HOW IT WORKS
How spectrum analysis works inFTIR: what it does and does not do
espectroFTIR is a support tool: it reproducibly organizes what the spectrum shows for you to review and assess. This page explains each stage, what results mean and where their validity ends.
What is ARIS©
ARIS© is the Algorithm for Rule-based Interpretation of Spectra, the core of espectroFTIR. It reads an infrared spectrum, describes the bands and says which functional groups and compound classes are compatible with what appears. The steps below show that path, from file to report.
What ARIS© does
Data reading and quality
Reads data files, spectrometer files or a plot image. Recognizes transmittance or absorbance and records the choice.
Checks quality. Evaluates range, sampling interval, noise, baseline, saturation and aqueous samples. If something compromises interpretation, it records a caveat rather than forcing a conclusion.
Band description and assignment
Describes the position, relative intensity and width of each band, with possible assignments and alternatives when more than one range applies.
Interpretation in context
- Compatible classes
Indicates functional groups and common organic classes: carbonyls, alcohols, aromatics, nitriles, nitro, halogenated compounds and others.
- Similar cases
Distinguishes cases such as carboxylic acid and amine salt, or mineral carbonate and organic carbonyl.
- Common materials
In common cases, indicates polymers, carbohydrates, starch, silica and other inorganic materials rather than forcing an organic classification.
- Mixture caveats
When one component dominates, warns that another minor component may be hidden.
- Compatible classes
Comparison with references
OptionalUsing references you upload or an open library, calculates numerical similarity indicators without an identity verdict.
Traceable report
Records the program version, code fingerprint, file SHA-256 and parameters used.
The same data and parameters, with the same version, allow the result to be reproduced.
Compatibility is not identification. The report supports professional assessment and does not by itself confirm the identity of a substance.
1. Reading and preparing data
The program reads data files (DPT, CSV, TXT, JCAMP-DX and spectrometer files, such as .0, .spa, .spc and .sp) or an image of the spectrum plot. For images, it calibrates the X-axis from tick marks and extracts the curve. Resolution depends on image dimensions and the range shown, so extraction is approximate. The Y-axis quantity (transmittance or absorbance) comes from the file or your input and is recorded in the report. The spectrum is then normalized to a 0–1 scale solely for detection and comparison; original data are retained.
2. Band detection
Bands are detected computationally by prominence. Three criteria apply simultaneously: minimum prominence of 3% of plot height, an adaptive floor of 6 times the noise estimated from the spectrum itself, and a minimum separation of 12 cm⁻¹ between bands. Spectrum edges, where bands may be truncated, are ignored. The report records the parameters used and measured noise so the result can be reproduced.
3. Band assignment
Each band is compared with general literature functional-group ranges (for example, PAVIA et al.), compiled in an internal database, with a tolerance of ±4 cm⁻¹. The table reports position, relative intensity, width and the main assignment, with alternatives when multiple ranges apply. The same table is available in the frequency table, where you can check each range.
4. Interpretation by region
The report organizes interpretation into four operational spectral windows: 4000–2500, 2500–1800, 1800–1500 and 1500–650 cm⁻¹. Rules describe compatible functional groups and compound classes, never the identity of a substance. They distinguish similar cases: carboxylic acid from amine salt, mineral carbonate from organic carbonyl, and carboxylate from nitro. When a carbohydrate such as lactose or mannitol dominates, the report warns that a minor component may be hidden. These windows belong to the report rules. The educational guides on this site use 2500–2000 cm⁻¹ for triple bonds and 2000–1500 cm⁻¹ for double bonds. An absent band is described as “not detected under the conditions and within the region evaluated”, never as proof that the group is absent.
5. Comparison with references (optional)
If you upload reference spectra or select one of the open libraries, the report provides numerical indicators: Pearson correlation between 650 and 1800 cm⁻¹ (after normalization and asymmetric-least-squares baseline removal), matching peaks and unmatched peaks. HQI is shown only if you confirm that the reference was measured with the same technique and instrument. Screening mode lists the most similar candidates without a compatibility verdict. Copies of the sample spectrum itself are flagged and do not count as agreement.
6. What the report contains
- Summary, annotated spectrum and band table.
- Interpretation by region and, when references are provided, comparison.
- Data-suitability assessment: range, sampling interval, noise, baseline and signs of saturation, with caveats when these affect interpretation.
- Evidence status: 0 means no bands were detected; 1, bands were described; 2, compatible functional groups; 3, compatible classes; and 4, measurable agreement with references. No level confirms the substance identity.
- Traceability: software version with a code fingerprint, the analyzed-file SHA-256 fingerprint and the parameters used.
Precision varies from class to class and with the technique: we measure each one with a confidence interval, and classes with little data or recurring errors get no reliability label, rather than looking safer than they are. Performance on real materials, including where ARIS© is wrong, is in the spectra compilation.
7. Limitations and responsibility
- The report is not an expert opinion or a conclusive identification. Conclusions and signatures are the responsibility of the person using it.
- It does not replace a confirmatory technique or a reference measured under the same conditions.
- Mixtures, additives, fillers and the matrix change the spectrum. A minor component may be hidden beneath a major component, such as an excipient.
- For images, calibration and curve extraction involve error; check the verification image.
- Spectra without the fingerprint region, with low contrast or heavy noise are flagged as inadequate data, with the reading limited to the bands described.
- Aqueous samples mask the bands of interest; the report warns when water dominates.
- Some classes still have no rule or have low sensitivity (for example, anhydrides, haloalkanes with CH₂ and ureas) and remain a limitation until there are data to fix them. There are also known false readings, such as a C=C or water band read as carbonyl. See the spectra compilation.
- Human review is part of using the tool: read the bands and caveats before using the text.
8. How the rules are evaluated
The rules have automated tests and are also compared with open-source reference spectra, including fiber, polymer and food libraries and the NIST/NIJ CADS dataset. These examples cover specific materials; they do not represent universal validation for every substance or sample. When an incorrect interpretation is identified, the rule is reviewed and a test is added. Development prioritizes avoiding unsupported assignments, even if some possible hypotheses are not suggested. Known limitations remain explicit: for example, a very minor component beneath an excipient is not detected.
Frequently asked questions
What is ARIS©?
It is the Algorithm for Rule-based Interpretation of Spectra, the core of espectroFTIR: it reads the spectrum, describes the bands and indicates compatible functional groups and classes.
Does espectroFTIR identify the substance?
No. It describes compatible functional groups and compound classes and provides similarity indicators against references. Comparison indicators alone do not confirm identity. The analyst is responsible for the conclusion, supported when necessary by a confirmatory technique and a reference measured under the same conditions.
Can I reproduce the result?
Yes. The report records the software version and code fingerprint, the analyzed-file SHA-256 and detection parameters. The same data with the same parameters give the same result.
What does the program do with poor-quality data?
It evaluates range, sampling interval, noise, baseline and saturation, and includes caveats in the report. When the spectrum does not allow reliable interpretation, the text states this rather than forcing a conclusion.
Should I send absorbance or transmittance?
Prefer absorbance: band height grows with the amount of substance and strong bands do not saturate, which makes the reading more stable. If your instrument only exports transmittance, send it as it is. The report and the plot keep the quantity of your file: if you send absorbance you get absorbance, and if you send transmittance you get transmittance. We do not convert the spectrum from one quantity to the other. If the file does not state the quantity, set it in the “Spectral quantity” field; a wrong setting changes the interpretation.
Ranges are typical tabulated values and vary with technique, physical state, matrix and instrument. This guide describes functional groups and compatible classes; it does not identify substances.