The infrared differential method is a standard technique for monitoring in-service fluid properties, and is described in multiple ASTM standards. This paper describes an impactful error in calculating differential spectra that should be broadly and urgently addressed by the industry.
An infrared differential spectrum is obtained from a computer subtraction of a sample spectrum from its reference spectrum. The error in the current method is primarily due to the baseline defined within these original two spectra. This article explains how this issue came to pass despite decades of intensive use. A definition of the issue is given, along with a method of correction. This method of correction involves applying a corrective baseline to the original before performing the subtraction, which means correcting the original spectra before manipulation. Examples of the existing issue are shown with the corrected differential spectra using this improved process. The unimproved definition of the baselines can cause incorrect results in the differential spectra, including a shift in differential baselines or modification in the differential spectrum peak sizes. All of these issues are discussed with examples given. The improved method described should be rapidly incorporated into differential spectra processes, including instrument software and analytic tools.
For many years, lubrication monitoring has proven to be a valuable tool for substantial cost savings through the reduction and root cause remediation of costly equipment failures. Two Fourier-transform infrared spectroscopy (FT-IR) methods of lubrication monitoring have been widely used during this period: the trending (direct read) method, and the differential (subtraction) method. The lubricant trending method monitors the addition and depletion of components over time, requiring multiple sample tests to monitor the trend in the data. The differential (subtraction) method requires a reference spectrum, often a new oil sample, to be subtracted from the in-service sample spectrum. This results in a differential spectrum containing information on the addition or depletion of lubricant sample components. Differential spectral analyses have long been an important technique that researchers have applied to aid in the interpretation of chemical processes occurring during in-service lubricant use. Differential spectral analysis has been used to simplify the resultant spectrum, and to make the changes in chemistry easier to observe, and therefore more actionable. Theoretically, only the sample components not undergoing changes are removed spectrally through the differential process.
When the differential method was originally developed, most infrared instruments were monochromator-based spectrometers with a front sample beam and a back reference beam that were alternatively scanned to obtain the energy spectrum of the sample. The two resulting spectra were subtracted from each other to yield the differential transmission spectrum of the sample in the front beam. The back beam was the reference beam, providing the operator information on how much total energy passed through the instrument. The front beam was the sample beam, and was subtracted from the back beam automatically in the instrument. The spectrum generated was the front beam, percent transmission spectrum of the sample, referenced to the back beam sample as follows.
Original %Transmission = energy transmitted through sample (front beam)/ energy transmitted through matched cell (back beam)
Computerization %Transmission = energy sample2/energy background – energy sample1/energy background
When a differential spectrum was desired, a reference sample was placed in the matched cell in the back beam. The spectrum then would be observed as an optical difference spectrum.
Improvements in infrared technology during the late 1970s led to computerization of infrared spectroscopy. During this time, FT-IR technology displaced common monochromator-based systems. Not knowing any better, we utilized the same practices in the emerging FT-IR devices as had been used in the monochromator-based systems. Many of these practices have never been reviewed nor updated in all these years. They were simply embedded into the software and assumed to be correct. Differential (subtraction) was one of those practices.
Over time, infrared instrumentation adopted the use of interferometers and the Fourier transform algorithm. Interferometer-based instruments combine the infrared energy from a moving mirror and a stationary mirror to determine the transmission spectrum of a sample. For this reason, a background spectrum must first be collected, which is digitally subtracted from the sample spectrum. In contrast to today’s highly stable instruments, in older days devices were not stable enough to have the single beam operation until FT-IR was implemented. With the computerization of the differential operation, we don’t do exactly the same operation. Modern devices don’t ratio the two energies; they subtract the background-referenced spectrum from another background-referenced spectrum. It is more of a subtraction of spectra; however, in this operation modern devices are not removing the effects of the changing baselines.
Initially, the difference spectrum was calculated with unprocessed spectra. When there are no baseline changes this subtraction is valid. However, with in-service fluids the baselines are typically changing therefore the current method needs to be further corrected.
To study this process, let’s start with synthetically generated spectra. These spectra will generate results that can be projected and are known. The spectra can be modified with the calculation estimated and compared.
