Introduction
One of the most common measurements made by the solar energy industry today is quantification of a material’s surface reflectance. These materials are as diverse as metal coatings, semiconductor coatings, anti-reflective coatings on window material, as well as the window material itself. These measurements are most commonly made between 300 nm and 1500 nm. This is where the solar cell is responsive to energy from the sun. Reflection comes in two varieties, specular and diffuse.
Specular reflection (part A in Figure 1) is generated by a smooth surface. The light ray’s angle of incidence is equal to the angle of reflection; therefore, specular materials frequently produce images on their surface (mirror). Specular reflectance is measured by a number of different types of accessories (VW, VN, IV, and Universal Reflectance Accessory).
Diffuse reflection (part B in Figure 1) is generated by a rough surface. Here the light ray’s incidence angle gives rise to a multiplicity of reflection angles; therefore, images are not produced. Diffuse reflectance is how people see the world. This is because the vast majority of objects in the world are diffuse reflectors. Diffuse reflection is measured by an integrating sphere, which comes in two sizes, small (60 mm diameter) and large (150 mm diameter).
One of the unique requirements of the solar industry is for specular sample to be measured on an integrating sphere. Why? Because the solar industry needs to measure total reflection (specular + diffuse) even if the sample is predominately a specular reflector. Integrating spheres are the only devices which excel at total reflection measurements.
The 60 mm Integrating Sphere
The 60 mm integrating sphere is a low cost reflectance accessory that measures both “diffuse only” and total reflectance. It consists of a hollow 60 mm sphere of highly reflective Spectralon polymer with two holes for the sample and reference beam to enter and two ports for placement of sample and reference material. Background collection (autozero) is performed by placing two Spectralon plates at the sample and reference ports. Since Spectralon has a diffuse reflectance of 99.0+ %R, the reflectance of spheres can be assumed to be close to absolute %R. Because of the size of the sphere and the lack of “baffling” of the detectors in the sphere from first bounce sample reflectance; this sphere type is subject to several types of spectral artifacts such as incorrect %R values and steps at instrumental filter and detector change points.
In the solar industry however, total reflection measurements must be obtained for highly specular samples. Under these circumstances the 60 mm integrating sphere can provide acceptable spectra for diffuse, specular, and combination diffuse/specular samples. We will show here that although the 60 mm sphere can be prone to artifacts, with the use of the proper background correction techniques, excellent and accurate specular reflectance data can be obtained on this type of sphere.
 |
| Figure 2 |
What happens when you measure a totally specular sample with little or no diffuse component in a sphere? In order to answer this question we need to understand what happens to the light inside the sphere. Figure 2 depicts how both diffuse and specular light reflect off the sample and into the sphere. The diffusely reflected light will evenly illuminate the entire area inside the sphere through 380 degrees (Figure 2, left). The specular reflected light, however, will strike only an area along the midline of the sphere in the vicinity of the transmittance port (Figure 2, right).
 |
| Figure 3A |
Photographs of this area, taken via a webcam inserted inside the sphere, are shown in Figures 3 A, B, and C. The photograph in Figure 3A shows the even illumination of the sphere from a 100% diffuse reflectance sample; however, if a specular sample is placed at the sample port a “hot spot” is formed on the wall of the sphere.
 |
| Figure 3B |
 |
| Figure 3C |
This “hot spot” is photographed in Figures 3B and 3C for a 10% R and a 90% R specular sample respectively. Note that there is little illumination of the sphere interior, only the concentrated “hot spot” shows up. The brightness of the “hot spot” is directly related to the reflectivity of the specular sample. Errors result from the fact that the background correction performed with the diffuse Spectralon plate illuminates the sphere differently (see Figure 3A) than the specular sample “hot spot” illumination (see Figures 3B and 3C). The resulting different sphere interior images that the detector measures between background and sample can cause errors and artifacts.
The Test Samples
Four sample types that span a wide range of specular reflectance intensities were investigated. A piece of polished aluminum represents high reflectance (around 90%R), a wafer of silica has medium reflectance (around 50%R), and a piece of dark glass as well as a clear glass slab with back side protection represent low reflectors (around 4%R). The samples were measured in the total reflectance mode on a 60 mm PMT/InGaAs integrating sphere. A fixed 2 nm slit was used in the UV/Vis range and the servo slit mode at medium gain was used for the NIR region. Data was collected every 1 nm with a data point collection time of 16 milliseconds.
