How Hard Is It to Measure Temperature Accurately in MPCVD?

I recently came across a paper from a Japanese AIST team in Diamond & Related Materials. It examined how accurately the monochromatic and two-color pyrometers commonly used in MPCVD reactors actually measure temperature in a plasma environment. The results are worth sharing.

Start with two numbers. At 90 Torr, 4000 W and a methane concentration of 9.1%, the two-color pyrometer read more than 1,200 °C above the actual substrate temperature. Under the same conditions, the monochromatic pyrometer was high by about 200 °C.

In other words, the two-color method, which is theoretically the more sophisticated approach, produced the larger error under this particular condition.

How Did the Authors Know the “Real” Temperature?

They used in-situ Raman thermometry as the reference. The method relies on the temperature-dependent red shift of diamond’s first-order Raman peak. A nanosecond pulsed laser and gated detection were used to filter out the plasma’s own background emission and isolate the substrate’s thermal signal.

Because it is not dependent on the same plasma-emission signal as radiation pyrometry, this method served as the paper’s reference measurement.

The experiment compared three conditions: no plasma, pure hydrogen plasma, and hydrogen-methane plasma. In the first two conditions, the pyrometers performed reasonably well. Once emissivity was properly calibrated, both monochromatic and two-color readings agreed with Raman thermometry. The large discrepancy appeared only after methane was introduced.

What Changes When Methane Enters the Plasma?

Methane plasma introduces two additional sources of light: emission from the C2 Swan bands and broadband, blackbody-like radiation from carbonaceous particles, a common by-product in methane plasmas.

These emissions overlap with the substrate’s own thermal radiation. The pyrometer receives the combined signal and cannot automatically determine which part came from the substrate and which part came from the plasma.

The two instruments make different kinds of mistakes.

The monochromatic pyrometer makes an additive error. Plasma light is simply added to the substrate’s thermal-radiation signal. The detected intensity becomes too high, so the calculated temperature rises as well.

The two-color pyrometer is supposed to work differently. It calculates temperature from the intensity ratio between two wavelength bands. This can, in principle, cancel out the unknown emissivity and make the method more resistant to some disturbances.

But the continuous radiation from a methane plasma becomes stronger toward shorter wavelengths. That changes the intensity ratio between the two detection bands. For blackbody radiation, stronger short-wavelength emission is exactly one of the signatures of a higher temperature. The algorithm can therefore mistake a plasma-induced distortion of the spectral shape for a hotter substrate.

The closer a selected band is to the C2 emission region, the more pronounced this misinterpretation becomes.

The Material Variable: Emissivity

The paper also contains a useful material detail. Bare diamond has very low emissivity in the near-infrared region, approximately 0.05. This is why the emissivity setting of a monochromatic pyrometer is so sensitive: a small change in the assumed value can shift the calculated temperature by several hundred degrees.

The authors also compared substrates with a graphitized treatment on the back. Their emissivity increased to roughly 0.7, and their spectral behavior became closer to that of an ideal gray body. The influence of plasma interference was also significantly smaller.

This is a reminder that temperature measurement is not only an instrument problem. The optical properties and surface treatment of the substrate are part of the measurement chain.

The Practical Question

MPCVD temperature control is often discussed as if the reactor simply has a temperature that can be read directly. In practice, the measurement depends on plasma chemistry, emission bands, particles, wavelength selection, emissivity calibration, viewing geometry and the substrate surface itself.

So which method do you use in an MPCVD reactor? A monochromatic or two-color pyrometer, Raman thermometry, or another approach? It would be useful to compare how different teams handle substrate-temperature measurement in real processes.