Access the full text.
Sign up today, get DeepDyve free for 14 days.
Implementing statistical training into undergraduate or postgraduate chemistry courses can provide high-impact learning expe- riences for students. However, the opportunity to reinforce this training with a combined laboratory practical can significantly enhance learning outcomes by providing a practical bolstering of the concepts. This paper outlines a flow chemistry laboratory practical for integrating design of experiments optimisation techniques into an organic chemistry laboratory session in which students construct a simple flow reactor and perform a structured series of experiments followed by computational processing and analysis of the results. . . . . Keywords Flow chemistry Design of experiments Optimisation Curriculum Hands-on learning Introduction groups that have used these setups to optimise chemical pro- cesses. Flow optimisation by Design of Experiments (DoE) is Flow chemistry is increasing in popularity as synthetic chem- a very useful and efficient method, where examples have been ists continue to discover the numerous advantages afforded to reported with varying experimental criteria, whether this be them by swapping their round-bottom flasks and condensers yield, purity, E-factor etc [8–12]. DoE as a technique can be for pumps and tubes . The rate of adoption of continuous used for a number of applications aside from chemical reac- flow chemistry is continuing to grow, as more enabling tech- tions, and has been reported often in the pedagogical literature nologies are developed and more research groups are building [13–16]. The teaching of this technique provides students with their own reactor platforms [2–4]. Whilst there are many in- a statistical basis for their experimentation, with the aim of stances of harnessing the capabilities of flow chemistry for solidifying transferrable skills that ultimately lead them to synthesis over traditional batch methods [5–7], there are also becoming well-rounded scientists. However, this optimisation resource is still under-utilised in a lab setting, with one-factor- at-a-time (OFAT) optimisation approaches often substituting Supplementary Information The online version contains as a method for chemical process optimisation and under- supplementary material available at https://doi.org/10.1007/s41981-020- standing [17–19]. The aim of this paper is to provide a flow- 00135-0. chemistry-specific optimisation case study, where students can learn DoE and statistics by performing a continuous flow * Richard A. Bourne R.A.Bourne@leeds.ac.uk based practical experiment, as well as experiencing and over- coming challenges that they would not usually encounter if Institute of Process Research and Development, School of Chemistry they were running a synthetic experiment in batch. and School of Chemical and Process Engineering, University of The OFAT approach is a common method of optimisation, Leeds, Leeds LS2 9JT, UK especially in academia, in which experiments guided by scien- Chemical Development, Pharmaceutical Technology & tific intuition are performed, by fixing all process factors except Development, Operations, AstraZeneca, Macclesfield, UK for one . These factors are experimental conditions (such as University of Southampton, University Road, Southampton SO17 temperature, reagent stoichiometry, reaction time etc.) which 1BJ, UK when combined, make a multi-dimensional space where there Department of Chemistry, Loughborough University, Epinal Way, are a large number of possible combinations of these factors to Loughborough LE11 3TU, UK 76 J Flow Chem (2021) 11:75–86 make up one experiment - this is called the parameter space, Design of experiments and is constrained by the lower and upper limits of each factor (e.g. max and min temperature). After the best value for one DoE is a statistical method of reaction optimisation that is factor has been optimised, another set of experiments are exe- often practiced in industry , but is less commonly used cuted to then optimise another factor, until all factors are in academia, where a OFAT approach is much more common. optimised and the scientist believes that they have arrived at Although OFAT can give an idea as to how particular factors the optimal reaction conditions [21, 22]. However, this method influence the yield of a reaction, the parameters of interest are gives an incomplete picture of the chemical process, as it dis- explored less comprehensively and no indication of how these regards any synergistic effects between any factors in the multi- factors affect each other and themselves at varying levels is dimensional and complex parameter space. This means that obtained. When using a DoE approach, however, the entire interactions between the experimental factors are not consid- parameter space is mapped in an efficient manner, which ex- ered. An example of this could be the difference between a plores multidimensional space at the same time. This is be- reaction at a high temperature with a short reaction time, com- cause in each sequential