The Problem
Figures 4A and 4B graphically display the problems associated with measuring specular samples on a small integrating sphere. The data are obtained with a autozero correction using the typical Spectralon plate.
 |
| Figure 4A |
The first problem is that the two highest %R samples, the NIST mirror (green) and the polished aluminum (red), have values over 100% R. This is an obvious physical impossibility and is due to the intense “hot spot” in a small, non-baffled sphere. The second problem is the step at the UV/Vis-NIR detector change for all samples. Note that if we calculated the size of the step as a percentage of %R, the step is the same size (between 5%R to 6%R) for all samples regardless of their %R intensity. Parameter juggling (UV/Vis silt vs. NIR gain) will not decrease or eliminate these steps. They are a physical reality of the specular “hot spot” inside a small sphere.
 |
| Figure 4B |
When we measure diffuse samples under the same instrument conditions, there are no problems and the data are excellent. As seen in Figure 5, the spectra of the white, gray, and black Spectralon plates lack the artifacts observed with specular samples.
 |
| Figure 5 |
The Possible Solution
Since we know that the artifacts due to specular samples are generated by the difference in internal sphere illumination between a diffuse autozero target (white Spectralon plate) and the “hot spot” illumination of a specular sample, would it be possible to improve the situation by using a specular autozero target. To investigate this we used a NIST front surface aluminum mirror for the autozero target and then measured the test samples under the same instrument conditions as the data above. The results are displayed in Figure 6. The artifacts that appeared when a white Spectralon plate was used for the autozero target appear to be gone when we use the specular mirror in its place; however, on closer inspection we see that we may have traded on set of artifacts for another. The %R values are all below 100 and the steps at the detector change are gone, but there appears to be a “bump” in the spectra at about 818 nm and are the %R values correct?
 |
| Figure 6 |
How could using a NIST mirror as an autozero target cause artifacts? At this point a little information on how a UV/Vis/NIR instrument autozeros is appropriate. When a UV/Vis/NIR instrument is turned on it is literally “as dumb as a stump”. A procedure must be performed to calibrate 100 % (or 0 absorbance) on the instrument. The autozero (background correction) sets the 100 % level for the instrument and is usually performed with a “blank” or sample that is 100% R. A white Spectralon plate fits this criteria well since Spectralon is over 99% R for its usable ranger of 250 nm to 2500 nm. So when Spectralon is used as an autozero target, 100% R is properly calibrated.
But what happens when an aluminum mirror is used for an autozero target? Figure 7 shows the spectrum for a front surfaced NIST aluminum mirror. As we can see the reflectivity is well below 100% R over the entire spectral range. This means that when this mirror is employed for the autozero procedure, a %R value of less that 100% R be set in the instrument’s calibration file to that value. This means that an error in photometric accuracy (%R) is commensurate with a mirror autozero correction. In addition, the spectral features of the mirror in the background correction will be introduce into spectra measured with that correction. As a result the sharp downward peak at 818 nm (or any other additional spectral features) appear as an artifact in all spectra measured with a mirror correction.
Fortunately there is a easy solution to both of these problems.
 |
| Figure 7 |
The Final Solution (%RC)
The solution is a simple mathematical calculation that uses the known values for the autozero target mirror to eliminate both the photometric and wavelength dependent artifacts (see article “Procedure for Creating %RC Correction Files in UVWinlab V6”). Equation 1 below displays the math used for this correction. The R100 spectrum corresponds the the autozero target mirror values, while the R0 spectrum corresponds to the spectrum obtained with a light trap at the sample port or open sample port. The R0 spectrum corrects for the small amount of atmospheric scattered light in the sphere.
 |
| Equation 1 |
When this methodology is employed and the calibrated mirror is measured for the autozero and then run as a sample, you obtain the spectrum seen in Figure 7. Here we do not see the familiar flat 100 %R spectrum obtained before, but rather the actual values for the calibrated (NIST) mirror. With the combination of a mirror used for autozero and the %RC correction math, a spectrum of a specular sample can be obtained that is both photometrically accurate and free of any wavelength dependent artifacts.
In Figures 8, 9, and 10 we see the spectra for the polished aluminum, silica wafer, and back side protected clear glass respectively. The spectra represented in each graph are:
Green Spectrum = white Spectralon plate target for autozero
Blue Spectrum = mirror target for autozero
Red Spectrum = calibrated mirror target and %RC correction used
 |
| Figure 8 |
 |
| Figure 9 |
 |
| Figure 10 |
