Since the very beginning of our journey with AdV @ DF, a peculiar behavior can be noticed in the power evolution during the lock acquisition. Just after we reach the dark fringe, the power as seen by the end PDs (B7, B8) begins to drop. This was true during O2, when we increased the power back in Nov 17, after the recovery from the monolithic installation and now with the increased input power. See first figure with some randomly picked transmitted-power evolution right after DF related to the aforementioned periods.
This does not seem to be related to some decrease of coupled input power (aka mismatch), since the backreflection off the PR (monitored by B2) does not increase. On the contrary, it decreases with similar trend as for B7 (top-right plot in the first figure), pointing to increasing losses in the ITF.
Triggered by Paolo persistent questions, we had a look at the trend since the recent increase of power (see figure 2). Paolo's sharp eye immediately noticed two things: 1) the usual drop of power is increased, and 2) the final power shows an slow decreasing trend.
Actually, 1) is definitely true. Figure 3 shows B7 power evolution after reaching the DF, normalized to its steady state.
We lose now 25-30% of power wrt the start of the lock, it was <15% back in November with similar input power, while it used to be around 7-10% with low power both during O2 and now.
So, since this light is not back-reflected off the ITF, where does it go? One possibility is of course some increased leak trough the dark port. In figure 4 we reported a typical evolution of powers just after reaching the DF. Indeed there is some increase of the fraction of transmitted power to the dark port. But it is not enough, as we show below.
Actually, we computed the carrier recycling gain in the PRC, as seen in trasmission from the arm cavities.
Figures 5-9 show the evolution of the carrier recycling gain just after reaching the dark fringe (left y-axis). On the same figures, we plotted the equivalent average losses in the arms needed to match the observed gain, as if no losses other than the dark port transmission were in the CITF (right y-axis).
The fact that we need to add increasing equivalent arm losses to match the rec gain means that this drop of power cannot be completely explained by the evolution of the dark port transmission.
Something else is going on there.
Why do we chose to introduce only additional equivalent arm losses and not, e.g., some change in the losse of CITF itself? The reason is a pure speculation: from picture 1, the drop of carrier power is present also when the sidebands (B4_112MHz) do not see big changes, so it seems more likely (?) that it is linked to something occurring in the arms, that are not probed by the sidebands.
Details:
Fig.5: random lock during O2. Car rec gain in the PRC starts from 42.5 (-> needed arm rtl = 65ppm) and drops to 39.5 (-> needed arm rtl = 74ppm).
Fig. 6: Power increase during Nov. 17. Car rec gain in the PRC starts from 40.5 (-> needed arm rtl = 67ppm) and drops to 35 (-> needed arm rtl = 86ppm).
Fig. 7: After monolithic susp installation, just before increasing the power, Jun 18. Car rec gain in the PRC starts from 40 (-> needed arm rtl = 70ppm) and drops to 36 (-> needed arm rtl = 84ppm). To be compared with fig.5.
Fig. 8: First lock after rising the input power, Jun 18. Car rec gain in the PRC starts from 38 (-> needed arm rtl = 76ppm) and drops to 30 (-> needed arm rtl = 110ppm). To be compared with fig.6.
Fig. 9: Randomly picked one of the few locks @DF, Jul 18. Car rec gain in the PRC starts from 37 (-> needed arm rtl = 78ppm) and drops to 28.5 (-> needed arm rtl = 115ppm). To be compared with fig.8.
Conclusion:
(1) A slow (~minutes) transient after reaching the dark fringe makes the arm transmitted powers to drop. The drop is not explained by either an increased mismatch of input beam (B2 is dropping as well) or increased transmission of the dark port (not enough to explain the whole decrease). The drop is larger when the input power is larger. Furthermore, this has slightly worsened over the past year.
(2) A possible interpretation is that the YAG power heats up the test masses and slightly changes the RoCs. This could bring some HOMs closer to resonance in the arm cavities (see for instance VIR-0423A-18) and, in turn, increase the rtl. The fact that a very slow trend is overimposed to this behavior is a bit worrying and could point to some slightly increasing absorption in the mirrors, making the process more effective. Of course this is a simple speculation to be further tested and investigated.
