from gas flow profiles produced using the flat plate (2.0 mm diameter orifice) xs/mm xp/mm FWHM/FWHMG Measured/ideal intensity 26.7 61.0 7.77 0.016 71.5 7.60 0.017 91.5 7.22 0.013 132.5 7.56 0.009 37.0 71.5 6.10 0.016 81.5 7.03 0.014 101.5 7.11 0.008 141.5 0.007 47.0 81.5 8.50 0.016 91.5 11.40 0.013 111.5 13.60 0.009 149.0 0.005 Flat plate Fig. 9 Gas flow profiles using the large diameter skimmer. Do= Finally, flow profiles at xsk–p#35, 45, 65 and 105 mm were 1.12 mm; DS=2.0 mm; xs=16.5 mm; P1=0.24 Torr. The horizontal lines indicate the FWHM for each curve. obtained using a flat plate with a 2.0 mm diameter orifice for each sampler–plate separation of 26.7, 37.0 and 47.0 mm.The results are summarised in Table 5. At xs=26.7 mm the gas flow profiles had centreline intensities of 0.013–0.017 times Table 4 Results from gas flow profiles produced using the large that of an ideally skimmed beam. The flat plate transmission skimmer (2.0 mm diameter orifice) was 0.024. This can be compared with the transmission of the xs/mm xp/mm FWHM/FWHMG Measured/ideal intensity small and large skimmers of about 0.3 and 0.5, respectively, at similar xs.The FWHM of the profiles produced using the flat 11.5 52.5 1.12 0.278 plate were 7.2–7.8 times the widths from a geometric 61.5 1.10 0.262 projection. 78.5 1.20 0.240 Moving the flat plate back to a position of xs=37.0 mm 121.5 1.38 0.161 produced gas flow profiles similar to those obtained at xs= 16.5 51.5 1.02 0.415 61.5 1.06 0.365 26.7 mm. The intensities ranged from 0.016–0.007 times that 79.0 1.13 0.326 of an ideally skimmed beam and the transmission was 0.022. 121.5 1.18 0.233 The FWHM, which were 6.1–7.1 times the widths from a 26.8 61.5 1.07 0.354 geometric projection, were marginally smaller than the profiles 71.5 1.07 0.350 produced using the flat plate at xs=26.7 mm. 92.0 1.06 0.327 When the flat plate was placed behind the Mach disk, 131.5 1.18 0.230 36.8 71.5 1.13 0.164 47.0 mm from the sampler, the centreline intensities remained 81.5 1.10 0.134 at about 0.013–0.017 times that of an ideally skimmed 101.5 1.21 0.127 beam.The transmission was found to be Trs=0.030. The 141.0 1.24 0.104 FWHM of the profiles obtained at xp=81.5, 91.5 and 149.5 mmwere 8.5, 11.4 and 13.6 times, respectively, the widths from a geometric projection. The impact pressure at xp= 81.5 mm was 0.51 mTorr, which is similar to the intensity of was 33 mTorr, which is much greater than the centreline 0.48 mTorr obtained by placing the small skimmer behind the impact pressures of 3.4, 7.1 and 13.6 mTorr produced at the Mach disk at xs=47.0 mm.same xp using the small skimmer at xs=6.7, 12.0 and 17.0 mm, respectively. Discussion At a sampler–skimmer separation of 26.8 mm the FWHM of the gas flow profiles were only about 1.1 times the widths The skimmer transmissions for all cases are given in Table 6. of a geometric projection. The centreline pressures were The wide, low intensity, gas flow profiles associated with the 0.354–0.230 times those of an ideally skimmed beam and were conventional interface design show that a highly perturbed approximately equal to those at the same xp for an xs of 17.0 mm.The skimmer transmission was 0.462. At xp= Table 6 Skimmer transmission and inverse Knudsen numbers 61.5 mm the centreline pressure of the gas flow profile using the large skimmer for xs=26.8 mm was 3.4, 2.0 and 2.2 times Interface xs/mm Trs Kn-1 larger than the pressures with the small skimmer for xs=12.0, 17.0 and 27.3 mm, respectively.Conventional 6.4 0.0836 5.47 Small diameter skimmer 6.7 0.0825 5.09 Moving the skimmer further from the sampler, to xs= 12.0 0.102 2.05 36.8 mm, produced gas flow profiles which had FWHM about 17.0 0.230 1.20 1.1–1.2 times larger than the widths from a geometric projec- 27.3 0.304 0.58 tion and centreline intensities 0.164–0.104 times those from 37.3 0.083 0.36 an ideally skimmed beam. The skimmer transmission was Large diameter skimmer 11.5 0.393 4.57 Trs=0.187.The centreline intensities for xs=36.8 mm, at 16.5 0.536 2.66 26.8 0.462 1.29 common values of xp, were only about 0.35 and 0.45 times 36.8 0.187 0.79 those at xs=17.0 and 26.8 mm, respectively. However, the Flat plate 26.7 0.024 1.31 intensities were more than twice the centreline intensities 37.0 0.022 0.79 produced at common xp using the small skimmer at nearly 47.0 0.031 0.55 the same position (xs=37.3 mm). J. Anal. At. Spectrom., 1999, 14, 9–17 15Beijerinck et al.7 have assessed the transmission probability of an expansion through a skimmer (experimental/calculated flow through