Getting decent spectra from crime scene samples is not as clean as the textbooks make it look
I spend most of my time between FTIR and Raman instruments, trying to figure out what someone left behind on a surface they didn't even know they touched. Spectroscopy In Forensic Science sounds like a glamorous intersection until you're actually sitting there at 11pm with a degraded sample and a instrument that needs wavelength calibration. It's real work. Here's how I approach it and where people usually get tripped up. When we talk about spectroscopy in this context, we're generally dealing with two main techniques. Fourier Transform Infrared spectroscopy (FTIR) measures how molecules absorb different wavelengths of infrared light, giving you a structural fingerprint. Raman spectroscopy measures inelastic scattering of photons and complements FTIR by picking up on different vibrational modes. Both are non-destructive, which matters when your sample is a few milligrams of white powder from a ziplock bag found in a suspect's vehicle. There's also mass spectrometry, but that's destructive and a different conversation entirely. We use it when the spectroscopic methods hit a wall, which is more often than the protocol wants you to admit.
The core principle is straightforward: every molecule has a unique interaction pattern with electromagnetic radiation. Your job is to capture that pattern cleanly and match it against a reference library. The part that's not straightforward is what happens between "capture" and "match."
The practical workflow I actually use
First, I assess the sample physically. Particle size, color, texture, substrate. If it's on glass or metal, FTIR transmission or ATR works fine. If it's on porous fabric or a dark surface, you're dealing with fluorescence issues that will wreck a Raman reading before you even start. I've seen analysts skip the visual assessment and go straight to measurement, then spend three hours wondering why their baseline is completely unusable. For FTIR with an ATR accessory, I place the sample directly on the diamond crystal and apply firm, even pressure. Five seconds per scan, sixteen co-added scans minimum. That gives you a signal-to-noise ratio worth trusting. Background collections need to happen fresh, not from yesterday when the humidity was different and the atmospheric CO2 peak was lurking in your spectrum. Raman requires a different mindset. Laser power needs to stay low enough to avoid heating or degrading organic samples but high enough to get a usable signal. I typically start at 50mW or less for unknowns. If the sample fluoresces, I switch to a longer wavelength laser, usually 785nm instead of the standard 532nm. Backscatter geometry helps with opaque samples on dark substrates. Confocal Raman lets me focus through packaging materials sometimes, which has saved me from opening evidence I wasn't supposed to touch yet.
Get the Full Details
A specific problem that almost cost me a case
Two years ago I was analyzing a pressed pill fragment recovered from a car glovebox. The outer coating was relatively clean. The core was heavily compressed and mixed with an excipient I couldn't immediately identify. FTIR gave me a decent match for acetaminophen but the binding agent was a mess. The library search kept suggesting microcrystalline cellulose, but the spectra didn't align perfectly. Peaks were shifted, likely from the compression process changing the crystalline structure. The workaround was straightforward but not obvious if you haven't run into it: I ran the reference standards through the same compression cycle using a manual pellet press with the same force. The shifted peaks then matched perfectly. The binding agent was KSM-1002, a specific grade of microcrystalline cellulose used in pharmaceutical manufacturing. Without that step, I'd have been unable to confirm the excipient match beyond reasonable doubt. Court doesn't care about "close enough."
Counter-intuitive things nobody tells you
More scans is not always better. At some point you're just accumulating instrument noise and thermal drift. After about 64 co-added scans on a good FTIR instrument, you're not gaining meaningful information, you're just burning time. I've seen analysts run 256 scans per sample and then wonder why their throughput is terrible. Library matching is a starting point, not a conclusion. A high similarity score from a commercial library like the Bio-Rad or Sadtler collection means the software found something close. It does not mean you have an identification. I always verify by checking that the key diagnostic peaks match in position and relative intensity, not just by the numeric score. Software will happily give you an 85% match to the wrong compound if your sample is degraded or contaminated. ATR depth of penetration varies by wavelength. This matters when you're analyzing layered samples or coatings. The effective sampling depth in ATR mode ranges from about 0.5 to 2 micrometers depending on the wavenumber and crystal material. A polymer coating on a drug formulation might dominate your spectrum while the active ingredient underneath contributes barely anything. Switching to transmission mode through a KBr pellet or using diffuse reflectance can give you a more complete picture.
Where these methods fail completely
Aqueous solutions are basically unusable with FTIR. Water absorbs infrared radiation so aggressively that you cannot get through any meaningful path length. If your evidence is a liquid sample or a residue reconstituted in water, you're looking at Raman instead, or you need to dry the sample down first. Drying changes the physical form though, so document everything. Pure Raman also struggles with heavily fluorescent samples. Black powder residues, charred materials, and certain dyes will produce overwhelming fluorescence that drowns out the Raman signal entirely. In those cases I fall back to FTIR with careful baseline correction, or I move to GC-MS if the sample amount allows destruction. No single technique covers every scenario. Quantification with FTIR is possible but finicky. Peak height and area correlate with concentration, but matrix effects, particle size variation, and ATR contact pressure introduce enough variability that you need proper calibration curves for every sample type. I've seen forensic reports claim percentage compositions from FTIR alone that later fell apart under cross-examination because the calibration was never validated.
What I do before I send a sample out for analysis
Document the condition. Photograph the sample in situ. Note the substrate, the quantity estimate, and any visible contamination. Check the instrument status log to confirm it passed validation that day. Run a known reference standard alongside your unknown. Keep a backup spectrum saved before any processing or baseline correction. Chain of custody documentation matters as much as the analytical result, and everything needs to be defensible in court. The spectroscopic data itself should include the raw spectrum, the processed spectrum, the library match results with similarity scores, and a notes section explaining any deviations from standard procedure. Reviewers and opposing experts will look at those details more carefully than your final conclusion. If you're just starting out with this, get comfortable with instrument maintenance. Wipe the ATR crystal properly between samples. Run regular background checks. Verify your wavenumber calibration with a polystyrene film at least weekly. Most bad spectra come from poorly maintained instruments, not from difficult samples. I would rather deal with a challenging matrix than another instrument that hasn't been calibrated because someone was too busy to do it.