In Figure 1, the sample spectrum has a narrower peak that is slightly shifted from the broader reference peak. Both spectra have a common baseline position, and baseline adjustment will not be needed. The differential spectrum of the two will produce a spectrum with positive (above zero) and negative (below zero) lobes, due to the shift and breadth difference of the peaks.
Figure 2 is a synthetic spectra example that uses a sample and a reference spectrum subtraction to generate a differential spectrum. In this example, the sample spectrum peak is wider than the reference spectrum peak and has a shoulder. The black lines are drawn to show where the differential peak in the spectrum should be measured.
Figure 3 and 4 show what happens to the differential spectrum as the two original spectra from Figure 2 are shifted upward by 0.2 abs–Figure 3, spectrum 5 shifted; Figure 4, spectrum 8 shifted. No baseline adjustment before the differential peak height or differential spectrum measured was needed, because they are synthetic spectra. The peak area lines are drawn in each figure in black.
Obvious changes are observed in the differential spectrum, especially when the absorbance shift occurs with the new fluid spectrum (Figure 4). Sample 5 in Figure 4 is simulating collection of new fluid with a dirty cell. One should first adjust for baseline, then calculate areas or heights (through the differential spectrum calculation or totaling of spectral areas). Calculation of differential areas can be obtained by subtraction of peak areas of the original new fluid from in-service spectra (where differential is positive).
However, in-service fluids can exhibit both a shift in absorbance (needing adjustment) and a curvature in the baseline (needing correction).
The curvature will often add to the shift in absorbance during subtraction. Subtraction needs to be handled with zeroing; having both adjustment and corrections performed on the individual spectra before they are differenced.
Curvature of the spectra can be very complex and difficult to handle. An older method of handling this issue appears to be successful. The method was to define a short frequency range, to assume linearity, and baseline adjust to this narrowed range. The historically chosen locations are listed in Table I.
These baseline frequencies should be applied to a baseline adjustment of both spectra before the differential is applied. This helps eliminate baseline curvature and works best when applied after the baseline absorbance adjustment is accounted.
Without baseline repair, knowledge of the baseline location becomes very difficult. This differential spectrum in Figure 5 showed a gain in oxidation at 1730 cm-1 and a loss in succinimide at 1706 cm-1, both of which are incorrect based on a zero baseline.
Differential spectral analyses without prior baseline zeroing has been found to have disagreement with spectra produced by prior baseline zeroing. This needs to be addressed. With a differential spectrum, a baseline that is not at absorbance-zero still has ingress components and instrumental factors remaining in the spectrum. A reference collected in the main beam of the spectrometer, or collected as is in conjunction with the sample spectrum, carry interference information with their baseline that needs to be separated from the chemical portion of the spectra. However, when there is no baseline correction applied to either of the two spectra before the differential-math, this is not done. This was not attempted within the older process, but it is in the new process.
When one studies differential spectra, focus on the effects of chemical changes take precedence over baseline changes. Peak changes come from chemistry changes. Baseline changes come from external factors like impurity ingress (dirt, soot, fluid color, and fluid haze, for example) and instrumental factors (foggy cells, pathlength changes, cell window sizes, detector and source energy instabilities, for example). Chemistry and external changes should be separated.
When performing the improved process proposed in this paper, an overall baseline absorbance adjustment is applied to both spectra. This brings the spectral baselines to approximately spectral absorbance zero. Then, to handle the spectral curvature, another narrow regional baseline is applied, similar to that given in Table I. This is projected on both spectra within the studied region to further refine the zero-absorbance baseline. The math difference can now be applied obtaining the difference spectrum. This yields an absorbance zero that is actually the zero baseline. The spectral area in the positive side of zero absorbance to be positive (gained in sample), and conversely that in the negative side of zero absorbance are negative (lost from the sample).
As an example, we take real in-service samples from a diesel engine. Typically, one starts, as with Figure 6, showing a shift in baseline of the in-service sample’s spectrum with respect to a new fluid sample’s spectrum. Some baseline shift can also be seen with the new spectrum. Both spectra require zeroing.
The baseline adjustment and correction are applied to both spectra to bring them each to an absorbance zero baseline (Figure 7).
Now, with the two spectra having their baselines brought to absorbance zero, the subtraction can be applied. In this example the oxidation region is being studied, thus the secondary baseline region that is applied was a minimum at 1850 and 1620 cm-1 to correct for curvature effects.