experiment, multiple factors are pared with a long reaction time - there may not be a linear changed at the same time. A comparison of the parameter relationship between these factors and the desired output, mean- space exploration by these two methods is shown in Fig. 1, ing that a change in temperature after the optimisation of the by observing three experimental factors for the optimisation of reaction time could lead to a suboptimal result [17, 23, 24]. the demonstration reaction: temperature (°C), residence time As research laboratories are diversifying their equipment, (min) and pyrrolidine equivalents. It is typically the case that by incorporating flow and automation technologies, it is also OFAT experiments mostly explore individual planes of pa- necessary for chemists to evolve at the same pace, by diversi- rameter space which makes it difficult to infer the overall fying their skillsets to fully harness the capabilities and ways space behaviour, whereas experimental designs can interpo- of working enabled by this new equipment. Synthetic chem- late factor interactions much more effectively. This is true ists are embracing facets of process chemistry, chemical engi- regardless of the number of experiments undertaken in either neering, analytical chemistry, and programming, to name a approach. The face-centred central composite (CCF) design few. Concurrently, better understanding and adoption of reac- shown in Fig. 1 splits each experimental factor into different tion optimisation methods should be implemented: OFAT op- levels. These levels: (-1, 0, + 1), are named by convention and timisations need to be replaced by more robust and more ef- correspond to the degree of the experimental factor, where − 1 ficient techniques [25–28]. is the lower bound of the parameter space, + 1 is the upper In this paper, the use of a structured experimental bound and 0 is the midpoint. For example, if the experimental design integrated with a flow chemistry platform is re- bounds for residence time were between 1 and 5 minutes, the ported. It is shown how this can be used as a teaching levels would be: 1 minute (-1), 3 minutes (0) and 5 minutes (+ resource to introduce students to performing flow chem- 1). These levels are defined to ensure that all areas of param- istry experiments and to better understand the type of eter space can be explored, regardless of the range of the data required for the optimisation of chemical processes. factors. We herein report a demonstration of Design of These structured experiments allow statistical models to be Experiments for teaching the next generation of chem- constructed from the experimental results, that accurately de- istry in a practical lab setting, whereby a chemical pro- scribe the changes in responses to experimental factor chang- cess with a number of possible products is optimised es. If a DoE analysis tool such as MODDE (from Umetrics) or for the highest yield of a particular product. The chem- Design-Expert (from Stat-Ease) is used, the generation of ical process chosen to be optimised was the S Ar reac- these models is performed easily and intuitively. Empirical tion between 2,4-difluoronitrobenzene, 1, and pyrrol- models, made up of experimental responses, can then be used idine, 2, to form the desired ortho-substituted product, to predict further experimental results based on how the model 3,and impurities 4–5,shown in Scheme 1. weights a particular input variable. These variables can simply be the experimental factors, but can also be interaction terms between the different factors, or squared interactions of the Learning objectives same factor. These interaction and squared terms indicate how experimental factors influence the reaction output, when & To set up a flow chemistry system to execute flow other factors are changed alongside them. In the case of our experiments. S Ar example, it may be insufficient to describe the experi- & To methodically plan flow experiments using DoE. mental data by simply incorporating model terms of ‘resi- & To statistically analyse DoE results and generate empirical dence time’, ‘temperature’ and ‘pyrrolidine equivalents’.It models for an experimental data set. may be a significant factor in the modelling of the data to & To use DoE models to optimise an S Ar flow process. include an interaction term between residence time and N J Flow Chem (2021) 11:75–86 77 Scheme 1 The S Ar reaction of NO N 2 interest, where the yield of the ortho-substituted product, 3,is to be optimised in a flow setup NO NO Desired product NO 12 5 temperature, meaning in real terms that there is a higher influ- generated model will then be used to portray the entire param- ence of residence time/temperature at higher residence times/ eter space, and hence identify the optimum operating condi- temperatures. Similarly, a squared temperature model term tions for the highest yield of the ortho-substituted product, 3. could better describe larger effects of temperature changes There are also other experimental designs one can consider when the temperature is generally