How to Digitize Spectra from Papers (Raman, IR, XANES)
⚡ Quick Answer
Yes. A spectrum is an XY curve: set X to wavenumber/wavelength/energy and Y to intensity/absorbance in Engauge Digitizer, trace the spectrum, and export CSV. Two gotchas: IR wavenumber axes usually run reversed (4000→400 cm⁻¹ left to right — place axis points honestly and it calibrates fine), and peak positions deserve zoomed-in manual reading since they’re the data that matters most.
Why Digitize Spectra
Literature spectra — Raman, IR, XANES, UV-Vis, NMR — are published as images, but modern materials research wants them as numbers: for machine-learning training sets, for comparing your measurement against a literature reference, or for extracting peak positions a paper discusses but never tabulates. Researchers have noted that literature plots are “not conducive to direct programmatic analysis” — digitizing bridges that gap.
Step-by-Step: Digitizing a Spectrum
- Import a high-resolution figure: Journal PDF at full zoom. Sharp peaks need sharp pixels — a blurry spectrum gives blurry peak positions.
- Set the X axis (watch the direction!): Raman shift usually increases left→right; IR wavenumber conventionally decreases (4000→400 cm⁻¹). Place axis points on labeled ticks and type the true values — Engauge handles reversed axes correctly.
- Set the Y axis: Intensity, absorbance, or transmittance — linear in most cases. Note the units; “arbitrary units” is fine as long as you record it.
- Trace the spectrum: Spectra are usually single clean traces — Segment Fill works well. For peak positions, switch to manual and click each maximum in the zoom window; auto-traced peaks can sit a pixel or two off the true top.
- Export and document: File > Export to CSV. Record the source paper, figure number, and any baseline assumptions — future-you (or your ML pipeline) will need them.
💡 Tip: If the figure shows multiple overlaid spectra, add one curve per spectrum in the Curves List before tracing. Trying to separate them after export is painful; naming them during tracing is free.
Honest Limitations
- Baseline ambiguity: published spectra are often baseline-corrected already; you digitize what’s drawn, not the raw measurement.
- Overlapping peaks: digitizing gives you the envelope, not the deconvolution — fitting overlapping peaks is a separate analysis step.
- Low-resolution figures: peak shoulders and weak features below a few pixels wide can’t be recovered faithfully.
Related Guides
How to Digitize Well Logs
Another reversed-axis case — the same calibration skills.
How to Read Values Off a Graph Accurately
Precision skills for peak positions.
Is Digitized Data Accurate Enough to Publish?
Validation evidence for figure-extracted data.
Frequently Asked Questions
Can I digitize a Raman or IR spectrum with Engauge Digitizer?
Yes. Spectra are XY curves with a twist: the X axis is wavenumber/wavelength (often reversed on IR plots — 4000→400 cm⁻¹ left to right) and Y is intensity or absorbance. Set the axes accordingly, trace the spectrum, and export CSV. Read peak maxima in the zoom window — peak positions are usually the critical data.
How do I handle the reversed X axis on IR spectra?
Place axis points honestly: click 4000 cm⁻¹ on the left and 400 cm⁻¹ on the right and type those values. Engauge calibrates from the values you give it, so a decreasing axis works fine — just like the downward depth axis on well logs.
What about overlapping spectra in one figure?
Add one curve per spectrum in the Curves List and trace each separately, exactly as with multi-curve XY plots. Use distinct colors per spectrum in your notes so the CSV columns stay interpretable.
Can digitized spectra be used for machine learning?
Yes — recovering literature spectra as training data is a growing practice in materials informatics. But document the source figure, resolution, and any baseline treatment: ML models inherit every digitizing artifact, so quality control matters more than quantity.
Ready to try it yourself?
Download Engauge Digitizer free and follow along with this guide.
Download Engauge Digitizer