The application of combinatorial or high-throughput methods to the discovery and commercialization of new heterogeneous catalysts requires a means of rapidly screening the performance of novel materials. To meet this requirement, UOP Honeywell and SINTEF have a long history of collaborating, developing and optimizing new and efficient tools capable of measuring key catalyst-performance variables. A critical enabler of combinatorial screening is high-speed analysis. However, in parallel testing units, complex analyses often demand a significant amount of time to manually reprocess files. In some cases, this could incur a reduction in testing rate. Furthermore, post-run reprocessing efforts frequently are complicated, tedious, time-consuming and prone to human error. In this article, we will illustrate an effective real-time automated solution to correcting retention time shifts across multiple gas chromatographs with multiple detectors that might lead to peak misidentifications. The solution directly integrates multivariate correlation-based chromatographic alignment software into the high-throughput test control software.
THE CATALYST DISCOVERY PROCESS
The search for a suitable prototype catalyst may require many experiments to screen different chemical elements, additives, substrates and treatments. In addition, evaluation of catalyst substances must take place at relevant temperatures and pressures using applicable chemical feedstocks. This often necessitates a very large number of tests. High-throughput testing and combinatorial approaches have emerged not only as a contributor to catalytic science but as an enabler to product commercialization. The sequential strategy requires a workflow process incorporating catalyst library design and construction, primary or secondary screening, and piloting. Figure 1 shows one of our high-throughput systems for catalyst screening.
High-throughput capabilities and infrastructure for catalyst discovery at UOP Honeywell utilize an integrated end-to-end methodology. It encompasses: a combinatorial synthesis of material libraries; catalyst preparation using ion exchange and impregnation; catalyst finishing such as oxidation, steaming and chlorination; catalyst pretreatments done either in-situ or ex-situ; and parallel testing of catalytic formulations to rapidly identify prototypes for commercialization. Specific program objectives have led to the development and application of innovative tools, novel experimental methodologies, predictive performance models, and optimization tools for heterogeneous catalyst discovery. The tools and methodologies routinely are used in catalyst discovery and technology development projects with broad economic and societal returns.
The current infrastructure supports a wide range of UOP Honeywell process-development, catalyst-development and commercial-support projects. By applying unique testing methodologies, the coverage of process technologies with assorted feedstocks and reactor phase regimes is extensive. Furthermore, the infrastructure is flexible to allow exploration of future process technologies.
The growth in the processing of high-molecular feedstocks with more-complex product distributions (Figure 2) presents challenges for high-speed analytical methods, making innovations necessary. For example, this paper discusses the development of automated chromatographic alignment tools to ensure proper peak identifications with minimal human intervention.
The defining feature of a high-throughput system is the quantity of analyses per unit time. In this case, the analysis is defined as a gas chromatograph (GC) injection. As those skilled in the art of high-throughput testing well understand, conducting a statistically meaningful experiment requires multiple injections. These are achieved by a combination of a process flow scheme that directs multiple reactors to an instrument and the use of several analytical instruments per test system. The length of the analytical method, product complexity, the requested test length, and the anticipated testing demand determine the numbers of reactors multiplexed onto a given analytical instrument and the total number of instruments per system.
The speed of analysis is a critical parameter and many tactics are used to reduce the analysis time. For example, the number of components individually identified and peak quality may be reduced to gain speed for screening. (More-detailed analyses can be done on selected samples in secondary or pilot testing.) Even when some loss of resolution is accepted for speed, the repeatability of the analysis must remain high. With several reactors being tested by one instrument, the injection-to-injection variability must remain smaller than the performance differences in the materials being tested. Likewise, analyses must be consistent across multiple instruments.
To maximize analysis speed, instruments often are pushed to their performance limits. Small variations in column oven temperature and pressure, due to short equilibration times for instance, result in small-scale peak retention time variability. At the same time, the method time often is compressed and peaks are close together, frequently with incomplete resolution so the chromatogram looks like a “peak forest” having long stretches without a return to baseline (see Figure 2). Realistic feeds and products may not have internal standards and some peaks may not always appear (across high and low conversion conditions) so that standard automatic peak-window-timing adjustment during integration fails. A recurring challenge is to reliably identify a peak critical to a key catalyst performance metric when it is located within a peak forest. The combination of tight peak packing, peak time variability, and the elution of a critical peak in a peak cluster often results in peak identification errors. Because a high-throughput test system may generate numerous chromatograms daily, even a small error rate requires significant post-run operator time for manual reintegration. If the error rate is too high, reprocessing time can exceed the original assay time, effectively reducing overall throughput. If the analysis errors are not manually corrected, the variability of the key performance metrics increases, reducing the statistical power of the screen assay and its ability to discriminate among tested materials.
