ADVANCES IN OPTICAL MANUFACTURING Measurement considerations when specifying optical coatings PETE KUPINSKI and ANGUS MACLEOD Design, specification, and procurement single layers of the con- of optical coatings all benefit when the stituent coating materials designer has a good understanding in the past and has deter- mined, within the mea- of measurement techniques and surement uncertainty of uncertainties. their spectrophotometer, Accurately measuring the performance that there is little scatter or absorption of an optical thin-film coating can be as in any given layer of the filter. challenging as designing and manufac- Once the coatings are deposited, if the turing it. Understanding measurement performance is tested and reported only techniques and uncertainty when spec- in transmission, the broadband mirror ifying and procuring optical coatings is and passband regions of the filter will important for optical system designers. both appear to be within the designer’s This article is meant as a brief intro- specified tolerance. With broadband duction to optical coating measurement dielectric mirrors, however, it is possi- for optical system designers, quality en- ble to have very different performance gineers, and purchasing agents. In the when measured in transmission vs. re- course of the article, we hope to touch flection. Broadband dielectric mirrors on several of the more common mea- create resonant structures within the surement questions that are asked of multilayer stack, leading to field ampli- coating engineers when optical systems fication at certain wavelengths. Even a are being specified. very small amount of loss in individual coating layers results in large losses for Understanding how the filter at these resonant frequencies. A dielectric multilayer stack coating, measuring 0.5 µm in total thickness, that performance is measured When measured in transmission, this has been intentionally released from the The following is an example of a simple resonant loss has a positive influence optic surface before imaging. The image was measurement oversight with a potential- on blocking, making the mirror appear taken at 200x using a Nomarski microscope. ly large effect on system performance. better than it actually is (see Fig. 1a). This example is meant to highlight When measured in reflection, this same the need to ask questions about how resonance creates an obvious and sig- coating laboratories. They are more things are measured when purchasing nificant ripple in reflectivity across the versatile and less expensive than la- a coated optic. visible spectrum, and consequently the ser-based measurement systems and, In this example, a system designer is coating doesn’t meet the specification when used properly, can be very ef- interested in including an optical filter for the mirror (see Fig. 1b). fective. Most coating companies use in the system that reflects a minimum commercially available spectrophotom- of 99% intensity in the 350–450 nm Measuring specular eters; standard features include grat- wavelength band and transmits greater performance ing-based monochromators and a ref- than 99% at longer wavelengths. The Spectrophotometers are the prima- erence beam to deal with light source coating engineer has measured loss in ry measurement tool in most optical and detector drift. Reprinted with revisions to format, from the October 2015 edition of LASER FOCUS WORLD Copyright 2015 by PennWell Corporation a) 100-measured % transmission 100 95 90 85 80 75 70 320 370 420 470 nm b) % reection 100 95 90 85 80 75 70 320 370 420 470 nm ADVANCES IN OPTICAL MANUFACTURING a) 100-measured % transmission b) % reection 2. Detectors are often spatially vari- 100 100 able. The effect of a small change in beam 95 95 position from reference to sample mea- 90 90 surement is small when measured in re- 85 85 flection and large in transmission for the 80 80 reason outlined in point 1. As the position 75 75 70 70 of the beam changes between baseline and 320 370 420 470 320 370 420 470 sample measurements, the accuracy of the nm nm measurement suffers. Beam-path changes FIGURE 1. Spectral measurements are shown for a multilayer broadband dielectric mirror in become more significant as sample thick- transmission (a) and reflection (b). Ripple in the mirror as measured in reflection is caused by ness, index of refraction, and measurement resonance and amplified loss. angle increase. The degree of beam-posi- tion change is described by Snell’s law of It is important to note that the follow- tightly controlled index of refraction as the refraction, the effect of which can be am- ing discussion is of general considerations high-signal reference instead of an open plified by multiple Fresnel reflections from for commercially available instruments. beam. Schott N-BK7 glass is common- opposing optical surfaces. For thick wit- In some cases, optical coating companies ly used because of its stability in refrac- ness samples mounted at high angles, it will develop their own spectrophotome- tive index. can become nearly impossible