I set up the temperature controller and oven. Still working on the mount for the objective and oven-waveguide.

Since the laser is at 5 inches H all optics were previously set up at the same height...

[JC, Shruti]

We removed the extra optics, cables, and fiber stuff from the WOPA area. While doing this we placed the waveguide (still covered) in one the cabinets labeled "optics cleaning". See attachments for after pictures.

We also put down the curtains and left the HEPA on. The curtains were almost sticky, so we took one of them to the 40m to test clean according to a cleaning procedure JC found online.

We also noticed a possible temperature sensor on the table (Attachment 3). I'm not sure where to read the data or whether it is useful to us so I may remove it.

Also while working down there one of the grey pipes on the south wall was rattling furiously but stopped after a while.

[Koji, Shruti, Vittoria]

We borrowed the tweezers from Cryo and CTN. We also borrowed the microscope from Cryo and Koji found a mini-USB to USB cable at the EE shop (since we couldn't locate one near the microscope). All of this is now at the Cryo Lab.

At the OMC lab,we garbed up and cleaned all the surfaces with IPA. Then we set up inside the HEPA enclosure.

All images and videos from the procedure can be found in the ligo.wbridge Google Drive.

1. Chip condition

We looked at the waveguide under the microscope and noted the top surface (that was all we were able to look at). See attachments 1,2 for the setup and attachment 3 for the microscope images. We did notice some defects in the cladding, but hopefully they are minor. Also not sure why one side looked blue (because of the wedge adn AR coating maybe??)

2. Practice

The waveguide should only be handled from the side surfaces labelled S3 and S4 here. The front and back surfaces are optical quality and the top surface is a shorter cladding. Because of this I practiced lifting a set screw a few times.

We set up the mount so that the width was a little over 1 mm, but not more than 1.5 mm (to avoid the waveguide from tipping over after placement on the mount).

While lifting the waveguide, it tipped over a few times in the film that it was packaged in. For this, we used a clean swab (optical quality) to life the top surface, and used one of the plastic tweezers as a wall to stabilize it.

3. Mounting

Finally, I picked up the waveguide and dropped it into the groove between the two portions of the clamp (see attachment  4 where it is slightly misaligned).

We then pushed the clamps further in and tightened the screw while holding them in place. (See attachment 5 and 6 for final setup). One portion of the waveguide does stick out a little, but we don't think that is a problem since it is well within the cover.

We wrapped it up in aluminum foil and it is now stored on the WOPA optical table in QIL (See attachment 7). Unfortunately, we did not save the film that the waveguide was mounted on  and only realised too late that we might need it in the future.

[Paco, Shoki]

Here are some of the design parameters for the pump-resonant SFG experiment. Signal is at 2050 nm, pump at 1064 nm and output is at 700 nm.

Crystal = MgO:PPLN

•     deff = 16 pm/V
•     αsig = 3e-4 / cm  # Absorption at signal wavelength
•     αpmp = 4e-4 / cm
•     αout = 5e-4 / cm 
•     R_AR = (0.007 + 0.004) # Covesion PPLN custom coating 2023
•     α_coat = 100 * ppm
•     α_inel = 15 * ppm
•     I_dmg = 500 * kW / (cm ** 2) # At 1064 nm

The phase matching condition for the three wavelengths is

Which using the following Sellmeier dispersion

to calculate the refractive indices gives a poling period requirement of 14.744 um. The length of the crystal is chosen to be 20 mm.

The MgO:PPLN SHG crystal for1350-1490nm pump crystals from Covesion get the following specs:
- 0.5mm thick, 10mm wide, 20mm long, PPLN crystal
- 9 gratings: 13.83, 13.96,14.08, 14.55, 15.10, 15.60, 16.10, 16.60, and 17.10 microns
- Grating aperture of 0.5x0.5 squared mm
- Both facets are 0.6° angle polished
- Custom AR coating:
    - Input facet: AR1064nm+2050nm/0°, 1064nm: R<0.8% and 2050nm: R<0.7%
    - Output facet: AR700nm+1064nm/0°, 700nm: R<0.4% and 1064nm: R<0.4%

Beam overlap parameter and quantum inefficiency

The ratio of signal to pump Gaussian beam waists is the primary figure of merit to approach 100% QE. Attachment #1 shows the contours of the input signal waist (ws) to pump waist (wp) in the usual size ranges. The paraxial approximation breaks for our crystal length if ws < 55 um, so letting ws = 60 um, and targetting a QE = 99.99% drives wp > 169.86 um, which in turn drives the pump power Pp > 595.8 Watts. This exceeds the damage threshold assumed above, so with a nominal ~600.0 Watts the requirement to avoid damage is actually wp > 194.75 um, or ~15% larger beam than originally assumed.

