Calibration of Fast Response Differential Mobility Spectrometers

Transcription

1 Calibration of Fast Response Differential Mobility Spectrometers Jon Symonds Cambustion Ltd, Cambridge, UK

2 Contents Introduction to fast response Differential Mobility Spectrometers (with reference to Cambustion DMS series) Data Processing and Data Inversion Size and Number Calibration of the Charging & Classification System Morphological Effects Calibration for Mass Measurement Sampling and Dilution Systems Traceability and Uncertainty

3 The Need for Fast Response Aerosols can change rapidly, SMPS scan can take 2 minutes. Fast response electrical mobility analysers: Electrical Aerosol Spectrometer, Tartu University / Airel Ltd Cambustion DMS500 TSI EEPS TSI FMPS Cambustion DMS50 DMS Series and EEPS especially aimed at measuring engine exhaust aerosols Adoption of such systems by automotive researchers order of magnitude accuracy no longer good enough But all such instruments currently compromise on sensitivity and spectral resolution over SMPS systems This paper uses the DMS series an example. Both these instruments are available with integrated sampling and dilution systems, so this paper considers the whole picture of calibration.

4 DMS Series Principle of Operation Unipolar diffusion charger Electrometer detection Sizing by charge : drag ratio electrical mobility Similar principle applies to TSI EEPS and FMPS DMS500: 10 Hz data, 200 ms time response, 5 nm to 1 µm or 2.5 µm DMS50: 10 Hz data, 500 ms time response, 5 nm to 560 nm, 12 V operable Fast Response Classification of Fine Aerosols with a Differential Mobility Spectrometer ; Reavell, K. Proc. AGM Aerosol Soc. UK. 2002

5 DMS Data Inversion Charging Model Classifier Model Empirical Calibration Transfer Function Calibrated Transfer Function Least Squares Minimisation with Smoothing 34/38/45 Channel Discrete Spectrum 22 Electrometer Currents Bayesian Algorithm Multi-Lognormal Parameterisation Mass Measured Noise Base Engine Air Flow

6 Particle Charging Modelled response to 100 nm NaCl particle with 1,2,3 charges entering the classifier Unipolar Diffusion Charger Particles gain net, multiple, positive charge from corona discharge Relative response (a.u.) 1 charge 2 charges 3 charges Bigger particles less mechanically mobile, but gain more charge Eventually, large particles become as electrically mobile as small particles: Mobility Inversion Inversion point moved to larger sizes by dropping pressure Cyclone important! Electrical Mobility (a.u.) Detector # Electrical Mobility of Particles in DMS500 1 atm pressure 1/4 atm pressure Size (nm)

7 Instrument Transfer Function Initially generated from Monte Carlo simulation. For a random particle of a given size: Random charge state is selected from a calculated probability distribution for that sized particle Entry point to the classifier randomly selected, and particle s trajectory calculated to predict the landing detection ring and measured current. Repeated across all sizes for many particles Empirically adjusted for every instrument during test using a linear transform Modelled Transfer Function Typical Adjusted Transfer Function Detection Ring Size (nm) Empirical Data Detection Ring Size (nm)

8 Size Calibration: PSL Duke Scientific (now Thermo Scientific) Polystyrene Latex Spheres NIST Traceable, traceability provided by microscopy Large surfactant / impurity mode makes unsuitable for smaller sizes: dn/dlogdp /cc 299nm PSL Spheres - Standard DMS500 Spectral Output 1.20E E+06 Standard Lognormal air Regulator HEPA Collison Nebuliser Diffusion Dryer Neutraliser DMS 8.00E+05 PSL suspension HEPA 6.00E+05 spill 4.00E+05 SURFACTANT PEAK PSL PEAK 2.00E E Dp (nm) Lognormal parameterisation used for ease of analysis & improved apparent spectral resolution

9 Gain Calibration CPC not a primary standard in photometric mode at higher concentrations required by electrometer based fast response instruments. Hence go back to methodology based on that recommended for CPC calibration; the use of a standard electrometer: Liu and Pui (1974). Useful primary standard if can be ensured that each particle is singly charged. 5.00E+04 Comparison of DMS500 and 3022 CPC 4.50E E E E+04 DMS CPC N/cc 2.50E E E E E+03 CPC PHOTOMETRIC MODE CPC COUNT MODE 0.00E Time / s