Figure 8 shows this differential spectrum with the new and in-service spectra both displayed. The differential spectrum (solid line-red) obtained without baseline zeroing of the original spectra is included for comparison.
Figure 8 shows the differential spectra before and after of the zeroing the original spectra. New; Reference fluid (new fluid, green); In-service, fluid taken difference (in-service fluid, gray); Diff3 calc (BL 1800-1670 cm-1), differential with 1800–1670 cm-1 baseline (blue); diff calc (no BL) differential with no baseline (red).
With proper modification of the spectra baseline, and since the subtraction and peak height (area) are just a math operation, the peak height (area) can be measured from the spectra measurements or by summing the positive or negative spectral regions of the individual zeroed spectra.
Sometimes, all we have is a simple baseline shift. When they are not defined properly in the spectra from the beginning, the baseline is really a guess. Figure 9 shows that one would incorrectly report nitration (based on 0.036 abs for no BL example and -0.0026 abs for the corrected sample) based on the peak height of the no-baseline example, although the American Society for Testing and Materials (ASTM) International D7624 (5) baseline zeroed sample shows there is no nitration in this sample. The operator of the physical system the in-service sample is drawn from would find it critical to know if nitration is influencing his system, which might, for example, be responsible for gas compression on a pipeline system of national importance.
Continuing with another example, Figure 10 shows how easily errors may occur. In this example, the un-zeroed spectra generate a baseline that isn’t absorbance-zero; thus, there is a tendency to assign this baseline to the wrong location. At the wrong baseline location, the negative region is estimated as larger than it actually is. In this example, there wasn’t actually any loss in the VII at 1744 cm-1 because it is wrongly defined. In addition, a region of the positive spectrum (1779-1744 cm-1) is incorrectly ignored.
When calculating the peak area, as recommended in ASTM D7414 (2), the two values differ (no-baseline area = 3.16 abs/0.1 mm, correctly zeroed = 3.49 abs/0.1 mm).
With Figure 11, the two differential spectra are obtained from the baseline of the two original spectra (diff calc [no BL] and diff5).
By studying Figure 11, it is clear the peak height of “diff5” (blue line spectrum) and the spectrum with no baselines, “diff calc (no BL)” (red line spectrum), yield entirely different values (reported based on no BL reporting 0.153 abs and diff5 reporting 0.053 abs).
The calculation process currently described in the ASTM literature for differential spectra property analysis does not account for baseline variability. The current ASTM process is an artifact of early developments in the field, and goes with the theory that baselines come from external sources and spectral peaks come from chemical sources. The current process dictates that the difference of the two spectra be obtained first, then a baseline is attempted to be calculated. Peak heights or peak areas are then measured resulting in significant error. The errors are based on the questionable baselines.
To eliminate these systemic errors in modern differential spectroscopy, a computer modified procedure for this difference calculation should be implemented, whereby the spectra are made to have equal baselines at zero first, with the difference and its calculations performed only after zeroing. This procedure produces differential spectra with known baselines at zero absorbance. Thus, all gained components produce positive value peaks and loss components produce negative value peaks. This procedure removes the error of improper baseline assumptions ubiquitous in modern spectroscopy. The importance of making this change cannot be overstated. The improved method has direct implications to the proper management and maintenance of a myriad of industrial and commercial equipment comprising the bulk of global production.
(1) ASTM International, Standard Test Method for Condition Monitoring of Oxidation in In-Service Petroleum and Hydrocarbon Based Lubricants by Trend Analysis Using Fourier Transform Infrared (FT-IR) Spectrometry; ASTM D7414-22, 2014, West Conshohocken, PA.
(2) ASTM International, Standard Test Method for Determination of the Oxidation of Used Lubricants by FT-IR Using Peak Area Increase Calculation; ASTM D7214-22, 2014, West Conshohocken, PA.
(3) ASTM International, Standard Test Method for Condition Monitoring of Phosphate Antiwear Additives in In-Service Petroleum and Hydrocarbon Based Lubricants by Trend Analysis Using Fourier Transform Infrared (FT-IR) Spectrometry; ASTM D7412-22, 2014, West Conshohocken, PA.