higher, meaning that tem- using depending on the outcomes that are desired, but these perature has a non-linear effect. are not covered in this paper . These interaction considerations can give a better descrip- tion of the experimental data, as the synergistic effects be- Necessary equipment tween the factors are also incorporated into the empirical mod- el. In this case, an empirical model is a purely statistical rep- In order to run the experiment as described, it is recommended resentation of the experiments and their outcomes, as opposed to have the following equipment and chemicals. A full list of to a physical model determined by the underlying chemistry. recommended vendors is located in the ESI. This model can then allow response surfaces to be plotted and & PTFE tubing, 1/16” internal diameter. optimum operating regions to be identified, by interpolating & Tubing fittings. the areas between the equidistant experimental points. & A tubing cutter. In this paper, we describe the use of a CCF design in the & Two syringe pumps, or equivalent. MODDE software. After already determining that the three & Three stirrer-hotplates to place water baths on. factors of residence time, temperature and pyrrolidine equiv- & Three water baths, 500 mL. alents are significant, the CCF optimisation design identifies & 2,4-Difluoronitrobenzene (CAS: 446-35-5). all interaction and squared terms between these factors. The & Pyrrolidine (CAS: 123-75-1). Fig. 1 A comparison of the parameter space exploration when conducting a OFAT optimisation alongside a structured DoE design, where • represents an experiment. Note also that a OFAT optimisation does not require a pre-determined number of experiments, and may or may not exceed the number of experi- ments in an experimental design 78 J Flow Chem (2021) 11:75–86 & Triethylamine (CAS: 121-44-8). etc. However, in our case the HPLC calibration was conduct- & Hydrochloric acid (CAS: 7647-01-0). ed in advance by a trained instructor but this could be a task & Common laboratory solvents: ethanol, water, isopropyl for the students as part of the experimental procedure. It is amine. recommended also for students to read introductions to DoE & Access to HPLC, or an equivalent quantitative analytical papers or seek advice from postgraduates or academic super- technique. visors prior to experimentation. Key introductory reading & MODDE Pro, or equivalent DoE software. could include references reported by Krawczyk et al.  andAggerwaletal. , as well as the book written by Antony  which are all useful resources. The experimental flow setup is shown schematically in Experimental setup Fig. 2, and pictorially in Fig. 3.Fourreservoirs were used, one containing 2,4-difluoronitrobenzene (0.1 M) and The experimental bounds for each of the factors are: residence triethylamine (0.11 M) in ethanol, then three other reservoirs time (0.5 to 3.5 minutes), temperature (30 to 70 °C) and equiv- containing triethylamine (0.11 M) and varied pyrrolidine con- alents of pyrrolidine (2 to 10). The concentrations of 2,4- centration (0.1 M, 0.5 M and 1 M) in ethanol. Each experi- difluoronitrobenzene and triethylamine are kept constant. The rationale behind these pre-determined experimental ment setup contains the 2,4-difluoronitrobenzene solution in one syringe, and one of the triethylamine/pyrrolidine in etha- bounds came from the kinetic understanding of the work re- ported by Hone et al. on the same reaction . The HPLC nol solutions in the second syringe, depending on the low/ medium/high equivalents of pyrrolidine that were investigated peak areas are converted to relative concentration percent for each of the species, each of which are reported as outputs for in a particular run. Harvard syringe pumps are used in each experiment to pump the solutions into a PTFE length of tubing that particular experiment. The run order of the experiments (1/16” internal diameter, 6.3 cm, equal to 1 mL volume), was randomised, to prevent any extraneous (uncontrolled) submerged in one of three water baths at 30 °C, 50 °C or variables affecting the results, shown in Table 1. 70 °C. Three water baths were set up so that there are no When running the experiments, undergraduate students can be placed into groups of 5 or 6. Recommended tasks within waiting times between experiments for the water baths to achieve the desired temperature. This is important as the lab the group can be split into: making up stock solutions, prepar- ing the tubing, connecting the tubing, running/timing the ex- time is the most crucial resource, and the experiments must be executed in a specific order; running each block of periments, experimental sampling, running HPLC analysis Table 1 The experimental conditions ran to perform the DoE study −1 −1 ID Run order Residence time /min Temperature /°C Pyrrolidine eq. Pump 1 Flow /mL min Pump 2 Flow /mL min Pump2Conc /M N1 3 0.5 30 2 0.096 1.920 0.1 N2 7 3.5 30 2 0.014 0.274 0.1 N3 12 2 30 6 0.038 0.463 0.5 N4 16 0.5 30 10 0.180 1.820 1 N5 2 3.5 30 10 0.026 0.260 1 N6 8 2 50 2 0.024 0.480 0.1 N7 13 0.5 50 6 0.150 1.850 0.5 N8 1 