Development of a New Reference Spectrophotometer. J.C. Zwinkels, D.S. Gignac

Development of a New Reference Spectrophotometer Development of a New Reference Spectrophotometer J.C. Zwinkels, D.S. Gignac J.C. Zwinkels, D.S. Gignac National Research Council of Canada, Division of Physics Ottawa, Ontario, Canada K1A OR6 National Research Council of Canada, Division of Physics Ottawa, Ontario, Canada K1A OR6 ABSTRACT ABSTRACT A new reference Spectrophotometer is being developed at the National Research Council of Canada (NRCC) for high-accuracy transmittance measurements over the spectral range 200 to 2500 nm. This computerized instrument is a single-beam design based upon a servomotor-driven double monochromator with a wavelength resolution of 0.022 nm. The other main components are: (1) two interchangeable sources (deuterium and tungsten-halogen), (2) two computer-selectable TE-cooled detectors (GaAs photomultiplier tube and PbS cell), (3) all-reflective input and exit optics, and (4) a large sample compartment. The significant feature of the optical system is a large, well-collimated measurement beam: the angle of convergence is 0.7 for a slit height of 7 mm, and the maximal beam size is 37 X 20 mm. This beam geometry eliminates the need for polarization corrections (using linearly polarized light) or compensation for spatially non-uniform detector responsivity (using averaging sphere). The paper describes the instrument design and presents data from preliminary performance tests. The systematic and random sources of error that have been investigated include: wavelength accuracy and reproducibility, bandpass, beam uniformity, polarization, stray light, system drift and linearity. A new linearity tester has been developed for checking the photometric accuracy. This automated device is based upon the double-aperture method but takes advantage of high precision piezoelectric motors to give a single adjustable aperture. Transmittance measurements of several neutral-density glass filters at 546 nm demonstrate that the photometric precision is better than 0.01% of the measured value and that the photometric accuracy is a few parts in 10 4. The wavelength scale is accurate to better than ±0.1 nm from 300 to 2500 nm, and is reproducible within ±0.03 nm. A new reference spectrophotometer is being developed at the National Research Council of Canada (NRCC) for high- accuracy transmittance measurements over the spectral range 200 to 2500 nm. This computerized instrument is a single -beam design based upon a servomotor- driven double monochromator with a wavelength resolution of 0.022 nm. The other main components are: (1) two interchangeable sources (deuterium and tungsten -halogen), (2) two computer -selectable TE- cooled detectors (GaAs photomultiplier tube and PbS cell), (3) all- reflective input and exit optics, and (4) a large sample compartment. The significant feature of the optical system is a large, well -collimated measurement beam: the angle of convergence is 0.7 for a slit height of 7 mm, and the maximal beam size is 37x20 mm. This beam geometry eliminates the need for polarization corrections (using linearly polarized light) or compensation for spatially non -uniform detector responsivity (using averaging sphere). The paper describes the instrument design and presents data from preliminary performance tests. The systematic and random sources of error that have been investigated include: wavelength accuracy and reproducibility, bandpass, beam uniformity, polarization, stray light, system drift and linearity. A new linearity tester has been developed for checking the photometric accuracy. This automated device is based upon the double- aperture method but takes advantage of high precision piezoelectric motors to give a single adjustable aperture. Transmittance measurements of several neutral- density glass filters at 546 nm demonstrate that the photometric precision is better than 0.01% of the measured value and that the photometric accuracy is a few parts in 104. The wavelength scale is accurate to better than ±0.1 nm from 300 to 2500 nm, and is reproducible within ±0.03 nm. 1. INTRODUCTION 1. INTRODUCTION Spectrophotometry involves the ratio measurement of two radiometric quantities. The accuracy of the measurement is defined not by a fundamental standard but by the design and calibration of the measuring instrument, and by the sample itself. However, most commercial spectrophotometers are designed more for ease of operation and reliability rather than the highest accuracy. Their double-beam design and restricted source, sample and detector compartments do not readily permit an investigation of their systematic errors. This limits their measurement uncertainty to typically a few parts in 10 3. In order to achieve higher accuracy spectrophotometric measurements at the level of 1 part in 104, it is necessary to build a reference instrument which is sufficiently simple in design and versatile in function that all systematic errors can be identified, separated, and accurately characterized. Spectrophotometry involves the ratio measurement of two radiometric quantities. The accuracy of the measurement is defined not by a fundamental standard but by the design and calibration of the measuring instrument, and by the sample itself. However, most commercial spectrophotometers are designed more for ease of operation and reliability rather than the highest accuracy. Their double -beam design and restricted source, sample and detector compartments do not readily permit an investigation of their systematic errors. This limits their measurement uncertainty to typically a few parts in 103. In order to achieve higher accuracy spectrophotometric measurements at the level of 1 part in 104, it is necessary to build a reference instrument which is sufficiently simple in design and versatile in function that all systematic errors can be identified, separated, and accurately characterized. At the NRCC, such a reference Spectrophotometer has been recently designed and constructed to perform regular transmittance measurements over the wavelength range 200 to 2500 nm. Regular transmittance is the undeviated component of transmitted flux, as contrasted to the diffuse component. This mode of measurement is used for transparent, nonscattering samples. However, a modular design approach has been used to allow ready extension to other measurement modes at a future date. This paper presents details of the instrument design, calibration procedure, and current performance capabilities. At the NRCC, such a reference spectrophotometer has been recently designed and constructed to perform regular transmittance measurements over the wavelength range 200 to 2500 nm. Regular transmittance is the undeviated component of transmitted flux, as contrasted to the diffuse component. This mode of measurement is used for transparent, nonscattering samples. However, a modular design approach has been used to allow ready extension to other measurement modes at a future date. This paper presents details of the instrument design, calibration procedure, and current performance capabilities. 