(4) ASTM International, Standard Test Method for Condition Monitoring of Sulfate By-Products in In-Service Petroleum and Hydrocarbon Based Lubricants by Trend Analysis Using Fourier Transform Infrared (FT-IR) Spectrometry; ASTM D7415-22, 2014, West Conshohocken, PA.
(5) ASTM International, Standard Test Method for Condition Monitoring of Nitration in In-Service Petroleum and Hydrocarbon Based Lubricants by Trend Analysis Using Fourier Transform Infrared (FT-IR) Spectrometry; ASTM D7624-22, 2014, West Conshohocken, PA.
(6) ASTM International, Standard Test Method for Condition Monitoring of Water in In-Service Petroleum and Hydrocarbon Based Lubricants by Trend Analysis Using Fourier Transform Infrared (FT-IR) Spectrometry; ASTM Work Item 50531, 2014, West Conshohocken, PA.
(7) ASTM International, Standard Test Method for Condition Monitoring of Phenol Antioxidant Additives in In-Service Petroleum and Hydrocarbon Based Lubricants by Trend Analysis Using Fourier Transform Infrared (FT-IR) Spectrometry; ASTM Work Item 73062, 2014, West Conshohocken, PA.
(8) ASTM International, Standard Test Method for Condition Monitoring of Glycol in In-Service Petroleum and Hydrocarbon Based Lubricants by Trend Analysis using Fourier Transform Infrared (FT-IR) Spectrometry; ASTM Work Item 61351, 2014, West Conshohocken, PA.
Dave Wooton has been utilizing FTIR for supporting in-service lubricants for over 45 years. As such, he was one of the original researchers to working with “Computer Assisted IR” and the first to publish which contained differential FTIR.He was also one of the co-founders of the ASTM D02.96.03 on “FTIR of In-service lubricants”.He also served as chairman for more then 10 years, where he was an author or co-author of several FTIR Standard Test Methods that are currently use worldwide to analyze lubricant.Dave Wooton is the owner and primary Researcher of a consulting company, Wooton-Consulting – which he founded in 1999. He can be reached at davewooton@wooton-consulting.com.
Cory Schomburg received his undergraduate degree from Texas A&M University and his Ph.D. from Louisiana State University specializing in analytical electrochemistry and nano-scale molecular spectroscopy. Before joining PerkinElmer, Dr. Schomburg has worked for several companies such as the USDA, The Texas State Chemists Office, ATMI, and Columbian Chemicals Company performing various roles ranging from laboratory technician to laboratory manager.Now as an instrument and applications specialist for PerkinElmer, Cory’s main function is to provide training and technical support for PerkinElmer chromatography, thermal and spectroscopic instrumentation. He can be reached at Cory.Schomburg@PERKINELMER.COM.
David Swanson has held positions with Shell, Bayer, and Hastings Deering in his native home of Australia to his current home in the United States throughout his 25 years in the fluid analysis industry. Specializing in lubricants and greases, David’s in-depth chemical knowledge allows him to determine the underlying conditions affecting the lubricants in a system. In his current position, he oversees the science team and quality system operations at POLARIS Laboratories, an independent fluid analysis laboratory in Indianapolis, Indiana. He can be reached at <dswanson@polarislabs.com>
FT-IR Spectroscopy for Microplastic Classification
December 19th 2024A new study in Infrared Physics & Technology highlights the pivotal role of Fourier transform infrared (FTIR) spectroscopy in identifying and quantifying microplastics, emphasizing its advantages, limitations, and potential for advancement in mitigating environmental pollution.
Advances in Mid-Infrared Imaging: Single-Pixel Microscopy Modernized with Quantum Lasers
December 10th 2024Scientists have developed a novel and creative mid-infrared (MIR) hyperspectral microscope using single-pixel imaging (SPI) technology and a quantum cascade laser (QCL). This innovation offers faster, more cost-effective chemical analysis compared to traditional methods, promising new frontiers in microscopic imaging.
The Advantages and Landscape of Hyperspectral Imaging Spectroscopy
December 9th 2024HSI is widely applied in fields such as remote sensing, environmental analysis, medicine, pharmaceuticals, forensics, material science, agriculture, and food science, driving advancements in research, development, and quality control.