2 50 6 0.038 0.463 0.5 N9 11 2 50 6 0.038 0.463 0.5 N10 17 2 50 6 0.038 0.463 0.5 N11 4 3.5 50 6 0.021 0.264 0.5 N12 9 2 50 10 0.045 0.455 1 N13 5 0.5 70 2 0.096 1.920 0.1 N14 14 3.5 70 2 0.014 0.274 0.1 N15 6 2 70 6 0.038 0.463 0.5 N16 10 0.5 70 10 0.180 1.820 1 N17 15 3.5 70 10 0.026 0.260 1 The pump flow rates and the concentration in Pump 2 was changed for each experiment to vary residence time and pyrrolidine equivalents, and the tubular reactor was placed in a different temperature water bath to vary temperature. The flow rates are calculated for a 1 mL reactor. Run order should be generated randomly. See Fig. 2 for further details. J Flow Chem (2021) 11:75–86 79 Fig. 2 A schematic of the experimental flow setup used for the S Ar reaction. The pyrrolidine concentration is changed for varying equivalent experiments, and the reactor is movedbyhandintodifferent water baths corresponding to the temperature that the experiment requires temperature experiments (for example, all 30 °C experiments gradient (51% water/49% acetonitrile, each reservoir contain- − 1 at once) could introduce extraneous variables, and must be ing 0.1% TFA) at 1.5 mL min flow rate for 2 minutes HPLC avoided. run time. It is beneficial to have short analytical methods to Each experiment was allowed to reach steady-state by allow fast analysis and turnaround between different sets of equilibrating for 2 reactor volumes, meaning that for each experimental conditions. flow experiment, a wait time of two residence times is neces- sary before collection of material for analysis. For example, if the residence time for the reaction is 0.5 minutes, then 1 min- Hazards ute of reaction mixture is purged to waste before steady-state is established. For each experiment, the desired temperature is Safety goggles and lab coats should be worn throughout the reached by placing the tubular reactor in a separate water bath course of the experiment. All handling of organic solvents and at the corresponding temperature. Samples can then be taken preparation of solutions should be conducted inside the fume from the end of the reactor, by immediately quenching a few hoods. Special care should be taken when handling concen- droplets of material into a vial containing a drop of hydrochlo- trated hydrochloric acid to quench the reaction in the HPLC ric acid at the outlet of the flow system. This can then be vial. If any reagent is spilled on the body, wash the area with diluted with methanol before transferring to analysis. These copious amounts of water for at least 15 minutes. Consult the samples can then be sampled by HPLC, or other analytical MSDS for the specific guidance on handling each of the techniques such as GC, requiring that quantitative yields of chemicals. After experimentation, any tubing can be washed each of the species can be obtained - this is shown in Fig. 2. by pumping isopropyl alcohol through the reactor for 10 re- HPLC analysis was performed using an Ascentis Express C18 actor volumes in order to keep the tubing, or it can be column (5 cm x 4.6 mm x 2.7 µm), using an isocratic flow discarded. Fig. 3 The flow setup used for the S Ar experimentation, where the tubular reactor is submerged in one of three different temperature water baths 80 J Flow Chem (2021) 11:75–86 Table 2 The experimental dataset generated from the running of the DoE for the S Ar reaction Run Run order Residence time /min Temperature /°C Pyrrolidine eq. (1) /% (3) /% (4) /% (5) /% N1 3 0.5 30 2 79.7 20.3 0.0 0.0 N2 7 3.5 30 2 36.3 60.0 0.0 3.6 N3 12 2 30 6 29.6 66.4 0.0 4.1 N4 16 0.5 30 10 52.7 44.6 0.0 2.7 N5 2 3.5 30 10 10.9 83.9 0.0 5.2 N6 8 2 50 2 34.0 62.0 0.0 4.0 N7 13 0.5 50 6 41.2 55.3 0.0 3.5 N8 1 2 50 6 13.8 80.9 0.0 5.3 N9 11 2 50 6 14.9 79.9 0.0 5.2 N10 17 2 50 6 14.9 79.9 0.0 5.2 N11 4 3.5 50 6 6.9 87.3 0.0 5.8 N12 9 2 50 10 9.1 84.9 0.4 5.6 N13 5 0.5 70 2 49.1 47.8 0.0 3.1 N14 14 3.5 70 2 11.8 82.2 0.4 5.6 N15 6 2 70 6 4.7 88.1 1.2 6.0 N16 10 0.5 70 10 15.8 78.8 0.0 5.4 N17 15 3.5 70 10 0.5 91.0 2.5 6.0 The replicates in this data set are experiments: N8, N9, N10 Analysis, results and discussion found in the ESI. MODDE then fits a saturated model for each of the responses given. A saturated model is where all model The full CCF DoE (shown in Fig. 1) was run using the experi- terms, including all interactions and squared terms, are includ- mental setup described in Figs. 2 and 3, which consisted of ed in the model. When a saturated model is initially generated, 2 2 running the experiments shown in Table 2. Three centre-point the R value is the largest value it can be. R is a percentage experiments were also run throughout the course of data acqui- measure of how well a given model fits the data, which is sition, to monitor the reproducibility of the experiments as time usually represented as a number between 0 and 1. When a passed. These repeated experiments, or replicates, ensure that model uses all of the possible terms available to it, the varia- any extraneous variables are identified (uncontrolled variables tion in the experimental response is best described. This