2. INSTRUMENT DESIGN 2. INSTRUMENT DESIGN A schematic drawing of the instrument is shown in Pig. 1. The double monochromator is a commercial item, but the rest of the system was developed at NRCC. The main features of the instrument are: A schematic drawing of the instrument is shown in Fig. 1. The double monochromator is a commercial item, but the rest of the system was developed at NRCC. The main features of the instrument are: (1) a single-beam design which allows a clear separation of its various systematic errors; (2) a well-collimated measurement beam which, for normal-incidence transmittance of a plane-parallel sample, unambiguously defines the path length through the sample, and eliminates the effects of instrument polarization and beam geometry errors such as beam displacement and defocussing; and (3) a large sample compartment to accommodate a variety of samples and sampling accessories. The large sample space, in combination with off-axis mirrors in the exit optics and good imaging properties, also reduces the probability of interreflections between the sample and the monochromator and/or detector. (1) a single -beam design which allows a clear separation of its various systematic errors; (2) a well -collimated measurement beam which, for normal- incidence transmittance of a plane- parallel sample, unambiguously defines the path length through the sample, and eliminates the effects of instrument polarization and beam geometry errors such as beam displacement and defocussing; and (3) a large sample compartment to accommodate a variety of samples and sampling accessories. The large sample space, in combination with off -axis mirrors in the exit optics and good imaging properties, also reduces the probability of interreflections between the sample and the monochromator and /or detector. SPIE Vol 1109 Optical Radiation Measurements 11 (1989) / 89 SP/E Vol. 1109 Optical Radiation Measurements II (1989) / 89

Sample Compartment Sample Compartment PbS Cary 14 Monochromator Gary 14 Monochromator He -Ne Loser I21 2.1 Sources 2.1 Sources Figure 1. Optical layout of NRCC Reference Spectrophotometer The UV source (200-350 nm) is a deuterium lamp with a 1 mm circular aperture (Cathodeon D902). It is operated at a current of -300 ma (30W) and its power supply is stable to 0.01 %. The VIS -NIR source (350-2500 nm) is a tungsten -halogen lamp (Q6.6A /T4 /1CL), operated under constant current regulation at 6.5A (200W) to 0.005 %. The coiled -coil filament is 10 mm long by 3 mm wide. 2.2 Input optics The condenser optics comprise a plane mirror, an 80 mm diameter spherical concave mirror (f =125 mm) and a 50 mm diameter cylindrical convex mirror (f = -500 mm) to image the source onto the monochromator entrance slit with a magnification of x2. This arrangement provides adequate illumination of the grating aperture with both the deuterium and tungsten- halogen sources. The overfilling in the case of the tungsten -halogen lamp is desirable in that it minimizes sensitivity to misalignment. The spherical concave mirror is oversize with an aperture stop to prevent the beam from catching the edge of the optics and producing stray light. The purpose of the cylindrical mirror is to correct for astigmatism. All mirrors are coated with aluminum and SiO. A computer -controlled shutter, immediately before the monochromator, is used to determine the dark current signal. There is also provision in the compact condenser optics for adding a light chopper for NIR operation and a polarizing prism for special polarization studies. 2.3 Monochromator Figure 1. Optical layout of NRCC Reference Spectrophotometer The UV source (200-350 nm) is a deuterium lamp with a 1 mm circular aperture (Cathodeon D902). It is operated at a current of -300 ma (30W) and its power supply is stable to 0.01$. The VIS-NIR source (350-2500 nm) is a tungsten-halogen lamp (Q6.6A/T4/1CL), operated under constant current regulation at 6.5A (200W) to 0.005%. The coiled-coil filament is 10 mm long by 3 mm wide. 2.2 Input optics The condenser optics comprise a plane mirror, an 80 mm diameter spherical concave mirror (f=125 mm) and a 50 mm diameter cylindrical convex mirror (f=-500 mm) to image the source onto the monochromator entrance slit with a magnification of x2. This arrangement provides adequate illumination of the grating aperture with both the deuterium and tungsten-halogen sources. The overfilling in the case of the tungsten-halogen lamp is desirable in that it minimizes sensitivity to misalignment. The spherical concave mirror is oversize with an aperture stop to prevent the beam from catching the edge of the optics and producing stray light. The purpose of the cylindrical mirror is to correct for astigmatism. All mirrors are coated with aluminum and SiO. A computer-controlled shutter, immediately before the monochromator, is used to determine the dark current signal. There is also provision in the compact condenser optics for adding a light chopper for NIR operation and a polarizing prism for special polarization studies. 