Fourier Transform InfraRed vs Grating InfraRed Instruments for Optical Measurements

Dr. Fred Goldstein in his 16 February 2004 Film Star News had the following comment: " appears that the current consensus on ‘FTIR vs Grating IR' is that FTIR instruments are unsuitable for optical measurements. If you disagree, or have evidence to the contrary, we'd like to hear."

Having been associated with FTIR since 1957, I feel compelled to respond to that comment. In general, I have to agree that the application of FTIR can lead to results that are prone to photometric errors. However, there are things that can be done to overcome the limitations of FTIR for optical measurements.

The wavelength (or wavenumber) scales of FTIR systems are usually impeccable because of the way they are sampled with reference to laser wavelength intervals. FTIR instruments also have advantages in the area of signal-to-noise (S/N) ratio for a given observation time; this can alternately be viewed as faster data gathering for a given S/N. This can be very helpful for chemists who's primary concern is the qualitative determination of trace elements in a background spectrum. However, from my experience, there are two potential sources for photometric error in FTIR systems; and the photometric scale is what is of much greater significance to optical measurements than most chemical endeavors. Chemical use of spectrophotometers is much greater than optical applications, and therefore the manufacturers orient their products to those needs for simple economic reasons.

One source of photometric error is the fact that all commercially available FTIR instruments are single beam spectrometers (to my knowledge). Single beam is only a problem if there is a significant change in the system's response in the time span from when a reference spectrum is taken until when a sample spectrum is measured. This might be due to drifts in the light source brightness, detector and electronics response, or changes in the transmittance of the optical path such as with atmospheric water and carbon dioxide changes within the instrument.

For these reasons, we developed and produced a true double beam FTIR spectrophotometer(1,2) with an integrating sphere in 1970. The first of these was used by NASA in the development of the Shuttle reentry tiles because of its ability to measure the IR emittance of almost any surface. The second and third instruments were used by the US Army and Air Force for IR signature measurement and other coatings work. The fourth was used by NASA in solar collector/absorber work and other materials development. A large industrial chemical and coatings manufacturer was the only non-government purchaser of one of these instruments. Unfortunately for our business, the idea was probably 20 years ahead of its time; and these instruments are no longer available. However, a lot of valuable work(3-6) was done with these instruments.

Subsequent to that period, I had occasion to need an IR instrument; and I tested and purchased a commercially available single beam FTIR. I'm willing to assume, unless I see evidence to the contrary, that the behavior of these instruments is sufficiently stable over some reasonable time for single beam measurements. Therefore, I typically run a reference spectrum with no sample and a 0% transmittance sample at the beginning and at the end of a series of measurements. If the starting and ending references agree to within an acceptable tolerance, I'm willing to assume that the intervening measurements were also stable enough.

The second and most critical source of photometric error that was apparent in this new instrument (and probably most other FTIR's) was the non-linear photometric response of the detector and possibly its associated electronics. Figure 1 shows an actual "interferogram" from an FT "instrument. Figure 2 is an that actual interferogram in the visible spectrum, for illustration. This raw data is the Fourier Transform of the spectrum that the instrument has observed. It is then transformed in the computer to give the observed spectrum. The central or "white light" peak is orders of magnitude larger than most of the observed signal. This peak can be said to represent the average signal over the whole spectrum. If the response of the system is not perfectly linear between this peak and the rest of the low level "wiggles" in the interferogram, the photometric scale of the transformed spectrum will be distorted (in error).

Fig. 1. Actual interferogram data plot from an FT.

Fig. 2. Actual visual spectrum interferogram.

My preferred check for linearity is to use samples of known transmittance in the band of interest. For the 2-16 micron region, I like plane parallel slabs of Ge, ZnSe, and CaF2 (or some other known lower index material). These have well know indices of refraction and therefore transmittance, and you can get them checked on a non-FTIR instrument whose photometrics you believe. I have typically found that there is a systematic nonlinearity in the results measured with the FTIR when used at full beam aperture and f/number.

One fix for this nonlinearity is to stop down the sample beam at its focus. If the beam illumination at the focus were very uniform, we could measure the linearity with apertures of controlled area. I would be cautious in this approach; uniformity is a risky assumption to make. However, this is a convenient way to reduce the signal level in a systematic way and then remeasure the linearity at each reduced level. Some level should be found where the linearity is acceptable. Admittedly, the advantages of S/N have been reduced in proportion to the area of the limiting aperture as compared to the unlimited beam focus; but this is probably a reasonable compromise for the sake of photometric accuracy.

There are other fine points of FTIR technology that might enter into the mix if not handled properly by the manufacturer, but I would guess that they will not show up once the linearity problem is properly handled.


It is my experience that the FTIR spectrometer can be used satisfactorily for optical measurements, once the procedures have been implemented to overcome its limitations.

References (If you would like copies of any of these papers, please contact the author at

1. "Fourier Transform Infrared Spectrophotometer For Transmittance and Diffuse Reflectance Measurements", Applied Spectroscopy, 30, 6 pp 593-601 (1976)[Pdf, 1276 KB]

2. "An instrument to measure spectral emittance from 2 to 20 micrometers," SPIE Vol 590 (1985) [Pdf, 805 KB]

3. "Independent Measures of IR Reflectance Yield Good Agreement",Applied Optics, 15, 1124 (1976)[Pdf, 45 KB]

4. "Total Reflectance Properties of Certain Black Coatings", SPIE Vol 384, 19-26 (1983)[Pdf, 1340 KB]

5. "Emittance and reflectance of various materials in the 2 to 20 micrometer spectral region," SPIE Vol 643, 93-100 (1986)[Pdf, 403 KB]

6. "Results of a Round Robin measurement of spectral emittance in the mid-infrared," SPIE Vol 807, 147-147 (1987)[Pdf, 535 KB]

If you would like a copy of a "white paper" on this subject and spectrometers in general, email your request to

P.S. If you are aware of anyone who might benefit from courses or books in Optical Thin Film Design and/or Production, we would appreciate it if you would refer them to our home page,

Return to Publications List
Return to Index