Performance characteristics of a new wide range, fast settling electrometer design for a residual gas analysis mass spectrometer

Size: px

Start display at page:

Download "Performance characteristics of a new wide range, fast settling electrometer design for a residual gas analysis mass spectrometer"

Nora Miller
5 years ago
Views:

1 Performance characteristics of a new wide range, fast settling electrometer design for a residual gas analysis mass spectrometer MKS Spectra Products, January 2010

Design considerations for RGA components The maximum linear outputs from either a micro-channel plate or single channel electron multiplier are at the A level, typically requiring a moderate value

of measurement Typically, signal measurements tend to be required at faster than 1Hz so it is desirable to have lower noise in the measurement.

2 Design considerations for RGA components The maximum linear outputs from either a micro-channel plate or single channel electron multiplier are at the A level, typically requiring a moderate value resistor in the feedback circuit of the electrometer amplifier (in the MKS case 50M ) The noise across this 50M resistor would be ~1e -14 A RMS for a 1 second integration leading to ~e7 dynamic range of measurement Typically, signal measurements tend to be required at faster than 1Hz so it is desirable to have lower noise in the measurement. This can typically be achieved by using a higher resistor value in the electrometer amplifier feedback circuit A 20G resistor would have a 4e-16 A noise RMS for a 1 second integration, however the maximum measurable signal would then only be 0.5 na before the amplifier saturated The traditional approach to signal measurement has been to use gain switching either by switching between scans or within a scan on a channel by channel basis The setting times required to drain saturated circuits these very low current levels when switching is typically not published by the amplifier manufacturers because they are very long (hundreds of ms) For these very low currents a logarithmic amplifier circuit is not feasible due to the poor drift characteristics and the capacitive effects of typical devices used Traditional dual range electrometer, requires either: Two separate scans to cover the complete dynamic range Or long settling time to discharge circuit as relays switch the resistor values

MKS s approach to an electrometer design The signal from the detector is available to two amplifiers, one with a 50M feedback resistor and one with a 20G feedback

feedback resistor amplifier Above the trigger level the diode proportions the signal between both the amplifiers where most of the signal is now measured by the low gain

amplifier reaches saturation on the maximum signal from the detector Below the trigger threshold, the software is programmed to ignore the low gain amplifier (and

3 MKS s approach to an electrometer design The signal from the detector is available to two amplifiers, one with a 50M feedback resistor and one with a 20G feedback resistor Each amplifier is sampled simultaneously by two ADCs A a fast acting diode circuit ensures that detector currents below a trigger level are available to the high feedback resistor amplifier Above the trigger level the diode proportions the signal between both the amplifiers where most of the signal is now measured by the low gain amplifier due to it s much lower resistor value The combination of the two ADC outputs is now the detector signal Due to the proportioning between the amplifiers, neither amplifier reaches saturation on the maximum signal from the detector Below the trigger threshold, the software is programmed to ignore the low gain amplifier (and therefore the noise generated from the 50M resistor) After the signal measurement a charge injection to the diode counters any residual charge resetting it before the next measurement

4 Microvision 2 proprietary technology to reduce settling time from large peaks 5.0 Graph showing detector settling time when jumping to a mass with no signal from different masses within the measurement range of the detector Starting signal 1E7 times higher than noise floor 4.0 Starting ignal 1E6 times higher than noise floor Starting signal 1E5 times higher than noise floor Part s per m illion of full range signal Starting signal 1E4 times higher than noise floor Statistical detection limit at specific measurement times Microvision 2 can jump from the highest peaks to the lowest without any danger of false positive measurements as the settled data for a no signal peak will always be below the theoretical detection limits Measurement starts after ms for smallest signals Settling and completed measurement time, milliseconds Signals close to the noise floor can still be reliably measured, even immediately after measuring very large signals

