SoFi - SourceFinder


SoFi (SourceFinder) is a program developed at PSI that facilitates source apportionment analysis based on PMF/ME-2 on any ambient data. SoFi comes in two versions, the standard version (freeware) and the Pro version (lisence-based).

SoFi standard

SoFi is intended to facilitate the source apportionment based on Positive Matrix Factorization (PMF) and the multilinear engine (ME-2) of measured ambient data. The software was developed for analysis of ACSM data. However, the algorithm can be applied to any kind of data and therefore, it can be distributed to all ME-2 / PMF users dealing with any kind of measurement data.

The standard version allows to explore and pretreat the PMF input, which includes, but is not limited to, selection of the effective PMF input (exclusion of time points and/or variables), down-weighting of variables and use a robust threshold. With the standard version it is possible to run unconstrained PMF as well as constrained PMF with either the a-value or the f-peak approach. The preparation of the external data can be easily done in SoFi. SoFi offers a very user-friendly PMF data analysis. With a few mouse-clicks, the user can plot time series, profiles, contributions, correlations with variables and externals and much more. More information can be found in the manual SoFi manual.

The latest SoFi standard version can be downloaded here SoFi standard.

SoFi Pro

The SoFi Pro module currently allows for the following extra features:
  • (resampling strategy) bootstrap application on PMF input for the assessment of the statistical error and subsequent analysis using the dynamic criteria-based feature
  • (criteria-based feature) Inspection and selection of PMF runs based on user-defined proxies/tracers
  • (statistics on average) average over several PMF runs and visual inspection of the solution
  • (rolling technique) user-based PMF sub-window moves over the entire PMF input allowing to model changing factor profiles. Especially relevant for long-term SA studies with profiles known to vary over time
  • (relative error scaling) manual and automated application of the C-value, when combining data from two and more instruments and subsequent graphical support when exploring these solutions
  • (additional averaging) hourly and/or daily average in SoFi for externals, PMF input and solution
  • (classes) variables and/or time points can be classified for further analysis, e.g. PMF with data from various stations at the same time or various size-fractions
  • (saving&loading utilities) saving and loading user-specific PMF input and constraints or user-defined criteria, graphical support for the quantification of the PMF error, statistics on the a value constraints.
The following table summarizes the costs for a SoFi license.
  1 PC per year 5 PCs per 1 years 1 PC per 5 years 5 PCs per 5 years
costs per PC and year / € 1000 750 750 500
The latest SoFi Pro version can be downloaded here SoFi Pro.

ME-2 engine

SoFi is based on the ME-2 engine. This download includes the full engine but not the license key. For the key, please contact Pentii Paatero ( The key has to be saved in the ME-2 engine folder: \ME2_engine\.

The latest ME2 folder can be downloaded here ME2 folder.


SoFi requires Igor 6. It has not been tested for Igor 7 or 8 yet and it can also not be guaranteed that it will run error-free with older Igor versions.
The manual for the released SoFi version can be found under the same link (see below).
download of ME2 folder (2016/03/11) here.

download of SoFi (standard and currently also Pro) (vers. 6.6, 2018/11/21)

For more information please contact Francesco Canonaco (


The main SoFi manuscript is a highly cited paper (102 citations, 2018/07/06) with more than 45 papers based on the SoFi software.

The following 3 papers represent the backbone of the SoFi approach and it is recommended to study them prior to the source apportionment analysis.

SoFi, an IGOR-based interface for the efficient use of the generalized multilinear engine (ME-2) for the source apportionment: ME-2 application to aerosol mass spectrometer data Canonaco F, Crippa M, Slowik JG, Baltensperger U, Prévôt ASH, ATMOSPHERIC MEASUREMENT TECHNIQUES 6, 3649 (2013). DOI: 10.5194/amt-6-3649-2013
Organic aerosol components derived from 25 AMS data sets across Europe using a consistent ME-2 based source apportionment approach Crippa, M., Canonaco, F., Lanz, V. A., Äijälä, M., Allan, J. D., Carbone, S., Capes, G., Ceburnis, D., M., D. O., Day, D. A., DeCarlo, P. F., Ehn, M., Eriksson, A., Freney, E., Hildebrandt Ruiz, L., Hillamo, R., Jimenez, J. L., Junninen, H., Kiendler-Scharr, A., Kortelainen, A.-M., Kulmala, M., Laaksonen, A., Mensah, A. A., Mohr, C., Nemitz, E., O'Dowd, C., Ovadnevaite, J., Pandis, S. N., Petäjä, T., Poulain, L., Saarikoski, S., Sellegri, K., Swietlicki, E., Tiitta, P., Worsnop, D. R., Baltensperger, U., and Prevot, A. S. H. ATMOSPHERIC CHEMISTRY AND PHYSICS 14, 6159 (2014). DOI: 10.5194/acp-14-6159-2014
Seasonal differences in oxygenated organic aerosol composition: implications for emissions sources and factor analysis Canonaco F, Slowik JG, Baltensperger U, Prévôt ASH, ATMOSPHERIC CHEMISTRY AND PHYSICS 15, 6993 (2015). DOI: 10.5194/acp-15-6993-2015