Flexible Efficient & Practical  


An IFTS is an electronically programmable filter which obtains spectra for every pixel in the field of view. The imaging performance of IFIRS is diffraction limited at 2 µm, and Nyquist sampled at 3 µm in the near-IR channel and 6 µm in the mid-IR channel. The SNR performance of a camera equipped with filters and an IFTS are equivalent when they have perfect detectors. However, an IFTS has an extra degree of flexibility because the pan-chromatic image (i.e., the image formed by adding all the spectral channels together) has a speed advantage over the filter camera equal to the number of bands observed. For example, the Hubble Deep Field was observed with four filters. The U+V+R+I image would have been twice as deep if obtained with an IFTS. A camera equipped with a set of narrow band filters will become readnoise and dark current limited.


The efficiency of an ideal four-port IFTS is 100%. Every photon is directed towards either one of two focal plane detector arrays. The switching probability is a function of the optical path difference and the wavelength of the light. An IFTS is a "black-box" which sorts photons according to their wavelength.

Thus when statistical noise dominates, the peformance of a 3-d imaging spectrometer based on a 2-d focal plane array is the same for all architectures, whether a camera with filters, a dispersive long-slit spectrometer, or an imaging Fourier transform spectrometer. Of course, this only applies when the same image cube is acquired with each instrument. To build up the data cube the camera must cycle a set of filters and the dispesive spectrometer must scan a slit across the field.

Ideal gratings do not have 100% efficiency. Even on blaze an ideal grating has an efficiency of < 90% due to light lost in unwanted orders. The exact number depends on the blaze angle, the order of interference, and the opening angle between the incident and diffracted beams. When averaged over the free spectral range this figure typically drops to 70%. Real grating spectrographs also have slit losses, light scattered by irregularities in the ruling, absorption in the grating coatings, and limited free spectral range. A slit must also be used in a dispersive spectrometer. A grating used in first order spans an octave in wavelength, so for diffraction limited operation the slit can be matched to the image size at only one wavelength. The slit is either too big or too small, which compromises the SNR. At short wavelengths excess background light increases the photon shot noise, and at long wavelengths slit losses reduce the signal.

An Ideal Beam Splitter

Real IFTS instruments are not 100% efficient because beam splitters are not perfect. A perfect beam splitter has a relfectivity of 50% and a transmission of 50%. This is impossible to satisfy over all wavelengths. The simplest beam splitter, shown on the left, consists of a quarter wavelength high refractive index material deposited on a substrate.  If R is the reflection coefficient, and T is the transmission coefficient, a perfect beam splitter has an efficiency 4RT = 1. The figure on the right shows the efficiency plotted versus relative wavenumber for a simple single dielectric layer beam splitter (see Born & Wolf p 61).

The reflectance of a beam splitter does not have to be precisely 0.5 to achieve high efficiency. If the beam splitter does not absorb, then 4RT = 4R(R-1), so if 0.28 < R < 0.72, the beam splitter efficiency, 4RT > 80%.

A simple beam splitter is efficient only over a finite wavelength range, and in a practical IFTS it is this component that determines the useful free spectral range. In this example, the free spectral range spans a factor of 6 in wavelength. Compare this with a grating, where the corresponding factor is (m+1)/m, where m is the order of the interference. The best a grating can do is when m = 1, and the spectrum spans an octave.

Real Beam Splitters

High efficiency mulitlayer beam splitters can be fabricated. The following figure shows an example of a beam splitter designed and built by Janos Technology Inc. for an FTS operating at 0.96-5.4 µm.  The modulation efficiency 4RT > 80% over the entire wavelength range.  A Fourier transform spectrometer equipped with this beam splitter would have a free spectral range approching the theoretical limit of an ideal beam splitter. Note, that unlike a grating, this beam splitter is only weakly polarizing.
For the mid-IR channel (5-15 µm) of IFIRS we have chosen a Bomem beam splitter as characteristic of a high efficiency component.


All reflective, diamond machined, metal optics will be used for IFIRS. This ensures high image quality and high throughput. The reflectivity of a single gold-coated surface exceeds 99% for wavelengths longer than 1.5 µm. The optical system of IFIRS consists of, a collimator, a camera, two fold mirrors, cube-cornrer reflectors (three reflections) and a dichroic to separate the near- and mid-IR light. The collimator and camera are both three-mirror anastigmats, giving a total of eleven reflecting surfaces. The corresponding throughput is plotted.


IFIRS will use InSb and HgCdTe focal plane arrays. The near-IR channel will be equipped with a InSb detector similar the 1024 X 1024 Astronomy Focal Plane Array (ALADDIN) manufactured by Raytheon. InSb arrays can be anti-reflection coated to give broad-band performance from 0.6  µm to the detector material band-gap at 5.6 µm.

Hg1-xCdxTe is used for the long wavelength channel. This detector material has the advantage that the band gap can be tuned by between 3-30 µm by varying the composition (i.e. x) . Since dark current is a strong function of temperature and the cut-off wavelength, the composition of Hg1-xCdxTe can be selected for optimum performance at the operating temperature of the passively cooled NGST instrument module (T ~ 30 K). This example shows the QE curve for 14 µm cut-off HgCdTe, which is typical of longwave detectors from the Rockwell Science Center. (The  true response is probably flatter than shown here - the large scale ripple probably an artifact of the spectrometer).

System Efficiency

The efficiency is determined by the combination of the throughput of the optics, the beam splitter efficiency, and the quantum efficiency of the detector. The combination of these three factors is displayed in the plot for the 0.96 - 5.4 µm Janos beam splitter and an SBRC InSb detector.
The efficiency is high and flat. The average in-band efficiency between the half-power points is 73%. Over most of the wavelength range, the detector dominates the losses. The ripples in the efficiency are due to the anti-reflection coating on the detector. A similar plot is shown for the mid-IR channel.


An IFTS is practical because it represents a marriage of proven technologies: Back to the Berkeley NGST page.