Hi all!
I am currently applying micro-ATR (Bruker Hyperion 2000 microscope) on microscopic, single-cell biological and palaeontological dinoflagellate cysts.
Apart from atmospheric and baseline corrections, Savitzky-Golay smoothing and extended ATR correction (OPUS software, for wavelength-dependent penetration depth), are there any other preprocessing steps you can undertake (and in what order) to improve further qualitative analysis of the data? Is there an 'optimal' protocol in existence?
Are there any differences with macro-ATR data (preprocessing)?
Thanks in advance.
Kind regards,
Pjotr
Maybe this article can help you: Article Using Fourier transform IR spectroscopy to analyze biologica...
Concerning interpretation, I would recommend it together with the following review:
Article The Bouguer-Beer-Lambert Law: Shining Light on the Obscure
Hi Pjotr,
Check out our last tutorial on spectral data analysis: https://www.nature.com/articles/s41596-020-0322-8
There are guidelines for pre-processing and multivariate analysis of this type of data.
Thomas Mayerhöfer, thank you for your suggestions. Baker et al. (2014) mainly talks about macro-ATR measurements. My samples, when squeezed, usually only cover about half of the area of the micro-ATR crystal tip (which is 100 µm in diameter) and might be prone to more pronounced 'physical effects' (not really scattering), due to a non-complete contact with the crystal (and maybe other factors as well?). When pressed, some samples become very thin (hard to say how thin exactly, also depends on the species at hand but possibly < 2 µm), so some of my data might also suffer from contributions from the SiO2 of the measurement substrate (= Au-coated mirror). I assume its hard to correct for these factors, so for smaller species (ca. 20 µm in diameter) clustering multiple specimens to ensure enough area covered and sufficient thickness might be the way to go here?
From literature it seems that ATR-data mostly benefits from SG smoothing, baseline corrections + normalization and/or EMSC-correction. These authors (Article Characterizing Aeroallergens by Infrared Spectroscopy of Fun...
) have used EMSC model parameters for taxonomic classification, which might suggest that some of these 'physical effects' in ATR are systematic and might be traced by EMSC.Very nice and comprehensive Bouguer-Lambert-Beer law overview, thank you for that as well!
Hi Camilo Morais , thank you for your answer. Is the pre-processing approach in this work specifically tuned to ATR data? Or to transmission (or transflection?) data as well? It wasn't fully clear when reading the paper. Thank you in advance.
Kind regards,
Pjotr
Pjotr Meyvisch Generally ATR has a higher resolving power than if the light is incident from air, meaning that the limits of what is large change and samples that are micro-homogeneous under incidence from air/vacuum may become micro-heterogeneous - this might have some influence on the spectra. Certainly establishing good contact is a must for ATR measurements. If your SiO2 substrate is coated by a metal film, a thickness of less than 50 nm is already sufficient to prevent any appearance of the SiO2 modes. Maybe you see a small interference effect, but it should be comparably small. Indeed, removement of all these factors would be quite challenging if not impossible. Clustering samples is in anyway a good idea.
Concerning EMSC correction I am very sceptical. EMSC assumes additivity of absorbance (perfect adherence to Beer's law) and we tried it for comparison when we put together Article Deep Learning for ‘Artefact’ Removal in Infrared Spectroscopy
. Its performance was not adequate, something which was also realized inArticle Extended Multiplicative Signal Correction for Infrared Micro...
. It certainly removes undulating baselines etc. but to speak of a real signal correction is an exaggeration. It was actually never proved that it is able to correct the absorbance signal, e.g. when it is disturbed by interference effects. For basic configurations is also not needed because we can correct the spectra fully analytical. If you ask me it is something that is hyped, but it cannot do magic in contrast to deep learning approaches.
Thomas Mayerhöfer Thank you for your swift and extensive responses. Looking forward to read more of your (future) work.
Concerning the SiO2 substrate contributions and minimal metal film thickness: do you know articles that have studied this? I only seem to find a paper of Bassan et al. (2013)
Article Substrate contributions in micro-ATR of thin samples: Implic...
but it mainly covers the issue of minimal required sample thickness.
Cheers,
Pjotr
Pjotr Meyvisch My pleasure! We have investigated the minimum necessary gold film thickness as a side project of Article Ultra Sensing by Combining Extraordinary Optical Transmissio...
The problem was that usually a gold film thickness of 200 nm was used to assure that light transmittance is zero, but since gold layers grow in a columnar way, this causes poor thickness homogeneity for 200 nm thickness. We found that 50 nm is by far enough. Since (one way) transmittance is then less than 1/1000, you definitely will not see any signal from SiO2. You can also search for the term "skin depth" which is quite low at infrared wavelength for gold, because it is the metal which is closest to being a perfect conductor (skin depth is zero, which means the EM wave does not penetrate the metal).
Best wishes,
Thomas
Many thanks Thomas Mayerhöfer , this is really helpful information.
Cheers,
Pjotr
- Badges
- Science topic