TECHNICAL NOTE Instrumental Limitations for High Absorbance Measurements in Noise UV/Vis/NIR Spectroscopy Author: Jeffrey L. Taylor, Ph.D. PerkinElmer, Inc. Shelton, CT Spectrophotometer Energy Using a Fixed Slit Width Along with instrumental stray light noise is the other major contributing factor limiting high absorbance measurements. As can be seen in Figure 1, the amount of radiant energy in a spectrophotometer varies greatly as a function of wavelength. The amount of noise in a UV/Vis/NIR spectrum is directly related to these energy levels. The ability to measure high absorbance values is critically dependent on the amount of noise close to the 0 %T axis. High absorbance measurements will be more difficult to obtain in low energy wavelength regions of the instrument. Key instrumental parameters associated with the control of spectrum noise are slit width, detector gain, and integration (response) time. LAMBDA 1050 Figure 1. Radiant energy in a spectrophotometer. Note: Low energy regions below 400 nm, around 860 nm, and above 3000 nm
Spectrophotometer Energy Using a Servo Slit at Set Gain One common methodology for improving noise through the use of instrumental parameters is to servo the slit, which varies the resolution of the measurement, as a function of wavelength. This technique has the advantage of maintaining a constant energy and noise level throughout the NIR wavelength range. The Servo Slit mode maintains 70% energy through almost the entire NIR region, with the exception of the long wavelength area above 2900 nm. The energy declines here because the slit is open to its maximum value of 20 nm. Once the slit has opened to this maximum size, the energy will start to decline. The Detector Range Problem and Noise As seen in Figure 2, as a sample absorbs more light, less energy falls on the detector. The higher the sample absorbance, the greater the noise contribution to the spectrum. At some point the signal disappears into the noise background of the detector. This is why many spectrophotometers cite noise specifications at several increasing absorbance levels. How a given spectrophotometer s software, firmware, processes noisy, high absorbance data is very important to understanding the photometric accuracy in samples of high absorbance. In Figure 3, the purple spectrum is around 520 nm, not only does the spectrum noise increase, it would appear that the noise is causing the absorbance to go off scale. The subject of what happens as a function of noise for a highly absorbent sample is in considerable detail to the right. The Problem With High Absorbance Spectrophotometers are unable to measure sample absorbance directly. They can only directly measure transmittance of a sample. Absorbance is a post-data acquisition processing step. This fact is critical in understanding high absorbance spectra. High absorbance samples should always be measured in %T mode, so that one can directly see what is happening to the signal in the instrument. The spectrum in Figure 4, is a 7-plus absorbance blocking filter. Note the %T values for the blocking regions. The values are so close to zero %T that some of the values are negative due to instrumental noise. This causes a slight data handling problem for the absorbance processing step. Beer-Lambert Law and Negative Percent Transmission Values At top left is a graphical representation of the relationship between percent transmission and absorbance values. The logarithmic nature of absorbance is apparent. On the bottom is the Beer-Lambert Law's conversion from percent transmission to absorbance. Remember those negative %T values due to noise for a highly absorbing sample? They represent a real problem. What is the logarithm of a negative number? The Log function Y = Log b (X) is the inverse of the exponential function X = b Y. Since the base b is positive (b>0), the base b raised to the power of Y must be positive (by>0) for any real value of y; therefore, the number x must be positive (x>0) as well. So the real base b logarithm of a negative number must be undefined. Figure 2. Servo Slit. Note: Water noise energy regions around 2600 nm, 1800 nm, and 1400 nm Figure 3. Didymium filters scanned on the PerkinElmer LAMBDA 950 without reference beam attenuation. Figure 4. 7-plus absorbance blocking filter. Note: The 10-6 % Transmission Y-axis range for this sample 2