that are being changed unknowingly, e.g. stock solution contam- means that as R tends to 1, more of the variation is explained ination or degradation). The outputsareshownasmolar percent- by terms in the model, as closer to 100% of the experimental ages, where the starting material 2,4-difluoronitrobenzene (1), variation can be attributed to specific terms. However, satu- the desired product (3), the para-substituted impurity (4), and rated models typically contain non-significant model terms 2 2 the di-substituted impurity (5). We assumed that each of the that lead to a low Q value. The Q value is the percentage materials have equivalent HPLC response and did not run prior of the variation of the response predicted by the model by calibrations with standards, although this could be done with using cross validation, represented as a number between 0 additional time. Molar percentages were calculated using inter- and 1; simply put, Q tells you how well the model can predict nal normalisation for each of the species, where the area of the new data. For a useful model, it is necessary to have a high R HPLC peak for the species of interest was divided by the total that explains the dataset well, as well as high Q that can summed HPLC area for each peak, multiplied by 100. This is interpolate new data points accurately. To achieve this, the showninthe equationbelow,where (x)isthe HPLC area for the model for each response must be edited as to remove any species of interest, and (1)/(3)/(4)/(5) are the HPLC areas of the nonsignificant terms. Figure 4 shows the coefficients plots species in this study: for a particular response, in this case the response for the amount of the desired product, (3), which graphically indi- ðÞ x cates each model term (x axis) and their respective signifi- Molar percentage ¼ 100 ½ ðÞ 1 þðÞ 3 þðÞ 4 þðÞ 5 cance (y axis). Each of the model terms are ‘scaled and cen- tered’, meaning factors with different units can be compared to determine the influence of model terms over the range of Using this dataset, MODDE can fit a model automatically the factors studied. using the ‘Analysis wizard’ tool. Full instructions can be J Flow Chem (2021) 11:75–86 81 Fig. 4 Thesignificanceof model terms on the response for the desired product, (3). a The difference between significant and non-significant model terms. b The saturated model, R = 0.990, Q =0.764. c The optimised model, R =0.986, Q = 0.894. Time = residence time, Temp = temperature, Eq. = pyrrolidine equivalents Each model term has a respective uncertainty (repre- This process is then repeated for the other responses of sented in the plot as an error bar), and if that uncertainty compounds (4)and (5), shown as Fig. 5a/b and Fig. 6 respec- overlaps with y = 0, then that model term can be deemed to tively. The response for (4), the saturated model (Fig. 5a)ap- be statistically nonsignificant, Fig. 4a illustrates this point. pears to describe the data well as the R is high, however, there This is because there is a probability that the relative effect are many non-significant terms. Because of this, the Q is of the model term could be zero. The saturated model for negative, meaning there is no acceptable degree of predictabil- the response of (3) is shown in Fig. 4b, where there are ity to be obtained from the model. As these non-significant several significant model terms and two non-significant terms are removed, even more terms become non-significant, 2 2 2 terms: Temp and Temp*Eq.. The R and Q measures until the only significant term that remains is temperature 2 2 are shown alongside the coefficients plot as a green bar (Fig. 5b), but the R and Q values are still very low. This is and a blue bar respectively. Upon removal of the two because the response for (4) remained largely unchanged non-significant terms, shown in Fig. 4c,the Q value rises throughout our experimentation, meaning it is difficult to from 0.764 to 0.894, meaning that the predictability of the model well, as there were no factors that could be shown to model is increased for an insignificant decrease in R . have a strong effect on the outcome of this response. 82 J Flow Chem (2021) 11:75–86 Fig. 5 Thesignificanceof model terms on the response for (4). a The saturated model, R = 0.864, Q = -0.200. b The 2 2 optimised model, R =0.246, Q = -0.047 Conversely, the model for the response of (5) was found to be As the models were further optimised to have the highest 2 2 2 excellent without any need for further optimisation - the R R and Q possible, the optimum operating conditions for the and Q measures were both high, and the saturated model production of the desired ortho-substituted (3), could then be contained no non-significant model terms. This means that identified. By selecting the ‘4D Contour’ option in MODDE, by using the same experimental data set, a secondary response the response for (3) can be interpolated across the entire land- can also be modelled and optimised for. This means that re- scape of the parameter space, providing a total insight into the sponse surfaces for (5) can