University of Reading · School of Pharmacy

Thermal Fingerprinting of Powder Beds

Thermolens captures a multidimensional thermal fingerprint encoding chemistry, physical form, and particle packing from a single non-destructive scan using angular-resolved IR imaging over a standard DSC.

See the data Get in touch
91.9% LOOCV accuracy
13 Materials tested
2D 2D fingerprint
Spatial thermal fingerprints of four pharmaceutical materials
Spatial thermal fingerprints of four materials. Each circular map reveals a unique directional temperature structure.

Raw Data

Thermal maps that reveal hidden structure

Spatial ΔT maps for four materials at two temperatures (~80 °C and ~120 °C). Each map encodes directional thermal structure invisible to conventional DSC.

Raw spatial delta-T heatmaps for four materials at 80 and 120 degrees C

The Technique

How Thermolens works

An IR camera mounted above a standard DSC captures spatial thermal maps during a controlled heating ramp, encoding structural information invisible to conventional calorimetry.

Thermolens physical setup and signal processing pipeline
1

Pour the powder

Minimal sample preparation — powder is loaded into a standard DSC pan. No grinding, no pressing, no sealing required.

2

Heat with IR imaging

A controlled heating ramp while the IR camera records pixel-level temperature across the powder surface.

3

Extract the fingerprint

Proprietary signal processing extracts a multidimensional thermal fingerprint unique to each material.

4

Classify and correlate

Simple classification achieves 91.9% accuracy with no machine learning required. Physical metrics correlate with molecular properties.

Validated Results

Quantitative performance

13 materials, 37 measurements, leave-one-out cross-validation. All analysis in the pre-melt regime (55–146 °C).

91.9% LOOCV classification accuracy
34/37 correct, 3 errors between
structurally similar pairs
r = −0.91 Fourier f₁₋₃ vs Flory-Huggins χ
(HPMC-AS, n=10, amino acids)
r = +0.86 dΦ/dT peak vs log aqueous
solubility (n=12)
LOOCV confusion matrix showing 91.9% accuracy
LOOCV confusion matrix. 34 of 37 samples correctly classified using simple similarity matching — no machine learning required. All 3 errors involve structurally similar amino acid pairs.

Orthogonal and complementary to existing techniques

FeatureDSCXRPDFTIRThermolens
Data dimensionality1D curve1D pattern1D spectrum2D spatial
Material discrimination✓✓✓✓✓✓✓✓ (91.9%)
Particle size sensitivity×××✓✓
Relative solubility pre-screening×××✓ (exploratory, r=+0.86)
Compatibility proxy×××✓ (exploratory, r=−0.91)
Sample preparationSeal panGrind + mountKBr / ATRPour powder
Non-destructive×~✓~✓✓

Applications

What Thermolens enables

From raw material identification to drug–polymer compatibility ranking — a single thermal scan provides multiple orthogonal readouts.

Validated · 91.9% accuracy

Material classification and batch release

Identify incoming raw materials from a single non-destructive scan. Discriminates structural isomers (isoleucine vs leucine) sharing molecular weight and functional groups. Metric limits flag batch differences even when XRPD appears identical.

Validated · 5–20% detection

Adulteration and contamination detection

Correlation with the pure reference fingerprint drops as contamination increases. Detectable change in the thermal fingerprint from as little as 5–20% adulteration for most material pairs tested.

Adulteration detection: beta-Alanine spiked with L-Glutamic acid
Exploratory · r=−0.91

Drug–polymer compatibility ranking

Fourier f₁₋₃ correlates with estimated Flory-Huggins χ across three polymer systems: HPMC-AS (r=−0.91), PVPVA 64 (r=−0.84), PEG 6000 (r=−0.81). A single scan may complement existing compatibility screening approaches (n=10, amino acids).

Exploratory · r=+0.86

Relative solubility pre-screening

dΦ/dT peak correlates with log aqueous solubility (r=+0.86, n=12) and molecular weight (r=−0.83). Non-destructive log solubility ranking from a thermal scan — an association that may inform early formulation screening.

Interested?

Get in touch

Thermolens is under active development at the University of Reading. We welcome academic collaborations, industry partnerships, and discussions about licensing.

✉
Dr Hisham Al-Obaidi
h.al-obaidi@reading.ac.uk
🎓
University Profile
School of Pharmacy, University of Reading
📜
Patent
WO2025068694A1 — National phase