An infrared spectral library for atmospheric environmental monitoring
Infrared (IR) spectroscopy is one of several powerful analytical techniques that is well-suited to characterizing atmospheric composition. A few applications of infrared spectroscopy include air-quality monitoring of building environs, automotive exhaust emissions, and ‘fence-line’ or open-path monitoring near industrial facilities and smokestacks. The development of both high-speed computers and lasers has made possible the commercialization of two kinds of atmospheric monitor: passive Fourier-transform infrared (FTIR), and active laser-based spectroscopy systems.
Though many of these spectroscopy systems are available off the shelf, the availability of comprehensive spectral libraries for atmospheric (gas phase) applications has lagged behind the instrumentation. While vapor-phase spectral libraries exist, they are generally non-quantitative, of low spectral resolution, or limited in their content and/or lack adequate documentation for proper use. This is a problem because, regardless of the application or specific instrumental configuration (Fourier-transform, dispersive, laser-based, etc.), a comprehensive reference library is critical to interpreting the spectral data. In addition, the library must be in a suitable format for ready consumption by any one of the dozens of software programs available for data analysis. For this reason, Pacific Northwest National Laboratory (PNNL), with the support of the U.S. Department of Energy (DOE), is developing a comprehensive infrared spectral library tailored to atmospheric environmental monitoring.1
The PNNL/DOE library is being created to meet the demands of current and future passive and active spectral sensor systems. We believe that we have, for the most part, resolved the issues of quantification, spectral resolution, library content and documentation. The library currently consists of over 400 vapor-phase chemical spectra, in their pure form, at 5, 25, and 50°C. Each reference spectrum is in fact a statistical fit to at least 10 individual spectra, taken at varying concentrations of a specific chemical.
According to the Beer-Lambert law, the spectral absorbance of a chemical is proportional to its concentration. This is, of course, wavelength dependent, and each chemical will present its own unique spectral ‘fingerprint’.There are several important advantages of fitting the reference data to Beer's law: improved signal-to-noise; a statistically meaningful error analysis with the ability to detect outliers; and, in many cases, the ability to detect the presence of impurities. None of these benefits is possible by looking at a single spectrum, or even two or three spectra.
Spectral resolution is often a property limited by the instrument used to perform a measurement. In fact, there is an ultimate spectral resolution above which no new information is yielded. Since the PNNL/DOE library is intended for tropospheric measurements near one atmosphere (760 mmHg), a spectral resolution of finer than 0.1cm-1 (wavenumber) is counterproductive (pressure or collisional broadening sets this limit). The finer resolution reference spectra are required for optimal analysis of laser-based sensor data.
Library content is a balance between wanting everything and limited resources (time and funding). For instance, the Sigma-Aldrich catalog lists some 30,000 chemicals for sale. Consequently, we have prioritized a list of approximately 3,000 chemicals according to probability of observation, importance to atmospheric chemistry/pollution, stability, and vapor pressure (at 25°C). For instance, water, carbon dioxide, nitrous oxide, and many other species are abundant in clean air and consequently are likely to be detected. The oxides of nitrogen (NO, NO2, HNO2, HNO3, N2O4, etc.) are both ubiquitous and play a critical role in the chemistry of the atmosphere.
Obviously, in order to observe a chemical in the atmosphere and as a vapor, it must exhibit some minimal vapor pressure. For instance, many hydrocarbons have significant vapor pressures at 25°C. On the other hand, most salts and metals do not have measurable vapor pressures unless heated to hundreds or thousands of degrees centigrade. We have set a somewhat arbitrary cutoff for analyzing only chemicals with a vapor pressure of 0.1mmHg (at 25°C) and greater. Then there are a large set of chemicals that are not necessarily important to atmospheric chemistry, but are of interest for health and safety reasons. This class of toxic chemicals includes hydrogen cyanide, arsine, cyanogen chloride, and many others.
Lastly, product documentation is paramount to both correct product use and continuity. Continuity refers to the ability of a user, 50 years from now, to be able to understand the history of how the library data was acquired. Each reference spectrum is accompanied by a metadata file (in portable document format, PDF) that contains information about sample origins including impurities, sample preparation, instrument configuration and post-processing information.
A typical library spectrum appears in Figure 1 and contains plots corresponding to vapor phase dichlorodifluoromethane or Freon-12. Although the spectrum extends from 600 to 6500cm-1, no features are observed above 3500cm-1 for this molecule.
The PNNL/DOE infrared spectral library is available for web download for a small fee.2
Accurate interpretation of infrared spectral data requires a comprehensive reference library. Atmospheric monitoring applications put specific requirements on this library including spectral resolution, quantitative accuracy, and chemical content. The spectral library described here addresses these and other important issues, and is tailored for both passive and active atmospheric (tropospheric) monitoring applications.
PNNL is operated for the US Department of Energy by the Battelle Memorial Institute under contract DE-AC06-76RLO 1830.
