To measure a surface temperature we use the radiation flux of thermic origin that all bodies emit. The physical law that governs this phenomenon is Planck’s Law that describes the behavior of an ideal entity called the blackbody. The set of curves in Figure 1 describes the spectral distribution of this radiation for several temperatures representative of those commonly encountered, particularly in metallurgy.
We see that the peaks of these curves are located in the infrared (λ>0,8µm) and that the hotter the body is, the more the curve shifts toward short wavelengths.
All equipment currently marketed work in the infrared, meaning close to energy peak, which allows reducing the cost of their fabrication.
Emissivity gets in the way
Unfortunately the blackbody is a purely theoretical entity for, in nature, no body is perfectly black or even grey. Planck’s Law gives only the maximum radiant energy that a body can emit. This value must be weighed by an inclusive factor between 0 and 1 called emissivity and symbolized by the Greek letter ε. To complicate everything, this factor which depends on the nature of the body to be measured, varies according to the condition of the surface (oxidized or not), the wavelength and even the temperature! If controlling an industrial process is desired in following the evolution of the heating of a body, a constant and precise temperature measurement is necessary. However, as we have seen, the result obtained by measuring the flux depends also upon the variation in emissivity. Up to now, there has been no reliable and exhaustive experimental data available that would allow an analytical representation of this factor, or even a reliable interpolation with all the parameters concerned. In the low-cost equipment that can be found on today’s market there is a pre- defined emissivity value, usually 0.8. On the more sophisticated equipment the user chooses an emissivity value that remains unchanged until the end of the running process. As a direct consequence, the measurement result is false and the estimation of the error value is very unpredictable since we have no idea of the amplitude of variation of ε !
Signal variation 100 times greater in ultraviolet
Ultraviolet pyrometry and its generalization, ultraviolet thermography, bring the solution because the measurement is done in a spectrum range where the signal variation with temperature is so large that it hides the consequences of an emissivity variation. Due to the linear scale used in Figure 1, there is no evidence of the steep slope of the curves on the left of the energy peak. In Figure 2, with a logarithmic scale, we zoom toward the short wavelengths for the two temperatures 800°C and 1000°C.
It can be seen that at 0.3µm, in the UV, a 200°C variation results in a multiplication by 1000 of the energy signal, whereas it is only multiplied by 10 at 1µm. The drawback is that the energy levels in the UV are 10-10 times lower.
A reliable solution, highly sensitive and sturdy
Photomultipliers are detectors that have been designed to deal with these extremely weak energy levels. We use photomultipliers in the photon counting mode whose intrinsic noise is only 10 photoelectrons/second and a 100 photoelectron/second signal (that is to say 10-16 joules) is good enough to get a significant measurement result. Due to the extreme sensitivity of our detectors we can work with a very narrow bandwidth, almost monochromatically. Since the signal doubles every 20°C, we obtain very precise results. Another advantage of such a measurement process is that it is fully digital, from the captor to the display of the result. Thus we get rid of the drift problem that is the main flaw with analog devices and we get an excellent repeatability of results. Moreover, these photon counting devices have proven to be very sturdy and they can work continuously in a harsh environment for years.