5 Application example 1: Transfer chamber RGA data quality Most transfer chamber slit valves are open a few seconds, during which time the partial pressure differences between chambers will balance The initial immediate change when the valve opens and the equalised partial pressures give key indicators of leaking mechanisms and pumping performances The detector must be calibrated so the highest partial pressure pulse is still less than the maximum signal so all events can be captured In the following data the maximum signal for Microvision Plus is 1.4E-5mbar but the maximum signal for Microvision 2 is >5e-4mbar Often more than one slit valve can be open at any one time or one valve can open within <1 second of another Adequate time resolution of data is required to correctly isolate a faulty component to save costly diagnostic time during which the tool is down Failures are often air and water leaks but process gas and wafer contaminant issues can also be seen A balance is required, more gases monitored = less time resolution Standard TOOLweb RGA conditions are to use a 1 to 70 mass barchart with a scan time of ~1.7 seconds

XFR chamber monitoring with 1 to 70 Barchart A B C Clipped signal on traditional RGA Model&Settings Mass 1 Mass 2 Mass 4 Mass 18 Mass 20 Mass 28 Mass 32 Mass 36 Mass 38 Mass 40 Mass 44 Mass 68

65% 1.18% 2.14% 5.91% 2.32% 4.04% 0.72% 12.01% 28.89% (Trace A) BKG 5.557E-11 3.5E-08 5.731E-10 3.63E-08 1.102E-10 9.175E-09 2.449E-09 2.828E-08 4.865E-09 4.276E-07 6.87E-10 3.419E-10 2.96E-11 LOD 2.

6 XFR chamber monitoring with 1 to 70 Barchart A B C Clipped signal on traditional RGA Model&Settings Mass 1 Mass 2 Mass 4 Mass 18 Mass 20 Mass 28 Mass 32 Mass 36 Mass 38 Mass 40 Mass 44 Mass 68 Combined mass MV+ (Acc3) Plateau Signal -4.43E E E E E E E E E E E E s %RSD % 1.28% 4.50% 1.65% 1.18% 2.14% 5.91% 2.32% 4.04% 0.72% 12.01% 28.89% (Trace A) BKG 5.557E E E E E E E E E E E E E-11 LOD 2.864E E E E E E E E E E E E E-10 MV2 (Acc3) Plateau Signal 5.245E E E E E E E E E E E E s %RSD 5.04% 1.98% 5.04% 1.15% 0.78% 1.52% 2.96% 1.17% 3.02% 0.47% 8.14% 16.16% (Trace B) BKG 4.028E E E E E E E E E E E E E-12 LOD 5.554E E E E E E E E E E E E E-11 MV2 (Acc1) Plateau Signal 5.463E E E E E E E E E E E E s %RSD 8.80% 3.54% 6.77% 2.77% 0.68% 3.48% 6.80% 3.34% 4.17% 0.46% 12.41% 36.09% (Trace C) BKG 3.993E E E E E E E E E E E E E-12 LOD 9.227E E E E E E E E E E E E E-10 Microvision 2 can run at the same speed as traditional RGA with x6 better detection and similar precision Microvision 2 can run 40% faster than traditional RGA with x3 better detection and similar precision

7 Often a fast event will occur while another slit valve is open Chart showing a pulse witin a transient event for Microvision Plus (1 to 70 amu barchart, 5ms delay, 7.5ms measurement) Chart showing a pulse witin a transient event for Microvision 2 (1 to 70 amu barchart, 10ms delay, 2ms measurement) Partial pressure, mbar 1.00E E E E E E-08 Scan time = 1.10secs DL = 4E-10mbar Mass 4 Mass 28 Mass 32 Mass 36 Mass 40 Partial pressure, mbar 1.00E E E E E E-08 Scan time = 1.01secs DL = 2E-10mbar Mass 4 Mass 28 Mass 32 Mass 36 Mass E E E E E-11 00: : : : : : :51.8 Time, mm:ss E-11 00: : : : : : :51.8 Time, mm:ss.0 By changing to a lower accuracy setting, faster data can be acquired The clipped maximum signal of traditional RGA is now more noticeable Detection limits at 2x worse for traditional RGA If the events are any faster then switching to peak jump mode will offer better time resolution

