Three dimensional (3D) ToF-SIMS depth profiles produce large and complex hyperspectral data sets. Interpretation requires that the complexity of these data sets is reduced. Individual ion peaks are often extracted and displayed in 3D or a handful of peaks are plotted in one dimension as a function of depth. This method works well for simple samples, or for samples with a well-defined structure. In the case of complex or unknown samples or for those with important spatial information in the x-y plane, these methods struggle to convey the depth of information captured within the data set. The specific choice of individual ion peaks has the potential to impart user bias and make a significant difference to the interpretation of results as only a tiny fraction of the original data has been considered or displayed.

We have developed an alternative method which considers all or an analyst-specified sub-set of ion peaks within a given depth profile. Unsupervised machine learning, specifically self-organizing maps with relational perspective mapping (SOM-RPM), is employed to create a colour-coded similarity map in which changes in colour are specifically graded to accord with changes in molecular state. The complete 3D depth profile can then be visualized, providing a unique picture of the local and global mass spectral relationships between individual voxels.

The visualisation using a similarity map makes distinct regions of depth profiles easily visible and repeated layers immediately recognisable. Interfacial regions are highlighted as distinct areas allowing for in-depth chemical analysis. SOM- RPM allows the spectra from each region to be extracted for identification and comparison to other regions, making it an excellent technique for exploring unknown or complex samples..

This work will present detailed studies of conducting polymer aerospace coatings and double silver low emissivity coatings, illustrating 3D depth profiles which consider the totality of the mass spectrum at every voxel. Using SOM-RPM to analyse the data yields an intuitive visualisation of 3D depth profiles, highlights any structural flaws such as pinholes, illustrates the degree of interfacial mixing and allows for in depth spectral analysis of selected regions. The SOM-RPM methodology has proven to be a robust technique that offers a substantial advance in this field.

One of the long-term objectives of ToF-SIMS research has been the high resolution 2D and 3D imaging of pharmaceuticals and biomolecules in tissues and biofilms at physiologically relevant concentrations. Although much progress has been made through advances in instrument design and development of cluster ion sources, the technique continues to be limited by low signal-to-noise ratio for many important systems. Improving signal-to-noise, and thereby image contrast, is one of the key challenges needed to expand the useful applications of ToF-SIMS. Although a variety of multivariate analysis (MVA) methods have proven to be effective for improving image contrast in ToF-SIMS, the distribution of important but low intensity ions can be obscured in the MVA analysis, leading to a loss of chemically specific information. Furthermore, the results from MVA methods can often be challenging to interpret.

We have developed an alternative approach in multivariate analysis of mass spectrometry images: inverse maximum signal factors (iMSF) denoising. Standard MVA methods produce scores and loadings which can be difficult to interpret. In contrast, the output from iMSF is a denoised image for each of the original mass peaks. To strengthen the approach, five tests have been developed to validate the denoised images. Results of denoising for 2D and 3D images will be presented. Using this approach, a signal-to-noise improvement of as much as two orders-of-magnitude has been demonstrated. This tool, however, can do more than just make high contrast images. Combining iMSF denoising with Pearson’s correlation coefficients can be used to assist with clear interpretation of the classical MVA results. In some cases, it is even possible to obtain MS/MS-like information that can assist in unambiguous compound identification. iMSF denoising can also be used as a core part of image fusion algorithms that combine mass spectral images with a variety of complementary imaging methods. This tool allows researchers to more quickly visualize, identify and validate key features in mass spectrometry imaging data sets. iMSF denoising is a powerful addition to the suite of image processing techniques available for studying mass spectrometry images.

In OrbiSIMS [1], secondary ions are accelerated by an extraction electrode and using a switching electrode can either pass directly to a time-of-flight (ToF) analyser or be deflected to a transfer system to an Orbitrap^TM analyser. In that case, a quasi-continuous stream of secondary ions is injected into a special ion trap where they revolve around a central spindle shaped electrode and oscillate along it with a frequency inversely proportional to the square root of the mass of the ion. An image charge is created in a pair of outer electrodes and is measured with time. This time-domain transient signal is converted to frequency (and hence mass) domain by a Fourier transform and has signal processing.

At NPL, we have established a metrology (measurement science) programme to study three key interconnected properties of OrbiSIMS: transmission, signal, and noise to improve measurement reproducibility. For example, we took advantage of the stable 30 keV Bi_n⁺ primary ion beam to report measurements of noise across a range of ion intensities and created a statistical model that was used to develop a data scaling strategy that accounts for non-uniform noise across a mass spectrum and has important implications for multivariate statistical analysis methods such as principal component analysis (PCA) [2]. As part of the transmission study, we conducted a systematic assessment of two key parameters, the target potential, V_T, and the collision cell pressure, P, in the transfer optics on the transmitted secondary ion intensities [3]. We revealed a complex behaviour, indicating the possibility for additional separation of ions based on their shape, stability, and kinetics of formation [4].

We use this measurement base to assess linearity of the signal intensity scale. We systematically varied the number of secondary ions sent to the Orbitrap and extracted metrics showing that non-linearity in signal arises from space-charge effects in the trap and is mass-dependant. These findings have direct implication on applications including depth profiling and imaging.

[1] M. K. Passarelli et al., “The 3D OrbiSIMS—label-free metabolic imaging with subcellular lateral resolution and high mass-resolving power,” Nat. Methods, no. november, p. nmeth.4504, 2017, doi: 10.1038/nmeth.4504.

[2] M. R. Keenan et al., “In preparation.”

[3] L. Matjacic et al., “OrbiSIMS metrology part I: Optimisation of the target potential and collision cell pressure,” Surf. Interface Anal., no. November 2021, pp. 1–10, 2021, doi: 10.1002/sia.7058.

[4] G. F. Trindade et al., “In preparation.”