First of all, we used and optical spectrometer to determine which proportion of pump and signal is in the injected beam.
the wavelength is THE SAME for the 2nd stage for the beam coming from the core or coming from the clad !
thus, it is impossible to use a spectrometer to obtain this proportion.

Thus, we made the assumption the pump beam (Sp) is unpolarized in contrary to the core beam (Sc) which is almost linearly polarized (Scp, Scs)

with a half waveplate and a PBS, we can define 2 cases in rotating the half waveplate at the output of the PBS :
maximum power : Scp + Sp/2 = a1
minimum power : Scs + Sp/2 = a2
The sum of both power a1 and a2 : a1+a2 = (Scp+Scs) + Sp is the total power of the beam before PBS.
The subtraction of power a1 and a2 : a1-a2 = Scp - Scs is independant of the pump power which can vary for example in opening an iris which normally block the most part of the pump beam.

if the core beam is perfectly linear, Scs=0 => a1-a2=Scp and Sp=2*a2
We did 2 measurements with a1=227mW, a2=67mW => Scp=160mW (Sp=134mW)
and with a1'=277mW, a2'=107mW => Scp=170mW (Sp'=214mW with a wider opened iris)

if these measurements and assumptions are correct, the real power injected to the cavity was 170mW (instead of 300mW).
with a coupling of 75% => Gain ~ 6.3k => Finesse ~ 19.7k !

to be confirmed...

Global locking almost clean : tranmission and reflection are quite noiseless except the peaks at every 500ms.

During the amplifier 2nd stage lock, we experienced 2 strange behaviors :

- short "unlocks" every 500ms exactly all the time, during the whole lock process

- some PZT (in green on the picture) shift exactly synchronized with these "unlocks" (see the red circle).
these shifts are triggered by some OneFive motor move (Frep adjustment).

We tried to disconnect software activity (use the disconnect button on the software), to disconnect USB cable, to disconnect power supply of the motor controller box but none of these actions had an effect
=> still short unlocks every 500ms...

The amplifier 2nd stage seems to have small fan for the power supply... we can try to disconnect it.

We tried to turn on the 3rd stage which uses several fans for cooling the pump diodes... thus, one can see immediatly a big increasing in the PZT signal which tries to compensate the vibrations and quicly one loose the lock.
we have to think on how to suppress these vibrations.

1st lock today with the second stage of the amplifier.
~ 300mW of injected power (but which proportion of the pump and of the signal ? => One must use a spectrometer to determine this proportion)
MAYBE the real injected power of the signal is lower => maybe the Finesse is higher !

Coupling ~ 75%

1.2 mW measured in transmission => ~800W inside the FP cavity
=> Gain ~ 3500 => Finesse ~ 11k !!!

Last Finesse measurement was about 19k ! :-(

The Onefive spectrum has shifted a little bit to lower lambda. It's still enough for the experiment (3mW injection of ampli intead of 1mW needed).

Fiber injection is finished with the new ThomX stretching CVBG in double pass (+50% injection power compared to K-BOX CVBG in simple pass).

- Output Onefive: 27 mW
- After Onefive + lambda/2 + isolator: 25.8 mW
- After Onefive + lambda/2 + isolator + lambda/2 + PBS + CVBG*2 + PBS: 9.6 mW
- After injection into fiber and EOM (+all preceding): 3.2 mW
- Input 99% of the amplifier: 2.86 mW

> 660 mW on 1% photodiode DET100 @1Mohm

FSR is 33.347 MHz @1e-8 mbar

We've added a clamp in the 3rd stage to prevent the pump fiber from moving and tilting (it was tilting to the left).

The ThomX M2 has been measured between 1.5-1.6 for x and y

The circuit breaker on ThomX cavity drops @1.5-1.8kW when we turn on 1/2nd stage of the amplifier or 3rd stage of the amplifier.
Amplifiers are now connected to K-BOX UPS.

We are buying a clamp ammeter for further investigations.

The cavity is right now at 6.5e-8 mbar (turbo pumps on for 1 night and ionic pumps for 3h), primary pumps and turbo pumps are cut, ionic pump are on and the pressure level is still going down. 
Therefore there aren't any big leak in the cavity and we can expect a final vacuum level close to e-9 mbar.

Frederic and Manu from vacuum group came this morning to fix again the pneumatic valves which had a leak again.

To complete these observations:

Finesse has been measured on thursday, right after cleaning the cavity with ethanol @6000.
Finesse has been measured on friday, after opening the cavity for few tens of minutes @12000-12500. The cavity was openned while locking and the transmission was continuously and ~linearly increasing.
Finesse has been measured today after ~50h pumping with primary vacuum @17000. It's back to the highest value we've ever measured on this cavity.

Parameters @17000 finesse:

Input power: 7.85 mW (Koheras + AOM + 2*EOM)
Coupling: 70-75% => Pin ~ 5.9 mW
Transmission through S2 and P4: 50 µW (transmission estimated to 1.5 ppm) => Pcav ~ 33 W => G ~ 5600
It is compatible with a simulation with nominal parameters for mirrors and adding 47ppm by mirror of losses (absorption and scattering) which indicates 17k of Finesse and 4100 of power gain.
in this simulation, the coupling should be 94% instead of the 70-75% measured. It is possible that the losses (47 ppm) are not the same on all mirrors.