Cavity

Refer to Attachment #2 for a drawing. For M2 and M3 we may consider these 3.05 mm thick bandstop mirrors. The specs are:
- HR (0 deg aoi) T<100 ppm @ 1064 nm
- AR (0 deg aoi) R<0.09% @ 700 nm and R < 0.05%@2050 nm which we have been allocated as signal Fresnel losses
- The material is fused silica, so we may assume a bulk extinction coefficient of 0.0001 and allocated as bulk absorption loss.

• With 10.0 W of seed power, the cavity gain requirement is G > 59.58
• Assuming a roundtrip pump loss of Lrt ~ 13415 ppm, the cavity Finesse at maximum gain will be F ~ 214
• For maximum Gmax = 68.50, an input coupler with T1 = 1.457% gives a circulating power ~ 15% above the minimum requirement which can account for parameter tolerances and mode-mismatch

Attachment #3 shows the gain curve as a function of the input coupler transmission.

We will need to fire an amplifier to get the 10 Watt seed. Only then may we try to reach > 95% QE.

After obtaining permission, we borrowed the beam profiler shown in the image attached. It will be in Bridge B102. If it is needed, contact me and I can get it to you. The profiler model is Beam'R2-DD

We briefly borrowed the laptop that was with the profiler but returned it after we installed the proper software onto the B102 computer. 

The HC Photonics waveguide with oven arrived today and I partially unpackaged it and left it in QIL. See pictures here.

Plan:

See Attachment 1 for a diagram of the planned setup with options to perform SHG, nonlinear gain measurement, and finally squeezing.

The first goals:

• Mount the chip (chip holder was provided separately)
  • It might be better to do it in a cleaner environment...
  • Need to figure out the right mount for the chip+holder+oven.
• Set up temperature control with oven
• Mode-match to the waveguide
  • two lenses and an objective to obtain a collimated beam of the mode field diameter entering the waveguide (this will probably have to be done iteratively using the SHG generated green beam). I will post a few solutions soon.

The West Bridge Wiki for WOPA has more information including the PM temperature, MFD, scanned waveguide manual etc

Google Drive folder with west bridge gmail has more files and pictures.

1. If it is possible, you should enhance conversion efficiency by optimizing waist size. In that situation, more precise specification of noise coupling can be measured.
2. The same measurements in single pass should be also done in cavity scheme, such as the followings:

• optimization of cavity scheme for enhancing efficiency
• measurement of noise coupling. 

In addition, conversion efficiency should be measured when input beam power (cavity finesse) is varied to confirm that pump deletion is occurring. 

In the Cavity scheme, there is no need to make the waist size as small as single pass. If you do so, beam intensity will exceed the damage threshold of the crystal.

→In the region that there is no pump depletion, SHG can be regraded as the same of SFG, so SHG’s characterization will be useful for SFG’s characterization.

As the reference of the present scheme, 532nm output at the peak efficiency was measured, which is shown in Fig.1. In this situation, TC200 setting temperature was 36.3°C, the laser source current was 1.503, and the laser source crystal temperature was 20.02°C. And 532nm PD output voltage then and its spectrum are also shown in Fig.2. It means most of output fluctuation came from 60Hz and its harmonic. In contrast to this, the 532nm output was sometimes unstable, which is shown in Fig.3. And 532nm PD output voltage then and its spectrum are also shown in Fig.4. The latter has more componets of 60Hz than the former, which semms to come from voltage supply or something like that. 

For specifying the noise coupling of SHG, in the noise injection at Laser AM, PM, temperature, these PSD were measured. These result will be useful for computing the noise budget of frequency cnoversion process.