also be predicted without any fur- chemistry that could not be achieved by other means such as ther experimentation. Interestingly, it is not possible to do the OFAT. This contour plot is shown in Fig. 7, which indicates same for the response for (4) due to the low formation of the clearly the yield of (3) that would be achieved with varying product, as there are no changes in the experimental condi- experimental factors. Figure 8 shows a similar plot on how the tions that lead to a significant amount of this product being yield of the di-substituted impurity, (5), also changes with generated. This manifests itself in the uncertainty of the model these differing inputs. It is important to note that to sensibly terms, as most of the error bars for these model terms intersect use contour plots, DoE model performance metrics such as R y = 0 and are therefore their relative effects are non- and Q must be good. This is also a significant point for the significant. student learning and can be adapted into leading questions J Flow Chem (2021) 11:75–86 83 Fig. 6 Thesignificanceof model terms on the response for (5), showing the saturated model with no non-significant terms, R = 0.997, Q =0.944 such as: ‘Using the 4D Contour Plot, predict the yield of the be refined to give further understanding and predictability. major product at x, y and z experimental conditions?’. This could be warranted if there were additional experimental The optimum operating region for the highest yield of the needs, such as productivity of material. This can highlight (3) have been identified using this DoE approach, whilst giv- areas where the highest yields are present in the shortest res- ing a full picture of the parameter space. The results show that idence time, by compromising higher yields for quicker prod- high temperature, high residence times and high pyrrolidine uct generation. equivalents lead to the highest yield of the desired product (3), Upon completion of the experimental work, students were as well as the highest yield of the di-substituted impurity (5). asked to prepare a report on their findings - this can be in a word There are still other aspects of DoE that can be explored, such document or a research article format. Conveying their ability as model validity and reproducibility, predicted kinetic plots to report on statistical models and find optimum reaction con- and ‘Sweet Spot’ visualisations and ‘Optimizer’ usage in ditions for the production of (3)servesasthe main assessment MODDE. These tools can use the same data set to give further criteria for this work, where > 90% of students were successful. process understanding, and the empirical model can be Correct assignments of the optimum parameter regions indi- exported to further explore responses such as E-factor, cates that they have performed the experiments correctly and space-time yield etc. The same data set can also be used to should be considered when grading the report. Further ques- build further models on multiple responses, each of which can tions can also be postulated to the students, such as ‘what are Fig. 7 The contour plot for the response of (3), showing how the yield of the ortho-substituted product changes with varying ex- perimental conditions 84 J Flow Chem (2021) 11:75–86 Fig. 8 The contour plot for the response of (5), showing how the yield of the di-substituted product changes with varying experimen- tal conditions the advantages of running this reaction in flow?’ and ‘why for a desired output. The effect of varying reaction conditions on perform a DoE?’. These questions can enhance the student the outcome of a chemical reaction is explored and therefore learning experience as they are asked to reflect upon their work allows better understanding of the reaction system than a OFAT directly. Sample questions and full answers with suggested approach. This particular experiment is run annually as part of the grading criteria are provided in the ESI. undergraduate chemistry course at the University of Leeds, but can be used as an exercise in teaching flow chemistry and opti- Student feedback misation to researchers at any level. Third year undergraduates that select this optional project learn the theory of DoE as part of This example has formed part of the EPSRC Dial-A-Molecule the pre-laboratory preparation - these theory PowerPoint slides are Summer School in 2018 and 2019 targeted at 1st year PhD provided in the ESI. Depending on the experience of the students, students. The summer school was a lively and interactive event the experiment can be altered to constrain what each participant and in addition to the experiment/analysis outlined also included will conduct experimentally and what is provided for them. The experimental setup requires low cost equipment alongside com- a series of lectures from academic and industrial experts. Furthermore, practical sessions on 3D printing and an evening mon laboratory analytical equipment, and the experimentation itself is suitable for undergraduates and upwards; all experimental session outlining Design of