Light, Bones and Time: Spectroscopy Edges Closer to Dating the Dead

When human remains are found reduced to bare bones, one of the hardest questions in forensic science becomes deceptively simple to ask and brutally difficult to answer: how long ago did this person die? Once soft tissues and body fluids are gone, the entomological clocks, microbial succession studies and biochemical assays that work so well on fresher corpses lose their grip. A new systematic review published in the International Journal of Legal Medicine has now taken stock of a decade of research into a technology that promises to fill this gap, and its verdict is both encouraging and sobering: light-based analysis of bone and teeth can indeed read the passage of time, but not yet reliably enough for the courtroom.

The review, conducted by an international team led by researchers at Universiti Malaya in Kuala Lumpur, followed PRISMA 2020 guidelines and was registered on PROSPERO. The team searched PubMed, Scopus and Web of Science for peer-reviewed studies published between January 2015 and December 2025 that used spectroscopic methods to estimate the post-mortem interval, or PMI, of skeletal remains. From 167 initial records, 88 duplicates were removed and, after screening, 19 studies made the final cut. Risk of bias was assessed with the Joanna Briggs Institute critical appraisal tool, and every included study scored 6.0 or above, indicating moderate to high methodological quality across the board.

The technologies at the heart of this field are Fourier-transform infrared spectroscopy, known as FTIR, and Raman spectroscopy. FTIR was the most frequently used modality, appearing in ten of the 19 studies, most commonly in its attenuated total reflection configuration, while Raman spectroscopy featured in seven. Both techniques probe the molecular structure of skeletal tissue, but in different ways. Infrared spectroscopy measures how chemical bonds absorb mid-infrared light, typically across the 400 to 4000 per centimetre range, revealing the composition of the mineral phase and the organic matrix. Raman spectroscopy, by contrast, detects the tiny shifts in wavelength produced when laser light scatters inelastically off molecular vibrations, offering the practical advantage of being largely unaffected by water, which makes it well suited to hydrated or minimally prepared specimens.

The molecular targets of these analyses are the building blocks of bone itself. Bone consists of poorly crystalline carbonate-substituted hydroxyapatite embedded in a matrix dominated by type I collagen, while tooth enamel is far more heavily mineralised and crystalline. As bone degrades after death, measurable changes accumulate: the crystallinity index shifts, carbonate-to-phosphate ratios drift, collagen-related amide bands weaken, and the balance between mineral and organic content changes. Studies in the review typically acquired around 120 scans per spectrum at resolutions of 2 or 4 per centimetre, then extracted peak intensities, integrated band areas and derived ratios. Energy-dispersive X-ray spectroscopy, the second most common technique, added an elemental dimension, tracking calcium, phosphorus and trace elements such as barium, zinc and iron whose distributions shift with depositional environment and time.

The evidence base is strikingly diverse. Sample sizes ranged from 6 to 145 specimens, drawn from human remains in most studies and from domestic pigs, Wistar rats and Eastern grey kangaroos in four. Nine studies combined forensic and archaeological material, with archaeological specimens originating from burial sites across Italy, Spain, Slovenia, Portugal and Austria, including remains recovered from World War II mass graves. The femur was the most analysed bone, followed by ribs, skull bones and tibiae, while dental studies focused on incisors and molars. Depositional environments spanned surface exposure, burial, indoor storage and even freshwater and saltwater immersion. Notably, 58 percent of the included studies built directly on the foundational work of Nagy and colleagues and Patonai and colleagues, published before 2015, underscoring how heavily the field leans on a small number of seminal papers.

Where spectroscopy shines brightest is in separating the old from the very old. Several studies reported consistent discrimination between forensic and archaeological remains, reflecting cumulative diagenetic changes that unfold over decades and centuries. Machine-learning models integrating spectral data achieved classification accuracies ranging from 58 to 100 percent. Among the standout results, an orthogonal partial least squares model built on Raman data achieved a root mean square error of cross-validation of 1.25 years with a cross-validated coefficient of determination of 0.89. A genetic algorithm-assisted PLS model trained on second-derivative FTIR spectra predicted PMI with an error of 54.36 days for buried samples and 68.48 days for unburied ones. A logistic regression model combined with receiver operating characteristic analysis correctly classified 87 percent of samples with a ten-year PMI, 76 percent at 25 years and 80 percent at 50 years.