Laser AM : This is by diode current in laser source. Its coefficient to modulation voltage is 0.1A/V, which can be converted to 0.075 W/V.

Laser PM : This is by PZT of cavity in laser source. Its coefficient to modulation voltage is 1rad/V, which was shown in 40m Elog 12077.

Temperature : This is crystal temperature operated by TC200's PID controller.

When each port wes injected, its injection and SHG output were measured. These datas were calculated , which are shown in Fig.1. As the reference, that of  in no injection is also plotted. That about temperature was measured in lower sampling rate because this measurement took more time than others.

Please clarify the definition of the quantum efficiency:

 in this plot is 532nm output power divided by squared 1064nm incident power. This is the conversion efficiency per watt. 

This could be an unrelated question:

I think that shot noise in 532nm is the half of that in 1064nm. If it is correct, shot noise including  is shot noise of 1064nm multiplied by eta.

Please clarify the definition of the quantum efficiency:
When you say "quantum efficiency," is it the number that is 100% when all the 1064nm photons are converted to 532nm photons? Or is this a conversion efficiency in terms of the power?
The former has the theoretical maximum of 50% because two 1064nm photons become a single 532nm photon. The latter has twice the number of the former definition.

My convention of "QE" is the photon number efficiency because I usually work with a single wavelength.

This could be an unrelated question:
How is the shot noise (or any photon number fluctuation) of 1064nm converted to 532nm? You have <n_1064> +/- dn_1064. Say dn_1064=sqrt(<n_1064>).
I know <n_532> = <n_1064>/2. What about dn_532? Is it sqrt(<n_532>) (poissonian), or anything else? Also if we include the efficiency of \eta how does it come in?

When the crystal temperature was modulated by using step function, SHG output and crystal temperature were measured, which are shown in Fig.1. In this figure, SHG output is converted to efficiency.

This modulation seems to have too more amplitude, so it is nessesary to make smaller one. Making amplitude smaller enable to modulate faster, which is also expected to realize some modulation like noise.

Under the assunmption that the temperature system has 2nd order lowpass, its behavior was measured. PID controller setting was changed from [P=250, I=30, D=0] to [P=250, I=50, D=0] because the response was not fast enough. It is not finalized yet. Some types of temperature controll were measured, which differ in terms of heating/cooling period and amplitude. These result are shown in Fig.2 and Fig.3. These result means that the modulation frequency is limited by room temperature. The heating gradient can be adjusted by the temperature setting and step time, but (at least) the gradient of cooling is constant around 36.5°C, and it is not possible to make it faster. 

The area where the 5 mHz modulation was observed in Fig. 2 is picked up and shown in Fig. 3. When the heating and cooling cycles are halved and the temperature setting is not changed, it is shown in Fig.4.

Given all this, it is guessed that the following steps should be followed to obtain the target modulation

・Determine the amplitude [°C] of the modulation

・The time needed for cooling can be obtained from modulation amplitude

・Find TC200's setting temperature that is able to heat the modulation amplitude in ths same time of cooling (if you cannot adjust the amplitude well, try changing the I-gain and its period). Heating gradient depends on I-gain and the heating period.

・There is a delay in both heating and cooling, so the timing needs to be adjusted while taking this into account. This alos depends on I-gain and  the amplitude [°C] of the modulation.

In this situation of room temperature, The coolig gradient is seemed to be fixed to -1°C/min. So to make heating faster by increasing I-gain is not likely to lead to increasing modulation frequency. If the modulation needs to be faster, the amplitude [°C] of the modulation has to decrease.

Fig.1 shows the estimation by using the present temperature system's features. Actually, this modulation may differ because there should be differences in response time and cooling due to room temperature. In addition, I-gain may need to be increased because the response time may not be fast enough.

 This remperature system's response function seems to be the entire system including the response of sensor, temperature transmission, PID controller, and so on. Its estimation was calculated by using MATLAB tfestimate function, which result is shown in Fig.5. It seems that this system has second-order low-pass filter around 0.01 Hz. At higer freqency band, it also has something like high pass filter. But it looks strange because temp sensor usually has no response to high frequency operation. For the moment, this system is assumed to be as a model with a second-order low-pass filter around 0.01 Hz. By using this, how to modulate crystal temperature will be determined.