skimmer) for a Campargue type source as a function of the inverse Knudsen number (Kn-1).The Knudsen number is given by Kn= l(xs) Ds (12) where l(xs) is the mean free path at the skimmer tip. They found that for a skimmer with a 0.5 mm diameter orifice the highest transmission (about 100%) was observed at the lowest inverse Knudsen numbers (Kn-1#1) and that the transmission decreased exponentially with increasing Kn-1 from about 0.8 at Kn-1=2 to about 0.1 at Kn-1=6. The Kn-1 calculated as in ref. 7, for the operating conditions used in this work, are given in Table 6. The transmissions found for the small skim- Fig. 10 Skimmer transmissions for the small (0.9 mm) and large mer show a decrease from Kn-1=5 to 0.6, i.e., for xs from (2.0 mm) skimmers and the flat plate (2.0 mm) at diVerent sam- 6.7 to 27.0 mm, in qualitative agreement with the data of pler–skimmer spacings, xs.The open circle shows the transmission Beijerinck et al.7 At larger xs (i.e., Kn-1< 0.5) background for a conventional interface. xs=6.4 mm; Ds=0.9 mm. penetration due to the encroaching Mach disk causes the transmission to decrease below the expected transmission beam is formed from this type of interface arrangement. Even probability based on Kn-1 calculated for a free jet. For the with a lower interface pressure (i.e., a higher speed pump) a large diameter skimmer the calculated Kn-1 at xs from 11.5 small sampler–skimmer separation produces a highly per- to 26.8 mm are 4.6–1.3.The large skimmer transmissions are turbed beam with essentially the same skimmer transmission. also similar to those reported by Beijerinck et al.7 for the same At larger sampler–skimmer separations and lower interface Kn-1. However, in this work, the small diameter and large pressures the FWHM and centreline pressure measurements diameter skimmers have diVerent transmissions for the same show that the beam quality improves significantly over the Kn-1.Apparently, for the ICP source the transmission depends conventional interface arrangement. Fig. 10 shows the skimmer not just on the inverse Knudsen number but also on other transmission versus xs. For the small skimmer the transmission parameters such as edge eVects at the skimmer. increases as xs is increased from 6.7 to 27 mm, although the For sampler–skimmer positions less than the optimum postransmission is still substantially lower than 1.From this we ition the beam is probably attenuated by shock formation in conclude that the beam formed using the small diameter front of the skimmer tip. For sampler–skimmer positions skimmer is less than ideal at all xs. greater than the optimum position the beam intensity should The FWHM, centreline pressures and skimmer remain unchanged. However, attenuation of the skimmed transmissions measured for beams produced using the large beam will occur as the skimmer approaches the Mach disk.diameter skimmer are generally much closer to those expected The distance of closest approach of the skimmer to the Mach from an ideally skimmed beam. The transmission reaches disk will depend on the finite thickness of the Mach disk and about 0.5 at xs=16.5 and 26.8 mm. At xs=36.8 mm the on the height of the skimmer cone.7,18 When the skimmer is skimmer transmission drops to 0.187.At this xs, the close too close to the Mach disk, penetration of the background proximity of the skimmer to the Mach disk may give rise to gas into the free jet results in scattering of the beam and loss penetration of the background gas into the free jet, leading to in intensity of the skimmed beam. increased beam scattering in front of the skimmer.4,5,7 Photographs of the shock wave around a skimmer in a free Conclusion jet can be found in ref. 18. The flat plate corresponds to a situation where the supersonic The most common interface design produces poor quality gas beams in the ion extraction region. The poor quality arises flow of the free jet forms a shock wave in front of the orifice and this arrangement is known not to form a beam. At xs= from a combination of a high gas density at the skimmer tip and a small sampling orifice. Increasing the distance between 27 mm the centreline pressures were 10% of the pressures with the small diameter skimmer.This demonstrates that although the sampler and skimmer decreases the degree of scattering, but does not completely alleviate the problem because further the beam is perturbed with the small diameter skimmer, some beam-like qualities remain. For example, ion kinetic energies scattering occurs due to the small size of the