5 Summary of the laboratory tests

These laboratory simulations were of interest to avoid uncontrollable disturbances arising in real observing variable seeing conditions, clouds, jitter, optical device alignment... Concluding remarks are as follows:

1:

From a relatively good consistency of the long exposure SNR with its theoretical expectation (Eq. 4), we have concluded that the AO gain and the intensity ratio have been accurately estimated;

2:

The measured dark-speckle SNR increases when the sampling approaches its optimal value (Eq. 1) and becomes closer to its theoretical expression (Eq. 5);

3:

The scaled SNR evaluated for the finest sampling ($1\times 1$ pixel/group) is similar to the long exposure SNR owing to the duality between 0 and 1-photon events. When the sampling is reduced, the scaled dark-speckle images clearly outperforms the long exposure;

4:

The saturation of residual speckles may tend to reduce the contrast of the faint companion;

5:

Present capabilities are limited by the residual static speckles. For accurate comparison with the dark-speckle model, we did not take into account these defects for the calculation of the SNR. In some conditions (see Sect. 4.3), a shape criterion can remove the brightest residual speckles which differ from a symmetrical Airy peak.

We have reported hereabove the detection of a companion 7 times fainter than the average halo (Sect. 4.3). Regarding the promising results obtained, the study suggests that even fainter companions could be imaged with the present instrument. The instrument sensitivity is derived from the results obtained in sample 3. In the same integration time and seeing conditions, theoretical study based on dark-speckle analysis leads to a limited brightness ratio of about $\Delta m=10$. Taking the global speckle noise into account decreases this limit down to $\Delta m=8$. Nevertheless, it is expected that deeper imaging can be achieved with longer integration and higher photon rate. The detection of substellar companions is currently constrained by fixed patterns on coronagraphic images. Precise map subtraction, dark-speckle analysis and further refinements ($\chi ^2$ test for instance) are required for extracting the faintest levels around stars.