4 First results

4.1 Visible interferences

The entire system was designed to obtain high rejection rates (typically 10⁴ to 10⁶) at a wavelength of 10.6 $\mu$m. However, the first channel to be set and aligned was the visible channel (used later as a metrology channel for OPD and tip-tilt control of the infrared beam). We made several measurements on that channel.

A first measurement was performed with a "light'' configuration (essentially beam-splitter and beam-combiner), without any other wavefront control device. With that configuration, it was not possible to obtain a stable flat field. We could only measure the contrast of the fringe pattern. Such a measurement was performed with a photographic method described at the ICSO'97 and it exhibited a rejection rate of several hundreds (Ollivier et al. 1997).

Figure 4: Interferometric extinction as a function of time obtained with the interferometer by moving OPD from a bright to a dark fringe, without spatial filtering (top left) and with spatial filtering by a 200 $\mu$m pinhole (top right), by a 50 $\mu$m pinhole (bottom left) and by a 20 $\mu$m pinhole (bottom right)

4.2 Infrared interferences

The infrared channel and all the sub-systems were completed and implemented on the optical bench by the end of September 1999. The contrast measurements began just thereafter. However, further developments are still necessary to utilize the entire potential of the test-bed.

All the sub-systems were used but manually controlled. The contrast measurements were obtained from the following procedure:

$\bullet$ The interference state control is obtained manually by moving the delay lines with the control software. As the system is not servo-controlled by the visible channel, occasionally one must correct the optical path difference in order to keep the destructive interference state, in spite of the atmospheric effects which are present when the protective enclosure is open.

$\bullet$ The lateral and angular superpositions of the beams are set before the measurement, and they are not modified thereafter. These settings are obtained by the superposition of the two beams on two different points: the beam-combiner, and a point located far beyond it. To allow these settings, we use software that measures the photometric centre of mass of the beams, and then tip-tilt mirrors adjust the position. This system is a simplified version of the system described before.

$\bullet$ The flux balance device is set in neutral position. The theoretical transmission is the same in the two arms of the interferometer. Because the beams are not perfectly aligned while they propagate in the different arms, the photometric imbalance is still about 5% which is outside the range of the settings. This imbalance is due to the fact that the visible and infrared beams are not always laterally superposed, because the ZnSe beam-splitters have dispersion. Thus, it is difficult to control the exact position and diameter of the thermal infrared beam over the whole optical path. As the optical sub-systems are compact, their entrance window is hardly bigger than the size of the beam itself and vignetting can occur. This point will be improved upon later. Flux imbalance, visibility (V) and rejection rate ($\rho$) are linked by the following relation, assuming no phase mismatching:

$$V = \frac{2\sqrt{x}}{1+x} = \frac{\rho - 1}{\rho +1}\cdot$$ (2)

Where x=I1/I2, and I1 and I2 are the intensities in each arm of the interferometer. Assuming x=0.95, the biggest rejection rate that could be expected is about 6000. This flux imbalance does not preclude rejection rates of 10³, that are sufficient to demonstrate the relative improvement an optical filter provides. That's why we did not spend too much time improving that point for the first measurements. However, this question of alignment is crucial for improving the absolute value of the rejection rate.

$\bullet$ The detection chain is composed of:

• a chopper that modulates the signal just after the source at a frequency of about 1 kHz;
• a $100\times100~\mu$m² single pixel HgCdTe detector cooled by liquid nitrogen, in a dewar that allows an angle of view of about 30 degrees;
• a pre-amplifier with a gain of about 10^-7A V^-1;
• a lock-in amplifier with a 1 s time constant;
• an AD acquisition card, recording the 0-10 V coming from the lock-in amplifier (that is why all the graphs presented here have the same Y-axis unit). The quantification step of the AD converter can be seen at low level signals.

$\bullet$ The contrast measurements were performed as follows:

• a background signal was recorded with the source off;
• successive bright and dark interferences were recorded during several minutes;
• the contrast was calculated from these data.

The following four graphs (Fig. 4) show typical results obtained from the records. They show successive bright and dark interference levels, and the depth of the extinction. These graphs were obtained both with and without optical filtering by a pinhole, whose diameter was decreased, thereby increasing the contrast of the interference pattern as theoretically predicted (Ollivier & Mariotti 1997). With the same wavefronts, showing a flux inbalance of about 95%, we increase the rejection rate from 40 (without filtering) to more than 1000, which is at present the best value of extinction obtained by such a technique in the thermal infrared. This result clearly demonstrates the efficiency of the spatial filtering.

However, the maximum extinction seen as better than 1000 in Fig. 4 is obtained within large fluctuations. These fluctuations are typically due to phase variations over the optical paths due to turbulence. Note that the protective enclosure was open at the time of the measurements.

The next step is to stabilize the null. As mentioned before, there are two ways to do this and they can be done either separately or simultaneously:

• reduce the turbulence over the setup;
  
• actively control the OPD over all the optical paths.

At that level of extinction, only the first way was necessary to stabilize the interference pattern. The protective enclosure was closed, and its top was heated during 12 hours using heating patches. To reach values of rejection of about 10³, we found that a helium-filled enclosure was not necessary, since turbulence effects are negligible at that level of extinction. However, a helium filled enclosure remains necessary when high absolute values of rejection (typically 10⁵-10⁶) are expected. The result is a cancellation of the turbulence over the setup as shown in Fig. 5 and the stability of the null. Note that the OPD is kept unchanged during the whole record.

Figure 5: Interferometric extinction obtained by changing the OPD with the protective enclosure closed and the top heated. The phase noise is strongly reduced and the rejection rate can be stabilized at a value of about 1000 without any active control of the OPD

A new series of developments and settings is under way to improve the rejection of this interferometer.