NIST: Spectrum of Platinum - Photoelectric Observations

3. Photoelectric Observations

To determine the relative intensities of lines emitted by the Pt/Ne lamp and to observe lines weaker than those recorded on the photographic plates, we recorded the spectrum by translating an exit slit and photomultiplier tube along the focal curve of the 10.7 m vacuum spectrograph. The entrance and exit slit widths were 0.050 mm. The line intensities were measured by photon counting. Signals from the photomultiplier were amplified and processed by a discriminator and logarithmic ratemeter. The analog output signal from the ratemeter was sampled at 1 Hz by a computer, which digitized and stored the data. This acquisition rate corresponded to a wavelength interval of 0.0086 Å per sample. Prior to each scan the analog response of the ratemeter was calibrated by using a pulse generator to simulate the amplified pulse signal from the photomultiplier tube. The response of the ratemeter was digitized and recorded for pulse frequencies ranging from 10/s to 10⁶/s by decades.

The resolution limit for the scans was about 0.07 Å. The spectrum was scanned in overlapping 650 Å segments, each segment corresponding to a different rotational setting of the grating. Each scan lasted 20 hours. Two scans were made for each region above 1685 Å, the first a normal scan and the second a scan at reduced sensitivity to record very intense lines that were saturated at normal recording conditions. The sensitivity was reduced by introducing a one decade offset in the logarithmic ratemeter. In addition, for the region above 2000 Å, the source intensity was attenuated by reflecting the lamp from an uncoated glass plate.

Four different Pt/Ne lamps were used in the course of the experiments. Two lamps were used for the photographic exposures. One of these and two additional ones were used for the photoelectric scans. The longest use of any lamp was during the photoelectric scans, where one of the lamps was run for about 250 hours. After this time the cavity of the cathode had become noticeably enlarged.

The position and intensity of each spectral line in the photoelectric scans was determined by using a computer line-finding algorithm. First, the recorded signal at each point in the spectrum was converted to absolute counts/s by using the calibration information mentioned above. Then these data were scanned by the computer to locate peaks in the spectrum. The position of each peak was determined by calculating the quadratically smoothed first derivative of the data in the vicinity of the maximum intensity point and linearly interpolating the zero crossing of the derivative. The wavelength was then calculated by making a linear fit of wavelength versus position for the local spectral region, using as standards four lines accurately measured from the photographic observations on either side of the line to be determined.

The raw intensity of the data from each scan was adjusted to produce a consistent set of values over the whole spectral region. First, using the measured intensities for lines of moderate strength in the overlapping regions of the various scans, a set of multiplicative factors was determined to bring the separate scans onto the same relative scale. Then the spectral response of the spectrograph/detector combination as a function of wavelength was calibrated by using accurate radiance values for about 80 lines of platinum measured by Klose [19] in a similar Pt/Ne hollow-cathode lamp. All of the spectral data were corrected for this instrumental response. Thus the intensities plotted in the atlas are on a true relative scale.

The number of lines observed by photon counting was much greater than observed photographically. Whereas the weakest photographic lines produced count rates of about 500 photons/s, lines having signals as low as about 10 photons/s could be observed photoelectrically. The most intense lines produced counts of about 2,000,000 photons/s. In all scans we observed a residual background count in excess of the photomultiplier dark count. This background was only a few counts/s at low wavelengths but increased to about 60 counts/s at the highest wavelengths. This increasing background is apparent in the atlas plots. The background count has been subtracted from the measured line intensities printed in the table so that the value reported accurately reflects the count rate due to the spectral line.