Adaptation of a High-Accuracy Spectrophotometer for Ultraviolet Work

Abstract

A high-accuracy spectrophotometer, originally designed for work at visible wavelengths, was modified to permit measurements in the ultraviolet without degradation of its original performance. This was accomplished by equipping the spectrophotometer with a stable deuterium arc source, a highly efficient averaging sphere with fluorescent wavelength converter, a new grating, and achromatic sample-compartment optics. The modified spectrophotometer will be used for the development of new Standard Reference Materials, as well as for materials research, in the region between 200 and 300 nm.

1. Introduction

A primary goal of the current NBS program in spectrophotometry has been the design of high-accuracy instrumentation for the certification of calibration standards and the development of improved measurement techniques. For transmittance measurements at visible wavelengths, this goal was accomplished by specially constructing two highly accurate single-beam spectrophotometers [1, 2],¹ which are presently used in the Institutes for Materials Research (IMR) and Basic Standards (IBS) for standards work at visible wavelengths (400 to 800 nm, approximately). The instrumental accuracy achieved with these spectrophotometers is 10⁻⁴ transmittance units (0.01 percent). Although similar in some design aspects, these two spectrophotometers differ significantly in others, and therefore can be employed for in-house comparisons to ensure the accuracy and consistency of all spectrophotometric measurements at NBS. The purpose of this work was to extend these capabilities into the ultraviolet spectral region by developing the means for modifying either spectrophotometer for applications down to about 200 nm.

The particular instrument that was modified was the IMR spectrophotometer described in reference [1]. As originally designed, this spectrophotometer employed a current-stabilized, 100-W tungsten ribbon lamp as the radiation source; a 1-meter, f/8.7 Czerny-Turner grating monochromator (1200 lines/mm, 500-nm blaze) with a fused-silica prism predisperser as the dispersing element; and an 11-stage, S-20 photo-multiplier tube attached to a 125-mm, BaSO₄-coated averaging sphere as the signal detector. The tungsten lamp is imaged on the predisperser entrance slit by a fused-silica lens. Two additional fused-silica lenses are used in the sample compartment to focus the monochromator exit slit at the sample, and to refocus it into the averaging sphere [3]. The photomultiplier signal is measured by means of a current-to-voltage converter and a digital voltmeter, interfaced with a computer.

It may be estimated that, for a spectrophotometer of this type, a clear-space signal of at least 10⁻⁷ A is required if measurements are to be made with a standard deviation of at least 10⁻⁴ transmittance units [4]. The actual clear-space signal current obtained with this spectrophotometer at different wavelengths and for typical values of tungsten-lamp power, monochromator slit width, photomultiplier anode voltage, etc. is shown as curve (a) in figure 1, and thus indicates an inadequate performance of the instrument in the spectral region below approximately 370 nm. In order to effect the desired improvement of ultraviolet performance, the spectrophotometer was equipped with a deuterium arc source, a re-designed averaging sphere, and a differently blazed grating. These components were chosen to achieve a wide spectral range (200 to 800 nm) of the modified spectrophotometer, requiring only a change back to the tungsten lamp for work above 400 nm. In order to facilitate the usage of the instrument throughout this extended range, it was necessary to replace the two sample-compartment lenses by fused-silica, lithium-fluoride achromats.

Figure 1.

Successive improvement of clear-space signal level of original spectrophotometer (curve a) obtained by: (b) deuterium arc lamp, (c) improved averaging sphere, and (d) new grating.

2. Modification of Spectrophotometer

2.1. Deuterium Arc Lamp

To a significant extent, the sharp drop of ultraviolet signal level shown by curve (a) in figure 1 is attributable to the tungsten lamp of the original spectrophotometer. A rapid decrease of radiant intensity with decreasing wavelength is typical for incandescent filament lamps, and therefore makes any source of this kind a poor choice for work at short wavelengths [5]. In contrast, a deuterium arc lamp exhibits an increase of radiant intensity with decreasing wavelength, and thus provides much better power levels in the ultraviolet.

A 60-W deuterium lamp with high-purity fused-silica window was chosen for the modified spectrophotometer. This lamp has a line-free continuous emission in the ultraviolet extending to below 180 nm, and is rated to have the same output at 200 nm as a 150-W xenon arc lamp. In comparison to a 250-W tungsten-bromide filament lamp, its output is about 1000 times greater at 250 nm, and 4 times greater at 300 nm. For work above 400 nm, tungsten lamps are preferable because of the decreasing intensity of the deuterium-arc spectrum and its line structure in the visible. The clear-space signal obtained after the deuterium lamp was installed in the spectrophotometer is shown as curve (b) in figure 1.

