---
title: How Thermal Conductivity Gas Analysers (TCD) Work
url: https://www.sarveshanalytics.com/learn/how-thermal-conductivity-gas-analysers-work.html
description: How thermal conductivity detectors measure hydrogen, helium, argon and other binary gas mixtures: heated element, conductivity values.
updated: 2026-11-03
source: html
---

Learn · Measurement principles

# How thermal conductivity gas analysers work, and where they fail

Thermal conductivity is the one property almost every gas has, which makes the method broad. It is also why it assumes only two components.

 Intermediate 4 min read Lesson 5 of 15[Measurement principles](https://www.sarveshanalytics.com/learn/track-measurement-principles.html)

Helpful first: [Gas analysis technologies compared: which sensor for which gas](https://www.sarveshanalytics.com/learn/gas-analysis-technologies-compared.html)

 Quick answer

A thermal conductivity analyser heats a small element and measures how quickly the surrounding gas carries heat away. Because gases differ strongly in thermal conductivity, the cooling of the element tells the composition of a binary mixture, for example hydrogen in nitrogen. It needs no reference gas, but a third component of different conductivity shifts the reading.

Updated 3 November 2026 · Figures from Sarvesh datasheets

**After this lesson you can**

- Explain how a heated element measures gas composition
- List typical thermal conductivities and which pairs are easy or impossible to tell apart
- State the binary assumption and what breaks it
- Explain why no reference gas is needed in ACCURRA-TCD

## The principle

A thin heated element, such as a platinum wire, sits in the gas. The element gives heat to the gas, and the better the gas conducts heat, the cooler the element runs. The element's electrical resistance changes with temperature, so a bridge circuit turns the cooling into an electrical signal. Compare the signal with the signal in a known gas, and you have a composition.

The heated platinum element in ACCURRA-TCD. The faster the gas carries heat away, the cooler the element runs.

## Thermal conductivity values

| Gas (near 25 C) | Thermal conductivity, W/(m K) | Relative to nitrogen |
|---|---|---|
| Hydrogen | 0.181 | about 7 |
| Helium | 0.151 | about 6 |
| Methane | 0.034 | about 1.3 |
| Oxygen | 0.027 | about 1.0 |
| Nitrogen | 0.026 | 1.0 |
| Argon | 0.018 | about 0.7 |
| Carbon dioxide | 0.017 | about 0.6 |

Hydrogen and helium are far above everything else, so each is easy to measure in almost any background. The exception is hydrogen and helium together: their conductivities are close, so hydrogen in helium (or the reverse) is a poor pair for this method. Oxygen and nitrogen are almost identical, so a thermal conductivity cell cannot separate air from nitrogen, or oxygen from nitrogen.

**Rule of thumb**

A binary pair is measurable when the two gases differ strongly in conductivity. The bigger the difference, the better the resolution.

## The binary assumption

The cell responds to the mixture's overall conductivity, not to individual gases. That equals a composition only if the mixture has two known components. A third gas of different conductivity moves the reading exactly as if the main gas had changed.

**Common mistake**

Reading hydrogen in a stream that also carries varying methane or CO2 and trusting the result. The instrument cannot tell. Check what else can enter the sample, or use a different principle, or a chromatograph.

## Worked example: how big is a third-gas error?

Take hydrogen in nitrogen, measured by conductivity. As a rough, linear estimate, each 1 % of hydrogen raises the mixture conductivity by about (0.181 - 0.026) / 100 = 0.00155 W/(m K). Now let 1 % of the nitrogen be replaced by methane, whose conductivity is 0.034. The mixture rises by (0.034 - 0.026) / 100 = 0.00008, which the analyser reports as about 0.05 % extra hydrogen.

With carbon dioxide instead of methane the error has the other sign: about minus 0.06 % hydrogen for each 1 % of CO2. Real mixtures are not quite linear, but the size is right: a few percent of a third gas moves a hydrogen reading by tenths of a percent. On a 97 to 100 % purity range that is the whole span; on a 0 to 100 % range it may not matter.

**Rule of thumb**

The narrower the range, the more a third component matters. State every gas that can be in the sample, and its range, when you specify.

## Reference gas and calibration

ACCURRA-TCD uses a platinum thermal conductivity cell and the catalogue states that it works with no reference gas, which removes a consumable and a source of drift. The calibration is set for the binary pair you specify. The catalogue lists binary mixtures of hydrogen, helium, argon, nitrogen, nitrous oxide, neon and krypton. The ACCURRA-TCD datasheet gives repeatability below 1 % of full scale, linearity below 1 % of range, drift below 1 % of full scale per week and a T90 below 10 seconds at 0.6 LPM.

Ask for the exact pair and range, because the usable ranges depend on them. Confirm the response time for your installation for your application.

