Greg: Many drastic process effects on pH measurement are not realized and rarely discussed. Often, the pH measurements most adversely affected are critical for product quality. Here, we gain first-hand plant experience addressing many of these issues, providing insight and a realistic viewpoint from John K. Emmons, analyzer specialist at a chemical plant in Texas.
John: Our basic problem is that the final product (a complex hydrocarbon) must be relatively free of both acid and base within certain tolerances. Either of these will result in reactive impurities. The intermediate process uses semi-aqueous solutions to produce the final product. These are solutions of aldehydes and water, and sometimes potassium hydroxide. The water concentration range is 10-90%, depending where you are in the process. The pH range is 6.5-13, with temperatures from 70-150 °F.
This process has depended on pH meters for control for many years even though the control systems have been sketchy. This plant was basically ignored until this product line became very important. Focus is now beginning to intensify.
About five years ago, I was assigned to assess and make recommendations to the existing pH analyzer systems, which I did. Adoption was slow due to, at that time, the unimportance of the product line. In one sample, I ran across a mixture of water, aldehyde and potassium hydroxide at around 150°F and a pH of 13 (sheesh!). The first thing I recommended is that they cool the sample or keep on changing the pH probe every 1-2 weeks as they had been. Calibration would be much easier also. The second thing was a little odd, but maybe you have heard of this phenomenon: the temperature slope for this solution was not standard. Usually, the Nernst slope increases with temperature. This sample did the opposite; it decreased with temperature. As the sample cooled, the pH went up instead of down with temperature compensation set manually to 25 °C. To get a little more data, I slowly heated and cooled the sample and measured the temperature and pH with manual temperature compensation set to 25 °C over a period of several hours. I took the temperature and pH readings and calculated a slope that was actually negative.
In this case, reducing the temperature to ambient greatly reduces the temperature effect, making calibration much more viable as well as greatly extending the life of the pH probes.
The effect of a “semi-aqueous” mixture on the solution pH and the pH electrodes is basically unknown; especially the interaction among the three compounds at different temperatures. For instance, if you calibrate with standard buffers at 25 °C, will this solution read a pH the same as a normal aqueous solution at the same temperature? It may be that consistent process readings that differ slightly from the “real” pH may be all that can be hoped for.
There is no real data as to what long-term effect this sample has on pH probes; we would need to gather more data to see (after we keep the process cool enough on our pH probes).
Laboratory measurements for this sample at room temperature show differences of +0.5 to +1.0 pH. This error is considered acceptable at this time, but is not acceptable to me, especially considering the possibility of other effects. Again, with the process temperature at ambient, I would expect these results to be much closer.
The corrected pH for this sample could easily be calculated using the new slope in the distributed control system (DCS).
Greg: I am impressed at your excellent decisions to move forward. Cooling the sample is a great idea for prolonging electrode life, especially for a pH approaching either end of the pH scale. You are correct in realizing the standard temperature compensation doesn’t work because the pH is near an acid or base equivalent acid dissociation constant pKa or the water pKw that changes with temperature. Often, the actual solution pH increases as the temperature decreases. Some smart transmitters offer a customized setting for compensation of solution pH with temperature, but I like your idea of doing it in your DCS so you can see the before and after pH, and better adjust it.
More difficult to compensate for is the increase in the pH measurement from the decrease in water concentration even if the acid and base concentrations are unchanged. The pH electrode is measuring hydrogen activity (the product of the activity coefficient and the hydrogen concentration). Hence, a decrease in the activity coefficient from a decrease in water concentration increases the measured pH even though the hydrogen ion concentration is unchanged. This change in measured pH with water concentration is most noticeable when the water concentration is below 60%.
For high water concentrations and for low salt, acid or base concentrations (low ionic strength solutions), the activity coefficient is 1.0 but decreases as the water concentrations becomes relatively low or the ionic strength becomes relatively high causing the measured pH to increase. Thus, I think you are on the right track of realizing that knowing the actual pH is not realistic, and the best you can hope for is that the change in pH with the change in acid or base concentration is relatively consistent. Hopefully, the change in the hydrogen activity coefficient is relatively small compared to your control range.
Doing a buffer sample pH calibration is a good idea to get the electrodes at the same starting point and to see if the electrode must be replaced. If the electrode efficiency has dropped below 80%, the pH offset is more than 2, or the response time is greater than 20 seconds, the pH electrode is approaching the end of its life. However, it may take the reference electrode several hours to equilibrate when it is moved between a process sample that is low in water or high in ionic strength and a buffer solution. Getting an exact calibration is not really required. If the reference is not at equilibrium, it will be difficult to accurately adjust the pH span and get consistent results better than 0.2 pH in low and high pH buffers.
Often, an offset adjustment is the obvious choice for bias errors, but can also be effective for small changes in efficiency since the change in span is mostly seen in the process gain that is already changing much more due to the nonlinearity of the titration curve.