The stability of the radiant-intensity output of the deuterium lamp was tested by performing regularly repeated measurements of the spectrophotometer signal over extended periods of time. The long-term stability of lamp power so observed may be seen from figure 2, showing a gradually decreasing signal drift at the rates of 1.5, 0.6, and 0.08 percent per hour after the lamp had been operated for 1, 3, and 5 h, respectively. The average short-term instability (random noise plus drift) of the deuterium lamp was found to be 0.02 percent for a total of 15 sets of 20 individual readings, taken at 5-s intervals and at arbitrary times during an 8-h period of lamp operation. Since all data obtained with this spectrophotometer are derived by averaging a large number of individual measurements of transmittance, performed in a time-symmetrical sequence which eliminates the effects of drifting, this stability of the deuterium lamp is adequate for routine measurements with a precision better than 10⁻⁴ transmittance units. The above results for the deuterium lamp compare favorably with the stability data found for a current-stabilized tungsten lamp [2].

Figure 2.

Long-term signal drift observed with deuterium lamp.

2.2. Averaging Sphere

The original averaging sphere of the spectrophotometer was designed as shown in figure 3a. It has an internal diameter of 125 mm, and circular entrance and exit ports 20 and 50 mm in diameter, respectively. The circular target, 35 mm in diameter, is located at the center of the sphere. The inside of the sphere, both sides of the target, and the thin target support rod are coated with several layers of specially prepared barium sulfate and polyvinyl alcohol coatings as developed by Grum and Luckey [6]. The photo-multiplier tube is attached to the sphere, so that its 50-mm cathode is located 16 mm behind the exit port. The efficiency of this sphere was determined by successively measuring photomultiplier signals obtained at different wavelengths with and without sphere, using the deuterium arc lamp. These measurements showed a steep decrease of sphere efficiency from about 15 percent for visible wavelengths near 500 nm to less than 0.1 percent at 200 nm, and thus indicated the necessity to equip the spectrophotometer with an improved, more efficient sphere [7].

Figure 3. Geometry of (a) original averaging sphere and (b) improved sphere.

A: entrance port, A′: exit port, D: detector, F: fluorescent dye cell, S: sphere wall, SH: shutter, T: target.

According to theory [8], the efficiency of an averaging sphere of the type considered is, approximately,

$${\varphi }_D/\varphi =[r^2/(1-r)](1-e/d)d^2/16s^2$$ (1)

where

ϕ = radiant flux into sphere,

ϕ_D = flux reaching detector,

r = wall reflectance (assumed close to, but not equal to unity),

s = sphere radius,

d = radius of sensitive area of detector (assumed equal to radius of exit-port)

e = distance between exit port and detector.

It may be seen from this equation that, to a large extent, the poor efficiency of the original spectrophotometer sphere is due to inadequate coupling of exit port and detector. The collection efficiency of the detector (i.e.; the fraction of the diffuse flux from the exit port which actually reaches the photocathode) appears in Eq (1) as the factor (1-e/d), which in this case has the numerical value 0.64 and thus shows that more than one third of the available flux is lost through the gap between exit port and detector. Therefore, a large increase in efficiency could be gained by modifying the sphere as shown in figure 3b. The sphere wall was tapered in a pear-like fashion and was extended into the detector housing, so that the exit port is butted directly against the photo-cathode. This simple modification of the sphere resulted in a large increase of efficiency (from 15 to 70 percent at 500 nm). However, the efficiency in the ultraviolet (now 2.5 percent at 200 nm) was still inadequate.

The averaging effectiveness of this sphere was tested by Mr. K. L. Eckerle of the Optical Radiation Section of NBS. The sphere was found to be insensitive to beam displacements from the target center of ±1.5 mm within a ±10⁻⁴ limit of signal variation. This is to be compared with a ±5 percent variation for a ±1 mm beam displacement on the same photo-multiplier used without the sphere.

The high efficiency of the modified sphere in the visible suggested that a near-optimal sphere geometry had been achieved, and that the limiting factor in the ultraviolet was now the reduced reflectance of the barium-sulfate sphere coating. In order to overcome this remaining difficulty, a fluorescent dye was used to shift the incident short-wave radiation into the longer wavelength region for which the reflectance of barium sulfate is high. This dye, which is contained in a cell placed in front of the sphere target as shown in figure 3b, had to be chosen such that virtually all of the incident flux at short wavelengths is absorbed and is converted into fluorescence with a quantum efficiency as nearly equal to unity as possible. In order to avoid impairing the high sphere efficiency at longer wavelengths, it was also required that visible light be not absorbed, so that it passes through the dye unaffected and is reflected from the target as before. The fluorescent material selected was a 0.923 g/l solution of 2,5-diphenyloxazole (PPO) in n-hexane in a 10 × 10 × 2 mm high-purity fused-silica cell. It is effectively transparent for wavelengths above 360 nm, and below 340 nm absorbs more than 99 percent of the incident radiation. Its estimated quantum efficiency is 0.9, with the peak emission occurring near 370 nm. The use of this dye proved highly successful, and resulted in a sphere efficiency of 20 percent at 200 nm.

