Plasma density is at present measured on FTU by a five-channel DCN (=195 m) interferometer  that successfully works at almost all densities of the FTU operation range. Some problems arise at a density higher than 2x1020 m-3, due to refraction, and during pellet injection, fringes are lost after fast density perturbation.
In order to get reliable measurements at very high density and during multipellet injection experiments, a single-channel two-colour interferometer with a CO2 ( = 10.6 m) laser as main wavelength and a visible HeNe (=0.63m) for compensating the vibrations has been developed and tested.
The phase shift introduced on a light beam, with wavelength , that crosses the plasma has two contributions: one from the plasma, proportional to , and the other from the vibrations of the mechanical structures, inversely proportional to , i.e.,
where L is the chord length, n is the line average density, and dvib is the total excursion of the optical path due to vibrations. Longer wavelengths are preferred as the disturbing effect of vibrations decreases and the plasma effects increase. On the other hand, the refraction effect due to plasma density variation increases with 2 and, at a very long wavelength, becomes intolerable even at low density, and measurements are no longer possible. The optimal wavelength is the longest compatible with the refraction; for FTU it is in the range of hundreds of microns.
For a shorter wavelength, e.g., 10.6 m, the disturbing effect of vibration is several times greater than the plasma phase shift, but this contribution can be compensated for by introducing another interferometer with a different wavelength that shares nearly all the optical components with the first interferometer.
If 1 and 2 are the two wavelengths and 1 and 2 the corresponding phase shifts, then density can be obtained applying (1) and eliminating the vibrations:
provided that the vibration displacement dvib be identical for both interferometers.
Another important feature of an interferometer for plasma density measurements is the frequency shift that must be introduced in one beam in order to apply the heterodyne detection technique. A rotating grating (Veron wheel ) that gives a Doppler shift up to 100 kHz is typically employed for DCN and HCN lasers (=193 m and =337 m, respectively). With alcohol lasers, such as CH3OH (=114 m), a frequency difference of some MHz is generated by tuning two different cavities pumped by the same laser (twin lasers ); while for a CO2 laser (=10.6 m), an opto-acoustic Bragg cell can be used to introduce a frequency shift of some tens of MHz [3,4,5]. This frequency shift fs determines the maximum rate of variation of the density that can be detected by the system without loosing fringes. In fact, to be unambiguously identified, the phase variation must satisfy the condition
which gives an upper limit for the density variation rate of
For comparison, table 2.II, reports this limit for the DCN (fs = 10 kHz) and CO2 (fs =40 MHz), together with other relevant quantities that must be compared with the density rise rate of 1.3x1024 m- 3 s- 1 expected for FTU with typical pellet injection.
A schematic layout of a two-color interferometer is shown in the figure.
The Bragg cell (BC) splits the laser beam into two parts and introduces a frequency shift in the diffracted beam. The two wavelengths are aligned together on a dichroic mirror that totally reflects wavelength 1 while it is transparent to wavelength 2. One beam (probe beam) crosses the plasma and is mixed with the other beam (reference beam) by a beam splitter (BS). Another dichroic mirror separates the two wavelengths before they arrive at the detectors. It is important to note that the only elements that are not shared by the two wavelengths are mirrors M1 and M2, which, however, do not give an important contribution to vibrations.
A Mach-Zender scheme has been mounted and tested in the laboratory on a granite bench (1x2.5 m) that holds all the optics apart from those (two mirrors) that will deviate the beam into the plasma . To reduce vibrations, the two mirrors are attached to the antivibration structure that holds the optics of the DCN interferometer. Two ZnSe windows, transparent both to CO2 and to HeNe beams, will be used for the vacuum insulation.
The layout of the interferometer is shown in the figure.
The total optical path is about 23 m. Only two focusing elements have been used as most of the probe beam is outside the optical bench. The first mirror (f= 1.5 m) makes a waist of 6 mm of diameter around the plasma center and the second (f=3 m) is placed on the top of the tokamak and reflects the beam back. The first mirror is also used to focus the beam on the detectors. The diameter of the CO2 beam is kept below 20 mm along the whole path, and mirrors of 50 mm diam are used to avoid diffraction. The HeNe beam diameter, however, is kept below 14 mm by means of an additional lens (f = 0.4 m) placed in front of the Bragg cell. For symmetry, the same optics have been used in the reference beam.
The CO2 detector is a HgCdTe room-temperature photoresistor; the HeNe is a PIN photodiode. The signal is amplified and compared with the LO signal, taken from the Bragg cell driver, by a phase comparator that outputs the sine and cosine of the phase difference. A fast ADC CAMAC module was used to acquire the phase comparator outputs. Data are then processed to get the plasma density.
The interferometer is routinely working on FTU tokamak since December 1996 and the density in several plasma conditions has been measured. The residual error on the density is 3x1018 m-3.
 D.Veron:"Diagnostic for fusion reactor conditions',Vol 1,Varenna 1982
 S.M. Wolfe et al. Appl Opt. 15, 2645 (1976)
 J.H. Irby et al. Rev. Sci. Instrum. 59,1568 (1988)
 T. N. Carlstrom et al. Rev. Sci. Instrum. 59,1063 (1988)
 P. Innocente S. Martini Rev. Sci. Instrum. 63, 4996 (1992)
 D. Frigione,O. Tudisco ENEA Report RT/ERG/FUS/94/20,Frascati Research center