orifice. The data here show that using a larger diameter skimmer placed at a match those expected from a skimmed beam.8,9,19 It is apparent that a disturbed beam is formed using the small skimming sampler–skimmer spacing greater than most conventional instruments employ eliminates the problem.This is only poss- orifice and modelling the gas flow downstream of the skimmer based on a simple extrapolation of the free jet flow upstream ible by lowering the interface pressure by about an order of magnitude. The resulting beam appears to be closer to the of the skimmer is not possible. The centreline gas flux increased as the small skimmer was moved from xs=6.7 mm to 17.0 mm conditions expected from the ideal skimming of a free jet.In contrast to Langmuir probe and laser induced because the density at the skimmer tip was being lowered. Any disturbance at the skimmer tip, such as shock waves or fluorescence measurements, impact probe measurements are relatively simple to implement and interpret. Impact probe scattering from the edges of the skimmer, produces less of a perturbation of the flow through the skimmer at the lower gas measurements have been used here to confirm the poor beam quality from a conventional interface, to elucidate the problems densities.With the larger diameter skimmer placed at xs= 11.5 mm or greater, the measured FWHM and centreline associated with the conventional interface design and to demonstrate the higher quality beam produced using a large orifice intensities indicate that a much better beam is formed. In this case, the eVects of a lower gas density at larger xs and a larger diameter and a large sampler–skimmer separation.In the future it would be of interest to measure electron and orifice diameter are such that, even if a shock is formed at the edges of the skimmer, the edges are further from the centreline ion densities with the large skimming orifice under conditions that appear to produce a much better quality beam. The and the amount of centreline scattering is reduced. 16 J.Anal. At. Spectrom., 1999, 14, 9–17VCH, New York, 1992, p. 613; (c) S.D. Tanner and D. J. Douglas, implications of these results for improving the eYciency of ion in Inductively Coupled Plasma Mass Spectrometry, ed. sampling are currently under investigation. A. Montaser, Wiley-VCH, New York, 1998. 9 H. Niu and R. S. Houk, Spectrochim. Acta, Part B, 1996, 51, 779. 10 H. Niu and R. S. Houk, Spectrochim. Acta, Part B, 1994, 49, 1283. Acknowledgement 11 B. S. Duersch, Y. Chen, A. Ciocan and P. B. Farnsworth, This work was supported through a Natural Sciences and Spectrochim. Acta, Part B, 1998, 53, 569. 12 D. J. Douglas, paper presented at the Annual Meeting of the Engineering Research Council (NSERC) SCIEX Industrial Federation of Analytical Chemistry and Spectroscopy Societies, Chair. Cincinnati, OH, October 1995, paper 216. 13 (a) G. R. Gilson, D. J. Douglas, J. E. Fulford, K. W. Halligan and References S. D. Tanner, Anal. Chem., 1988, 60, 1472; (b) S. D. Tanner, Spectrochim. Acta, Part B, 1992, 47, 809. 1 D. J. Douglas and J. B. French, Anal. Chem., 1981, 53, 37. 14 (a) S. D. Tanner, L. M. Cousins and D. J. Douglas, Appl. 2 R. Campargue, Rev. Sci. Instrum., 1964, 35, 111. Spectrosc., 1994, 48, 1367; (b) S. D. Tanner, D. J. Douglas and 3 R. Campargue, Entropie, 1969, 30, 15. J. B. French, Appl. Spectrosc., 1994, 48, 1373. 4 R. Campargue, J. Chem. Phys., 1970, 52, 1795. 15 J.-F. Ying and D. J. Douglas, Rapid Commun. Mass Spectrom., 5 R. Campargue, Thesis, Faculty of Sciences, University of Paris, 1996, 10, 649. 1970. 16 J. B. French, in Molecular Beams for Rarefied Gas Dynamic 6 R. Campargue, in Proceedings of the 6th International Symposium Research, ed. W. D. Nelson, AGARDograph 12, NATO-AGARD of Rarefied Gas Dynamics, ed. G. Trilling and H. Y. Wachman, Fluid Dynamics Panel, Paris, 1966. Academic Press, New York, 1969. 17 K. R. Enkenhus, PhD thesis, University of Toronto, 1957. 7 H. C. W. Beijerinck, R. J. F. Van Gerwen, E. R. T. Kerstel, 18 (a) G. E.McMichael and J. B. French, Phys. Fluids, 1966, 9, 1419; J. F. M. Martens, E. J. W. Van Vliembergen, M. R. Th. Smits and (b) A. L. Gray, J. Anal. At. Spectrom., 1989, 4, 371. G. H. Kaashoek, Chem. Phys., 1985, 96, 153. 19 J. E. Fulford and D. J. Douglas, Appl. Spectrosc., 1986, 40, 971. 8 (a) D. J. Douglas and J. B. French, J. Anal. At. Spectrom., 1988, 3, 743; (b) D. J. Douglas, in Inductively Coupled Plasmas in Analytical Spectrometry, ed. A. Montaser and D. W. Golightly, Paper 8/06205F J. Anal. At. Spectrom., 1999, 14, 9–17 17