IBD: Calibrating Optical Thickness

From UCSB Nanofab Wiki
Revision as of 21:39, 21 September 2023 by John d (talk | contribs) (added category Process)
Jump to navigation Jump to search

Basic method for calibrating optical thickness, for DBR/multi-layer optical coatings (2 alternating films only). Commonly used for SiO2/TaO DBR mirrors/filters.

  • Get dep rate & refractive index (RIX) of individual SiO and TaO films, using single-deps and J.A. Woolam Ellipsometer or equivalent tool. Approx. rate from previous user is also acceptable.
  • Get RIX at the target wavelength, using "Derived Params" on JAW or Cauchy equation A/B/C params. For example, if targeting a DBR centered at 1550nm (target λ=1550nm), you will want to know the RIX at 1550nm specifically.
  • Calculate approximate dep. time to achieve a 1/4-wave thickness at the target wavelength, for each film (SiO and TaO).
  • For example, if the dep. rate of SiO2 measured at 5.2nm/min and RIX is n1550 = 1.494, then
SiO 1/4λ thickness: d1/4λ = 1550nm / 4 / 1.494 = 259.4nm
SiO 1/4λ time: t1/4λ = 259.4nm ÷ 5.2nm/min = 49.88min = 2993.077 sec
(and then do the same for TaO)
  • Deposit a fabry-perot 1/2-λ of one film (SiO or TaO) cavity onto a Silicon piece, using the above 1/4-λ deposition times (aka. "SiO/4" or "TaO/4" in the following), to calibrate that film's optical-thickness.
  • eg. for SiO Fabry-Perot cavity (aka. SiO-FP):
SiO/4 + TaO/4 + SiO/4 + TaO/4 + SiO/4 + TaO/4 + (SiO/4 + SiO/4) + TaO/4 + SiO/4 + TaO/4 + SiO/4 + TaO/4
  • Here we used only 3 periods of DBR on either side of the 1/2-λ cavity, to speed up the deposition.
  • Measure the reflectivity on Filmetrics F10-RT.
  • Locate the wavelength of the minimum (dip) in the optical spectrum. Spectrum will typically be very broad, due to omitting many of the surrounding DBR layers for speed.
  • For example, the trough might show a minimum at 1600nm instead of the targeted 1550nm.
  • Correct the film's optical thickness as so:
t1/4λ * λtarget / λmeasured = corrected t1/4λ
2993.077 sec * (1550nm / 1600nm) = 2899.543 sec for SiO/4 layers
  • Use this corrected time for all SiO/4 layers.
  • You only need to edit the One "SiO2_dep" step in the IBD recipe, which will also change all "SiO2_dep" steps in the recipe.
  • Here is an example of a Fabry-Perot cavity that was targeting a 1050nm center-wavelength, but measured a 1100nm trough:
    Simulated plot of SiO-FP reflectivity, for a 1050nm target but measured trough at 1100nm
    Example: Targeting a 1050nm SiO-FP, the measured trough location is at 1100nm.
    In this case, one would apply a correction to the SiO/4 time by multiplying by (1050÷1000).
  • Do the same Fabry-Perot correction for the other film, in this case do an "TaO-FP", and apply the new TaO/4 time to the recipe.
  • Perform a test-DBR deposition onto Silicon, eg. 9 periods (less than full, which could be 15 periods or more), and measure on Filmetrics F10-RT, to confirm that center wavelength is in the right spot. An example DBR test-dep targeting 1050nm looks like this:
    Example reflectivity spectra of the 9-period DBR test-dep, showing 1050nm center-wavelength in the middle of the DBR high-reflectivity spectrum.
    Example reflectivity spectra of the 9-period DBR test-dep.
  • Perform full-DBR deposition onto production parts. Include a flat Silicon witness for measuring the final DBR reflectivity spectrum.
  • Sources of error: thick, stressy films exhibit the stress-optic effect, in which compressed films (closer to the substrate) will often show a reduction in RIX. In going from a 9-period DBR to a 18-period DBR, you might see a ~10-20nm blue-shift. Some users for whom such a shift is outside the device tolerance will do a full DBR dep, then apply a % reduction to all dep times to further dial in the DBR reflectivity band.
  • The same method can be used to calibrate optical-thickness for arbitrary multi-layer optical filters.

Note, you can verify/better understand the above method using any electromagnetic thin-film simulator. For example, EMpy has a simple transfer-matrix example for doing this, along with RIX models contributed by Demis. Demis created the above simulations using EMpy.

Developed by Demis D. John, Bob Farrell, Dustin Kleckner, ~2008-2010. This is the same method used by UCSB VCSEL groups years earlier, for calibrating VCSEL MOCVD/MBE growths.  Please consider our publication policy if you publish papers using this information.