## Ranges from the datasheet

The usable range depends on the pair. A few lines from the ACCURRA-TCD datasheet:

| Measured gas | Background | Ranges |
|---|---|---|
| Hydrogen | Nitrogen, air, CO2 or methane | 0 to 100 %, 0 to 3 % or 97 to 100 % |
| Hydrogen | Oxygen | 0 to 100 %, 0 to 4 % or 97 to 100 % |
| Hydrogen | Helium | 20 to 100 %, 20 to 40 % or 85 to 100 % |
| Argon | Nitrogen, air or oxygen | 0 to 100 %, 0 to 5 % or 96 to 100 % |
| Nitrous oxide | Nitrogen or oxygen | 0 to 100 %, 0 to 5 % or 96 to 100 % |

Hydrogen in helium starts at 20 % because the two conduct heat so similarly. The suppressed ranges, such as 97 to 100 %, serve purity measurement; the low ranges, such as 0 to 3 %, serve leak and blanket checks.

**Worked example**

Drift below 1 % of full scale per week on a 0 to 3 % range is up to 0.03 % hydrogen a week. If the alarm sits at 2 % and the plant can tolerate 0.1 % error, a weekly check keeps the error inside it with room for the other sources.

## Where it is used

- Hydrogen in nitrogen in annealing and sintering furnaces: [Continuous annealing and galvanising line atmospheres (HNX)](https://www.sarveshanalytics.com/applications/continuous-annealing-and-galvanising-line-atmospheres-hnx.html).
- Hydrogen purity in electrolysers and bulk hydrogen: [Hydrogen purity](https://www.sarveshanalytics.com/applications/hydrogen-purity-analyser.html).
- Helium and argon purity: [Gas purity certification and cylinder filling](https://www.sarveshanalytics.com/applications/gas-purity-certification-and-cylinder-filling.html).
- Hydrogen in the float glass bath atmosphere: [Float glass bath atmosphere](https://www.sarveshanalytics.com/applications/float-glass-bath-and-furnace-atmosphere.html).

## Installation points

- Run the recommended 0.6 LPM, regulated externally; the analyser has a built-in flow indicator. The datasheet T90 below 10 seconds is at that flow.
- Keep the sample dry. Water vapour has its own conductivity and acts as a third gas.
- Where the sample carries hydrogen in a classified area, use the flameproof ACCURRA-Hz with the 1030 hydrogen model, rated for Gas Group IIC: [Hazardous area gas analysers explained](https://www.sarveshanalytics.com/learn/hazardous-area-gas-analysers-explained.html).
- Calibrate with the binary pair the analyser is set for: zero in the pure background, span in a certified mixture near the top of the range.

**In short**

- Heat loss from a heated element tells you the gas's thermal conductivity.
- Hydrogen and helium conduct heat far better than other gases, so each is easy to measure, except against the other.
- Nitrogen and oxygen conduct almost equally, so TCD cannot separate them.
- A third gas of different conductivity breaks the binary assumption.

## Check your understanding

Answer in your head or on paper first, then open the answer. Recalling an answer teaches more than rereading.

**Why can a TCD measure hydrogen in nitrogen accurately?**

Hydrogen conducts heat about seven times better than nitrogen, so the signal changes strongly with composition.

**What happens if methane leaks into a hydrogen-nitrogen sample?**

The cell sees a changed conductivity and reports a wrong hydrogen value, because it assumes only two components.

**Can a TCD tell air from pure nitrogen?**

No. Their thermal conductivities are nearly identical.

## Key terms

- [TCD (thermal conductivity detector)](https://www.sarveshanalytics.com/learn/glossary.html#t-tcd-thermal-conductivity-detector)
- [Binary mixture](https://www.sarveshanalytics.com/learn/glossary.html#t-binary-mixture)
- [Reference gas](https://www.sarveshanalytics.com/learn/glossary.html#t-reference-gas)
- [Cross-sensitivity](https://www.sarveshanalytics.com/learn/glossary.html#t-cross-sensitivity)

## Frequently asked questions

**Do thermal conductivity analysers need a reference gas?**

ACCURRA-TCD does not, according to the catalogue.

**Which gases can a thermal conductivity analyser measure?**

Any binary mixture whose components differ in conductivity. The ACCURRA-TCD catalogue lists hydrogen, helium, argon, nitrogen, nitrous oxide, neon and krypton.

**Can it measure oxygen?**

Only in a binary background of a gas with very different conductivity, such as oxygen in hydrogen or in helium.

**Why does moisture matter?**

Water vapour has its own conductivity, so wet samples shift the reading. Dry the sample or hold its moisture constant.

## Related products

[

### ACCURRA-TCD

Thermal conductivity analyser for binary mixtures of H2, He, Ar, N2 and others, with no reference gas. View ACCURRA-TCD

](https://www.sarveshanalytics.com/binary-tcd-thermal-conductivity-analysers.html)

## Where this applies

- [Hydrogen purity](https://www.sarveshanalytics.com/applications/hydrogen-purity-analyser.html)
- [ACCURRA-TCD](https://www.sarveshanalytics.com/binary-tcd-thermal-conductivity-analysers.html)

## Where to go next

- [Measuring hydrogen: purity, mixtures and safety limits](https://www.sarveshanalytics.com/learn/measuring-hydrogen-purity-and-hydrogen-in-gas-mixtures.html)
- [Cross-Sensitivity and interference in gas analysers](https://www.sarveshanalytics.com/learn/cross-sensitivity-and-interference-in-gas-analysers.html)