2.3 Monochromator The monochromator is a commercial item (Gary Model 14) 1 with a prism predisperser (f=30 cm) for spectral order sorting and a 600 line/mm echelette grating (f=40 cm), which both operate with an aperture ratio of f/8. This monochromator is noteworthy because of its wide spectral range, high resolution, low stray light, and wavelength scale reproducibility. The entrance, intermediate, and exit slits are simultaneously adjustable from -0.01 to 3-0 mm and the slit height is fixed at either 7 or 20 mm. To reduce the effects of stray light and slit curvature mismatch, a slit height of 7 mm is routinely used. The reciprocal linear dispersion is 3.55 nm/mm at 550 nm, ie. giving a nominal bandpass of 3-55 nm with a 1 mm slit width. The monochromator is a commercial item (Cary Model 14)1 with a prism predisperser (f =30 cm) for spectral order sorting and a 600 line /mm echelette grating (f =40 cm), which both operate with an aperture ratio of f /8. This monochromator is noteworthy because of its wide spectral range, high resolution, low stray light, and wavelength scale reproducibility. The entrance, intermediate, and exit slits are simultaneously adjustable from -0.01 to 3.0 mm and the slit height is fixed at either 7 or 20 mm. To reduce the effects of stray light and slit curvature mismatch, a slit height of 7 mm is routinely used. The reciprocal linear dispersion is 3.55 nm /mm at 550 nm, ie. giving a nominal bandpass of 3.55 nm with a 1 mm slit width. 90 / SPIE Vol. 1109 Optical Radiation Measurements II (1989) 90 / SPIE Vol. 1109 Optical Radiation Measurements II (1989)

The monochrornator monochromator sine-bar mechanism is driven by a leadscrew coupled to a servomotor (PMI). The wavelength position is read by an absolute encoder (Litton, 17 bit) which is controlled by two anti-backlash gears with a gear ratio of 3.5:1. The larger gear is mounted on the shaft of the leadscrew and is rotated 360 about every 20 nm. This arrangement gives a wavelength resolution of 0.022 nm. A coupling plate also mounted on the shaft of the leadscrew provides thermal isolation from the motor drive, thereby minimizing instrumental wavelength shifts due to leadscrew expansion. 2.4 Exit optics The light beam emerging from the monochromator exit slit is diverted by a plane mirror to a 60 mm diameter spherical concave mirror tilted at at 15 15 (f(f=300 mm). This produces a large, highly-collimated beam. The maximum beam dimensions are 37 mm high by 20 mm wide, and the degree of convergence for a 7 mm monochromator slit height is is 0.7 0.7. A different size and shape of beam irradiating the sample are obtained with a precision limiting aperture. The sample compartment is 25 cm long. The usual sample holder is an automated stepper motor-driven filter wheel with six positions and a resolution of of 0.009. This can accommodate either 50 mm square or circular filters, from 1 to 10 mm in thickness. A novel automated linearity tester, described in Section 3.6, can also be inserted into the measurement beam. At the far end of the sample compartment, the light beam is diverted by a plane mirror to a second spherical concave mirror (f=300 mm) also tilted at 15 and to a stepper motor-driven plane mirror which automatically selects the appropriate detector during the spectral scan. 2.5 Detectors The UV-VIS detector (200-900 -900 nm) nm) is is a a photomultiplier () with a GaAs cathode (RCA 4832; 12.7x5 mm cathode with 9 circular-cage dynodes) used behind a ground Suprasil diffusing plate (50 mm diameter). Because the monochromator exit slit, not the sample, is imaged onto the detector, and the beam is well-collimated, a diffuser- combination compensates for slight beam displacement due to insertion of imperfect samples. This particular was chosen for its stability, its flat spectral response over a wide wavelength range, and its very low dark emission. To further increase the sensitivity and lower the dark current noise, the is TE-cooled to 258 K. To improve linearity of response, 110 V Zener diodes are placed between the cathode and first dynode and between the last dynode and ground, and the anode current is is taken directly to to a a current-to-voltage converter with a 10 MO Mft feedback resistor. The high voltage power supply is regulated to 0.001?. %. A typical maximum operating signal is 8V (0.8 ya) pa) for a bias voltage of 1000 V. Measurement precision is improved by integrating the signal with a 5 1/2 digit digital voltmeter (DVM). The average of 600 readings corresponds to to a a 1010-sec integration period. For the NIR range (850-2500 nm), a lead sulfide (PbS) detector (Quantum Detector Technology, 8x8 mm), operated with a 100V bias voltage, is used. It is TE-cooled to 258 K and the source is chopped at 90 Hz, to reduce the influence of the PbS's high dark level. The modulated signal, across a 1MS2 IMfl load resistor, is amplified by a wide-ranging (10 nv to 500 mv) lock-in amplifier. The lock-in or or the DVM perform analogue or digital signal averaging, respectively, under computer control. 2.6 Automation A block diagram of the signal processing in the NRCC Reference Spectrophotometer is shown in Fig. Pig. 2. The instrument operation is automated by by a a microcomputer (Hewlett-Packard 9817) programmed in BASIC. Software has been written to evaluate the various instrumental systematic errors, and to perform transmittance measurements, data storage on 3 1/2" diskettes, and statistical analysis of results. Measurement parameters that are under software-control include: wavelength range, wavelength step for pause and measure mode, detector settling time before measurement, number of readings, and measurement sequence for dark, 100% 100? and sample readings. SPIE Vol. 1109 Optical Radiation Measurements II ll (1989) / / 91