There are various solutions to shifting retention time in GC analysis. In our efforts over the last few years to develop more-advanced platforms for combinatorial testing of catalysts, UOP and SINTEF have turned to LineUp from Infometrix. This software utilizes a multivariate correlation method, known as the correlation optimized warping algorithm or the COW approach , to adjust the chromatogram retention axis to more closely resemble that of a target.
Figure 3 is a simple illustration of the process using LineUp. The software is provided with a target file and a source file; it then outputs a new result file in which the retention times match the target. The user can tune some parameters to optimize the alignment. For example, “alignment range” can limit what parts of the chromatogram to adjust. In process-intensive cases, this obviously can result in faster alignment by avoiding parts of the chromatogram without peaks or traces similar to the target. Further description of tuning the LineUp process appears in the literature  and the software’s user manual .We have demonstrated success with the GC analysis software platform EZChrom Elite from Scientific Software (acquired by Agilent in 2005) in such a way for many years. This system has an ActiveX automation interface, giving the programmer access and control over all aspects of the GC analysis; the software also has drivers for GC instruments from numerous vendors.
Many of our combinatorial test systems implement several gas chromatographs to handle analyses from numerous reactors within an acceptable timeframe using several stream-selection valves to sample the reactors. Peaks crucial to evaluation of catalyst performance and calculation of key metrics — such as conversion, selectivity and yield — must be properly identified across every chromatogram. Consistent and repeatable data across all instruments, detectors, and reactors are essential; we have demonstrated the efficacy of applying customized automated chromatographic alignment to high-throughput testing.
Figure 4 shows a variability gauge chart illustrating the compression of retention times, comparing data before and after processing with new control software. The peak of this particular product component, undecane, historically has proven problematic in the high-throughput test using standard software without the alignment routine. It often is convoluted within a forest of isomerized product peaks. Retention time compression generates more consistent and correct identification of peaks that frequently are extremely difficult to distinguish in complex chromatograms. The chart in Figure 4 reveals significantly improved detector-to-detector retention time variability, thereby demonstrating successful multidetector functionality.
Additionally, another permutation of these data illustrates improved reactor-to-reactor retention time variability within a single detector channel, as evidenced by the lower standard deviations in the aligned segment compared to the original segment. Even the slightest shift in peak retention time may result in misidentification. Reactor-to-reactor retention time compression is critical for peak identification consistency within a high-throughput test system. These data further support the efficacy of the alignment routine by converging retention times for more reliable peak identification and integration within a multireactor, multidetector system.
A POWERFUL TOOL
Development of the next generation of control software that can execute chromatographic retention-time alignment routines during a run is essential in providing an effective real-time automated solution to reprocess vast amounts of data generated by high-throughput testing. It is especially important for processes with complex product streams having hundreds of peaks and compressed method times. In complex chromatograms, peaks critical to catalyst performance are correctly identified across multiple instruments and detectors. The proof of concept has demonstrated the efficacy of our approach to correctly resolve complex chromatograms. In doing so, demands on personnel to reprocess data have been greatly reduced and enhanced testing efficiency and data quality realized.
MARK KRAWCZYK is a principal R&D engineer at UOP LLC, a Honeywell company, Des Plaines, Ill. CHARLES P. MCGONEGAL and MATTHEW J. SCHMIDT are senior R&D scientists at UOP in Des Plaines. MIKE MCCALL is principal R&D scientist and J. W. ADRIAAN SACHTLER is senior principal R&D scientist at UOP in Des Plaines. MARTIN PLASSEN is research scientist, ARNE KARLSSON is chief scientist and ELISABETH M. MYHRVOLD is senior engineer at SINTEF Materials and Chemistry, Oslo, Norway. E-mail them at Mark.Krawcyyk@honeywell.com, Charles.McGonegal@honeywell.com, Matt.Schmidt@honeywell.com, Michael.McCall@honeywell.com, Adriaan.Sachtler@honeywell.com,
Martin.Plassen@Sintef.no, Arne.Karlsson@Sintef.no and Elisabeth.M.Myhrvold@Sintef.no.
1. Nielsen, N-P. V., Carstensen, J. M.and Smedsgaard,J., “Aligning of Single and Multiple Wavelength Chromatographic Profiles for Chemometric Data Analysis Using Correlation Optimised Warping,”J. Chromatogr. A, Vol. 805, pp.17–35 (1998).
2. “LineUp User Guide,”V. 3.5.0, Infometrix, Bothell, Wash. (September 2012).