OR WAIT 15 SECS
In situ reaction monitoring using spectroscopy can be very useful in an industrial laboratory, but there are many factors to take into account. Here are seven essential steps to follow when considering or implementing the use of this approach.
1. Discuss with the Process Owner and Project Team Whether In Situ Spectroscopy Is the Right Tool for the Task
In situ spectroscopy is often the only viable option if the reaction or process falls into the following categories:
There are also many scenarios when it is advantageous to use in situ spectroscopy, such as:
If your reaction or process does not fall into any of the above categories, then ex situ or off-line analysis may already be sufficient for your purpose. Don't try to use in situ spectroscopy when simple ex situ analysis would already provide the right answer.
2. Feasibility Study or Proof-of-Concept: Choose the Right Techniques by Measuring Relevant References and Process Samples
Assuming your team decides that the project really will benefit from the use of in situ spectroscopy, the next question is usually which in situ techniques should be used, provided that you do have various instrument options available. There are many different in situ spectroscopy tools available now; however, we will only focus on mid-infrared (IR), near-infrared (NIR), and Raman spectroscopy. Each technique has its own pros and cons, and each has its many varieties of sampling modes (see Table I). This table shows the various measurement techniques and associated sampling modes, with caveats. It is critical to get a solid understanding of the process or reaction and ask the following questions (an incomplete list):
The answer to these questions will play a key role in selecting the right technique and sampling accessory.
Once these answers are identified, it is then time to let the data speak. Collect reference spectra from neat starting material and products, and also from typical process samples if they are available and if it is safe to do so. You can do this ex situ using your benchtop instruments. Along with your team, decide what the most critical information is so that they can decide on a suitable course of action. Vibrational spectroscopy techniques tend to be very information rich; different techniques shed light on many different things, such as composition and concentration, H-bonding interactions, speciation, and so forth. Make sure that you do not get overwhelmed by all the information, and stay focused on identifying the key information that truly needs to be measured. Sometimes you may find that you will need to combine multiple tools. For example, you may need to use Raman to determine the peroxide O-O consumption, but also NIR to track the formation of low levels of O-H. Hopefully at the end of this stage (feasibility study, or proof-of-concept), everyone on the team will have a reasonable expectation of what is possible, what are stretch goals, and what are unreasonable wishes.
3. Start with a Calibration Run or a Real Run
Now that you know which in situ spectroscopy tools to focus on, it is time to get some real data. Based on the complexity of the project, we recommend one of the following two approaches:
For a relatively simple reaction system that the team already has a decent understanding of and they know what to expect, start with a few calibration experiments. Keep it simple. For example, in a reaction A+B → C+D in solvent S, measure solutions with known concentrations of A, B, C, and D in S. Use your spectroscopy knowledge to decide whether a single point calibration (at the maximum expected concentration of each species) is sufficient, or if you want to use multiple calibration points (more on this in the next step). Sometimes you can completely skip this step if the project team is only looking for a trend, not full quantitation. However, having such calibration spectra will give you a solid data set regarding the relative spectral responses from all the species.
For a complex system where no one knows what to expect, start with measuring in situ spectra from a real reaction. Having such spectra from a real reaction system will immediately show the team whether the changes involved are subtle (expect a challenging project), or significant. This set of data also holds the key regarding whether any intermediate species may be involved (look for new peaks absent in the starting materials and the products, or peaks in the residual spectra if you use classical least square analysis). With information from this data set, you can go back and prepare a well-designed calibration set.
Notice that this article is written to be instrument-agnostic. Follow your instrument vendor's recommendation in data acquisition. The best practice varies based on techniques and instruments. For IR, make sure you start with a clean ATR crystal and collect background in air (instead of in solvent). For Raman, it never hurts to collect a spectrum of the room light and of the probe itself (sapphire, a frequently used material in Raman probes, has several peaks near 418, 578, and 752 cm-1).