Thermoelectric Cooler Regulated H.V. Supply UV/VIS Detection Printer Plotter GPIB Current to Voltage Converter Shutter Drive 5 f'z digit DVM GPIB Micro Computer GPIB 17 Bit Abs Encoder Interface Encoder I Wavelength I Drive Wavelength Drive Thermoelectric Cooler Lock -In Amplifier GPIB (1M RAM) Servo Drive Amplifier Bios Voltage Supply Preamplifier NIR Detection GPIB Motor Control Motor Motor Filter Wheel Detector Interchange Mirror Detector Interchange Mirror Figure 2. Chopper Reference Motion GAx s 1 Signal Controller Motor Linearity Tester Axis 2 Linearity Tester Figure 2. Schematic diagram of signal processing in NRCC Reference Spectrophotometer Schematic diagram of signal processing in NRCC Reference Spectrophotometer 3. ERROR EVALUATION In order to determine the accuracy and precision attainable with the new reference spectrophotometer, the following instrument errors were evaluated: wavelength accuracy and reproducibility, bandpass, beam uniformity, polarization, stray light, system drift, noise and non -linearity. The details of this characterization are presented below. 3.1 Wavelength scale and bandpass 3- ERROR EVALUATION In order to determine the accuracy and precision attainable with the new reference spectrophotometer, the following instrument errors were evaluated: wavelength accuracy and reproducibility, bandpass, beam uniformity, polarization, stray light, system drift, noise and non-linearity. The details of this characterization are presented below. 3.1 Wavelength scale and bandpass The wavelength scale was calibrated with 31 atomic lines from Hg, Cd, Cs and He spectrum lamps. The intensity data for each atomic line was recorded at small wavelength steps (0.03 to 0.1 nm) and with a narrow instrument bandpass (<0.3 nm). The positions of the peak maxima were estimated to the nearest 0.01 nm by fitting a triangular line shape to the data. The reproducibility of this wavelength calibration procedure was typically better than ±0.02 nm from 200-800 nm and better than ±0.03 nm from 800-2500 nm. The worst reproducibility of ±0.055 nm occurred in the region 1800-2000 nm. The wavelength calibration data recorded with the detector and with the PbS detector were separately fitted to first degree polynomial functions. These results are shown in Fig. 3- In both cases, the average agreement between calculated and measured wavelength readings is better than ±0.1 nm. The largest discrepancy of 0.20 nm occurs at 250 nm. The wavelength scale was calibrated with 31 atomic lines from Hg, Cd, Cs and He spectrum lamps. The intensity data for each atomic line was recorded at small wavelength steps (0.03 to 0.1 nm) and with a narrow instrument bandpass (<0.3 nm). The positions of the peak maxima were estimated to the nearest 0.01 nm by fitting a triangular line shape to the data. The reproducibility of this wavelength calibration procedure was typically better than ±0.02 nm from 200-800 nm and better than ±0.03 nm from 800-2500 nm. The worst reproducibility of ±0.055 run occurred in the region 1800-2000 nm. The wavelength calibration data recorded with the detector and with the PbS detector were separately fitted to first degree polynomial functions. These results are shown in Fig. 3. In both cases, the average agreement between calculated and measured wavelength readings is better than ±0.1 nm. The largest discrepancy of 0.20 nm occurs at 250 nm. The spectral resolution or bandpass was determined by measuring the full-width at half-maximum height of three Hg atomic lines (253.6, 546.07, and 1014.0 nm) as a function of monochromator slit width. The measured bandpass differs from the nominal bandpass (given by the product of the reciprocal linear dispersion and slit width) because of slit curvature mismatch, diffraction, and optical aberrations. However, at larger slit widths (0.1 to 2.0 mm) the relationship between nominal and actual bandpass was linear. At very small slit widths (<0.1 mm), limiting spectral resolutions of 0.2, 0.1 and 0.1 nm were obtained at 253.6, 546.07 and 1014.0 nm, respectively. The spectral resolution or bandpass was determined by measuring the full -width at half- maximum height of three Hg atomic lines (253.6, 546.07, and 1014.0 nm) as a function of monochromator slit width. The measured bandpass differs from the nominal bandpass (given by the product of the reciprocal linear dispersion and slit width) because of slit curvature mismatch, diffraction, and optical aberrations. However, at larger slit widths (0.1 to 2.0 mm) the relationship between nominal and actual bandpass was linear. At very small slit widths (<0.1 mm), limiting spectral resolutions of 0.2, 0.1 and 0.1 nm were obtained at 253.6, 546.07 and 1014.0 nm, respectively. The temperature dependence of the wavelength scale was checked by monitoring the peak position of the 546.07 nm Hg line as a function of monochromator temperature. Over a typical temperature range of 22.0-24.0 G, there was no detectable wavelength shift. The temperature dependence of the wavelength scale was checked by monitoring the peak position of the 546.07 nm Hg line as a function of monochromator temperature. Over a typical temperature range of 22.0-24.0 C, there was no detectable wavelength shift. 92 / SPIE Vol 1109 Optical Radiation Measurements ll (1989) 92 / SPIE Vol. 1109 Optical Radiation Measurements II (1989)

3000 3000 T T T T 2500 2500 I I T~ 2800 2800 Wavelength Calibration for UV-VIS Range Wavelength Calibration for UV -VIS Range 2200 2200 Wavelength Calibration for NIR Range Wavelength Calibration for NIR Range O) c r-» T3 (0 (D GC cn C.,y co a) 2600 2600 2400 2400 1900 1900 en c T3 1600 CO 1300 1300 2200 ~ 2200 2000 2000 200 + Hg ocd nhe ACs I T 360 520 680 840 200 360 520 680 840 Wavelength (nm) Wavelength (nm) 1000 1000 700 700 1000 800 + Hg QHe ACs I I 1200 1600 2000 1000 800 1200 1600 2000 Wavelength (nm) Wavelength (nm) 2400 2400 (a) (a) (b) (b) Figure 3. Least-square fits to wavelength calibration data: ; (b) NIR range with PbS. Figure 3. Least- square fits to wavelength calibration data: (a) UV -VIS range with ; (b) NIR range with PbS. 3.2 Beam uniformity 3.2 Beam uniformity (a) UV-VIS range with The uniformity of the beam in the sample compartment was determined by mapping the signal intensity as a 2.5 mm circular aperture was scanned across the beam in a horizontal and vertical direction. To ensure a uniform detector response, a Si photodiode with an opal glass'diffuser was used. A typical result is shown in Fig. 4. Beam uniformity was checked at other wavelengths and slit width settings and found to exhibit similar profiles. In all cases the uniformity was better in the vertical than horizontal direction with a maximum variation of -10% along a 32 mm beam height. The dip in the middle of the horizontal beam profile is most likely due to some blemish, eg. scratch on the input or monochromator optics. The uniformity of the beam in the sample compartment was determined by mapping the signal intensity as a 2.5 mm circular aperture was scanned across the beam in a horizontal and vertical direction. To ensure a uniform detector response, a Si photodiode with an opal glass'diffuser was used. A typical result is shown in Fig. 4. Beam uniformity was checked at other wavelengths and slit width settings and found to exhibit similar profiles. In all cases, the uniformity was better in the vertical than horizontal direction with a maximum variation of -10% along a 32 mm beam height. The dip in the middle of the horizontal beam profile is most likely due to some blemish, eg. scratch on the input or monochromator optics. 