How it all began

Imagine for a second that you are a research chemist, having the task of finding new reactivities using a specific catalyst. Let’s say you already have a starting point at hand, for example the reaction shown here:

A typical reaction in Organic Chemistry: Lots of lines and letters.

Let us take all of this apart for a moment. Looking at the scheme, we can see there are two reagents (the ferrocene derivative and the alkyne), a precatalyst (the Co complex) two additives (the silver salt and the pivalic acid) and the solvent, in this case 1,2-Dichloroethane. We also have process parameters such as temperature (100 °C) and time (1.5 h), and we measured one response (yield, 84 %).

A lot to look at, right? Where do these conditions even come from? Well, usually we look at the literature to find similar reactions which have already been done by others, and we gradually deviate from them to see if we can find something new. For example, a search for the CAS number of the Cobalt complex from the above scheme (35886-64-7, conducted on 2020-01-15) yields 194 results. So, strictly speaking, we have to look through 194 papers or reviews to have a comprehensive overview of our field. And this is still very few compared to other, more popular fields. Scrolling through just a few of the 194 papers, we quickly end up with something like this in our heads:

Conditions from literature for our example reaction, and just a fraction of them.


Confusing, don’t you think? How does one continue from here? Of course, you can take some inspiration from the literature and just start with something. If you are anything like me however, you will quickly get overwhelmed by the sheer amount of possibilities. How in the world does one prioritize here? Also, setting up these reactions is not trivial. If you are quick, you can conduct about one or two of those reactions per day, including workup. However this leaves you absolutely no time to do anything else, since you will be running around in the lab all day, weighing and pipetting reagents, standing next to your column while you watch the solvent drip into the test tube (and life out of you), waiting… so much for creative time and concentrated literature research. hat now?

«In any given moment, we have two options: To step forward into growth or to step back into safety.»

– Abraham Maslow

Luckily, at that time we stumbled upon a neat little book called “The Goal – a Process of Ongoing Improvement” by the late Physicist Eli Goldratt. Written as a novel from the perspective of a plant manager, whose factory is on the verge of being closed by management due to inefficient processes, the book introduces the “Theory of Constraints”. Goldratt states that each process must be limited by one or more bottlenecks, otherwise the throughput of the system would be infinite.

The flow of blue stuff is constrained by the width of the narrowest part of… this thing in the middle.

The book also suggests a five-step improvement cycle to get rid of the bottleneck and hence increase the system’s throughput. While being developed for production plants with many employees, we wondered whether this theory could also be applied to a doctoral thesis. So we first wrote down the typical workflow of a doctoral student in organic chemistry:

Just a single grain of salt.

Now let us think about what the actual essence of a doctoral thesis is (or at least should be): Having new ideas based on what is already known, designing a concept based on those ideas, investigating, coming to sound conclusions and overall learning to manage your project in an efficient manner. In reality though, most of your work time is devoted to manual labor as you set up new, unknown reactions while delegating the manual work for known reactions to your students. Now don’t get me wrong here: Being able to handle corrosive, toxic chemicals and fragile glassware is a crucial part of a chemist’s job and should absolutely be learned and trained frequently! The issue we see is when it takes over most of your work time, leaving less capacity for actual thinking and creative work, which, after all, should be the main chunk of your day’s activities as an aspiring researcher, manager, world improver, …!

«Perfection is not attainable, but if we chase perfection, we can catch excellence.»

– Vince Lombardi

So, if we assume that a doctoral student on average has more ideas than he can try out experimentally during his work time, we can agree that manual lab work is usually the bottleneck for an organic chemist on his way to those (in)famous two letters and a dot. Luckily, As we implied, „The Goal“ does not just leave you in the dirt after highlighting all your shortcomings. Quite the contrary: It actually offers a five-step routine to optimize your process and get rid of those bottlenecks! This routine looks something like this:

Five steps to eliminate them all.

