Raman Spectroscopy

7 Raman Spectroscopy Nancy L. Jestel SABIC Innovative Plastics, Advanced Analytical Chemist, Global Spectroscopy Functional Leader, New R & E, Selkirk, NY, USA 7.1 Attractive Features of Raman Spectroscopy Raman spectroscopy is particularly well suited for use in process monitoring and control. This chapter discusses Raman spectroscopy ’ s attractive features as well as alerts the reader to aspects that may present challenges. The fundamental principles of the technique are reviewed. The reader will learn about instru- mentation and options in order to make the most appropriate choices. Special aspects of performing quan- titative Raman spectroscopy are discussed since these are required in many installations. Applications from many diverse fi elds are presented. The reader is encouraged to examine all of the areas since there are good lessons and stimulating ideas in all. Raman spectra encode information about a sample ’ s molecular structure and chemical environment. Fortunately, that information is readily accessible since Raman spectroscopy is an extremely fl exible tech- nique, both in what it can measure and how the measurements can be made. 7.1.1 Quantitative i nformation Under constant experimental conditions, the number of Raman scattered photons is proportional to analyte concentration. Quantitative methods can be developed with simple peak height measurements [1] . Just as with infrared calibrations, multiple components in complex mixtures can be quantifi ed if a distinct wave- length for each component can be identifi ed. When isolated bands are not readily apparent, advanced mul- tivariate statistical tools (chemometrics) like partial least squares (PLS) can help. These work by identifying all of the wavelengths correlated to, or systematically changing with, the concentration of a component [2] . Raman spectra also can be correlated to other properties, such as stress in semiconductors, polymer crystal- linity, and particle size, because these parameters are refl ected in the local molecular environment. Process Analytical Technology: Spectroscopic Tools and Implementation Strategies for the Chemical and Pharmaceutical Industries, Second Edition Edited by Katherine A. Bakeev © 2010 John Wiley & Sons, Ltd. ISBN: 978-0-470-72207-7 196 Process Analytical Technology 7.1.2 Flexible s ample f orms and s izes u sed a s a ccessed without d amage Samples for Raman analysis may be solids, liquids, or gases, or any forms in between and in combination, such as slurries, gels, and gas inclusions in solids. Samples can be clear or opaque, highly viscous or liquids with lots of suspended solids, though certain solid particle sizes may be problematic. Little or no sample preparation generally is required for Raman experiments, which is particularly valuable in a process instal- lation. Generally the sample is not destroyed or altered during the experiment, except for a few materials, such as black or fl uorescent samples, that absorb enough laser light to be thermally degraded or are otherwise photoreactive at the laser wavelength. Even if a sample is heated by the laser, this problem often can be mitigated by using a larger sample volume and/or improving sample stirring or rotation. Raman spectroscopy has no inherent sample size restriction, large or small, but in practice, it is fi xed by the optical components used in each particular instrument design. The lower bound is set by the diffraction limit of light, roughly a few cubic micrometers depending on the numerical aperture of the optics used and the laser ’ s wavelength [3] . The upper bound typically is controlled by practical considerations, such as ensuring suffi cient laser power density at the sample. For example, Raman can be used to monitor the pro- gression of batch, semibatch, or continuous reactions, from small - scale like the curing of an epoxy coating on a microscopic substrate to large - scale, like a polymer produced in a 25 000 gallon (94 635 L) vessel. Macroscopic to microscopic samples can be measured with the appropriate selections of laser wavelength, laser power, and optics. 