Fouling is another issue that you need to watch out for. Keep in mind that some fouling on the probe body (which is typically made of stainless steel or Alloy C276) is usually not a big problem, as long as the tip of the probe (diamond, silicon, sapphire, ZnSe, quartz) remains clean. Position the tip at a high-shear zone to minimize the chances of fouling, and to maximize your chances of representative sampling.
4. Decide on How Much Data to Acquire and How to Analyze the Data
All modern in situ spectrometers come with software that allows continuous spectral collection. How often to acquire data is something your team should be able to easily decide. If the reaction is completed within a few minutes, you will want to collect every few seconds (and hope you have sufficient signal to noise). If the reaction takes a few days, then don't acquire spectra too frequently, to avoid accumulating unnecessarily large amounts of data. If you are not sure where to begin, try acquiring at 1 min per spectrum and adjust as you go.
Very soon, you will have a large amount of data in hand, and it is then time to think about the best way to analyze them. There are numerous software packages out there, and most likely the instrument software you used in collecting the data may also have some built-in data analysis functions. Again, let your needs drive your actions. For the simplest cases, where only trends are needed to make a decision whether to call a reaction complete or wait for a few more hours, a simple peak height or area analysis (preferably after baseline correction) may already be sufficient. There is no need to overcomplicate things.
When severe spectral overlapping is present (which is almost always the case for NIR), you will want to take advantage of chemometrics. Classical least squares (CLS) is easy to use, but it is easy to make mistakes with it. Make sure that you account for the temperature and matrix effects by using pure component spectra collected at conditions as close to the actual reaction conditions as possible. Partial least squares (PLS) is very popular, but takes more effort to calibrate and behaves more like a black box, in comparison to CLS. There are also many other useful tools, such as multivariate curve resolution (MCR) or indirect hard modeling (IHM), but these are beyond the scope of this article.
Once you finish the spectral analysis using any of the above mentioned methods, you are ready to convert the spectral response to properties your team actually cares about, namely concentration data. This is when you use the calibration curve you developed in Step 3. It is often a good idea to use an internal standard for normalization purposes (a peak from a species that is supposed to be constant, or at least well-behaved). For fixed pathlength NIR transmission measurement and for ATR–FT-IR, you may be able to rely on the absolute absorbance if the temperature is constant.
5. Validate In Situ Spectroscopy Results
Validation takes time and effort. Your team should decide on whether this is necessary, based on the specific needs of the project. For example, if the requirement is to determine only the rough half time of a reaction and you have a well-resolved peak (such as the -NCO peak), then you probably don't need any validation. However, if the goal is to use the in situ results to obtain a quantitative reaction endpoint, or to use the in situ spectroscopy data to develop a rigorous kinetics model, then it is very important to validate the in situ results with some other primary analytical techniques such as GC, LC, or NMR.
6. Implement the Method and Monitor Reactions and Processes
If Step 5 yields great confirming results, then you are ready to implement the method. As a spectroscopist, I find this is the best time to let go and let the project team utilize this newly developed in situ spectroscopy technique in any way they like. Try to train the users (maybe a synthetic chemist or an engineer, or even a technician or an operator) based on their experience level so that they feel comfortable using this new tool. Make sure that the real-time results (hopefully not only just spectra, but also concentration trends) are visualized to their liking. Check back from time to time to make sure that the method is used appropriately, and update it when needed.
Again, always watch out for signs of probe fouling: material deposited onto the tip, stagnant signals when you expect changes, a decrease in solvent signals, or an increase in signal from the dispersed phase.
7. Ready to Move onto the Next Stage: Process Analytical?
It is one thing to develop an in situ spectroscopy method for a laboratory application; it is even more challenging to implement this in a manufacturing plant. There are more constraints and requirements that have to be met before our new method can be deployed in the plant. Installation costs, ease of maintenance, propensity for fouling, and safety concerns all need to thoroughly be discussed with your colleagues in the plant and on the process analytical team before further action is taken.
Xiaoyun (Shawn) Chen is a Senior Research Scientist with Dow Inc., in Midland, Michigan. Direct correspondence to: XChen4@dow.com