7 (0 O) rt C/D Figure 4. Beam uniformity in sample compartment at 500 nm (slit horizontal, and vertical, o to instrument optical axis, Figure 4. Beam uniformity in sample compartment at 500 nm (slit horizontal, and vertical, o to instrument optical axis. -18-14 -10-6 -18-14 -10-6 -2 2 6 r -2 Position (mm) Position (mm) I 10 I 14 18 10 14 18 SPIE Vol. 1109 Optical Radiation Measurements II (1989) / 93 SPIE Vol. 1109 Optical Radiation Measurements II (1989) / 93

3.3 Instrument polarization 3.3 Instrument polarization The degree of polarization, 2 P was determined over the spectral range 400 to 800 nm with a linear polarizing filter (Oriel model 27300) in the sample compartment and with a polarization-insensitive detector (Si diode with opal glass diffuser). The value of P varied from.18 to 0.46, depending upon the wavelength. Calculations show that the transmittance error for a nonabsorbing glass filter (n=1.5) due to P=0.50 and an angle of incidence of 0.7 (convergence angle for 7 mm slit height), is less than 2x10~5 transmittance units. Thus, in the visible range, it is not necessary to perform transmittance measurements in linearly polarized light to correct for the effects of instrument polarization. The degree of polarization,2 P was determined over the spectral range 400 to 800 nm with a linear polarizing filter (Oriel model 27300) in the sample compartment and with a polarization- insensitive detector (Si diode with opal glass diffuser). The value of P varied from.18 to 0.46, depending upon the wavelength. Calculations show that the transmittance error for a nonabsorbing glass filter (n =1.5) due to P =0.50 and an angle of incidence of 0.7 (convergence angle for 7 mm slit height), is less than 2x10-5 transmittance units. Thus, in the visible range, it is not necessary to perform transmittance measurements in linearly polarized light to correct for the effects of instrument polarization. 3.4 Stray light 3.4 Stray light The isochromatic stray light error, ie. the transmitted radiation which bypasses the sample, was determined by measuring the transmittance of an opaque aluminum mirror (50 mm square) covering the sample beam. This test should exacerbate the error by introducing additional scattering from interreflections between the monochromator optics and the mirror sample. For a typical 100% level of 8V (0.8 ya), the isochromatic stray light error was a few parts in a million, although dependent on the detector settling time (eg. 3«7X 10~6, 2.6xlO" 6 and 2.2xlO~ 6 for 10, 20, and 40 sec delays, respectively, from 100% reading). The isochromatic stray light error, ie. the transmitted radiation which bypasses the sample, was determined by measuring the transmittance of an opaque aluminum mirror (50 mm square) covering the sample beam. This test should exacerbate the error by introducing additional scattering from interreflections between the monochromator optics and the mirror sample. For a typical 100% level of 8V (0.8 RA), the isochromatic stray light error was a few parts in a million, although dependent on the detector settling time (eg. 3.7x10-6, 2.6x10-6 and 2.2x10-6 for 10, 20, and 40 sec delays, respectively, from 100% reading). The heterochromatic stray light (HSL) error, ie. the transmitted radiation different from the nominal instrument bandpass, was determined by measuring the transmittance through one and two thicknesses of a red filter (Schott RG665) at wavelengths below its cutoff. Since the HSL is a function of the source and detector as well as the monochromator, the sample, and the operating conditions, the results are only valid for the particular conditions of the test. 3 In the case of the Si detector (1 mm slit width), the far-hsl error (at 450, 500 and 550 nm) was found to be 1 to 2 parts in 10 6, but the near-hsl error (at 600 nm) increased to several parts in 10 5. With the detector (0.15 mm slit width), the far- and near-hsl errors were both 4 to 5 parts in 10 6. The heterochromatic stray light (HSL) error, ie. the transmitted radiation different from the nominal instrument bandpass, was determined by measuring the transmittance through one and two thicknesses of a red filter (Schott RG665) at wavelengths below its cutoff. Since the HSL is a function of the source and detector as well as the monochromator, the sample, and the operating conditions, the results are only valid for the particular conditions of the test.3 In the case of the Si detector (1 mm slit width), the far -HSL error (at 450, 500 and 550 nm) was found to be 1 to 2 parts in 106, but the near -HSL error (at 600 nm) increased to several parts in 105. With the detector (0.15 mm slit width), the far- and near -HSL errors were both 4 to 5 parts in 106. 3.5 System drift 3.5 System drift The principle disadvantage of a single-beam instrument design is that fluctuations in the source output and detector response are not dynamically compensated. However, if the system drift exhibits a linear systematic bias, it can be effectively eliminated by performing' the measurements in a time-symmetrical sequence, eg. 100% reading, sample, sample, 100% reading, and averaging the results. To check if this assumption was valid, the system drift was measured for the various source-detector combinations used to cover the range 200-2500 nm. The signal was monitored every minute (10-sec integration) for a period of one hour. In all cases, the data could be satisfactorily fitted to a straight line. The slope gives a measure of the systematic drift and the residual least-squares error gives a measure of the random noise. Typical results for the tungsten-halogen lamp and PbS detector are shown in Fig. 5- The principle disadvantage of a single -beam instrument design is that fluctuations in the source output and detector response are not dynamically compensated. However, if the system drift exhibits a linear systematic bias, it can be effectively eliminated by performing the measurements in a time -symmetrical sequence, eg. 100% reading, sample, sample, 100% reading, and averaging the results. To check if this assumption was valid, the system drift was measured for the various source -detector combinations used to cover the range 200-2500 nm. The signal was monitored every minute (10 -sec integration) for a period of one hour. In all cases, the data could be satisfactorily fitted to a straight line. The slope gives a measure of the systematic drift and the residual least- squares error gives a measure of the random noise. Typical results for the tungsten- halogen lamp and PbS detector are shown in Fig. 5. 