Since the lowpass frequency should depend on the step start temperature and the amount of temperature change, each response was calculated for each step. In addition to that, the measured sensor temp was reproduced by applying an arbitrary 2nd-order low-pass filter to the step function, which is shown in Fig.2. This should allow more detailed estimation of the modulation waveform. Although some of them has the offset difference, it doesn't has effects so much.

From these result, the present temperature sysytem, including TC200's PID controller(P=250, I=30, D=0), temp sensor, how to put sensor, and cooling by room temperature, has the following features:

・2nd-order lowpass filter around 5×10^-3 Hz

・temperature change more than 2°C with some overshoot (it dosen't happen so much when 1°C)

・taking some minutes for temperature settling, but its flctuation looks slow and small.

This calculation was'nt PSD by using pwelch. Its corrent version is shown in Fig.1.

For power supply this circuit, E3649A(Agilent) was brought from CNL to QIL with Paco's approval.

The temperature sensor was calibrated using the temperature displayed on the TC200 and the sensor output voltage when the temperature was set from 31°C to 40°C in 1°C increments, which was set as the displayed temperature was finally settled at 36.3°C. This resut is shown in Fig.1. By using the TC200 and the sensor output voltag, the coefficients were obtained, which is shown in Fig.2. The temp sensor by converting output voltage in Fig.1 is shown in Fig.3.

The setting of TC200's PID controller is P=250, I=30, and D=0.

And then the behavior of temp sensor was measured when the temperature setting was changed like a step function, which is shown in Fig.4.

Temp sensor circuit was updated like Fig.1. Its operation has been checked but not yet calibrated.

The settings of laser driver were laser current = 1.49A and crystal temp = 20 °C. Firstly, in the situation that laser was shutted, the injected noise(stdev=0.5V) and PD output were measured. And then PSD was calculated from them. These results were shown in Fig.1 & Fig.2.

Two types of how to put sensor in oven was tested. Type 1 and 2 are shown in Fig.1 and Fig.2. Type 1  is separeted from oven a little more than 2, which is guessed because how to fix type1 was a little weak.

Each measurement was done at both types.  Temp controller (TC200) has temp sensor (PT100). By using PT100, this temp sensor was calibrated in two types, which result are shown in Fig.3. The reasons of differences between estimation and result are guessed something such as some +offset of OP27(means sensor temp tends to be lower than actual temp). This was measured by Data logger of Moku, which sampling rate was 10kSa/s(too large, so the followings were 1kSa/s.)

When PT100 showed nearly room temp or higher a little, TC200 was turned on and started to controll temperature by 36.3 °C. PID-controll's settings of TC200 were P=230, I=220, D=40, which were set before this experiment. Target temp was set to 36.5 °C. The reason why target temp was set to 36.5 °C was because it tended to be settled at +0.2 higher than target temp at this settings. This results at â‘  are shown in Fig.4. This behaviour was +0.2 higher than the temp displayed on TC200.

And then Fig.5 shows that the temp behaviour was monitored when TC200 set temp changed steeply. The temp displayed on TC200 started from 36.5°C to 37.3 and from 37.3°C to 35.4. But the sensor temp was not like that. It is strange, so it will be measured again at next.

In the situation of using SR560, the same measurement of Fig.5 was also done, which result shows in Fig.6. It seems that the sensor temperature changed lower than the set temperature because it was damped by AC coupling of SR560. It means that it is nesesary to detect temperature without AC coupling.

Temp sensor is shown in Fig.1. Resistor and capacitors parameter were measured actually. Gain coefficient to temp was estimated as 9.82mV/K.

Before calibration, the TF SR560 was measured and shown in Fig.2. It means that AC coupling one is better than another for remove DC.

It worked well by using AC coupling of SR560. Calibration was not yet. SR560 was brought from CNL to QIL.

It is enough to measure AC components of temperature modulation, so it is not nessesary to update this circuit.

Instead of its update, some AC coupling measurement should be used by SR560 and so on. 