The marked enhancement of signal level obtained with this improved sphere is shown by comparing curve (c) in figure 1 with curve (b). The measured efficiency of the sphere is roughly constant and equal to 20 percent between 200 and 300 nm, followed by a gradual increase to a constant level between 60 and 65 percent for wavelengths from 400 to beyond 650 nm. A further increase of sphere efficiency was obtained in a final design by reducing the sphere diameter and further optimization of the dye [8].

It was also ascertained that the fluorescence wavelength converter did not affect the linearity of the detection system. Theoretically, the radiant intensity of fluorescence from the dye can be expressed as

$$I_e=Q{I}_a,$$ (2a)

where Q is the quantum efficiency of the dye and I_a is the intensity absorbed. The latter is given by Beer’s law as

$$I_a=I_o(1-10{}^{-ϵcb})$$ (2b)

where I₀ is the initial intensity, and where ϵ, c, and b denote molar absorptivity, concentration, and path length, respectively. Thus, I_e is proportional to I₀ under the applicable assumptions that Beer’s law is valid and that the quantum efficiency of the dye is independent of intensity. This conclusion was verified by measuring the linearity of the detection system at one wavelength (250 nm) for which the dye is effective, and at another (400 nm) for which it is not. These two measurements, which were made using the double-aperture method [9], yielded indistinguishable results within the limits of experimental uncertainty, and also agreed with previously performed linearity measurements of the same photomultiplier tube at 575 nm [1].

2.3. Grating

The previously obtained clear-space signal level, curve (c) in figure 1, was still judged inadequate for accurate work at short wavelengths. Since the original grating of the spectrophotometer (blazed for 500 nm in the first order) had been chosen for work in the visible, it was obvious that a further improvement could be effected by substitution of another grating. This new grating has 600 lines/mm, is blazed for 200 nm in the first order, and resulted in the final clear-space signal current plotted as curve (d) in figure 1. Since the signal level near 200 nm can easily be raised by adjusting the photomultiplier gain, the original goal of achieving a clear-space signal of at least 10⁻⁷ A throughout the ultraviolet had thus been accomplished.

The loss of signal level in the visible due to the change of gratings [curves (d) and (c) in figure 1] is relatively small, and is more than offset by the higher efficiency of the new sphere [curves (c) and (b)]. Thus the modified spectrophotometer can be used above 400 nm by merely interchanging deuterium and tungsten lamps, without need to substitute gratings.

2.4. Achromatic Sample-Compartment Optics

In view of the wide spectral range of the modified spectrophotometer, the chromatic aberration of the original sample-compartment lenses proved to be a limiting factor of instrumental performance. Significant losses of signal level were caused by the fact that these lenses, when initially focused by eye, were out of focus in the ultraviolet and thus caused less-than-optimal illumination of sample and detector.

If the monochromator exit slit is focused at the sample with unit magnification at a wavelength λ₀, the diameter of the blur circle in the sample plane for a wavelength λ < λ₀ can easily be shown to be equal to

$$\delta =D[2f({\lambda }_0)/f(\lambda )-1],$$ (3)

where f is the focal length and D is the effective diameter of the focusing lens. For the particular lenses of the original spectrophotometer [f(200 mm) = 168 mm, f(500 nm) = 200 nm, D = 38 mm], and assuming that the focusing was done at 500 nm, the computed blur-circle diameter at 200 mm is 14 mm. This numerical example illustrates the large defocusing caused by chromatic lenses in the sample compartment, although it should be pointed out that due to the opposing effect of the second lens a smaller blur circle is incurred at the detector. Nevertheless, it was found empirically that the ultraviolet signal level of the spectrophotometer could be improved as much as 30 percent by repositioning the sample-compartment lenses.

The most effective way to overcome this deficiency of the spectrophotometer would have been to use mirror optics, but since this would have required a redesign of the sample compartment and its enclosure it was more expedient to employ achromatic lenses. The lenses selected are f/3 lithium-fluoride, fused-silica (non fluorescent) achromats with 200 mm focal length, best correction for longitudinal chromatic aberration between 200 and 400 nm, and with anti-reflection coating for 400 nm.

On account of the limited spectral regions for which the deuterium and tungsten lamps of the modified spectrophotometer are employed, it was not found necessary to use an achromat in the source compartment as well.

3. Conclusion

As a result of the work reported here, the modified high-accuracy spectrophotometer is now used, throughout the spectral region between 200 and 800 nm, for the certification of solid and liquid Standard Reference Materials for routine applications in spectrophotometry, as well as for materials research aimed at establishing inorganic and organic molar absorptivity standards.

The modified spectrophotometer may also serve as a model for improved commercial instrumentation. In particular, it is anticipated that the development of a highly efficient averaging sphere for visible and ultraviolet wavelengths will lead to a more widespread use of spheres in routine spectrophotometry.