3.6 Non-linearity 3.6 Non -linearity Photometric accuracy requires linearity of the photometric scale, ie. that the output quantity of the detector is exactly proportional to the input quantity over a specified range. However, all spectrophotometers exhibit nonlinearity errors. Accuracy is therefore limited by the precise Photometric accuracy requires linearity of the photometric scale, ie. that the output quantity of the detector is exactly proportional to the input quantity over a specified range. However, all spectrophotometers exhibit non - linearity errors. Accuracy is therefore limited by the precise 4.4 4.3 > E 4.2 > 4.4 4.3 4.2 4. 1 4.1 4.0 4.0 Figure 5. I I I I Straight line fit (DRIFT -- 0.005 % /min) ^Straight line fit (DRIFT 0.005%/min) 0 12 24 36 48 Time dv 0.01% 12 24 36 48 Time (minutes) (minutes) 60 0.002 0.002 0.001 > E 0.000 `-* -0.001 10-0.002-0.002 Figure 5. System drift and noise as a function of time for tungsten-halogen lamp and PbS detector (258 K) at 1300 nm. System drift and noise as a function of time for tungsten -halogen lamp and PbS detector (258 K) at 1300 nm. 94 / SPIE Vol. 1109 Optical Radiation Measurements ll (1989) 94 / SPIE Vol. 1109 Optical Radiation Measurements II (1989)

evaluation of this error. error. To non-linearity To determine the non -linearity correction, correction, we we built built aa new completely completely automated automated linearity tester4. tester4. The The device device is is aa modified modified version of of the the double-aperture > 6 which double- aperturemethod methodof of light light addition5 addition516 which uses uses aa single single adjustable adjustable aperture aperture instead instead of apertures to to measure measure 2:1 2:1 flux(4)) flux(<j>) ratios. ratios. This of two two fixed apertures is made possible possible by by This procedure is the the large, large, uniform uniform nature nature of of our our sample sample beam beam and and by by the the recent recent availability availability of of high high precision precision piezoelectric piezoelectric (PZT) (PZT) motors motors to to precisely precisely define define the the measurement measurement apertures apertures A, A, B, B, and and A+B A +Bfor forarbitrary arbitrary sizes sizes and and therefore therefore flux levels. levels. This, turn, eliminates eliminates the the need need of This, in turn, of external to characterize characterize different different flux external optical optical attenuation to flux levels. levels. AA photograph of of the new NRCC NRCC linearity linearity tester tester is Is shown shown in in Fig. Fig. 6. 6. It It consists of of two two blades and closed closed independently independently by blades which which are are opened opened and by directly directly coupled coupled PZT motors motors driven by aa driven by dedicated controller (Burleigh dedicated programmable controller (Burlelgh Instruments). Instruments). The The motors are clamped clamped to to aa mounting one above above the the other centrally with mounting plate plate and and located located one other and and centrally with respect respect to to the the instrument Instrument optical axis. axis. The The blades blades can can be be adjusted adjusted in in 0.5 0.5 ym pm Increments increments for for aa total total travel of 25 25 mm travel of each. An each. the drive shaft of An Incremental incremental encoder encoder on the drive shaft of each each motor motor reads an reads the the position to to an accuracy of of 11 um. pm. The The accuracy accuracy and and precision precision of of the the new new linearity linearity tester tester were were checked checked with with aa highly highly linear linear and SI photodiode photodiode (EG (EG&G and stable Si &G UV4444 BQ).7 BQ). 7 At 546 nm, nm, over a photometric range At 546 range of of 100% 100$ to to 10JE (100$ T=4.1 10% TT (100% T =4.1pA), pa), the the diode diode was was found found to to be be linear linear to to 22 parts parts in in 101* 104 with with a repreductibility 5, which reproducibility of of aa few few parts parts In in 10 105, which is is consistent consistent with with the the reported reported linearity linearity of of thi s de this detector tec tor in in the the visible v i s Ib 1 e.7.7 The non-linearity factors, 5 aa%%, 9 for, determined The non -linearity correction correction factors,5 for AA == BB == 1/2 1/2 $4), determined for for the the detector at 546 nm nm and and for for aa full full-scale detector at -scale (T-1.0) (T =1.0)anode anode current current of of 0.8 0.8 pa, pa, are are shown shown graphically in In Fig. Pig. 77 versus versus transmittance transmlttancet(4)). T(«f>). A smooth curve curve has has been been drawn drawn through through the the measured had an average reproducibility reproduclblllty of 11 to measured points points which had to 22 parts in 104. 10**. 0.7 0.6 0.5 0.4 b 0.3 0.2 0'. 1 0.1 0.0 0.0 00.0 0 0.2 0.2 0.4 T Figure 7. J. Figure Figure 6. 6. Photograph of Photograph of adjustable adjustableaperture linearity tester aperture with PZT PZT-controlled -controlled blades. 4. 0.6 0.8 11.0 0 (0) Measured non non-linearity error, Measured -linearity error, of versus versus transmittance, transmittance, a% of T(<j> ) for for T(1.0) T( 1.0) =0.8 T(4) =0.8 pa and nm. Data obtained obtained with, XX=546 =546 nm. with, x and x and without, without, o N.D.1.0 filter in in beam. beam. filter TRANSMITTANCE MEASUREMENTS the preceding error evaluation, evaluation, the the combined combined effects From the effects of of wavelength wavelength scale scale error, error, beam non-uniformity, non -uniformity, instrument instrument polarization, polarization, and and stray stray light, light, should should contribute contribute less part less than than 11 part in 104 I0 k to the total measurement uncertainty. uncertainty. Therefore, Therefore, no no correction correction is is applied applied for for these these errors. However, system specific errors. system drift, drift, noise, noise, and -linearity are and non non-linearity are significant. significant. The The procedures for correcting for procedures for these these latter latter errors errors are outlined below. below. SPIE Vol. SPIE Vol. 1109 1109 Optical OpticalRadiation RadiationMeasurements MeasurementsII Il/19891 (1989)/ / 95 95

