Alluxa’s thin film optical filter technology has advanced to the point where the spectral slopes and blocking levels are challenging to even the best metrology equipment and techniques. This article discusses the issues and provides solutions to measure the spectral response of this new range of high performance filters.
Spectral Measurements and Limitations
An optimum instrument for measuring the characteristics of advanced optical filters would consist of a bright monochromatic light source, tunable in wavelength over the region of interest, and a detector with high sensitivity (i.e. low noise floor) and high dynamic range.
For some wavelengths, instruments that approach these ideal requirements are readily available. For instance, lasers and detectors are available for the 1.5 micron wavelength region used for telecommunications. These devices enable measurements of changes in transmission of 100 dB (10 OD) or more with wavelength accuracy measured in picometers.
However, for many wavelengths of interest, particularly in the visible range, practical and affordable tunable lasers are unavailable or are available only over a very narrow wavelength range. Therefore, the standard measurement technique is approximating a monochromatic source using a broadband light source and filtering the light to a narrow wavelength range to resolve the features of interest.
Commercial instruments using this approach usually use a tunable monochromator to filter the spectrum of an arc lamp or light bulb integrated with a semiconductor or photomultiplier tube detector. Figure 1 shows a schematic diagram of such an instrument.
The use of a diffraction grating-based monochromator combined with a suitable choice of broadband sources and detectors, enable a single spectrophotometer instrument to cover the entire wavelength region from the UV to the near IR.
Figure 1. Spectrophotometer schematic
This wide wavelength range and the high performance of today’s commercial spectrophotometers make them popular measurement instruments for production use by optical filter manufacturers. Yet, the recent enhancements in optical filter performance have pushed these instruments to the limit of their ability, forcing users to work around their limitations.
However, these limitations are concerned with the spectrophotometers strength, i.e. its wide wavelength range enabled by the use of a broadband light source.
This article separate these performance limitations into two main areas: limitations in the spectrum of the filtered source (such as brightness and spectral width) and limitations in the measurement light beam output from the monochromator due to the optics that manipulate it.
The first limitation is common. The narrower the filtration of a broadband light source, the dimmer the measurement light beam. Increasing wavelength resolution reduces the dynamic range of the measurement (assuming detector noise floor and source brightness do not change).
Generally, optical filters are needed to block light to below a part in one million of full transmission over a particular wavelength range. That corresponds to an optical density of 6 OD or a 60 dB change in light level. New and more capable filters can increase this to 70 dB, 80 dB, or 100 dB and greater attenuation.
The current high performance commercial spectrophotometers achieve blocking measurements of this magnitude, while maintaining spectral resolution in the nanometer range. As an example, Figure 2 displays the broad blocking ranges measured for a series of visible bandpass filters that are used in scientific applications for fluorescence emission and excitation detection. Alluxa’s thin film filter technology enables the wide square passband regions with steep sides to quickly drop 60 dB below the peak.
Figure 2. Visible bandpass filters with wide deep blocking ranges
The measurement difficulties become more acute in applications needing high dynamic range, high wavelength resolution or both simultaneously. The types of filters that require this measurement ability include those with rapid changes in transmission, such as narrow bandpasses, deep notches, and very steep edge filters.
Optimizing Dynamic Range and Wavelength Resolution
The dynamic range of measurement can be improved for filters where high wavelength resolution is not needed. To enable high OD measurements, the slits of the monochromator may be opened wide enough to permit sufficient light. This works when the wavelength region of high transmission is small in comparison with the blocking regions as shown in Figure 3.
Figure 3. Triband bandpass filters with deep wide blocking regions between relatively narrow passband regions
The noise floor of the measurements is less than -80 dB or 8 OD. In this measurement, the slit width of the monochromator was opened wide enough (2 nm) and the scan speed was slow enough that adequate light reached the detector to enable the measurement of the very deep blocking.
However, the wavelength resolution is seriously compromised with these settings, as shown by the rounded top of the passband. Another measurement of the same filter made with a narrow slit width (0.1 nm) shows the true shape in the high transmission region, as shown in Figure 4.
Figure 4. Triband bandpass filters measured with high wavelength resolution
However, if the blocking region is deep and narrow, then leakage from scattered light can degrade the measured performance substantially. For instance, Figure 5 shows the measurement of a narrow notch filter with high transmission extending widely on each side.
Figure 5. Narrow deep notch with wide transmission window
One of the inherent measurement limitations of these instruments is the scattered light from the monochromator. It may only be significant when measuring certain types of filters.
With this filter, it is important because any scattered light in the high transmission wavelength regions can reach the detector and add to the detected signal due to the narrow blocking region. This restricts the minimum light level the detector sees when in the blocking region, reducing the dynamic range of the detector.
It can be useful to add extra attenuation to the reference path of the spectrophotometer, when the light levels after the sample reach low values. The reference path signal (Figure 1) is compared to the sample path signal in making the measurement. If only one path is highly attenuated and the other is not, then the imbalance of the signal levels on the detector(s) can add to the noise level, reducing the dynamic range.
This can be seen in the notch measurement illustrated in Figure 6, where the two measurements performed were identical, barring the insertion into the reference beam of a broadband neutral density filter with an attenuation of 2 OD.