Temperature sensor shown in ① of Fig.1 and Fig.2 was made and tested at room temperature. Its output at room temperature(〜25℃) was about -3V, shown in Fig.3, which was the same estimated in advance. But it was difficult to identify temperature change because its coefficient of output voltage versus temperature was not enough, which estimation was 10mV/°C. It means that the sensor needs more gain, which can be obtained by replacing resistor in front of output. It will also cause offset voltage increasing, so its offset should be changeable. Considering these, the scheme shown in ② of Fig.1 is better than ①. It has two variable resistances, one is for offset and another is for gain. So the circuit scheme ② will be as the next step.

For the characterizing the noise coupling of crystal temperature, it will need some modulator and sensor about crystal temperature.

Crystal oven controller TC200(Thorlabs) can be used as temperature modulator by python code, which have it to do modulation by pulse width modulation.

Temperature sensor is consist of AD590JH and some amp circuit. Fig.1 shows some plan of circuit scheme. They are not finalized yet.

Ref: [1]AD590JH data sheet

Laser output in front of the doubling oven was measured to obtain its relation to laser source current. This was not behind laser shutter but in front of the doubling oven, so it maybe contain some loss than the shutter.

Its result and fittng curve are shown in Fig.1. It gives 0.75 [W/A].

To do list for noise injection for noise coupling at single pass;

(i) Laser source (AM→Current laser diode, PM→Laser cavity PZT and T→Temperature laser crystal)

(ii) Crystal oven temperature(controlled by TC200 thrlabs)

For this mesurement, Moku:Go(labeled "MG2") was brought from 40m to QIL.

Preparing for these detection, QE η was measured again, which had the peak 0.28%/W at 36.3 °C. The noise injection will be done around this peak.

This calculation was mistaken. The coefficients not for Si but InGaAs was used, which was not correct. The revised version attached is 43% smaller than before.

The setup was decided to 50μm waist size at z = 0.031. This is the same of November 1th's one[Ref]. In this setup at 36.7 °C, the output green laser (532nm) was 1.4mW when input power was 506mW, which means QE was 0.55%/W. This configuration will be used for next step, which is the specification of noise couipling at single pass SHG.

As other beam waist desin, lens(f=150) was put at 672mm and lens(f=300) was put at 1739mm. The base of doubling oven was located at 2170mm. These distance is all from laser shutter. Then its beam shape is shown in Fig.1. The waist size was 42μm which is smaller than before, but beam divergence is bigger. So a part of input or output beam might be blocked. 

QE was also decrease, which is shown in Fig.2. It seemed to be from beam divergence. The optimaization is semmed to need more precise measurement step and logic, which will be done back to 11/1's setup.

Before detection, output beam was separeted by HBS(Thorlabs HBSY051 - Ø1/2"). About this HBS, 0.7% of 1064nm transmits. But its separetation was seemd to be not perfect, so the SHG(532nm) measurement was done by two types of PD(Si, InGaAs). These spec sheets was shown in [1], [2]. Si PD can detect both of 1064nm and 532nm, so Si PD output is not pure 532nm. In contrast, InGaAs PD output is only 1064nm. By substract 1064nm input power at InGaAs from both input power at Si PD, pure SHG output can be detected.

  , in consideration with each Gain at each wavelength at each PD. Although Si PD output might contain some 1064nm in my assumption, the measured ratio of 1064nm by 532nm at Si PD is ~ 1%. It is not problem.

In this way, QE measured at different temperatures is shown in Fig.2. The highest was 0.6 %/W at T = 36.3 ℃. Its estimated efficiency was ~ 3 %/W, so this situation was not optimized yet.

Ref: [1] Si PD(PDA10A)     [2] InGaAs PD(PDA10CS)

For smaller waist, a part of f seup was revised. From laser shutter, lens (f=150) is located at 0.773m and lens (f=300) is located at 1.662m. 

For calculation, the point of z = o was set at 1.866m. The front edge of doubling oven is located at z = 0.28. The beam shape in this situation is shown in Fig.1. Using data point, beam shape curve was made by least-squares method fitting. The rectangle in figure is crystal. This waist size is 50μm in Fig1. But this is not considered about crystal's refractive index. The actual waist size in this situation was 67.6μm by calculation. This is not optimized yet.

