Problem
Mismeasurement of pH is an underappreciated cause of irreproducibility in our chemistry and biology experiments. It’s too bad that color-changing pH strips don’t have the precision we need for most of our work in the lab: they are simple, fast and fun to use (especially with kids or at the dinner table). The digital pH meters and electrochemical probes we use instead aren’t just annoyingly slow and complicated, they can be downright anxiety-inducing – especially when everyone in the lab shares responsibilities for using and maintaining the same equipment. (“Can we really trust our pH measurements when our pH station so often looks like a warzone? Is it OK that someone left the probe sitting in a calibration buffer instead of its storage solution all weekend? Why does the probe seem to be taking even longer to calibrate today? Are my co-workers’ experiments any good if they’ve been forgetting to open the fill hole before making their pH measurements? Is my own confusion regarding the proper care of this pH probe jeopardizing our team’s success?”)
Getting every scientist and engineer their own personal pH probes is not a good solution, as it negates the one silver lining of the communal pH station-as-warzone: the increased likelihood that feedback will find its way to the members of our team who have forgotten key details of their complicated pH measurement training (e.g., see here and here). How can we ensure that feedback is generated in as clear and timely and actionable a way as possible?
Solution
Many labs are already on the right track by keeping a handwritten calibration log next to their shared pH meter, but pencil-scratched data don’t provide feedback by themselves. If we instead capture the same data in this Excel template, feedback is provided automatically and as soon as the data for each calibration event have been recorded:
- Is the calibration slope and offset within the manufacturer’s specification limits? If not, the corresponding cells turn red, and the workbook advises probe maintenance or replacement.
- Was there an unusual amount of drift since the last calibration event? If so, the corresponding point on the control chart turns red. There might be a trivial explanation — e.g., has it been an unusually long time since the last calibration? Alternatively, the last person to use the pH probe (see the log) might have suffered much of that drift in the course of his own experiment without realizing it! Be sure to communicate this concern right away to our potentially affected teammate. If repeatedly the largest drifts are associated with a particular operator, re-training should help to get different operators on the same page regarding our lab’s detailed pH measurement protocol.
You can also download here a version of the same Excel workbook that has been pre-populated with simulated data so that you can get a sense of how its chart and logic work.
Insights
- pH matters – a lot! – for so much of what we do in biological and chemical R&D. A single unit of pH can make reactions that should have gone to completion in hours go nowhere in days.
- The man with two clocks never knows what time it is… similarly, our on-line and off-line pH probes will invariably measure different values for the solutions in our reactors and our representative samples of those same solutions. It is common practice to assume the off-line probe is the accurate measurement since it is the one that has been calibrated more recently. However, a recurring theme in this article is that the pH measurement process is unavoidably complicated, and re-training is frequently needed. So sometimes what appears to be drift in the on-line pH probes is actually a difference in the way two operators (and sometimes the same operator!) are calibrating the off-line pH probes. A tell-tale sign that these differences exist is when, not coincidentally, our off-line probe tells us that 100% of our on-line probes drifted in one direction on one day and then in the opposite direction on the next. This kind of over-calibration of our on-line pH probes can make our experiments noisier despite our intentions to have the opposite effect.
- One source of pH measurement error is temperature. Just 5°C will shift the pH of Tris-buffered solution by as much as 0.15 units. E.g., see here. Therefore, we need to be sure both (1) to specify the temperature of pH measurement in our protocols and (2) to let solutions equilibrate at that specified temperature prior to making our pH measurements.
- The pH of our solutions can change quickly and unexpectedly if they have not been buffered effectively. The easiest way to mis-measure pH is thus to assume that whatever pH we measured at the beginning of our experiment remained unchanged throughout our experiment. So it’s a good idea to measure pH at the end of our experiments, not just at the beginning: if the observed change in pH is less than twice the value of sigma short-term calculated by this pH calibration log (cell C11 when Cell C2 = “pH 7 drift”), it may be a spurious signal attributable to drift in the pH probe; but if the observed shift is larger, it is more likely a real change in pH that could explain a significant amount of the real variation we observe in our experiments.
- Just because we’ve added a buffer to our solution doesn’t mean we are buffering effectively. Most buffers lose most of their buffering capacity when pH has shifted 1 unit from its pKa (which is temperature-dependent). The exceptions are buffers with multiple ionizable groups and thus multiple pKas. Here’s a great resource for learning more about buffers.
- Control Charts are awesome and broadly applicable to the work we do in R&D! Anytime we do the same thing twice – like measuring the pH of a calibration solution – we ought to get the exact same answer, even though some degree of noise is inescapable. But never mind what we’ve read in our manuals or of other people’s experiences on-line – just how much noise can we expect for our research processes performed in our lab? And at what point is our observed variation not merely noisy, but a signal that today’s data should not be compared with yesterday’s data? Control Charts answer these important questions and so much more… stay tuned for more articles on this subject.