Since spectra can not have holes in the wavelength axis due to undefined absorbance values, something must be done. The only solution is to substitute a real value that is on the same order of magnitude as the noise data. While this data can serve as a placeholder, it is inserting an unmeasured, artifactual value into the spectrum. Figure 5. The relationship between absorbance and percent transmission. A = 2 - log 10 %T Smoothing to Eliminate Negative %T Values One methodology that can be employed to eliminate the negative %T values is to minimally smooth the original percent transmission spectrum. This smoothing averages the data above the 0 %T threshold thereby eliminating the negative values in the native spectrum. This result can be seen in Figure 6, obtained from that 7-plus absorbance blocking filter. The original raw data, red spectrum, is subject to a 5 point smoothing which results in the blue spectrum where all %T values are positive. The resulting absorbance transformation would then be free from any artifactual data due to undefined absorbance from negative %T numbers. Since the correct minimal level of smoothing is basically a trialand-error procedure, this type of methodology is difficult to automate properly in instrument software. Total Beam Block in the UV Region So what happens in a spectrophotometer when the beam is completely blocked? The answer depends on how the data are processed before presentation in a spectrum. Some spectrophotometers smooth the spectrum internally before presentation. PerkinElmer is committed to never altering raw data so that the user is unable to see the exact nature of the original data. A highly absorbing sample or blocked beam spectrum in a PerkinElmer spectrophotometer will always be very noisy and centered around zero percent transmission. Since minimal energy strikes the detector at high sample absorbance, the only signal generated is random detector noise. The intensity of that noise will depend on the detector gain setting. The spectra here are all measured in the UV region. In Figure 7, the top spectrum in blue is from a 7 absorbance neutral density screen; whereas, the red spectrum is the result of a totally blocked beam. Note the abundance of negative percent transmission values in the blocked beam spectra. When the spectra are all converted to absorbance (bottom graph), the blocked beam spectrum displays noise truncation at around 8 absorbance, but the neutral density screen appears normal. This truncation is the direct result of the insertion of static 8 absorbance values for negative %T values. Figure 7. Spectra measured in the UV regions a 7 absorbance neutral density screen (blue) and the result of a totally blocked beam (red). Figure 6. A 7-plus absorbance blocking filter in the UV region. Figure 8. Spectra converted to absorbance. 3
Total Beam Block in the Visible Region Next, we will consider the visible region. This region typically has the highest relative energy in a UV/Vis/NIR spectrophotometer. In this case, three samples were measured. Their absorbance spectra can be seen in Figure 9. In Figure 10, samples are a 5.8 absorbance (green) and 7.2 absorbance (blue) neutral density screen, as well as a totally blocked beam (red). Note the increasing noise as the sample absorbance increases. This is due to the abundance of negative percent transmission values. The blocked beam spectrum appears smoothed, but in actuality it is the number of substituted 8 absorbance values for undefined negative percent transmission numbers that causes the spectrum to look noise free. Figure 9 displays the associated raw percent transmission spectra for the above mentioned absorbance spectra. Total Beam Block in the NIR Region The last region, the NIR, employs a different lead sulfide detector rather than the PMT used in the UV/Visible region. This solid state PbS detector has a different noise profile and is not as sensitive as the PMT detector. Consequently, it has a larger noise envelope and smaller absorbance dynamic range than the PMT. This can be seen Figure 11 and Figure 12 for a totally blocked beam (blue) and a 5.4 absorbance screen sample (red). Note that in the top percent transmission spectra the blocked beam noise appears to be slightly higher than the zero axis, thereby resulting in a smaller proportion of negative percent transmission values. This translates into a much nosier absorbance spectrum in Figure 12. The smaller absorbance range and difference in noise profile for the NIR detector are clearly be observed here. Figure 9. Raw percent transmission versus absorbance spectra. Figure 11. Comparison of blocked beam spectrum in blue and a spectrum of a 5.4A screen sample in red. Figure 10. Noise as the sample absorbance increases. Figure 12. Negative percent transmission values result in increased noise in absorbance spectrum. 4