PPKTP crystal length is 30 mm. Recalculated optimum beam waist wp is 31.236 μm.

The KTP (CW) laser damage threshold at 1064 nm is 260 MW/m² at reference [3]. The laser intensity at the experiment last week is 12 MW/m². Using  optimized paremeters, the laser intensity is 163 MW/m², which is behind laser damage threshold.

Ref: [3] KTP damage

[Shoki, paco]

We made a plan on the next steps for this setup:

1. Check the length; is it 20 or 30 mm?
2. Recalculate optimum w0 inside of the crystal to maximize upconversion efficienciy (L/2*zR = 2.84)
3. Confirm the KTP (CW) laser damage threshold at 1064 nm is well above the peak intensity for the calculated w0 and pump power of 500 mW.
4. Scan temperature vs conversion efficiency --> Fix setpoint to max
5. Scan w0 vs conversion efficiency --> Mode match to max
6. Scan input polarization (HWP angle) vs conversion efficiency --> Fix angle to max

At this point, we will have characterized the single pass upconversion quantum efficiency of SHG (degenerate SFG). The next step is to estimate thermal, PM and AM (intensity) noise couplings to the overall quantum efficiency (PD output).

Ref: [0] Izumi's thesis, [1] Ray transfer matrices, [3] KTP damage

The lens (f=51.5 [mm]) was put at z=0.438 [m] for focusing to PD. The dichroic mirror(HBSY051) was placed bihind the oven. After that, Si PD of PDA20C52 (direction of transmitting) was 75mm far from the lens and InGaAs PD of PDA10A (direction of reflection) was 50mm far from the lens. It is not really necessary that the dicroic mirror and two PDs have to be fixed to this position (Fig.1).

The laser power was 47.6 [mW] (laser source current was fixed to 0.9 [A], crystal temperature was fixed to 20℃). But this power was not sufficient for SHG, so the laser power was reset to 210 mW (laser source current was fixed to 1.1A, crystal temperature was the same). When the input power was 215 mW and crystal temperature was 36.7℃ (setting was 36.9℃), SHG output was 5.4 [mV] at Si PD. Si Photodiode(PDA10A) has responsivity 0.25 [A/W] at 532nm and transimpedance gain 1×10⁴ [V/A]. Output voltage 5.4 [mV] means 2.16 [μW]. This QE is 0.00103 %.

Some result is shown in the following. Its optimization will be done next week.

Laser current = 1.1 A Input power = 210.0 mW PD output = 5.4 mV Power measured after SHG = 2.16 μW QE = 0.00103 %

Laser current = 1.2 A Input power = 323.0 mW PD output = 11.2 mV Power measured after SHG = 4.48 μW QE = 0.00139 %

Laser current = 1.3 A Input power = 435.0 mW PD output = 56.0 mV Power measured after SHG = 22.4 μW QE = 0.00515 %

Laser current = 1.4 A Input power = 558.0 mW PD output = 93.0 mV Power measured after SHG = 37.2 μW QE = 0.00667 % 

Crystal was put far from laser, so lens (f=1000) was placed 1 inch in front of the first folding mirror and lens (f=300) was placed 70cm in front of the third folding mirror for focusing roughly (Fig1&Fig2).

Beam shape was measured as x = 0 is at the third mirror (on the far side of Fig2). Distance between laser shutter and the third mirror (x=0 point) is 1865mm. Fig.3 shows the data points and fit curves. Blue means x and red means y. The beam waist was located at x = 0.3[m]. Considering them, the doubling oven was installed as its center located at x = 0.3 [m].

The laser power was 47.6mW when laser source current was fixed to 0.9A and crystal temperature was fixed to 20℃.

I measured beam shape setup (Attached Fig1 & Fig2). I put doubling oven and crystal around x = 0.25m in Fig2.

I checked the laser power is 41.5mW when laser source current is 0.9A. One silicon PD(Thorlabs, PDA10A) and two InGaAs PDs(Thorlabs, PDA20CS2 and PDA10CS) were brought from CTN to QIL as per Paco's suggestion.

Si Photodiode(PDA10A) has responsivity 0.25A/W at 532nm. InGaAs Photodiode(PDA10CS) has responsivity 0.7A/W at 1064nm.

