Article

Artificial Intelligence (AI) Provides Ease-of-Use in TOF-SIMS Data Reduction

Surface Analysis Spotlight: TOF

by Greg Fisher TOF Scientist

Artificial intelligence (AI) is used increasingly in almost every aspect of life. Physical Electronics has implemented AI to address a customer pain point, the most challenging aspect of mass spectrometry imaging by TOF-SIMS, in data reduction. We employ an AI algorithm that uses a supervised approach for both feature identification and classification. The algorithm provides an AI Score to classify the spectral calibration as high or low confidence, allowing users to identify potential problems with the data.

In the data example below (Figure 1), the AI algorithm reveals in the 71% confidence score that using the “traditional” calibrant peaks does not produce a mass scale calibration of high confidence even though the mass deviation of each calibrant is < 1 mamu. There are a number of reasons for this outcome; perhaps the peak counts are too low, or the peak shape is off due to preferential charging or an interference. What we know is that we need to try another approach. As shown in Figure 2, a calibration using atomic peaks generates a mass scale calibration of high confidence (98%) even though the mass deviation of some calibrants is > 1 mamu. Without the AI, we would always select the calibrants that provide the lowest possible residual; but doing so would not necessarily provide the best possible calibration over the entire analytical mass range.

To get the latest PHI software features, contact your sales or customer service representative.

Figure 1 – The ion peaks of C₂H₃⁺, C₃H₅⁺ and C₄H₇⁺ do not produce a calibration of high confidence, though the residuals are low.

Figure 2 – The ion peaks of Si⁺, K⁺ and Mo⁺ produce a calibration of high confidence, though the residuals are less than ideal and are certainly higher than the residuals generated using the C₂H₃⁺, C₃H₅⁺ and C₄H₇⁺ ion peaks.