Table 1. Statistics on noise envelop in three major spectral regions. Spectrum Mean %T Absorbance P to P Noise Root Mean Square Noise Blocked Beam UV 3.4 x 10-6 7-8 6.0 x 10-6 1.53 x 10-8 Blocked Beam Visible 5.6 x 10-7 8-9 1.3 x 10-6 2.2 x 10-9 Blocked Beam NIR (PbS) 2.9 x 10-4 5-6 7.0 x 10-4 1.5 x 10-9 The Statistics of a Blocked Beam A completely blocked sample beam should result in an envelope of random noise centered around the zero percent transmission axis. The size of this noise envelope will depend on the instrumental detector and the spectral region being measured. Table 1 displays statistics on this noise envelop in three major spectral regions, the ultra-violet, the visible, and the near infrared. As expected, the values in the table reflect the relative energy differences in the three regions. The region able to measure the highest absorbance (lowest noise) is the visible, followed by the UV, and then the NIR. This is most apparent from the root mean square calculation of the noise as well as the peak to peak noise envelope. The mean of the percent transmission yields a rough estimate of the maximum absorbance limit. Scanning Versus Static Wavelength Measurements Noise in a spectrum is usually greater than the noise associated with a static wavelength measurement. While it is true that by increasing the data point response time of a scan the noise level is reduced, there will always be a minimal noise level associated with a moving monochromator resulting in gated data acquisition from the detector. Therefore, the best quality high-absorbance measurement will often result from a static wavelength measurement. The spectrum in Figure 13 is the 7-plus absorbance blocking filter reference previously. Note the three noisy high absorbance blocking regions. Measuring a selected wavelength from each of these regions blocking regions in a static mode should give insight into just how accurate high absorbance measurements are. Static Single Wavelength Measurements Table 2 displays static wavelength measurements obtained from a 7-plus absorbance blocking filter. Wavelength data from the three blocking regions were measured at two different slit settings and the mean absorbance values were calculated from 10 individual measurements of 10-second integration time duration. In general, the measurements performed with the 5 nm slit had lower standard deviations for the three wavelengths than the lower energy 1 nm slit configuration. The reduction in noise level is about one order of magnitude improvement from 1 nm to a 5 nm slit. The standard deviation is approximately 1% to 3% of the absorbance value at 1 nm slit and about 0.1% to 0.3% for a slit of 5 nm. This is excellent precision for a high absorbance measurement. The last two entries in the table represent a stacked sample+attenuator experiment to demonstrate that this filter had not maxed-out the ordinate dynamic range. Figure 13. A 7-plus absorbance plus blocking filter in the UV region. Table 2. Static wavelength measurements obtained from a 7-plus absorbance blocking filter. Wavelength Slit Width Absorbance Mean Standard Deviation 255 nm 1 nm 7.27 0.173 305 nm 1 nm 6.71 0.0669 350 nm 1 nm 6.97 0.113 255 nm 5 nm 7.34 0.0123 305 nm 5 nm 6.83 0.00620 350 nm 5 nm 7.18 0.00910 350 nm No Attenuator 350 nm 0.6A Attenuator 5 nm 7.12 0.0081 5 nm 7.7 0.059 5
High Absorbance Dynamic Range Test Highly absorbent samples always challenge the stray light and noise capabilities of the instrument. Any measurement above 7 absorbance relies on the sample to be its own stray light filter. There is a simple test that can validate the applicability of the instrument in making reasonable measurements at elevated absorbance levels. By stacking a 25 %T (0.6 abs.) neutral density screen with the sample, the spectral absorbance should be uniformly offset by the absorbance of the added screen. As seen in Figure 14, the stacked screen/sample spectrum (blue) is offset from the original sample (red) spectrum. A spectral subtraction of the two spectra yields a fairly flat, uniform spectra with a 0.72 absorbance mean. If the subtraction spectrum is reasonably flat, the spectral detail preserved in the screen stacked spectrum, along with the stacked spectral offset linear with respect to the screen absorbance, the measurement is likely valid. Figure 14. A 7-plus absorbance plus blocking filter in the UV region. PerkinElmer, Inc. 940 Winter Street Waltham, MA 02451 USA P: (800) 762-4000 or (+1) 203-925-4602 www.perkinelmer.com For a complete listing of our global offices, visit www.perkinelmer.com/contactus Copyright 2014-2016, PerkinElmer, Inc. All rights reserved. PerkinElmer is a registered trademark of PerkinElmer, Inc. All other trademarks are the property of their respective owners. 011683_02 PKI