4.1 Measurement procedure 4.1 Measurement procedure In order to remove the influence of linear drift on the results, the measurements are performed in a time-symmetrical sequence. For example, for a four-filter measurement on the automated 6-position filter wheel, positions 1 and 4 are used as clear spaces, and the following series of readings are taken: In order to remove the influence of linear drift on the results, the measurements are performed in a time -symmetrical sequence. For example, for a four -filter measurement on the automated 6- position filter wheel, positions 1 and 4 are used as clear spaces, and the following series of readings are taken: Pass 1 : 0, Fo, F1, F2, 0, Fo, F3, Fy, Fo, 0 Pass 2 : F0, 0, F4, F3, Fo, 0, F2, F1, FD, 0 where Pass 1 involves rotation of the filter wheel 360 clockwise and Pass 2 involves rotation 360 counter -clockwise; 0, Fo, F1, F9, F,, and F4 are signals for the dark current, clear space, filter 1, filter 2, filter 3, and filter 4, respectively. The data is processed by taking the average value of 0 and Fo on either side of an individual filter reading, F (j =1 to 4), giving 0 and Fo, then computing the filter transmittance, Ti according to: T = j Fo -0 The corresponding filter transmittance values for the two passes are then averaged. 4.2 Reproducibility The effectiveness of this time -symmetrical measurement procedure was evaluated by repeat transmission measurements on four neutral- density glass filters of nominal transmittances 10, 25, 50 and 70% (Schott NG4 and NG11, 50x50 mm, 1-5 mm thick). These results are reported in Table 1. The measurements were performed at 546 nm with a spectral bandpass of 0.5 nm and a temperature of 23 ±1 C. The filters were aligned normal to the beam and measured with the maximum beam size of 37x20 mm. Four independent measurements were performed on each filter, where each run represents the average transmittance of the two -pass measurement sequence described in Section 4.1 with a 10 -sec integration interval for each individual reading. The average % relative standard deviation for the four filters is 0.006 %. Table 1. Reproducibility of Transmittance Measurements on Four Glass Filters at 546 nm Run 1 Run 2 Run 3 Run 4 Pass 1 Average % Standard Deviation Pass 2 0, P., F,, P n, 0, F9, P., Pn, where Pass 1 involves rotation of the filter wheel 360 clockwise and Pass 2 involves rotation 360 counter-clockwise; 0, FQ, F X, F^, F-, and F^ are signals for the dark current, clear space, filter 1, filter 2, filter 3> and filter 4, respectively. The data is processed by taking the average value of 0 and FQ on either side of an individual filter reading, Pj(J=l to 4), giving 0 and FQ, then computing the filter transmittance, T according to: The corresponding filter transmittance values for the two passes are then averaged. 4.2 Reproducibility The effectiveness of this time-symmetrical measurement procedure was evaluated by repeat transmission measurements on four neutral-density glass filters of nominal transmittances 10, 25, 50 and 70$ (Schott NG4 and NG11, 50x50 mm, 1-5 mm thick). These results are reported in Table 1. The measurements were performed at 546 nm with a spectral bandpass of 0.5 nm and a temperature of 23 ± 1 C. The filters were aligned normal to the beam and measured with the maximum beam size of 37x20 mm. Pour Independent measurements were performed on each filter, where each run represents the average transmittance of the two-pass measurement sequence described in Section 4.1 with a 10-sec integration interval for each individual reading. The average % relative standard deviation for the four filters is 0.006$. Table 1. Reproducibility of Transmittance Measurements on Pour Glass Filters at 546 nm Run 1 Run 2 Run 3 Run 4 Average % Standard Deviation Nominal % Transmittance 10 25 50 70 10.396 10.395 10.395 10.397 Nominal % Transmittance 10 25 50 70 10.396 10.395 10.395 10.397 10.396.009 24.914 24.914.002 49-993 49-992 49-990 49-993 49.993 49.992 49.990 49.993 49.992.003 70.181 70.194 70.198 70.185 70.181 70.194 70.198 70.185 70.189.011 10.396 49.992 70.189.009.002.003.011 4.3 Non-linearity error correction 4.3 Non -linearity error correction The transmittance error correction, AT which must be added to the measured apparent transmittance, T was computed from the non-linearity error correction factors, a% in Pig. 7 according to the procedure of Clarke. 6 This yielded the smooth curve shown in Pig. 8. A maximum correction value of 3.64xlO~ 3 transmittance units occurs at a transmittance level of 40$. The transmittance error correction, AT which must be added to the measured apparent transmittance, T was computed from the non -linearity error correction factors, a% in Fig. 7 according to the procedure of Clarke.6 This yielded the smooth curve shown in Fig. 8. A maximum correction value of 3.64x10-3 transmittance units occurs at a transmittance level of 40 %. 4.4 Results 4.4 Results The accuracy of the non-linearity correction curve, AT(<j>) derived for the (Fig. 8), was independently checked by comparison with transmittance measurements using a highly linear Si diode (EG&G UV444 BQ with ground Suprasil diffuser). Six neutral-density glass filters (combinations of Schott NG4 and NG11), ranging from 10 to 70$ transmittance, and 1 to 5 mm in thickness, were measured on the reference spectrophotometer using the and the Si diode. In both cases, the measurements were performed at 546 nm with a 0.5 nm spectral bandpass, and a beam size of 37x20 mm. The was operated at the conditions corresponding The accuracy of the non -linearity correction curve, AT($) derived for the (Fig. 8), was independently checked by comparison with transmittance measurements using a highly linear Si diode (EG &G UV444 BQ with ground Suprasil diffuser). Six neutral- density glass filters (combinations of Schott NG4 and NG11), ranging from 10 to 70% transmittance, and 1 to 5 mm in thickness, were measured on the reference spectrophotometer using the and the Si diode. In both cases, the measurements were performed at 546 nm with a 0.5 nm spectral bandpass, and a beam size of 37x20 mm. The was operated at the conditions corresponding 96 / SPIE Vol. 1109 Optical Radiation Measurements Il (1989) 96 / SPIE Vol. 1109 Optical Radiation Measurements II (1989)