7.1.3 Flexible s ample i nterfaces Raman spectroscopy offers a rich variety of sampling options. Fiber optics enable multiple remote measure- ment points, up to about 100 m away, collected simultaneously or sequentially with one instrument. Measurements can be made noninvasively or in direct contact with the targeted material. Different probe options are discussed and pictured in Section 7.4.5 . With noninvasive or noncontact probes, there are several inches of air between the front lens of the probe and the sample. This confi guration is particularly valuable for very dangerous or radioactive compounds, corrosives, or sterile or packaged items. Samples can be measured directly through clear glass or plastic bottles, bags, blister packs, or ampoules. Powders are most easily measured with a noncontact probe. Samples in opaque containers may be measured with wide - area illumination (WAI), spatially offset (SORS), or transmission Raman systems. However, SORS and transmission confi gurations currently are not readily available commercially, though that is anticipated to change soon; those used in literature reports generally were custom - modifi ed from existing commercial systems. Contact or immersion probes can be placed directly in the process. The fi ber optics are protected behind a suitable window, such as sapphire, quartz, or glass, with the process material touching the window. Slipstreams are not required, but can be used. These probes are engineered to me et almost any specifi ed chemistry or condition, including high or low temperature, high or low pressure, caustic or corrosive, viscous, slurries, or submerged. However, a noncontact probe might be a better solution if the process components, such as catalyst dust, molten polymer, or powder mixtures, are likely to build up on the window and attenuate signal. Immersion probes with solvent wash capabilities are available but may add to the installation ’ s complexity. If a probe design without such washing capabilities is selected, it is important to ensure that the probe can be removed and cleaned when necessary. The relative time, effort, and expense involved in manual versus automatic options should be compared before choosing. For smaller samples, Raman spectra can be collected through a microscope. Process microspectroscopy systems, such as might be used for semiconductor chip manufacturing or pharmaceutical high throughput Raman Spectroscopy 197 screening (HTS), generally are fully automated. The resolution of the system is determined by the optics and laser wavelength, but generally is around a few micrometers. The same sample interface does not have to be used at each point, though there are limits based on the system ’ s optics. The fl exible sampling confi gu- rations and long fi ber lengths possible give Raman spectroscopy considerable advantages over mid - infrared (MIR), and even near - infrared (NIR) spectroscopy for process applications. 7.1.4 Attractive s pectral p roperties and a dvantageous s election r ules Raman spectra with sharp, well - resolved bands are easily and quickly collected over the entire useful spectral range, roughly 50 – 4000 cm − 1 , though the exact range depends on instrument confi guration. Bands can be assigned to functional groups and usually are easily interpreted. For quantitative work, simple univariate calibration models often are suffi cient. Calibration models can range from 0.1 – 100% for most compounds, and can reach detection limits of 100 parts per million (ppm) for strong or resonance enhanced Raman scat- terers. For low level concentrations, it can be benefi cial to use signal enhancement techniques, such as surface enhanced Raman spectroscopy (SERS) substrates, changes in laser wavelengths to activate resonance Raman (RRS), or the combination of the two, surface enhanced resonance Raman spectroscopy (SERRS) [4 – 6] . The differences in selection rules between Raman and infrared spectroscopy defi ne the ideal situations for each. Raman spectroscopy performs well on compounds with double or triple bonds, different isomers, sulfur - containing and symmetric species. The Raman spectrum of water is extremely weak so direct meas- urements of aqueous systems are easy to do. Polar solvents also typically have weak Raman spectra, enabling direct measurement of samples in these solvents. Some rough rules to predict the relative strength of Raman intensity from certain vibrations are [7] : • totally symmetric ring breathing >> symmetric > asymmetric • stretching > bending • covalent > ionic • triple > double > single • heavier atoms (particularly heavy metals and sulfur) > lighter atoms. A practical rule is that Raman active vibrations involve symmetric motions, whereas infrared active vibra- tions involve asymmetric motion. For example, a symmetric stretching or bending mode would be Raman active, but an asymmetric stretching or bending mode is IR active. Thus, it is clear why highly symmetric molecules, particularly homonuclear diatomic species, such as – C – C – , – C = C – , – N = N – , or – S – S – , generate such strong Raman scatter and are correspondingly weak in the infrared. Some classic strong Raman scat- terers include cyclohexane, carbon tetrachloride, isopropanol, benzene and its derivatives, silicon, and diamond. By comparison, some strong IR absorbers are water, carbon dioxide, alcohols, and acetone. 