10 air Pump Regulator DMA / Electrometer-Based Calibration Setup MFC HEPA DMS Filter in Faraday cage Collison Nebuliser H 2 SO 4 (aq) or NaCl (aq) Regulator Neutraliser Dummy Neutraliser Diffusion Dryer HEPA spill +1e Nucleation Apparatus 3081 DMA HEPA Path for H 2 SO 4 spill in Neutraliser (bipolar charger) Cannot ensure single charge for largest particles, therefore rely on extrapolation of model for > 300 nm dn/dlogdp /cc 1.60E+05 Electrometer dn/dlogdp /cc 2.50E+06 Ensure "cut" is made to RHS of spectrum to reduce the chances of larger, multiply charged particles passing the DMA. Weak solution used to reduce mean size of broadband aerosol. Tandem DMA used for very broad initial distributions (e.g. soot) 1.40E E E E E+04 Discrete Lognormal Lognormal narrower, ~same areas 1.50E E E E E E E Dp (nm) 0.00E Dp (nm)

11 Aerosol Sources Spherical Calibration GDI and Nucleation Mode Size Range Aerosol Generation Method Measure Gain? 5 50 nm Sulphuric Acid Nebuliser & Nucleation Rig (next slide) Yes nm Sodium Chloride Nebuliser Yes (or 2500) nm Polystyrene Latex Spheres (PSL) Nebuliser No Soot Calibration Diesel Accumulation Mode Soot from Propane Flame (mini-cast), nm (later )

12 Nucleation Source (H 2 SO 4 ) Air Diffusion Dryer T Heated Tube Acid Soln. In Nebuliser Spill HEPA F HEPA HEPA F T Residence Tube Secondary Dilution Re-nucleation tube dn/dlogdp /cc Size Spectral Density 2.50E E+07 F rotameter T thermocouple 1.50E E E E Dp (nm)

13 Effect of Morphology (1) What size is this? DMS originally calibrated with spherical particles Compare DMA (mobility) sizing with DMS (electrical mobility) sizing for Diesel Agglomerates Test apparatus Vehicle (steady state) DPF Feedgas Exhaust DMS 4:1 Dilution Air Sample Head HEPA Flow DMS 500 Filter Meter 7 slpm Heated 1 Atm 1 slpm 3080 DMA 8 slpm HR diluter OFF

14 200 Effect of Morphology (2) kph 4th Gear 70 kph 5th Gear 140 DMS Mean Diameter DMA Mean Diameter

15 Effect of Morphology (3) DMS500 Mean Particle Charge mean charge NaCl / ejector pump DEHS / peristaltic pump Diesel, 4th Gear 70kph Diesel, 5th Gear 70kph mobility diameter (nm)

16 Effect of Morphology (4) Differences observed with DMA cut soot under a spherical calibration : % % % Gain Difference (versus electrometer) Size Difference (versus DMA) Solution is to empirically calibrate with soot for use with Diesel emissions. Only calibrate accumulation mode of lognormal fit with soot; use this output for solid particle number. % differerence % 50.00% 0.00% % Dp / nm Multiple charging / DMA size range / source particle size range makes this only workable up to ~ 300 nm (sufficient for most engine work)

17 Results after soot calibration 4.5E E E+06 Ave Acc N/cc CPC + VPR Ave DMS Spherical Calibration Ave Acc N/cc DMS Agglomerate Cal mean DMS accumulation ~ CPC +9% mean DMS spherical cal ~ CPC +44% Comparison of DMS with PMP system with Diesel soot; with and without soot calibration. N/cc/s 3.0E E E E E E E+00 Transients Idle Transients Cold Start Warm up Fast Idle Fast Idle 4000 rpm Transients "MOT" However, correlation for Gasoline Direct Injection (GDI) works best for original spherical calibration Need separate calibrations for GDI and Diesel soot in instruments with corona chargers. Particle Number/ Second 6.E+11 5.E+11 4.E+11 Particle Number from DMS500 3.E+11 Particle Number from PMP 2.E+11 1.E+11 0.E Time (s)