I measured the beam through KTP crystal, but I didn't find up-converted 532nm. It maybe come from mis-alignment or smaller beam waist than requirement.

Also the beam profiler (Dataray Beam'R2-DD) and a laptop for it was brought from 40m to QIL as per Paco's suggestion. Optical components are not aligned yet. And I borrowed a mirror and lens(f=300) from Gyroscopeoptics shelf.

For experiment to get High QE, we start to make single pass SHG using PPKTP as a simple setup. This purpose is measuring its efficiency η and specification about some noise coupling.

I moved a Thorlabs piezo controller from QIL south optics table to Cryo lab. The controller output was already unplugged, so I only had to disconnect the input and power. 

We discussed selecting a coating for the inner shield with well-measured emissivity up to 30 um. I found a NASA black coatings summary document; the first 3 attachments are coatings of interest taken from there (there are several other coatings measured - I'll summarize these soon).

For P-type doped silicon, I found a plot from this page relating the absorption coefficient (α) to wavelength, for various doping concentrations [Attachment 1]. The curves correspond to the following doping levels:

1 -- 4.6·10¹⁷ cm^-3;
2 -- 1.4·10¹⁸ cm^-3;
3 -- 2.5·10¹⁸ cm^-3;
4 -- 1.68·10¹⁹cm^-3.

This is consistent with our intuition that the more highly doped, the greater the absorption and thus the emissivity. Our P/B doped sample sample has a resistivity of 0.001-0.005 ohm-cm; this corresponds to a Si P-type doping concentration range of (1.16e20 - 2.07e19) (calculated here). The closest curve to our sample is (4) in green. 

The peak wavelength of 123K radiation is ~30um, not included in the x-axis range. If we assume this increasing relationship continues to the right, we can imagine an absorption coefficient around ~10⁴ cm^-1 for our sample. But so far I haven't found a resource to support this. 

For thin wafers, emissivity scales with absorption, so we can assume the doping concentration would have the same relationship with emissivity. 

Can you post here some mode or plots of how the emissivity should change vs T for a few different doping levels? i.e. what should we expect for this wafer. Maybe that's in the paper your writing already...

[Radhika, JC]

We borrowed cryo varnish from cryo lab. We varnished an RTD to a new sample: a doped Si wafer [Attachments 1,2]. See the previous ELOG for motivation/next steps.

The mylar foil wrapped around the flexible copper strap had been slowly tearing over time, so today JC used another piece of mylar to patch up the original wrapping [Attachment 3].

We finally mounted the wafer onto the steel wire cradle [Attachments 4, 5] and closed up after testing the rest of the RTD/heater connections. Vacuum pump turned on at 3:00pm, cryocooler turned on at 3:40pm.

The last work done on Megastat was a full cleanup of the Aquadag foil. The state of the cryostat was left as in Attachment 1. 

We return to optimizing Megastat for emissivity testing of 3" wafers. Previous modeling and emissivity estimation resulted in overestimates for bare, undoped Si emissivity. The model included a conductive cooling term that ended up being non-negligible for previous data. As a result, I wanted to brainstorm/test alternative wafer mounting that minimized conductive contacts.

The suspension setup for large, cylindrical samples uses steel wire loops. I used the existing 4-40 screws/nuts around the covered apertures of the inner shield to extend the steel wire across the chamber. I used 4 pieces of wire, 2 in each direction to create a cradle or net for the wafer. It took me a couple iterations to get the right tension in the wires. At the end I tried resting a wafer sample on top [Attachment 2] and it held quite stably. The cradle area is large enough for the wafer and its attached RTD leads to rest without observed instability. While this is not an earthquake-proof setup, neither was the previous mounting setup since the wafer could have slid off the ball bearings. I'll want to place a test sample and do a cooldown cycle to make sure that it can widthstand the pressure/temperature cycling. 

I removed the black baseplate for the ball bearings and left the chamber in the state shown in Attachment 3. The cryo varnish in QIL has fully dried up, so we need to obtain more (either from another lab or order more) so that RTDs can be attached. I want to do a test cooldown of any sample to determine if a fully radiative model can accurately fit the data. If this is true, we would reduce model complexity and the number of parameters for MCMC to deal with. Then, I hope that even if the full enclosure area is not black, we can fit for a net effective emissivity of the environment (and the model dependency on this parameter would be minimal). 