Observation made with IR Beam Viewer: 

The observation of P1 and S3 simultaneously while locking shows much higher diffusion on the ULE mirror. The diffusion on the two other ULE mirrors (S2 and P4) shows similar diffusion as S3.

After mirrors were going back from LMA for cleaning and thermal anealing, they have been installed in FP cavity for Finesse measurement.

after getting strange results when the cavity is closed (very low Finesse...maybe too much ethanol inside cavity procucing some lineic losses), we get 12k-12k5 Finesse with primary vacuum (~4 mbar) and a clean lock (very flat transmission signal and quite noiseless reflection signal)
we let the cavity pumped all the WE to see if we can measure a difference or not at lower pressure.

19/09/2019 : email from Chrisophe Michel about AR coating of LMA mirrors after reading that Ta2O5 layers can be damaged in vacuum if not protected by a SiO2 layer :

"oui tous nos coating standards HR se terminent par une couche de protection SiO2 Lambda/2.
Dans certains cas (spécifications du projet) l'épaisseur de la couche de silice de sortie peut être ajustée. 
Par exemple pour matcher une transmission très précise.
Pour les coating AR la couche de sortie est en silice mais pas lambda/2.

20/09/2019 : new email from Chrisophe Michel after he sent back the mirrors (LMA made a cleaning and thermal annealing) :

"le colis avec les 4 optiques, nettoyées 2 faces après traitement thermique et caractérisation, a été envoyé cet après midi par DHL (tracking number 4995967550).

Dans le colis j'ai ajouté le miroir en saphir C16115-11.

Les mesures donnent des absorptions entre 0.12 et 0.5 ppm (nous avons réussit à faire fonctionner de nouveau le banc mais il est mode dégradé avec une incertitude plus grande sur les mesures mais l'ordre de grandeur est bon et correspond aux valeurs mesurées en 2016 qui était de 0.24 ppm).

Pour les pertes par diffusion voila ce que Laurent à mesurer (les valeurs sont identiques à celles mesurées en 2016 sauf pour le C16116/11):

- C16116/11: 9 ppm sur diamètre 10mm (6 ppm mesurées en 2016); 4.5 ppm sur diamètre 6 mm; T=1.5 ppm @ 1064 & 1031 nm (simulée)

- C16116/12: 8.5 ppm sur diamètre 10mm; 4.5 ppm sur diamètre 6 mm; T=1.5 ppm @ 1064 & 1031 nm (simulée)

- C16116/13: 14 ppm sur diamètre 10mm; 10 ppm sur diamètre 6 mm; T=1.5 ppm @ 1064 & 1031 nm (simulée)

- C16115/11 (saphir): 9 ppm sur diamètre 10mm; 4 ppm sur diamètre 6 mm; T=148 ppm @ 1064 nm soit 120 ppm simulée @ 1031 nm"

Today again, same problem with axis 9 blocking the controlling software during the mirror alignment procedure.

after opening the motor 9-16 controlling crate, we saw 2 screws (RX- and TX+ of the controller) which were badly screwed.

we firmly screwed them... could it be the cause ?

after reconnecting everything, the software launches properly... and everything is back to work... wait and see.

Several experiments have been done in order to check the potential source of dust in the ThomX cavity. For each step, we look at a clean surface, put the cavity under vacuum, go back at ambient pressure and check back the "clean surface".

1st: Primary pump on
2nd: Primary pump + Turbo pump on
3rd: Primary pump + Turbo pump + Ion pump on
4th: Primary pump + Turbo pump + Ion pump on + 10 "open/close" displacement of the pneumatic valve
5th: Primary pump + Turbo pump + Ion pump on + 1 full range displacement of P1 z motor
6th : Primary pump + Turbo pump + Ion pump on + RGA

None of these experiences have shown good correlation with dust generation. 
The RGA has shown reasonable values.
Conclusion is that dust was already in the cavity and was only displaced by all combined motors, pneumatic valves, cavity opening and closing movements.

Christophe MICHEL told us more precise informations about spin coater cleaning.

They always clean with pur water and cotton tipped sticks.
First a spray of water, then they put the wet cotton tipped sitck onto the mirror from the center to the edge ~x3 while continuing to spread water. They finish by putting more water without the cotton tipped stick.

WARNING: WATER HAS TO BE CHANGED EVERY DAY 

They said they only use clean water but in case of grease/oil onto the mirror they would use first acetone then clean water.

The cavity has been cleaned then put under primary vacuum (primary pump without vacuum flow control). Dusts presence was checked at air and no significant changes have been observed.

1ST image before cleaning
2ND image after cleaning
No changes after coming back from primary vacuum

Mirrors have been sent today to LMA for cleaning and maybe annealing (recuit). 3 ULE of ThomX and the fused silicate for injection (trans @140ppm ).