7.1.5 High s ampling r ate Acquisition times commonly vary from seconds to minutes, often with negligible time between acquisitions, even when measuring multiple locations simultaneously (multiplexing). The dedication of different areas on the charge coupled device (CCD) detector to each measurement point makes this possible. The detectors used for MIR and NIR instruments cannot be multiplexed in the same fashion and must measure multiple samples sequentially. 198 Process Analytical Technology 7.1.6 Stable and r obust e quipment Most process Raman instruments have few, if any, moving parts and thus are quite stable and robust. This reduces the number of parts likely to need replacement. The laser and detector shutter are the most common service items. Special utilities and consumables are not required. Proper enclosure selection allows instru- ments to survive harsh environments, such as installation outdoors or around vibrations. An instrument built to accurate specifi cations and well installed is unlikely to require much care or daily attention. However, it is important to remember that there is no such thing as a maintenance - free instrument. Limited additional maintenance will ensure robust performance for years. 7.2 Potential Issues with Raman Spectroscopy Like all techniques, Raman spectroscopy has its limitations and disadvantages. 7.2.1 High b ackground s ignals Fluorescence is one of the biggest, most frequent challenges to collecting quality Raman spectra, but does not impact MIR or NIR techniques. Common approaches to mitigate the problem include using longer laser wavelengths and summing numerous short acquisitions. Fluorescence problems are not always predictable and often occur in samples not typically described as fl uorescent. If the source of the fl uorescence can be identifi ed, sample treatment options may help. However, there are impressive examples of fl uorescence problems being reduced when the transmission mode is used, so instrument changes may need to be con- sidered also; see Section 7.5 [8] . Regardless, fl uorescence remains one of the fi rst issues a feasibility study must address [9] . It is quite diffi cult to collect Raman spectra of black materials. The sample is usually degraded, burned, or otherwise damaged from absorption of the intense laser energy. Any Raman signal is masked by strong blackbody radiation. Many of these samples are equally challenging for MIR or NIR. Raman spectra of highly colored materials can be similarly diffi cult to collect, though the technique has been used to great advantage in examinations of paintings and pottery [10 – 12] . 7.2.2 Stability Subtle changes to the position and shape of sharp Raman bands are indicative of small changes in the local chemical environment. This makes the technique very sensitive and suitable for challenging chemical prob- lems, but it also puts substantial demands on instrument stability. Small changes in laser wavelength or instrument environment could appear as wavenumber shifts and be mistaken for chemical changes. Instrument fl uctuations can impact quantitative results considerably [13] . Besides using the highest quality, most stable instrument available for such applications, there are other approaches that can be used to reduce the potential for problems. One of the most common approaches to enhance stability is to mathematically align reference or pseu- doreference peaks, such as from a solvent or unchanging functional groups, prior to use in a calibration model. In other cases, it can help to mathematically reduce peak resolution slightly, analogous to using peak areas instead of peak heights. If an instrument has not yet been purchased, changes to the grating to increase resolution by sacrifi cing some spectral coverage may help minimize sensitivity to the fl uctuations. Signals from solvent also can be used like an internal standard [14] . While these approaches may seem like hurdles, they are analogous to learning to handle NIR spectra with multiplicative scatter correction or second deriva- Raman Spectroscopy 199 tives. Similar problems plague NIR instruments, but are less noticeable on the broad, ill - defi ned peaks. While Raman instrument stability is excellent for many applications, users must remain alert to the possibility of a problem in long - term use and develop strategies to mitigate this risk. 7.2.3 Too m uch and s till t oo l ittle s ensitivity Raman spectroscopy ’ s sensitivity to the local molecular environment means that it can be correlated to other material properties besides concentration, such as polymorph form, particle size, or polymer crystallinity. This is a powerful advantage, but it can complicate the development and interpretation of calibration models. For example, if a model is built to predict composition, it can appear to fail if the sample particle size dis- tribution does not match what was used in the calibration set. Some models that appear to fail in the fi eld may actually refl ect a change in some aspect of the sample that was not