Yet the review also exposes a critical weakness: the first 90 days after death remain stubbornly opaque. Studies evaluating shorter post-mortem intervals frequently reported weaker or statistically insignificant associations, with differences between sample groups sometimes failing to reach significance at p greater than 0.05. The authors attribute this to the interplay of intrinsic biological factors with extrinsic environmental conditions, which produce different kinetics of bone degradation in the earliest stages. In that window, decomposition-related biochemical changes are simply too subtle to rise above biological and environmental noise. This matters because the early post-mortem period is precisely where conventional methods still work, meaning spectroscopy currently struggles most where it is least needed, and performs best where no alternative exists.

Environmental context emerged as one of the principal sources of variation. Ultraviolet radiation, burial conditions, aquatic immersion and storage conditions all altered the rate and extent of physicochemical change in skeletal tissue. One study showed that environmental conditions changed the expression of commonly used FTIR markers, while another reported distinct elemental profiles depending on depositional environment, and a third linked changes in elemental composition to soil interactions and fracture-related alterations in bone hydration. These findings align with established taphonomic principles, in which temperature, moisture, microbial activity, oxygen availability and soil chemistry collectively shape skeletal diagenesis. The implication is clear: spectral markers cannot be interpreted in isolation from the environment in which remains were deposited, and future predictive models must incorporate environmental variables alongside spectral parameters.

Biological variability adds a second layer of complexity. Animal models, particularly pigs, are widely used as analogues for human decomposition, but the review found that spectral patterns in animal remains were not always concordant with those in human samples, especially for phosphate-related parameters and mineral-to-organic matrix indices. Differences between skeletal tissues also matter: one study observed a decrease in the carbonate-to-phosphate ratio in teeth that ran counter to trends commonly reported in bone, likely reflecting the differences in mineral content, crystallinity and hydroxyapatite structure between the two tissues. Limited evidence even hints at sex-related differences in the magnitude of spectral change, although the overall direction of decomposition trends appears consistent between males and females.

The path forward, the authors argue, lies less in discovering new biomarkers than in standardising and validating the ones already known. Carbonate-to-phosphate ratios, crystallinity indices, collagen-related bands and mineral-to-organic matrix ratios have shown reproducible temporal trends across independent studies and now deserve large-scale, multicentre validation on human remains under realistic depositional scenarios. Machine learning will be central to that effort, but the review cautions that most existing models were built on small datasets and lack external validation, and that the limited interpretability of some algorithms may hinder forensic admissibility under legal standards such as Daubert, which demands tested, peer-reviewed methods with known error rates. The reviewers also call for multimodal integration, combining spectroscopy with micro-computed tomography or X-ray diffraction, and for greater attention to dental tissues, which remain underexplored. Spectroscopy, in short, has proven it can read the chemical autobiography written into bone. What remains is to teach it to testify.

Subject of Research: Spectroscopic estimation of the post-mortem interval of skeletal remains

Article Title: Post-mortem interval estimation of skeletal remains using spectroscopic analysis: a systematic review

Article References: Thrukkumar, S., Rahmat, R. A.-A., Soon, L. P., Hamdan, S. N. S. M., Rizky, B. N., Kurniawan, A., Humphries, M., Sani, S. F. A., Ibrahim, N., Rahim, A. H. A., Suzuki, T., & Linacre, A. (2026). Post-mortem interval estimation of skeletal remains using spectroscopic analysis: a systematic review. International Journal of Legal Medicine. https://doi.org/10.1007/s00414-026-04010-w

Image Credits: AI Generated

DOI: 10.1007/s00414-026-04010-w

Keywords: forensic science, post-mortem interval, spectroscopy, FTIR, Raman spectroscopy, skeletal remains, bone diagenesis, machine learning, chemometrics, taphonomy, forensic taphonomy, systematic review