This is my weapon

You want a nerdfit? I’ll show you a nerdfit.

It’s 1:00 pm and you’re sitting in the instruments lab, just you, your sample, and this monster of a machine. You place your vial on the rack, and hit the start button. The auto-sampler mechanical arm moves in to grasp your sample, making you feel like you’re an important scientist. The syringe extracts a small volume of the sample from the vial, about 5-10 nL or so. Then the liquid begins to move through the column, and the uv-vis starts spitting out data. 1 minute, nothing before the dead time, good. Then noise. Noise. Noise. Noise. Noise. Noise. Noise. Noise. Noise. Noise. Noise. Noise. Noise. Noise. Noise. Noise. Noise. 10 Minutes in, still noise. Fuck. You switch to a gradient. Noise. Noise. Noise. Noise. Noise. Noise. Noise. Noise. 30 Minutes in, nothing. Your sample wasn’t concentrated enough. Fuck, back to the lab to prepare another sample. 3 Hours later you finally get useful data. For a science so exact on paper there can be a ton of variables in the lab.

High Performance Liquid Chromatography (HPLC) is a specific form of liquid chromatography which is an incredibly useful analytical technique for detecting the presence of chemicals in unknown solutions. The most basic form of liquid chromatography relies on size exclusion of molecules. This is done by using a resin for a medium in your column, which allows large molecules to pass through intrabead space quickly, while keeping the smaller molecules in the column longer. This is useful if you’re trying to separate substances that vary greatly in size, like oh say, dextrin and sodium chloride. However size exclusion only works to a certain point, once molecules become close in size we can no longer separate them with this method. We can increase resolution by increasing the interaction of the molecules with the column, thus increasing separation. This is achieved by using a substance with no intrabead space, and interactivity is determined only by the relative polarity of the molecules, rather than their size. This is much more useful because minute differences in polarity are more exploitable for analysis on a column with only a lattice to move through. The only problem with this is that if we were to rely on only gravity to move the molecules through the column, it would take days, months, even years for them to move through. Thus we introduce pressure to the equation. By increasing the pressure of the column to 100-150 bars, we can force the molecules through the column, while still using polarity as an highly precise exclusion method.

And this is where the behemoth you see above comes in. The relative cost of a lab ready HP 1100 Series machine, approximately $100,000. For this you get your choice of pumping systems, your choice of uv-vis spec, an automated sampler, a vacuum degasser, the column of course, the HP control system, and the nice plastic trays on top to hold your polar/nonpolar solutions. A good deal all in all. But what they don’t advertise is the pain and suffering you signed yourself up for.

You see, HPLC is an extremely precise analytical method, so precise in fact, that the data it spits out is going to vary from machine to machine. This is because your data is determined by the resolution in your column. The resolution in your column is determined by the number of theoretical plates in the column, and the number of theoretical plates in the column is directly related to the structure of the column. No two columns are exactly the same at a molecular level, and HPLC is precise enough that it makes a difference. On top of this fact the number of theoretical plates on the column decrease with time/usage, basic wear and tear. This means that whenever you do an analysis of a sample in a lab it must be done on the same machine, which can be problematic when your machine decides that it just doesn’t want to work on Tuesdays and Fridays and sometimes on Mondays from noon-ish to five-ish. If the LHC has taught us anything it’s that no matter how expensive, how precise and complicated our technology is, that it is still susceptible to consumer level problems of electronic mood swings.

An HPLC machine not working is only the beginning of the headaches however, there can be and often are many more problems when it is working. For example, we determine a “dead time” on our column by running Dimethyl sulfoxide (DMSO) through it. DMSO is an incredibly polar molecule and since we often use a reverse phase column, where our solid phase is non-polor,  this means that it will basically not interact with the column at all. This dead time determines the quickest a molecule will run through the column. Sometimes however, we get an absorbance peak before the dead time, which could mean that the molecule in question is even more polar than DMSO, however most of the time we can determine that this is not the case simply by observation of its structure. When this happens it means that the solution is too polar, and that there will be poor resolution in our data. And so we have to start again, changing the solution we run through the column (which is often a mix of water and MeCN).

An example of a DMSO peak

Sometimes it’s not even the problems that are the source of the headache, sometimes the column is working perfectly without issue, and that can be in and of itself a source of misery. Coming back to the fact the we often use a  reverse-phase column, if a substance moving through the column is really, really, really non-polar it can stay on that column for hours, days, weeks, even months before showing up. This can result in one of two things. First, you never detect in in your run because you don’t have weeks. Second, you get unexplainable peaks in your run because someone else’s sample from a couple weeks ago just turned up, leading you on a wild goose chase.

All these problems have to do with the column itself, but often it’s not even the column where there is a problem. For example, when you run something across the column and the only thing you get is noise (often shown by cyclic peaks <5 mAU), it probably means your sample was too dilute, and you must now re-concentrate the sample, which in most cases means preparing your sample again from scratch.

An example of noise from a dilute sample

Also, what happens when you have a molecule that just won’t dissolve? Recently a couple of fellow students and myself were testing food for the presence of various pesticides. One of those pesticides was Imidacloprid (IUPAC 1-[(6-chloro-3-pyridinyl)methyl]-N-nitro-2-imidazolidinimine). Looking at the solubility data, this stuff is just basically insoluble in most substances, with the exception of dichloromethane. Without thinking I prepared a solution of imidacloprid in dichloromethane. What’s easy to forget is that while dichloromethane is a very good organic solvent, at the same time is very volatile. After running the solution once and getting a nonsensical absorbance peak, I was tired and decided to sleep on it. When I came back the next day the dichloromethane had partially  evaporated and eaten away the cap of the vial the solution was in, thus ruining the sample. I had just wasted 200 µL of an analytical standard that costs 52.50$ for 100 mg.  Now if I had more time I could run a rubber/teflon standard across the machine and it wouldn’t be a problem, but the project was due in a couple of days we had other standards to test.

WHAT ARE YOU!?!?!?!?!

There is one final kick in the pants though, when it comes to HPLC. This comes in the form of what we call “negative data.” You see, just because when we run a solution through a column looking for a certain absorbance peak and it’s not there, that doesn’t mean that that particular molecule is not present. What it does means that its below detectable levels at a specific wavelength. You see, no matter how good your column is the actual detection is done by a uv-vis spectrometer at the end of the machine. As such detection is still limited. If we are looking for one molecule we would run it through a uv-spec to find what λmax is for that molecule, then we set the spec on the column to that wavelength. However, if you were looking for multiple molecules, they may not share the same absorbance values. As such you would have to find the appropriate λmax for all the different substances you are testing for, and do a run on the HPLC at each wavelength. On top of this after all is said and done and there is no absorbance peak in your sample, you still can’t conclude that it is not there. You must then dilute your standards until you get zero absorbance at each respective wavelength, and then finally you can come to the conclusion “We did not detect any of substance X in the sample.” So pretty much, negative data is a bitch.

What all this boils down to is hours upon hours of time spent preparing samples, determining λmax, waiting for a machine to spit out data, only to find that that you screwed up, and have to go back and change something. HPLC may be an incredibly useful tool for detection of molecules, but that doesn’t mean that it won’t send you into the royalist of the nerdfits.