suffi ciently varied or represented in the calibration set. It is important to identify any differences between laboratory and plant conditions and perform a series of experiments to test the impact of those factors on the spectra and thus the fi eld robust- ness of any models. This applies not only to physical parameters like fl ow rate, turbulence, particulates, temperature, crystal size and shape, and pressure, but also to the presence and concentration of minor con- stituents and expected contaminants. The signifi cance of some of these parameters may be related to the volume of material probed, so factors that are signifi cant in a microspectroscopy mode may not be when using a WAI probe or transmission mode. Regardless, the large calibration data sets required to address these variables can be burdensome. Raman band intensity is a known function of temperature, but affects high and low frequency bands unequally. Quantitative results will be biased if calibration and process samples are measured at different temperatures. A separate temperature measurement may enable the results to be corrected. Even qualitative interpretations of band changes, such as inferring that a species has been consumed in a reaction, may be skewed by temperature - related changes in population distributions. Similarly, Raman spectra are very sensi- tive indicators of hydrogen bonding changes, which are affected by temperature. It is quite possible to accidentally turn the spectrum and instrument into an extremely expensive thermometer. It is possible to obtain temperature and composition from a single spectrum if carefully done, which is an advantage or disadvantage depending on the application. Regardless, temperature must be considered during the develop- ment and implementation of a calibration model. These issues are not unique to Raman spectroscopy, but users may not be as aware of potential problems as for other techniques. Good understanding of the chemistry and spectroscopy of the system being modeled along with well - designed calibration experiments can help prevent problems. The power of the technique can be a hindrance if not recognized and managed. Conversely, Raman spectroscopy has been described as an insensitive technique. Except with resonance enhanced materials, very strong scatterers, or special enhancement techniques, Raman spectroscopy is gener- ally considered to have a detection limit around 0.1% [15] . The technique often is described as being poorly suited to detecting small changes in concentration and this has hindered the technique ’ s use. It is hard to generalize since the limit depends on the situation and equipment confi guration. It is important to test sen- sitivity and detection limits for target applications during feasibility studies and to recognize the demands those requirements put on equipment. 7.2.4 Personnel e xperience In the past, process Raman applications largely were developed and implemented by formally trained Raman spectroscopists, aided by process engineers and other personnel. However, there are far fewer Raman spec- troscopists than chemical, process, or production engineers or other types of personnel. Increasingly, these 200 Process Analytical Technology professionals are identifying Raman spectroscopy as a potential sensor for their systems, arranging for or conducting feasibility studies, specifying equipment, and overseeing installation, model development and validation, and routine operation of the system. Instrument manufacturers are recognizing this shift and beginning to offer systems and services tailored to support this type of customer. However, it still may be necessary to work with a Raman spectroscopist, either an employee or a consultant, on sophisticated, cutting - edge applications; more straightforward applications also may go faster and smoother with technically experienced assistance. 7.2.5 Cost Instruments are available at more price points and from more vendors now than a decade ago, enticing ever more users to try Raman. A 2007 overview of Raman instrumentation lists prices from US$12 000 – 500 000 depending on confi guration and category (handheld, laboratory, or process) – an astonishing range [16] . Generally, handheld or portable instruments are least expensive. For process analyzers, the confi guration differences include the number of measurement points required and distance of each from the analyzer, the kinds of probes used, the demands on an enclosure from weather and electrical classifi cation, the system used to communicate with a facility ’ s host computer, and required redundancies. In a chapter on process Raman spectroscopy, Pelletier wrote that the ‘ analyzer may cost from $50 000 to $150 000, and the integra- tion of the analyzer into the production facility can cost as much as the analyzer itself ’ [15] . As the