8 By choosing to measure only the gas peaks of interest Chart showing a pulse witin a transient event for Microvision Plus (10 mass peak jump, 5ms delay, 3.6ms measurement, default zero for each scan) Chart showing a pulse witin a transient event for Microvision 2 (10 mass peak jump, 10ms delay, 2ms measurement, default zero for each scan) 1.00E E E-05 Mass 4 Mass 28 Mass 32 Mass 36 Mass E E E-05 Mass 4 Mass 28 Mass 32 Mass 36 Mass 40 Partial pressure, mbar 1.00E E E-08 Scan time = 0.32secs DL = 1.4E-9mbar Partial pressure, mbar 1.00E E E-08 Scan time = 0.19secs DL = 2.7E-10mbar 1.00E E E E E-11 00: : : : : : :51.8 Time, mm:ss E-11 00: : : : : : :51.8 Time, mm:ss.0 At fast peak jumping acquisition speeds with full dynamic range The clipped maximum signal of traditional RGA is now more noticeable Traditional RGA can not resolve all the changes in partial pressure Microvision 2 is 1.5x faster with 5x better detection limits

9 Conclusion for XFR type data Microvision 2 can capture data for standard TOOLweb RGA applications with six times better detection limits Microvision 2 can capture up to five times more data points leading to better definition of transient events without any loss in detection limit or coverage of gas species Microvision 2 data acquisition parameters can be optimised for up to 10x faster scan times with 8x better detection limits when peak jumping for specific gas species only

10 Application example 2: Transient gas pulses - ALD Atomic layer deposition uses bursts of gas into a large chamber to deposit single molecule thick layers of reactant species onto a substrate During each inlet phase it is preferred that the reactant gas species should be introduced at exactly the correct level to prevent wasting expensive pre-cursor chemicals and extending the time between PM cycles The recipe can be changed to vary the ALD pulse width to introduce the correct level of reactant if the source of the reactant is degrading After each inlet pulse the decay of the reactant gas in the chamber and the monitoring of product gases from the surface reaction give indications of the quality of the ALD process

11 Data acquisition experiment to prove the effectiveness of Microvision 2 for ALD To simulate the effects of an ALD chamber with pulsed gases, a 10 litre vacuum chamber was turbo pumped to 1e-7 mbar using an open ion source Microvision 2 and an ALD valve fed with 0.5mbar of air The ALD valve was pulsed every 10 seconds and the pulse width varied from 80 to 300ms. The effect of the chamber was to broaden the gas pulse to less than 1 second (50% peak height) and reduce the pressure into a range where an open ion source RGA could be used In normal ALD conditions the chamber would be much larger, with much greater pumping capacity and a differentially pumped RGA inlet with a closed ion source would be used (such as a Vision 2000-C).

12 Data quality concerns For ALD there is interest in both the major gas species (the reactants) and trace gas species (reaction byproducts and chamber contaminants) so the ideal RGA is capable of measuring the widest dynamic range with each scan As the species of interest are transient and short lived, the highest quality of information requires a high rate of data collection Change masses quickly with no loss in stability Quickly stabilise on the new signal Quickly switch between gain ranges for the widest measurement range Have a high stability zero so it is possible to run without having to spend a long time on each scan re-zeroing

13 Data collections parameters Peak jump was used to maximise the time spent measuring species of interest Masses 28 (N2), 32 (O2) and 44 (CO2) Masses 36, 38 and 40 Argon isotopes Scan time for six masses was 70ms This is possible as the RF electronics are temperature controlled to allow accurate mass positioning in <1ms The dynamic range at this speed is possible as the detector electronics have a very fast diode based gain switching system for high and low signal For the first data set shown, one zero was taken at the start of the data collection and data collected for 2 hours Baseline changed by <0.5ppm of full scale This is due to having temperature controlled electrometer

