GHS Mixture Classification, Explained

Imperium SDS guides · 2026-08-23

GHS classification is the step that turns a formulation into hazard communication: it decides the pictograms, signal word, and hazard statements that appear in Section 2 of the SDS and on the label. For a pure substance, classification is largely a lookup-and-evaluate exercise. For a mixture — which is what most commercial products are — it is a calculation, governed by rules that are precise enough to be computed. This guide explains how that calculation works conceptually, without pretending to be the regulation itself.

Substances vs mixtures

A substance is classified from its own data: test results, authoritative literature, and — in the EU — any legally binding harmonised entry (more on that below). A mixture is classified in a defined order of preference:

1. Test data on the mixture as a whole, where it exists. If the finished product has a measured flash point, that measurement — not any ingredient's — determines its flammable-liquid category.
2. Bridging principles, which let you extend a classification from a tested, sufficiently similar mixture (dilution, batching, concentration changes) without new testing.
3. Calculation from the ingredients — the workhorse case, using concentration cut-offs and additivity rules on the classified components.

A recurring real-world error is mixing these levels up: taking a component's classification (say, a concentrated supplier raw material) and stamping it onto a product that contains the component at a fraction of that strength. The rules exist precisely so that a 5% solution and a 35% solution of the same substance are classified as what they each actually are.

Concentration cut-offs

For most health and environmental hazard classes, the calculation method asks: is a classified ingredient present at or above the cut-off (threshold) concentration for that hazard class and category? Below the cut-off, the ingredient does not, by itself, trigger that classification for the mixture; at or above it, it does — sometimes at a lower category than the ingredient's own. The generic cut-offs differ by hazard class and category — the most serious chronic hazards, such as carcinogenicity, use much lower cut-offs than, say, skin irritation — and the exact values can differ between jurisdictions' adoptions of GHS. Requirements vary by jurisdiction — verify the current rules in the regulation you're classifying under (HazCom 2012 in the US, CLP in the EU, WHMIS 2015 in Canada; see our jurisdictions guide).

Additivity and acute toxicity: the ATE_mix concept

Some hazards don't work on a simple present/absent threshold, because the effects of several ingredients add up. Acute toxicity is the classic case. GHS handles it with the acute toxicity estimate (ATE): each acutely toxic ingredient contributes according to both its concentration and its potency, and those contributions are combined — a reciprocal-weighted sum across ingredients, computed per exposure route — into an ATE for the mixture (ATE_mix). The resulting value is compared against the category boundaries to classify the mixture.

Two properties of this approach matter in practice:

• Potency dominates. A small amount of a highly toxic ingredient can move the mixture's category more than a large amount of a mildly toxic one — which is exactly the toxicological intuition the formula encodes.
• It is a computation, not a judgment call. Given the same ingredient ATEs and concentrations, everyone should get the same ATE_mix and the same category.

Additivity-style approaches also apply to some other classes (for example, skin corrosion/irritation and serious eye damage/eye irritation sum corrosive and irritant components against thresholds), while other classes are evaluated ingredient-by-ingredient. Which method applies is defined per hazard class in the regulation.

Specific concentration limits (SCLs)

Generic cut-offs are defaults. Where regulators have concluded that a particular substance is more (or less) potent than the default assumes, they can assign a specific concentration limit — a substance-specific threshold that overrides the generic one. Under EU CLP, SCLs (and M-factors for aquatic toxicity) are published as part of the harmonised classification entries in Annex VI.

This is why famously "banded" substances classify differently at different concentrations. Hydrogen peroxide is the textbook example: at high concentrations (say, 35% and above) it is a serious oxidising and corrosive hazard; at household-strength dilutions it drops to irritant-level effects; and at low enough concentrations it may not be classified at all. The bands are not folklore — they follow from concentration limits attached to the substance's harmonised entry. Sodium hydroxide, sulfuric and hydrochloric acid, ammonia, hypochlorite bleach, and formaldehyde behave the same way: the concentration you use, not just the identity of the substance, determines the classification. A classifier (human or software) that ignores SCLs will systematically over-classify dilute products and can occasionally under-classify where a limit is stricter than the generic default.

Harmonised classifications as authoritative entries

In the EU, Annex VI of the CLP Regulation carries the harmonised classification and labelling list: for the substances it covers, the listed classification (with any SCLs and M-factors) is legally binding — a supplier must apply at least those entries, rather than deriving their own weaker view from the literature. That makes Annex VI a natural authority tier for classification data: where a harmonised entry exists, it beats an aggregation of scraped or vendor-reported hazard codes. (Self-classification still applies for endpoints the harmonised entry doesn't cover.) The current list is maintained by ECHA; the underlying GHS text is published by UNECE. The US and Canada do not have an equivalent binding substance list — classification there is the manufacturer's responsibility using the criteria — which is one more reason the same product can carry different classifications in different markets.

Why determinism matters

Everything above is rule-shaped: ordered data-source preferences, thresholds, additive sums, substance-specific overrides. That means mixture classification is, at its core, a deterministic function — same composition in, same classification out. This property is worth insisting on in any tool you use, for three reasons:

• Auditability. When a regulator or customer asks why a product carries (or lacks) a hazard statement, "here are the ingredients, the data sources, and the rule that fired" is an answer. "The model said so" is not.
• Reproducibility. Reformulate, re-run, and only real changes change the output. A classifier with randomness in it produces phantom diffs that waste review time and erode trust.
• Reviewability. A deterministic draft gives a qualified reviewer a stable object to check. Generative AI can be genuinely useful here — as an inspector that critiques a computed classification against the rules — but it should never be the thing deciding the classification. That's the division of labor Imperium SDS is built around: deterministic classification from official data, with AI confined to review.

Classification is one input to the finished document; for how it fits into the full 16-section sheet, see how to write a GHS-compliant SDS — and remember that whichever path produced the classification, a qualified person still reviews the result before it ships.

This guide is general educational information, not legal or regulatory advice. Requirements vary by jurisdiction and change over time — always have safety data sheets reviewed by a qualified person before use. See our Terms of Service.

Author SDSs the reliable way. Imperium SDS computes GHS mixture classification deterministically from official regulatory data, renders the full 16-section document for US, Canadian, EU, and Mexican requirements in eight languages, and keeps AI strictly in the reviewer's seat. See plans · More guides