J. Kissel

Recall the SPI pathfinder is pathfinding more than just the future fully assembled sensor array's roll in improving seismic isolation, but also -- at a low-level -- pathfinding some new individual types optomechanical components for LIGO. This aLOG covers the latest in the path for the Schaefter + Kirchhoff (SuK) fiber collimators.

The first of SPI's Schaefter + Kirchhoff (SuK) fiber collimators (DCC:S0272502, ICS:S0272502) is now Class B clean after having gone through one round of massaging with isopropyl alcohol and air-baking (see CNB:2187), at the planned time and levels for the future Class-A vacuum-bake (peak temp 85 [deg C]; ramp up 6 hours, hold for 48 hours, ramp down for 6 hours).

The first open question was "will the lens remain intact and undamaged through the bake." the definitive answer: YES -- the lens remained whole and, at least visually, without defect.
While I didn't use a optical fiber microscope, the lens appears very much intact with no obvious cracks or glinting. See the attached Visual Inspection.

Next up was to check it's optical performance as (hopefully) a more quantitative measure of performance change (hopefully not peformance degradation).
Prior to the bake, I'd set the lens position on the collimator to achieve some level of collimation, and characterized the beam profile -- see LHO:84825.
After the bake, I used the same characterization setup -- augmented only to keep the now Class-B fiber collimator clean (see LHO:86299) -- to remeasure the beam profile to see if the collimator still projected the same beam quantitatively. 

The beam remains as I had collimated it, with the waist within 100 [cm] of the prebake position.
I did *not* adjust the tuned "before" pre-bake lens position at all prior to taking the "after" data.

Check out the 2025-08-07_spifc_S0272502_beamprofile_fit.pdf attachment. 
    - First page compares the two data sets, X axis (parallel to the optical table surface plane) on the top panel, and Y-axis (perpendicular to the optical table plane) on the bottom panel. You can see that the change in waist position / size is 
        Using Delta = (2025-08-06)' - (2025-06-03), and % Diff = [(2025-08-06)' - (2025-06-03)]/(2025-06-03)

        z0x' - z0x = +0.0544 [m] (+4% difference)
        z0y' - z0y = -0.1103 [m] (-8% difference)

        w0x' - w0x = -12.14 [um] (1.3% difference)
        w0y' - w0y = -10.17 [um] (1.1% difference)
Excellent. 

Because the X waist position moved in +Z, and the Y waist position moved in -Z, the beam has become a bit more astigmatic. This is evident in the "Far Field" projection of the model on pages 3 and 5. But, given that the measurement setup is limited with the furthest data point being z(meas_max) - z0 ~ 5.41 [m] - 1.4 [m] ~ 4.0 [m] away from the waist, I wouldn't claim that the measurements perfectly constrain the far-field behavior. Recall we *want* the waist to be at z=0, at the fiber collimator and this collimator's lens position was *not* tuned to that simply because of user error. I'm fully confident I *can* set the lens position such that the waist *is* at the fiber collimator, and after doing so the 5.41 [m] NanoScan position will get a bit more "in to" the far field, which will hopefully better constrain the model, and thus get a better handle on how astigmatic the beam gets after baking (or even *if* it gets consistently astigmatic).

If we define the dimensionless astigmatism parameter, A, as the (zRx - zRy) difference in Rayleigh range, divided by the (zRx + zRy) sum (with the Rayleigh range defined by the fit waist, zR = pi * w0^2 / lambda), then the change (% Difference) in astigmatism is only
    2025-06-03    A  = +2.3983
    2025-08-06    A' = +2.3409 

    (A' - A)/A       = 0.02494 = 2.5%
which seems totally tolerable from pre- to post-bake. Regardless, for the remaining to collimators that we've yet to bake, we'll be setting the collimation (lens position) post-bake anyways, so how it changes across a bake is moot.
Whether the absolute value of A ~ +2.35 is tolerable an open question, that we'll work to answer in the mean time.

Pages 2 and 4 show the model zoomed in to the data between z = 0 and 6 [m]. Here you can also see that the fit doesn't *perfectly* match the data their either, so there's another systematic grain of salt to take with the assessment of change.

Minor note: I was not consistent with the orientation of the collimator w.r.t. to the optical table surface; I oriented the collimator 90-deg from the 2025-06-03 vs 2025-08-06 data (because I found out / rediscovered that the 2025-08-06 position is how you align the p-pol transmission with the optical table; see Figure 5 of T2400413). I've flip-flopped the X and Y axes data for the waist size in the 2025-06-03 data to account for this (which is why the careful reader would notice a difference between this entry's version of the 2025-06-03 results from that in LHO:86342).

But -- all in all -- this looks good enough! 

I've taken these results to indicate that we can move forward with the full Class-A clean-and-bake. 

All three collimators, (S0272502, S0272503, S0272504) in the queue now -- see CNB:2243.