Looking back at our typical workflow diagram, the first step in our dreaded manual labor cycle is the actual setup of experiments, which includes numerous steps. An attempted summary of those steps is shown here:

Twist the stopcock, switch on the heatgun, heat evenly for a few minutes, …

So, this is what we usually need just to “get the reaction going”. These are purely mechanical, repetitive tasks, for which no “intelligent” input is necessary at all – once learned, muscle memory takes over and that’s it. Now what’s next? Surely, we want to know some things about our reaction, for example whether it has even worked at all – in the end that’s what we are here for! But what are our options?

Well, the quickest (and probably most used) way to determine whether a reaction went smoothly to the desired product or just spit out some decomposition products (while still consuming all our starting material, of course…) is TLC, short for thin layer chromatography. It’s remarkably simple: Put a tiny amount of your reaction mixture on a silica-coated sheet of aluminium, dip it into a suitable solvent (mixture), wait a few minutes, visualize the spots if necessary and there you have it! You immediately see if your starting material has been consumed and if there are any new, promising spots (and how many). While the mechanical procedure itself is straightforward, it does require a trained chemist to take a look at the results and decide how to proceed.

At least we don’t need UV!

Assuming all is well and the reaction worked (which is not the usual case when you try new things, by the way), the following steps now depend on what you are trying to do: Are you trying to isolate the molecule in its pure form, or do you want to quantify the reaction, i.e. determine how well it performed in terms of yield, conversion and so on? Either way, there is more work to do!

Will he though?

If you are interested in the isolation of your new, precious compound, you should have some sort of purification protocol in mind. This protocol can include steps like filtration, evaporation of volatiles or extraction in a separatory funnel, just to name a few. Often, the last step will be a column chromatography or a recrystallization, which will hopefully yield your spectroscopically pure compound, ready for that full characterization. Hurray! Note however, that we are mostly in the area of standard-protocol-robot work again – once you know what parameters you need for your extraction, they will stay the same forever (for that reaction, at least)!

Do it once, be bored forever!

If on the other hand you are only interested in the numbers, sometimes you don’t really need to go through all that. With the right methods in place, there are many cases where you can just take your crude reaction mixture, add some sort of internal standard and plug the whole bunch into an HPLC machine with a UV detector or measure an 1H-NMR spectrum and integrate those protons. Sample preparation usually includes dispensing and filtration steps, again nothing particularly enlightening. Our marvelous brains are only challenged again, when it comes to analyzing those NMR or HPLC/UV results, and making sense of them.

Left: Move your hand in the same way lots of times. Right: Actual complicated chemistry stuff for chemists.

So what now? Well, according to Goldratt, the first step in the overall process is to identify the bottleneck – in our case the manual work required for synthesis. The second and third steps would be to exploit those bottlenecks, and to subordinate everything else to that. At this point, the analogy already gets a bit tricky, because we are not dealing with an entire production plant, but a single person here. And trust me, we can safely assume that doctoral students are already running at their maximum capacity, all the time. To subordinate everything meant we had to stop all manual chemistry work until our problems were solved. As you might imagine, this was not a decision made light-heartedly. Having chosen this path, we arrived at step 4: Elevate the bottleneck. This can partly be done by getting good students to help out with practical work (thank all of you!), but this is not efficient enough and will only get us so far.

To solve this problem in a more long-lasting way, we chose 2 different strategies, which have been well-known in the chemical industry for a long time, namely:

1. Find a way to choose our experiments more carefully, so we get more information out of them, or

2. Increase our synthetic throughput using machines

Our Savior in times of dark – the Chemspeed ASW 2000P.

To find out how we ended up with a fully configured parallel sythesizer capable of conducting 64 experiments at once, and what we have done with (and to) it so far, follow this link. To read about ways to design your experiments according to statistical methods, potentially saving you millions of runs while delivering greater amounts of information, click here.

Now if you are remembering what step 5 is: It tells you to iterate over the first 4 steps over and over, because there will always be a bottleneck. This is where we stopped for now however. Implementing an entirely new way of thinking about chemistry and the underlying practical work is kind of exhausting, and we are quite happy with where we got. We are constantly refining everything though, playing around and trying different use cases to really understand what’s going on here. In the end, all of this is still “Neuland” for us as well.