FAQs AMS Data Analysis

From Jimenez Group Wiki
Revision as of 08:53, 20 March 2011 by DonnaS (talk | contribs) (→‎Example of interferences: frag_CO2[44])
Jump to: navigation, search

The purpose of this page is to serve as a repository of FAQs of AMS data analysis.

Can you explain in detail how the Frag Table works?

The Frag Table gives the mathematical formulation of the apportionment of unit resolution sticks to aerosol species. The word 'Frag' is shorthand for fragmentation. The AMS measures aerosol components that have been fragmented due to the evaporation and ionization processes within the instrument. This fragmentation happens in a predictable, quantifiable way, and the fragmentation table encapsulates all the information we know about this fragmentation, with the goal of providing estimates of chemical species concentrations.

As a practical matter, the AMS software code performs a mathematical multiplication of the unit resolution difference (Open beam - Closed beam) signal with coefficients determined in the frag table. The frag table itself is a group of 1 dimensional text waves in Igor that is automatically loaded and set to default values in the AMS software. Each row in each frag wave corresponds to an m/z value. The software interprets the entries in these text waves and generates numerical coefficients for the matrix multiplication. The frag table came about primarily from the work of James Allen and is detailed in the paper Allan, J.D., et al., Technical Note: Extraction of Chemically Resolved Mass Spectra from Aerodyne Aerosol Mass Spectrometer Data, Journal of Aerosol Science, 35: 909–922, 2004 and can be downloaded here.

There is no difference between the default fragmentation table for the quadrupole-AMS and the ToF-AMS. Because of the ToF-AMS measures all m/z values in one time step, while the quadrupole-AMS does not, users typically do need to smooth the time series signal in time to examine fragmentation effects.

The default frag table has been optimized for common ambient atmospheric conditions. A few of the fragmentation entries are known to change between instruments, instrument configurations, instrument deployments, instrument tunings, and ambient levels. These few fragmentation table entries need to be adjusted for each data set that is analyzed. The standard set of entries requiring adjustment is given in the field data analysis wiki. Special sampling conditions such as those for non-ambient air mixtures or excessively high conditions ( > 50 ug/m3) often require additional verification.

The Syntax of the frag table

An example of a typical frag table entry is 'frag_air[14]' which has a default value of '14,-frag_nitrate[14]'. This statement should be interpreted as follows: There is a text wave, called frag_air (the wave preloaded with the AMS software) and the [14] describes the fragment, or contribution corresponding to m/z 14. The wave frag_air describes all the fragment contributions to the 'species' we call air. Air is composed primarily of N2, O2, etc. and frag_air describes all the signals we expect to see at m/z 28 (N2), m/z 32 (O2), etc. For any m/z that we do not expect a contribution, i.e. at m/z 100, the entry is blank. All frag table entries use the following syntax:

  • Commas indicate addition of terms
  • Integers indicate the unit resolution stick value
  • Asterisk or '*' indicate a multiplicative coefficient
  • Dash or '-' indicate a subtraction

The default value of frag_air[14] = 14,-frag_nitrate[14] means that the amount of air that we see at m/z 14 is the total amount at 14 less the amount of aerosol nitrate we see at m/z 14. At m/z 14 we expect to see most of the signal to be N+. This N+ can be due to N2 fragmenting and a very small amount of N+ can be due to aerosol nitrate, NO3, being fragmented. The fact that 'N+' is not mentioned explicitly, and others like it, is a cause some bewilderment. But most frag entries can be similarly translated to their chemical formula analogs. This default also highlights the recursive nature of frag table entries: one entry refers to another, which in turn refers to another, etc. However, the software relies on the fact that the recursion stops at some point, and the software will not give results if this condition is not met. This recursive feature also means that changes to one entry may have an impact on other values. For this reason the frag table should not be edited or changed casually. The few frag entries that are requires to be examined and changed for any data set are listed in the field data analysis wiki

When one does change a frag table entry, a common and useful practice is to keep a copy of the original entry, but preface it with a 0* (zero times) so that it won't have any mathematical effect. For example the deafult O_16[16] entry is 0.353*frag_air[14]. To change this frag entry and retain a text copy of the original one could change this entry to be 0.34*frag_air[14], 0*0.353*frag_air[14], (where user has determined that the correct coefficient is 0.34, see below).

Aside from the main AMS aerosol species of organics, nitrate, sulfate, chloride, and ammonium, there are other frag_X waves, such as frag_air, frag_O16, frag_CO2, frag_K, frag_PAH, etc. These waves are used in helping to apportion the aerosol species by specifying or highlighting these contributions.

Types of frag table entries

The frag table entries accounts for isotopes, interferences from other species, and empirically-measured fragmentation. For new or beginning users, the array of values within the frag table can be overwhelming, but all entries are due to three types of contributions:

  • isotopes (one entry is a isotopic fraction of another),
  • interferences (several species contribute to the same m/z) and
  • fragmentation (the breaking up of the non-ionized entity into patterns of often smaller ions).

Example of isotopes: frag_K[41] from frag_K[39]

The default frag_K has only two contributions (two non-blank entries): frag_K[39]=39 and 0.0722*frag_K[39]. This means that all of the signal at m/z 39 is apportioned to potassium. A small fraction, the isotopic fraction 41K to 39K should exist at m/z 41. The naturally occurring fractional percents of K are: 41K=93.2581, 40K=0.0117 and 39K=6.7302. To calculate the amount of 41K we should see from 39K we get 6.7302/93.2581 = 0.0721675 ~=0.0722. The frag table does not consider the isotope for 40K because it is thought to be too small to matter.