I I I I I I I 4.0 3.0 3.0 3 2.0 r- 1.0 1.0 Figure 8. Transmittance correction, AT computed from non-linearity data of Pig. 7. Figure 8. Transmittance correction, AT computed from non -linearity data of Fig. 7. 0.0 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.3 0.9 to Pig. 7. These results are summarized in Table 2, where both the raw and linearity-corrected are included. The average agreement between the Si diode and linearity-corrected values for the six filters is 2.5 parts in 10 4. This low overall figure indicates that the new adjustable aperture linearity tester and the -diffuser combination provide reliable and accurate results. to Fig. 7. These results are summarized in Table 2, where both the raw and linearity- corrected are included. The average agreement between the Si diode and linearity- corrected values for the six filters is 2.5 parts in 104. This low overall figure indicates that the new adjustable aperture linearity tester and the -diffuser combination provide reliable and accurate results. Table 2. Comparison of Percent Transmittance (W) for Six Glass Filters at 546 run Using Si diode and detectors %T Si diode (measured) (corrected)* Difference Table 2. Comparison of Percent Transmittance ( %T) for Six Glass Filters at 546 nm Using Si diode and detectors %T Si diode (measured) (corrected)* Difference 10 17 25 35 50 70 10.589 17.096 25.224 35.471 50.296 70.414 10.396 16.828 35.150 49.992 70.189 10 10.589 10.396 17 17.096 16.828 25 25.224 35 35.471 35.150 50 50.296 49.992 70 70.414 70.189 10.596 17.102 25.241 35-512 50.335 70.447 ^corrected for non-linearity transmittance error, AT, in-fig. 8 *corrected for non -linearity transmittance error, AT, in Fig. 8.007.006.017.041.039.033 10.596.007 17.102.006 25.241.017 35.512.041 50.335.039 70.447.033 5. CONCLUSIONS 5. CONCLUSIONS A new high-accuracy spectrophotometer for performing regular transmittance measurements over the 200-2500 nm spectral range has been developed. This automated single-beam instrument possesses a highly-collimated beam geometry which eliminates the need of polarization correction and of an averaging sphere. Error evaluation is well-advanced, particularly in the visible region, where the overall photometric accuracy is estimated to be a few parts in lo 4, with an average reproducibility of 0.006$ of the value for four repeat measurements. The wavelength scale is accurate to better than ±0.1 nm from 300 to 2500 nm, ±0.2 nm from 200 to 300 nm, and is reproducible within ±0.03 nm. A new high- accuracy spectrophotometer for performing regular transmittance measurements over the 200-2500 nm spectral range has been developed. This automated single -beam instrument possesses a highly -collimated beam geometry which eliminates the need of polarization correction and of an averaging sphere. Error evaluation is well- advanced, particularly in the visible region, where the overall photometric accuracy is estimated to be a few parts in 104, with an average reproducibility of 0.006% of the value for four repeat measurements. The wavelength scale is accurate to better than ±0.1 nm from 300 to 2500 nm, ±0.2 nm from 200 to 300 nm, and is reproducible within ±0.03 nm. A new automated linearity tester with an adjustable, precisely-defined aperture has been designed and built. Its reliability and accuracy have been checked with a highly linear Si diode. In the visible, the largest systematic error limiting the accuracy of the reference spectrophotometer is non-linearity of the which is of the order of several parts in 10 3 for a 1 11A photocurrent. A new automated linearity tester with an adjustable, precisely- defined aperture has been designed and built. Its reliability and accuracy have been checked with a highly linear Si diode. In the visible, the largest systematic error limiting the accuracy of the reference spectrophotometer is non -linearity of the which is of the order of several parts in 103 for a 1 ua photocurrent. Further work is underway to refine the linearity tester design and to select a more linear that will provide a measurement accuracy of better than 1 part in 10 4 transmittance units in the VIS range. In the UV and NIR, error evaluation is continuing. The extension of the instrument's measurement capabilities to diffuse reflectance in the UV-VIS range has also begun. Further work is underway to refine the linearity tester design and to select a more linear that will provide a measurement accuracy of better than 1 part in 104 transmittance units in the VIS range. In the UV and NIR, error evaluation is continuing. The extension of the instrument's measurement capabilities to diffuse reflectance in the UV -VIS range has also begun. 6. ACKNOWLEDGEMENTS 6. ACKNOWLEDGEMENTS The authors wish to thank Dr. I. Powell for helpful discussions and for the design of the input and exit optics. We also wish to acknowledge the contributions of M. Kottler to the design of several of the mechanical components of the system. The authors wish to thank Dr. I. Powell for helpful discussions and for the design of the input and exit optics. We also wish to acknowledge the contributions of M. Kottler to the design of several of the mechanical components of the system. SPIE Vol. 1109 Optical Radiation Measurements ll (1989) / 97 SPIE Vol. 1109 Optical Radiation Measurements II (1989) / 97