Aug
19

The Importance of Method Development

I have actually been working on the same experiment for nearly two years. Although some specific variables have changed and the methods have evolved over time, I have still been running some variation of “photolyze liquid and measure concentration” since I joined the lab. This has shown me the huge difference that experimental conditions can have on your results. Anything can and will be a confounding variable, and sometimes things will be different and you will never find out why. Working on the same experiment for so long has taught me a lot about the art and science of method development.

When we started this experiment we worked with larger volumes of solution in a quartz jar. The solution was prepared in a separate container and then transferred to the jar. We would put the jar in front of the lamp and its contents would photolyze. Then we would take small aliquots (samples) of the liquid at 30 minute intervals and place the sample in cuvettes so that we could measure its concentration using the UV-vis. We would then wash out the cuvette and repeat every half-hour. This process took a lot of time and involved a lot of changing containers. Every additional step in a process introduces more error into the result, and we saw this reflected in the data we collected last summer. The constant disruption and transfer of the liquid solution meant contamination was often a problem. Since we work with very low concentrations, the concentration would sometimes be affected by residual water in the cuvette, loss of material in the pipettes we used, or even just evaporation in the transferring process. While we expected the data to reflect a slow and steady decrease in absorbance, we often had data points that bucked this trend, and our data would often show the concentration rising slightly instead of falling.

This data was not a true representation of what was happening in the system. Instead it was a product of the large errors we accumulated with all of the steps in our method. We knew that it was very unlikely that brown carbon was both being produced and destroyed in the way the data was suggesting. Even more importantly, the trends we were seeing were not reproducible between experiments. Data points would jump around in one spot one day and behave normally the next.  Over the following year we made many helpful improvements to make our data more reliable and reproducible. One big change that we made is reducing the amount of liquid transfer. We started photolyzing the solution in smaller volumes, directly inside the quartz cuvettes, so instead of taking liquid in and out we could simply pull out the entire container and measure our concentration without ever disturbing the solution itself. This made a huge difference in the consistency of our results. Another big change we made was investing in a new instrument. Our new UV-vis has a much larger signal to noise ratio and further reduced the errors we were seeing. By the middle of this summer, we had produced what I would call the perfect data set.

By the perfect data set, I mean it was extremely reproducible and clean. We ran some replicates of the same experiment and got results that agreed within a fairly small margin of error. To put it numerically, we went from errors of 30 minutes to errors of 4 minutes. That’s pretty impressive! And we achieved that difference entirely by tweaking our experimental methods.

Comments

  1. I found this post fairly interesting because one of my lab colleagues actually encountered the similar problem as you had. In our lab we use cyclic voltammetry to investigate catalytic ability of metal complexes. In a perfect experiment, the scan should be smooth curve without any disturbance. But in a two-week time period, he continuously got abnormal humps on the curves, which meant he had some impurities present in the experimental system. He tried many times to see which factor contributed to the presence of mysterious humps, but until the end of summer research the humps were still there. Your experience truly inspired us to think differently about our own trouble-shooting.

Speak Your Mind

*