As an aside, for almost all cases, the potassium signal is due to an instrumental artifact and is not due to aerosol. For most cases the assumption that all of the signal at 39 is due to K is an approximation; if the instrument is a V/W ToF, one should examine masses at m/z 39 to test this assumption. When examining m/z 39 in Pika, it one should be aware that surface ionization will show a different peak shape than for ionization of atmospheric aerosols.

Example of interferences: frag_CO2[44]

The AMS measurement has an interference due to gas phase signal. Because we know gas phase atmospheric composition well, we can subtract the gas phase contribution from the total signal to get an aerosol-only measurement. Frag_CO2 indicates the amount of gas phase CO2 that was measured. It has been isolated from the frag_air wave to highlight it's importance and necessity of adjustment. The default frag_CO2 wave has only one contribution: frag_CO2[44] = 0.00037*1.12588*1.28*1.14*frag_air[28]. This example also highlights the syntactical use of multiple factors to help remind users of the formula derivation. Frag_air[28] corresponds to N2 signal, the largest signal measured. The default frag table entry indicates that the amount of CO2 expected is a linear multiple of the amount of N2 that is measured. The factor 1.12588*1.28*1.14 are factors that account for the relative ionization efficiency of CO2 (w.r.t. to nitrate), the ratio of nitrogen/air, and an empirical factor to mainly account for the better focusing of CO2 (vs. N2) on the molecular beam formed at the exit of the lens (plus any other effects such as small differences in ion transmission efficiency). The first factor, 0.00037 roughly corresponds to expected ambient CO2 gas phase levels of 370ppm. The amount of gas phase CO2 detected in the ambient sample can change significantly with location and time of year, and so this is one of the entries that requires the user to adjust the default values to correctly reflect sampling conditions. See below for further details.

Example of estimation of an ion from a typical fragmentation pattern: frag_org[64]

The default entry of frag_organic[64]= 0.5*frag_organic[50],0.5*frag_organic[78]. In this case the amount of organics at 64 is estimated by the amount of organics at m/zs separate by +/- 14 amu, a CH2 fragment. This is based on the empirical observation that for many aerosol organic components, a picket fence of ion intensities appears for ions separated by 14 amu (a CH2 fragment), with intensities decreasing linearly with m/z. The 0.5 coefficients are estimates based on that observation. In general these estimates produce very good results, especially for many ambient datasets., However, significant adjustment may be necessary is for very high organic contributions, such as when sampling fresh biomass burning or combustion exhaust, or for chamber studies without an inorganic seed or NOx.

The standard frag table produces unusual results, should I modify it and how?

  • In some cases additional adjustments are needed for specific experimental conditions. Examples include changes to the frag table of SO4 and NO3 when sampling smog chamber SOA without a seed or NOx (which is known to not have those components), or biomass burning aerosol for which the SO4 and NO3 subtractions break down. If you are in this situation, keep reading. Note that if you have HR data, you can measure some of these things directly and that's always better than modifying the Squirrel (UMR) frag table. This is part of the "Users' Responsibility", which we have discussed at length at Users Meetings etc.
  • The AMS organic mass is calculated by applying the fragmentation matrix to the total spectrum. The fragmentation matrix is defined from the fragmentation table, as discussed in this paper (which you need to read if you want to be able to think intelligently about this topic):
  • The frag table was constant for field data from ~2003 and until 2008. The only one change for field data was proposed by the Aiken et al. paper below, after seeing that there was some CO+ at m/z 28 arising from ambient organics, and that the H2O+ arising from ambient organics was lower than assumed in that Allan et al. 2004 paper. We did this explicitly so that the total organic mass calculated was constant with either the "old" or the "new" frag table. (The shape of the spectrum at 16-18 and 28 does change, but the choice is obvious as soon as you see the spectrum). That paper is (see the section titled "Improved Fragmentation Table for Ambient Organics"):

How do you account for variable CO2 in the sampled air during analysis ?

  • The AMS samples particles 10 million times (1e7) more efficiently than gases, due to the lens and differential pumping system. E.g. N2 is about 1 kg/m3 of air, but in the AMS we measure it as 100 ug/m3 of equivalent aerosol signal, or 1e-7 of the actual concentration.
  • A typical CO2 concentration is of the order of 400 ppm of air, which is of the order of 400 mg/m3, which creates to about 40 ng m-3 of equivalent aerosol signal. This average value is generally subtracted with the Squirrel frag table and the HR frag table, as discussed in this section of the ToF-AMS Field Data Analysis Guide.
  • If CO2 levels in your experiment are pretty constant AND the amount subtracted is a small fraction of the aerosol CO2+ signal, that constant subtraction may be sufficient.
  • If CO2 levels vary a lot in your experiment, or reach much higher values (as when sampling concentrated combustion exhaust), AND/OR if the gas-phase CO2+ subtraction is a substantial (> 10%, or whichever criterion you establish based on your desired precision and accuracy), then you need to implement a time-dependent subtraction based on a CO2 gas-phase measurement. Some guidance for implementing this can be found at the software wiki.
  • If the CO2 gas-phase concentrations are very high (percent) you may be able to measure them with the AMS PToF mode, which also may allow separation of the gas vs particle phase components.
  • For other gas-phase species such as CO, N2O, CH4, at their typical ambient concentrations they produce too small of a signal to see it in the AMS. If you are sampling air with large concentrations of those gases (at least tens of ppmv), then you can apply the same principles described above for gas-phase CO2.
Retrieved from "http://cires1.colorado.edu/jimenez-group/wiki/index.php?title=FAQs_AMS_Data_Analysis&oldid=8342"