Figure 6. Effect of reference beam attenuation on noise level of deep blocking filters
Dealing with Beam Geometry
Another limitation is due to the beam geometry of the measurement system and is related to the size of the beam and angular content. In a high performance measurement system for telecommunications filters discussed earlier, fiber optic collimators provide intense, collimated beams (F/#100 or more with beams less than one millimeter in diameter) that usually eliminate beam geometry artifacts.
In the case of monochromator-based systems operating over a broad range of wavelengths, there could be severe limitations caused by beam geometry.
The beam of light used to measure the sample has to be large and, as a result, will contain light propagating at more than one angle. This makes it possible to collect as much light as possible from the monochromator to allow high OD measurements.
Typical values for f/# are around 5 to 8, but this can vary in the vertical and horizontal axes. For instance, the maximum vertical angle of incidence on the filter to be measured may be more than seven degrees while the horizontal angle is around three degrees.
Major measurement errors can occur because thin film optical filters shift in wavelength with angle of incidence. Filters designed to be used at angle such as dichroic beamsplitters are even more sensitive to measurement beam angular content, as the rate of spectral shift with angle of thin films increases almost linearly with angle of incidence.
The standard way to resolve this issue is to restrict the measurement beam to a smaller range of angles by adding apertures into the beam. For instance, Figure 7 shows a measurement of a dichroic edge filter made with and without apertures.
Figure 7. Effect of Apertures on Edge Filter
The rounding of the edge filter shown is not caused by the light at the same narrow wavelength coming in at a relatively large range of angles and not by a wide range of wavelengths (the slit width used was 0.1 nm). The wavelength resolution is independant of the resolution determined by the monochromator slit width.
By reducing the range of angles in the measurement beam, the effective wavelength resolution is increased. In the case of narrow filters, this can lead to severe distortion in the measurement as shown for a ~3 nm wide bandpass in Figure 8.
Figure 8. Distortion of narrow bandpass filter as seen by measuring with and without apertures
As previously mentioned, when the optical filter is tilted, which is common for dichroics and edge filters that are intended to be used at an angle, this effect becomes greater. A thin film optical filter shifts to shorter wavelengths when it is tilted and the shift increases with angle roughly as the square root of the difference of the effective index of refraction squared and the sine squared of the angle.
Therefore, the larger the tilt, the greater is the shift in wavelength of a filter for a particular angular spread in the measurement beam.
Figure 9 shows an example of the difficulties of measuring filters at an angle. This dichroic or edge filter was measured using an aperture at the entrance to the sample chamber in a Cary 5000 spectrophotometer that had an opening of about 2 mm diameter.
An aperture of a similar size was used at the output of the sample chamber at three positions: centered on the measurement beam, one diameter lower, and one diameter higher.
Figure 9. Wavelength shift due to placement of output aperture in measurement beam. The filter was mounted at a tilt of 45 degrees to the measurement beam
This measurement shift of a few nanometers was often unimportant in the past with its lower performance requirements. However, with the current high performance filters and tighter manufacturing tolerances, this measurement shift decides whether a part satisfies or fails customer specifications.
MultipleBandpass Filter Measurements
The measurements become very difficult when the steep edges and deep blocking are combined several times within a single filter in new class of multiband filters pioneered by Alluxa. For instance, Figure 10 shows a high performance triband filter in comparison with a theoretical design on a linear scale.
Figure 11 shows the same filter on a log scale, demonstrating that the blocking measurement is limited. The limit could be improved by opening the monochromator slit width and decreasing wavelength resolution, but this can only be improved in the region between the passbands.
The blocking value proximal to the high transmission region would still be unresolved because of the loss in wavelength resolution. Measuring the blocking amount proximal to the high transmission regions tests a spectrophotometer measurement system to its limit, but it is very important to the filter user.
Figure 10. Triband optical filter in the visible showing high transmission and steep slopes
Figure 11. Triband optical filter showing limit of blocking measurements
In cases such as this, when the performance of even the best spectrophotometer is not adequate to measure the blocking level of these types of filters, an accepted industry approach is to use the “Slope Method” as shown in Figure 12.
Figure 12. Use of the Slope Method to determine wavelength of -70 dB (7 OD) blocking level
This consists of considering the theoretical design and, using the knowledge of the measurement method, such as the rounding of the filter edge at the top, to fit the slope of the theory to the slope of the measured trace. If the fit is good, the slope of the theory can be applied to extend the measurement into the noisy region.
Given a good understanding of the way that potential errors in the optical filter would present themselves in the measurement and with additional measurements to remove the possibility of pinholes and other issues, this technique can extend the measurement limit by another 10 or 20 dB (1 or 2 OD).
Some uncertainty exists in such an approach, but the alternative of developing costly measurement equipment and techniques in-house, even if it is possible for the particular wavelengths and blocking levels involved, resulting in more expensive filters, has made this technique great value.
The performance of advanced thin films, as demonstrated by Alluxa’s thin films, challenge even the best metrology equipment and techniques. This article describes several simple methods, such as reference beam attenuation, appropriate use of spectral slit widths and scan speeds, and placement of apertures in the sample beam.
These techniques, combined with the knowledge of the theoretical filter performance, can enhance the accuracy of almost all spectrophotometer measurements of today’s most advanced thin film filters.
This information has been sourced, reviewed and adapted from materials provided by Alluxa
For more information on this source, please visit Alluxa