I moved one Moku:Go to cryo. Cryo lab now has 1 Moku:Go.

We borrowed 1x HP 8560 spectrum analyzer from QIL to Cryo

[Radhika, JC]

Today we decided to clean up Megastat and prepare it for the new coldplate insert (design almost finalized). The idea behind the insert is to paint it with Aquadag and thus have a high-emissivity surface above the cold plate (since we cannot paint the cold plate with Aquadag directly). The insert gets rid of the need for aquadag-coated Al foil that made the chamber super dirty (aquadag flakes everywhere, including RTDs and sample). The goal for today was to remove as many traces of the aquadag as possible from the cold plate and shields. 

Attachment 1 shows the initial chamber contents. We then did the following:

In this state the chamber has minimal aquadag residue from foil. We loosely closed it up; when we are ready to prepare for the next cooldown, we will need to:

Storage organizing session Jun 13, 2023
I'm posting the log here because there is no other appropriate place to post.
JC, Shruti, Aaron, Yahonathan, and Rana worked on organizing the WB SB storage cage.

• The left half was cleared at the beginning
• Wire shelving was built in the cage
• Placed the heavy vacuum stuff at the bottom, then placed the electrical instruments on the shelf. We still have a lot of space to put. Victory! (Attachment 1)
• We have some items to be transported to the 40m, including a picomotor driver that would be used for the 40mBHR (Attachment 2)
• We found a potential clean room table set. Bring the set to the 40m so that we can use it once the large optical table is removed. (Attachment 3)
• The previous bottom wooden pallet was removed. We will toss it. (Attachment 4)

Action Item:
Having one or two flat/piano dollies in the cage would be great.

[Yehonathan, Shruti]

Attachment 1:

• The blue dotted curve is a fit to measured data extrapolated to higher power. With the previous setup, we were not yet limited by loss (since we didn't see the level saturation), but by a combination of nonlinear gain and input coupling/power. The x-axis is the power measured exiting from the laser.
• For the same 75% loss that we fit to data, for 4x the current nonlinear gain we expect to reach the loss limit for measurable squeezing at ~200 mW (assuming the same input coupling)
• If we want to go beyond 4 dB, we need (1) the total losses in the output chain to be <40%, (2) the combination of nonlinear gain and input coupling to be more than 3x what we currently have, in order to see squeezing with our current laser whose max power is <150mW

[Yehonathan, Shruti]

Direct measurement of squeezed spectrum incl dark noise (Attachment 1)

With the noise locking on for the squeezed quadrature (red) and anti-squeezed quadrature (blue), we tried to measure squeezing in whole spectrum using the spectrum analyzer on the Moku:Go that was 'out of loop'. We do see some squeezing and anti-squeezing down to 30 kHz somewhat until where the PD dark noise dominates the noise spectrum.

Squeezing level with pump power

Attachment 2 shows noise traces when no green power is sent to the waveguide (shot, gray trace), when locked to the anti-squeezed state (blue) and locked to the squeezed state (red). The shot noise abruptly jumped while taking the 100 mW measurement so we had to use the value seen previously at lower pump powers at 100 mW and the later one at 125 mW.

In Attachment 3, we fit the data with the beam-splitter loss model and obtain a value of 75% for the loss. Our previous estimate for the loss was 65% = 100[1 - (0.5  * 0.9 * 0.8)]% where 0.5 is the coupling between the WG and fiber, 10% at the beam splitter, and 80% quantum efficiency at the PD. Notebook with the analysis can be found here.

In Attachment 4, setting the loss=0, we see the squeezing that is actually generated within the waveguide. At 125 mW incident power this is 2 dB.

If we do go to free space, we expect to not have the coupling loss of 65% into the waveguide that we observe now (we now expect 70% coupling into the fiber and 50% between the fiber and waveguide), and therefore should generate 2 dB * sqrt(125/(0.35*125)) ~ 3.5 dB in the waveguide assuming perfect coupling. Not that great of an improvement.