to get a rea- ters that have measurement advantages Uncertainty in wavelength for commer- sonably accurate transmission measure- over more general instruments for their cial systems is typically less than ±0.3% ment (a witness sample is typically a small, particular application. for ultraviolet (UV) through near-infrared flat plate of glass that is coated along with A large range of accessories are available (NIR) wavelengths. It is relatively straight- the rest of the components, serving as a for commercial instruments. Each of these forward to check wavelength calibration testable example and record of the partic- accessories has advantages and disadvan- using lamp emission lines and/or trace- ular coating run). Measuring transmission tages for different measurement types. A able glass standards doped to give sharp through a lens using a spectrophotome- full review of these accessories and their absorption peaks in the wavelength range ter is challenging for the same reasons. strengths and weaknesses is beyond the of interest. The challenges outlined above can be ad- scope of this article; we will instead focus dressed by adding an integrating sphere on trying to provide an understanding of Spectrophotometer-based to the detector, using thin witness sam- some general considerations when mea- transmission measurements ples, and decreasing the distance from the suring with spectrophotometers. One of the fundamental challenges of op- sample to the detector. Larger integrat- To make a measurement with a spec- tical coating measurement is that it is dif- ing spheres typically achieve better results trophotometer, known reference measure- ficult to measure AR coatings and mir- because of averaging of light over many ments with intensities less than and great- ror coatings accurately and precisely in bounces within the integrating sphere be- er than that for the surface under test are transmission and reflection, respective- fore it strikes the detector. They typically made across the wavelength band of in- ly. The following points highlight two contain baffles to eliminate direct bounc- terest. A detector-specific equation is then of the larger sources of error when mea- es to the detector and a smaller port-win- used to correlate measured signals at each suring AR surfaces in transmission using dow fraction to reduce the proportion of wavelength to sample intensity within this spectrophotometers: light leaving the sphere.2 range. Uncertainty associated with this 1. For AR surfaces measured in re- Using larger integrating spheres can have calculation is often referred to as error flection, a small error in photometric drawbacks, however, when measuring with from detector nonlinearity.1 In the simplest accuracy equates to a typically insig- a small beam (required for small samples case, reference measurements are taken nificant measurement error of coating or high angles), in that less light reaches with a blocked beam (0% intensity) and performance. For example, a 1% error the detector, leading to potential signal-to- open beam (100% intensity). in measured reflected intensity (R) for noise issues. With a very good setup on a Photometric, or ordinate, uncertainty a surface that only reflects 0.1% of in- commercial instrument, a best-case mea- is partially a function of the range used cident light equates to an inaccuracy in surement uncertainty for a flat, thin AR- for the reference measurements. For ex- measurement of 0.001%R for the sur- coated witness sample at normal incidence ample, when measuring surfaces with face under test. The same error in trans- is typically in the range of ±0.1%T. antireflection (AR) coatings in reflection, mission (T) would be much more signif- Polarizers and variable-angle detector it is common practice to reduce uncer- icant; 99.9% transmission × 1% error modules can be purchased for commer- tainty in detector nonlinearity by using equates to an inaccuracy in transmis- cial spectrophotometers. Variable-angle a single-surface reflection of glass with sion of nearly 1%T. reflectance accessories do an adequate job ADVANCES IN OPTICAL MANUFACTURING of measuring AR performance in reflec- associated with reflectance standards. skewed significantly when using poorly tion. A measure of the Brewster’s angle Ideally, a traceable calibrated mirror collimated test beams. Cropping the ap- can be used to gauge uncertainty in an- is used for the reference measurement. erture of the beam on noncollimated sys- gle and polarization; with a good setup, Degradation of these reference mirrors tems helps, the tradeoff being a reduction these sources of uncertainty are typically over time because of metal oxidation and/ in signal to noise. small compared to the photometric sourc- or mechanical degradation must be con- Also, for high-precision filters, the sig- es of error. A good summary of sources sidered. Total uncertainty in measuring nal bandwidth as dictated by the grating of uncertainty in transmission measure- high reflectors depends on all the terms and slit width can become limiting.