Experiments by designing and mak- ing paper helicopters and optimising the helicopter geometry results can be obtained in a 2–3 hour lab session. This experiment demonstrates that the outlined statistical were also conducted, simply to solidify the concepts of DoE and their applications to various real-life scenarios. These exer- modelling methodologies provide a greater insight into process cises create an equal baseline of background knowledge that optimisation than can be achieved by a OFAT approach and drives the use of the DoE methodology, which forces hypotheses represent some of the most efficient and effective data analysis to be made from understandings of factor selection and level techniques to explain the chemistry and identify regions of setting, rather than undisclosed assumptions based on prior ex- interest. As many exercises in undergraduate courses are based periences. The content was very well received, with 88% on the around synthetic batch experiments, this continuous flow ex- periment can be incorporated into the course as a different ap- feedback rating the course as “Good” or “Very Good”,and the students particularly valued the combination of practical and proach to carrying out a synthetic reaction and obtaining reac- tion data, and simultaneously provide an opportunity to learn theoretical examples detailed in this publication. about statistics and optimisation techniques. This also enable the students to work as part of a group to design and perform the Conclusions experiment, working towards a common goal, broadening their skills and encouraging new ways of thinking. It has been shown that by using a simple continuous flow setup, It is the hope of the authors that as the skillset required of a chemist is diversifying and expanding in sync with the in- consisting of syringe pumps, water baths and a method of quan- titative analysis, alongside a methodical experimental technique creasing capabilities and technologies of a ‘typical’ laboratory - so will the teaching of both chemistry and optimisation such as DoE, that multistep chemical processes can be optimised J Flow Chem (2021) 11:75–86 85 8. Holmes N et al (2016) Online quantitative mass spectrometry for techniques for process development. We are making strides the rapid adaptive optimisation of automated flow reactors. React globally in a positive and constructive way towards laborato- Chem Eng 1(1):96–100 ries which contain scientists with a wide variety of skillsets, 9. Ingham RJ et al (2014) Integration of enabling methods for the and this paper aims to serve as a guide to teaching a number of automated flow preparation of piperazine-2-carboxamide. Beilstein J Org Chem 10(1):641–652 these key skills, i.e. continuous flow synthesis, statistical data 10. Mostarda S et al (2014) Glucuronidation of bile acids under flow analysis, experimental design and reaction optimisation. The conditions: design of experiments and Koenigs–Knorr reaction op- evolution of curricula, the paradigm shift of academic labs and timization. Org Biomol Chem 12(47):9592–9600 overall increased awareness of other methodologies means it 11. Cyr P et al (2013) Flow heck reactions using extremely low loadings of is now a very exciting time to be in a chemistry setting: where phosphine-free palladium acetate. Org Lett 15(17):4342–4345 12. Nieuwland PJ et al (2011) Fast scale-up using microreactors: pyrrole being a chemist is more than just being a chemist. synthesis from micro to production scale. Org Process Res Dev 15(4): 783–787 Acknowledgements The authors thank the School of Chemical and 13. Penteado JC, Masini JC (2011) Exploring liquid sequential injection Process Engineering and the School of Chemistry at the University of chromatography to teach fundamentals of separation methods: A very Leeds for their support, and EPSRC and AstraZeneca for their funding fast analytical chemistry experiment. J Chem Educ 88(2):235–238 and support. We would like to thank the following people for helping to 14. Hope WW, Johnson C, Johnson LP (2004) Tetraglyme trap for the make the 2019 Dial-A-Molecule event such a success: Anna Slater determination of volatile organic compounds in urban air. Projects (University of Liverpool), Dan Tray (GlaxoSmithKline), Adam Price for undergraduate analytical chemistry. J Chem Educ 81(8):1182 (University of Loughborough), Laurence Coles (Added Scientific), 15. McAllister GD, Parsons AF (2019) Going green in process chem- Matt Penny (UCL) and John Blacker (University of Leeds). istry: optimizing an asymmetric oxidation reaction to synthesize the antiulcer drug esomeprazole. J Chem Educ 96(11):2617–2621 Compliance with ethical standards 16. Krawczyk T, Słupska R, Baj S (2015) Applications of chemilumines- cence in the teaching of experimental design. J Chem Educ 92(2):317– Conflict of interest On behalf of all authors, the corresponding author states that there is no conflict of interest. 