If the offset is adjusted based on a process sample, precautions must be taken to prevent the sample liquid temperature or composition from changing due to cooling, evaporation of reagents or solvents, dissolution of solid or gaseous reagents, reaction or absorption (e.g., carbon dioxide from the air or ions from a glass beaker for pure water solutions). This normally requires that the pH electrode be inserted in the sample immediately after the sample is taken.
If process samples are taken to the lab and measured, the results will not match what you have in your DCS from field measurement because the sample pH and temperature is changing, and lab meters tend to have only standard temperature compensation.
How much is your water concentration changing? What is your pH control range? What is your sample temperature range after cooling?
John: Our laboratory is very good, but, as you say, getting a representative sample and measuring it in the lab is problematic. We have similar problems with high-purity boiler return condensate: CO2 in the air lowers the pH quickly by making carbonic acid if the sample has a large air space in the sample vessel and if the sample is not analyzed quickly at the lab. Even then, there can be problems.
This semi-aqueous sample is even worse because we don’t have any data on the concentration ranges of the components. Aldehyde in the water will evaporate rapidly because it has a very high vapor pressure that quite possibly changes the pH due to increasing water concentration as aldehyde evaporates. Also, aldehyde in contact with air oxidizes into organic acid which will lower the pH; this reaction is slow, but of unknown impact on pH. Current laboratory procedures are based on anecdotal data and best guess only. This could easily be a research project, and in my opinion, should be.
Here are some answers to your questions: How much is your water concentration is changing? Unknown. There has never been a comprehensive assessment of composition over different process conditions. What is your pH control range? Depending on the process, it is 6.5-7.5, 7.5-8.5, and 10.5-12.5 pH. What is your sample temperature range after cooling? I am recommending cooling to between 60 and 80 °F.
Also, thanks for reminding me about reference stabilization between disparate samples. That could be a very important consideration with changing composition in these samples, and one more thing that needs study.
Greg: The sample changing in concentration is a problem that is much more common than people realize. It looks like you really have to depend on buffer calibration and hope that for the same point and time in a process, the changes in water and aldehyde concentration does not cause pH changes beyond your control range.
A flowing junction reference electrode in the actual installation and for testing process samples manually in the field quickly right at the installation point would reduce the liquid junction potential. By using a closed container with electrode connections and by being quick, the resulting reduction in exposure to air and elimination of the reference junction offset equilibration may enable a grab sample check. However, a quick measurement at the sample point may not be practical or safe. Also, the maintenance of an electrolyte reservoir for a flowing junction reference electrode is not liked.
I wonder if a conductivity measurement could infer the changes pH from the changes in water concentration. Maybe your lab could investigate this.
The article “How to Measure pH in Mixed and Nonaqueous Solutions” by Martin S. Frank in the June 1995 issue of Today’s Chemist at Work provides significant insight into the principles and magnitude of the effect of solvent and water concentration on pH.
John: That article you found was excellent; it fits nicely with what I think I know about our problem (phrasing intentional). I also noticed the date; June 1995. Time and again, problems that are “new” have probably been assessed and analyzed somewhere, sometime in the past.
When I was first employed longer ago than I care to admit, we used pressurized flowing references in our high-purity water pH measurements, and refilling the probes was on our PM schedule. I looked into conductivity briefly; that probably could use a closer look. I also looked into ion selective probes (we had a sodium probe at one time; again longer ago than I want to say), but on some of the applications, the pH is too high. Also, with what we know about the difficulties of pH in these solutions I can only imagine what can go wrong with ion selective probes.
Would a semiconductor pH probe like the ion-sensitive field effect transistor (ISFET)-based probe be effective in this solution? I know that they’ve made advances, but I know next to nothing about them except what I learned a few years back and that was not encouraging. I also tried an optical system. It worked well for 10 minutes until the aldehyde stripped off the luminescent dyes on the probe.
Greg: The data I have says the ISFET does not work well below 3 pH or above 11 pH, and is generally less accurate than a glass measurement electrode. Plus, it still has the problems associated with the effect of water concentration on the actual pH and on the reference electrode liquid junction potential.
Conductivity might not be affected as much as pH by water concentration, and will increase with acid or base concentration, but as reagent is added, the conductivity probably increases as well, meaning it does not work as typically envisioned in feedback control. A conductivity measurement upstream and downstream of reagent addition might verify the incremental change in acid or base concentration if you are always on one side of peak in the conductivity-versus-ion concentration curve. Maybe conductivity just offers a second opinion to help decide if the pH electrode is confused or not responding well.
The most intensive and extensive resource by far on all aspects of what determines process pH and its measurement, including two chapters on effect of solvents and ionic strength, is Determination of pH: Theory and Practice, second edition, by Roger G. Bates (1973). Unfortunately, like most incredibly useful books (e.g. all of Shinskey’s books), it is out of print. It is the source of all of my deep knowledge.