14 Example traces for varying ALD pulses

15 Data collected during an 80ms pulse of an ALD valve 1.00E ms 300ms pulse pulse has has broader broader peaks peaks which which are are not not proportional proportional in in peak peak height height to to the the 80ms 80ms pulses pulses A A data data integration integration method method is required to give more is required to give more accurate accurate measurements measurements for for variable variable ALD ALD pulse pulse widths widths Almost Almost 66 decades decades of of signal measurement signal measurement range range Maximum Maximum signal signal 1E 1E-5-5 mbar mbar Noise Noise between between pulses pulses for for mass mass is is 2E 2E mbar mbar Mass E-08 Mass 32 Mass 36 Mass 38 Mass E-09 Mass E E E-12 28: : : : : : : : : : : Time (mm:ss.000) Data collected during a 300ms pulse of an ALD valve 1.00E E E-07 Partial pressure, mbar 1.00E-07 P artial pressure, mbar Example data for 80ms and 300ms ALD pulses 1.00E-06 Mass E-08 Mass 32 Mass 36 Mass 38 Mass E-09 Mass E E E-12 30: : : : : : : Time (mm:ss.000) 31: : : :09.696

16 Comparison of peak shapes for different ALD pulse times and schematic of the data integration method Comparision of O2 measured peak for different ALD valve pulse times in ms. Chart showing the integration technique used for ALD pulse measurement 1.00E E-06 Mass 32 (80ms) 1.40E-06 Mass 32 (100ms) Mass 32 (150ms) Mass 32 (200ms) Mass 32 (80ms) Mass 32 (300ms) Partial pressure, mbar Partial press ure, mbar 1.20E-06 Mass 32 area Background 1.00E E E E E E E E E E+00 00: E-10 00: : : : : Time (mm:ss.000) 00: : : : : : : : : : : : Time (mm.ss.000) Integrating Integrating over over the the peak peak area area and and subtracting subtracting the the baseline baseline offset offset under under the the peak peak will will lead lead to to aa linear linear relationship relationship between between the the measured measured partial partial pressures pressures and and the the ALD ALD pulse pulse times times Integrating Integrating under under the the peak peak is is only only possible possible with with fast fast enough enough data data acquisition, acquisition, aa peak peak must must have have 77 or or more more points points over over the the top top 90% 90% of of the the peak peak for for best best accuracy. accuracy. This This data data has has nine nine points. points.

17 Graph showing pulse peak areas for N2 and O2 with varying pulse widths Graph showing pulse peak areas of 40Ar and CO2 with varying pulse widths 1.60E E-05 Mass 28 Mass E E-05 Mass E E E E-06 Pulse peak area Pulse peak area Mass E E E E E E E E E E Comparison of peak area and peak height for ALD chemical delivery 1.20E E-05 Interated value for several pu lses different different ALD ALD pulse pulse widths widths Comparison Comparison of of peak peak area area vs. vs. peak peak height height for for accurately accurately checking checking gas gas species species Peak height gives non-linear data 200 ALD valve pulse width, ms ALD valve pulse width, ms Comparison Comparison of of peak peak areas areas for for 150 Mass 32 Mass 32 height 1.00E E E E E E E ALD valve pulse width, ms

18 Control of the ALD process Relies on knowing the relative levels of chemicals in each pulse It is critical that peak ratios do not change with pulse width so correct mixture changes can be made if required Chart showing ratios of peak areas for ALD pulses of different width 3.00E+02 Ratio 40/36 Ratio 32/44 Ratio 40/ E+02 Peak area ratios 2.00E E E E E ALD valve pulse width, ms

19 Conclusion for ALD type data Microvision 2 can capture data for multiple peaks on fast transient signals 6 gas partial pressures with no re-zero With 14Hz acquisition rate With almost 6 decades of dynamic range No baseline drift over several hours With linear correlation between ALD pulse width and peak area With equal gas peak ratios for all ALD pulse widths

HPR-30 Vacuum Process Gas Analyser A differentially pumped RGA system for vacuum process monitoring

HPR-30 Vacuum Process Gas Analyser A differentially pumped RGA system for vacuum process monitoring HPR-30 vacuum process gas analyser To analyse processes with high dynamic range operating at pressures