18 Calibration Artefacts: Multiple charging For a given size, each possible charge state produces a mobility response on the rings; data inversion deconvolutes this to give particle size spectrum Response to monodisperse (DMA cut ) salt aerosol below: dn/dlogdp /cc Size Spectral Density DMS Response to 30 nm aerosol 2.50E+05 Current / fa 2.5E E E E E e charges +1 e charge Inversion 2.00E E E E E Detection Electrometer # 0.00E Dp (nm) Inversion problem hardest in nm region, beyond that effect is blurred by mobilities becoming closer together with increasing mean charge. For work requiring high accuracy in this region with narrow aerosols (e.g. gas turbine studies), need careful micro-calibration at many sizes in this region to avoid multiple peaks transferring to spectrum

19 Relating Size to Mass: The CPMA Classifies by Charge:Mass ratio, as DMA does for Charge:Drag ratio. Opposing electrical and centrifugal forces Development of APM (K. Ehara et al.), but with inner and outer electrode rotating at different speeds, to create a stable field and higher throughput of particles (Reavell & Rushton)

20 Measuring GDI Particle Mass (1) 1. Start with standard PSL particles of known size & density to calibrate system 2. Select size of particles with DMA ( size bandpass filter ) 3. All particles leaving DMA are charged 4. Measure number of particles (Condensation Particle Counter) 5. Select mass of particles with CPMA ( mass band-pass filter ) 6. Measure number of particles leaving CPMA (with 2 nd CPC) 7. Ratio CPC readings to get penetration whilst varying CPMA voltage transfer function 8. Peak voltage gives peak mass. System is now therefore calibrated with PSL 9. Repeat at the same size points with engine exhaust, peak in transfer function gives particle mass at that size. Penetration 97 nm N 0.4 Modelled TF Voltage / V m c nev = ω r ln 2 2 c c c ( r / r ) 2 1

21 Measuring GDI Particle Mass (2) 1E Particle Mass (kg) 1E E-19 M = D p 2.65 Effective density (kg m -3 ) E Particle Diameter (nm) Particle diameter (nm) Density of particles emitted from a gasoline direct injection engine, J. Symonds, P. Price, P. Williams and R. Stone, ETH Conference on Nanoparticles, 2008 Plot mass versus DMA cut size Gradient on log-log plot gives fractal density factor relating diameter to mass (would be 3 for spherical particles), D f = 2.65 for these GDI particles Particles get less dense as they get bigger, due to open structure Equivalent measurements show D f = 2.3 for Diesel Therefore GDI particles structure less open than Diesel Probably due to infill by volatile material

22 Mass Calibration: Diesel Particles 1 Dyno engine: prototype cal DPF weights 0.1 Tunnel or Exhaust Mass Conc by DMS / µg/cc 0.01 CVS filter paper measurements Production vehicle DPF weights Tunnel Mass Conc by Filter or Exhaust Mass Conc by DPF / µg / cc

23 DMS500 Sampling & Dilution System Required for direct engine exhaust sampling: Primary dilution stops condensation Secondary dilution reduces required cleaning

24 Annular type primary diluter Primary Dilution Raw Sample Flow Dilution Flow 1000nm Cyclone Critical flow restrictor: 4:1 pressure drop Dilute Sample Flow Raw Sample Flow (RSF) = Dilute Sample flow (DSF) Dilution Flow (DF) Dilution Factor (DF) = DSF/RSF = DSF/(DSF DF) Need accurate co-calibration (in series) of both flow DF = 5, 2% difference in flow 8% gain error