17. Lendrem DW et al (2015) Lost in space: design of experiments and scientific exploration in a Hogarth universe. Drug Discov Today Open Access This article is licensed under a Creative Commons 20(11):1365–1371 Attribution 4.0 International License, which permits use, sharing, 18. Lendrem D, Owen MR, Godbert S (2001) DoE (design of experi- adaptation, distribution and reproduction in any medium or format, as ments) in development chemistry: potential obstacles. Org Process long as you give appropriate credit to the original author(s) and the Res Dev 5(3):324–327 source, provide a link to the Creative Commons licence, and indicate if 19. Czitrom V (1999) One-factor-at-a-time versus designed experi- changes were made. The images or other third party material in this article ments. Am Stat 53(2):126–131 are included in the article's Creative Commons licence, unless indicated 20. Aggarwal VK, Staubitz AC, Owen MR (2006) Optimization of the otherwise in a credit line to the material. If material is not included in the mizoroki-heck reaction using design of experiment (DoE). Org article's Creative Commons licence and your intended use is not Process Res Dev 10(1):64–69 permitted by statutory regulation or exceeds the permitted use, you will 21. Lovallo D, Kahneman D (2003) Delusions of success. Harv Bus need to obtain permission directly from the copyright holder. To view a Rev 81(7):56–63 copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. 22. Mynatt CR, Doherty ME, Tweney RD (1977) Confirmation bias in a simulated research environment: An experimental study of scien- tific inference. Q J Exp Psychol 29(1):85–95 References 23. Owen MR et al (2001) Efficiency by design: optimisation in pro- cess research. Org Process Res Dev 5(3):308–323 24. Wahid Z, Nadir N (2013) Improvement of one factor at a time 1. Plutschack MB et al (2017) The hitchhiker’s guide to flow chem- through design of experiments. World Appl Sci J 21:56–61 istry. Chem Rev 117(18):11796–11893 25. Weissman SA, Anderson NG (2015) Design of experiments (DoE) 2. O’Brien M et al (2017) Harnessing open-source technology for and process optimization. A review of recent publications. Org low-cost automation in synthesis: Flow chemical deprotection of Process Res Dev 19(11):1605–1633 silyl ethers using a homemade autosampling system. Tetrahedron Lett 58(25):2409–2413 26. Elliott P et al (2013) Quality by design for biopharmaceuticals: a his- torical review and guide for implementation. Pharm Bioprocess 1(1): 3. Christensen M et al (2019) Development of an automated kinetic 105–122 profiling system with online HPLC for reaction optimization. React Chem Eng 4(9):1555–1558 27. Tye H (2004) Application of statistical ‘design of experiments’ methods in drug discovery. Drug Discov Today 9(11):485–491 4. Noël T (2019) Flow into the chemistry curriculum. Chemistry World. https://www.chemistryworld.com/opinion/flow-into-the-chemistry- 28. Murray PM et al (2016) The application of design of experiments curriculum/4010382.article (DoE) reaction optimisation and solvent selection in the development of new synthetic chemistry. Org Biomol Chem 14(8):2373–2384 5. McQuade DT, Seeberger PH (2013) Applying flow chemistry: methods, materials, and multistep synthesis. J Org Chem 78(13):6384–6389 29. Engineering Statistics Handbook: 220.127.116.11.3. Comparisons of re- sponse surface designs. https://www.itl.nist.gov/div898/handbook/ 6. Wegner J, Ceylan S, Kirschning A (2012) Flow chemistry–akey pri/section3/pri3363.htm. Accessed online: 20/03/2020. enabling technology for (multistep) organic synthesis. Adv Synth Catal 354(1):17–57 30. Hone CA et al (2017) Rapid multistep kinetic model generation from transient flow data. React Chem Eng 2(2):103–108 7. Baumann M, Baxendale IR (2015) The synthesis of active pharma- ceutical ingredients (APIs) using continuous flow chemistry. 31. Antony J (2014) Design of experiments for engineers and scientists. Beilstein J Org Chem 11(1):1194–1219 Elsevier, Amsterdam 86 J Flow Chem (2021) 11:75–86 Connor J. Taylor is currently a Richard A. Bourne is currently an postdoctoral research associate Associate Professor at the at Astex Pharmaceuticals in University of Leeds. He complet- col l aborati o n w ith the ed a M.Chem degree at the University of Cambridge. He University of Nottingham (2004) completed an M.Chem degree and his PhD under the supervision at the University of Leeds of Prof. Martyn Poliakoff, CBE, (2017) and his PhD under the FRS. He is now a Royal supervision of Richard Bourne Academy of Engineering Senior and Thomas Chamberlain Research Fellow working on the (2021). During his PhD he development of new sustainable founded the process optimisa- processes with focus on continu- tion company, Compunetics, ous flow routes to pharmaceutical in partnership with the and fine chemical products. His University of Leeds and main- group is based within the tains a strong interest in optimisation, kinetic analysis and reaction Institute of Process Research and Development (IPRD) at the modelling. University of Leeds, a joint institute between Chemical Engineering and Chemistry.
Journal of Flow Chemistry – Springer Journals
Published: Feb 4, 2021
Access the full text.
Sign up today, get DeepDyve free for 14 days.