25 Sample Line Losses 100% Sample Line Efficiency: Diesel Exhaust Particles (Re = 1600, P = 1 bar, T = 383 K, l = 4 m, i.d. = 4.7 mm) Penetration 90% 80% 70% 60% 50% 40% 30% 20% 10% Experimental Turbulent Laminar 0% Particle diameter nm dn/dlogdp /cc Effect of Sample Line Losses on Typical Diesel Spectrum 7.00E+07 Sample Line Losses Fit Turbulent Model given in Hinds even for laminar flow! 6.00E E+07 No Losses 2.5 m sample line 8 m sample line Symonds, Olfert & Reavell, 2007 Kumar, Fennell, Symonds & Britter, E E E E E Dp (nm)

26 100% 90% 80% Rotating Disc Diluter Calibration Losses thought to be mainly particle diffusion to disc pocket walls: Worse for small particles Worse for higher dilution ratios (disc slower more time to diffuse) Technically possible to apply size dependent correction, but would be impractical for every instrument. Broadband NaCl aerosol used to calibrate and check, size similar to engine soot +5% +5% -20% New R83 relative diluter acceptability limits, applied to DMS diluter 1000 Cambustion Rotating Disc Diluter Calibration Efficiency 70% 60% 50% 40% 30% 20% -30% Efficiency of DMS diluter at DF=100 New R83 mean correction efficiency, applied to DMS diluter Original absolute min efficiency proposal for Reg 83 Actual DF (75 nm NaCl) 100 y = x R 2 = % 0% Size / nm Theoretical DF

27 Traceability Size (PSL) Electron Microscopy Length scale (possibly other nanospheres ) Size (DMA) Physical characteristics of DMA, but ultimately final check from PSL sizing (as in ISO15900:2009) Electron Microscopy Length scale Number Electrometer Known current source (e.g. Keithley 5156) Known voltage source Josephson Junction Standard (relates frequency to voltage) - Caesium Standard Known resistance Quantum Hall Effect Standard Mass flow meter Piston Prover Length scale Clock Caesium standard Pressure / temperature standards Primary dilution traceable to mass flow meters Secondary dilution calibration only requires instrument to be linear in gain

28 Uncertainty Estimates Size from DMA / PSL ~ Coefficient of variance of 5%, 95% CI = ±10% (assume 2k ) Gain (no dilution) ~ CoV of 10%, 95% CI = ±20% Gain (secondary dilution) ~ CoV of 10% (classifier) + CV of 10% (diluter) ~ CoV of 14% (assume independent), or 95% CI = ± 28% Gain (primary & secondary dilution) ~ CoV of 10% (classifier) + CoV of 10% (2 nd diluter) + CoV of 8% (1st diluter, assuming 2% error in flows) ~ CoV 16%, or 95% CI = 32% + sample line losses In practice, generally much better agreement with PMP systems for particle number concentrations than this is achieved. Of course, these systems are also subject to their own uncertainties Mass Uncertainty = size CoV of 5% 3 (each dimension is not independent) ~ 15%, + gain CoV of 10%, so CoV is at least 18% (95% CI ~ 36%) even with no dilution.

29 Summary of DMS Series Calibration Size Calibration: Ultimately traceable to PSL spheres Directly via DMA using NaCl, H 2 SO 4 or Soot aerosols Number Calibration H 2 SO 4 or NaCl or Soot particles charged and size selected with DMA Concentration measured with electrometer and mass flow meter Morphological Effects Need soot calibration for Diesel engines, up to 40% error in number concentration if not used Dilution and Sampling Systems Effect on any measurement system can be significant & sobering..

30 Acknowledgements Team DMS R&D : Kingsley Reavell, Chris Nickolaus, Tim Hands, Nick Collings, Mark Rushton, Andy Livesey, Andrew Ellison & James Burrell (Cambustion) Team DMS Calibration : Justin Hunt & Joe Evans (Cambustion) CPMA Research: Jason Olfert (Universities of Cambridge & Alberta) CPMA Data from GDI engine: Philip Price, Richard Stone (Oxford University), Paul Williams (Manchester University) Sample Line Losses: Prashant Kumar (University of Surrey), Paul Fennell (Imperial College), Rex Britter (University of Cambridge), Jason Olfert Cambustion J6 The Paddocks 347 Cherry Hinton Road Cambridge CB1 8DH United Kingdom jps@cambustion.com