Because ammonia gas diffuses rapidly, the response differs little from the immersion method. The air gap does not add significant resistance to mass transfer.
An on-line ammonium measurement systems for fermentation has been reported ( Hill and Thommel, 1982; Thompson et al., (1985). Their technique used direct immersion of the electrode into a sample reservoir after pH adjustment, which gives the fastest response times and highest sensitivities. This method proved unreliable for continuous measurement in our system because surfactants present in the fermentation broth shortened the membrane life of the electrode. Surfactants degrade the hydrophobicity of the membrane causing a drastically reduced membrane life.
To overcome these problems, the electrode was suspended with a tiny air gap
above the sample stream in a small, sealed mixing chamber.
Because ammonia gas diffuses rapidly, the
response differs little from the immersion method. The air gap does not add
significant resistance to mass transfer.
The sample stream was aseptically withdrawn from the fermentor. Two different methods were used. For yeast cultures, a simple air break was sufficient for sampling. To halt metabolism, the pH was adjusted immediately after the air break. For A. eutrophus, precipitation problems forced the use of a recycle filtration sampling system (0.4 µ m crossflow filter Model CFP-4-D-3, AG Technology, Cole-Parmer). The cell-free stream was pH-adjusted just prior to the mixing chamber. All connecting tubes were minimized in diameter and length to reduce transportation lag.
A 0.1N caustic stream was used for pH adjustment. It was pumped simultaneously with the sample stream by using a multihead peristaltic pump, operated at approximately 20 rpm (Ismatec minicartridge 8 channel pumphead, Orange/Orange Three-stop Pharmed tubing, Masterflex 1-100 rpm variable speed drive, Cole-Parmer) resulting in a two streams of equal volume. The two streams mix in-line prior to entering to the mixing chamber. The electrode is an Orion ammonia electrode Model 95-12 with a replaceable membrane. The mixing chamber is glass with inlet and outlet tubes set low in the chamber. The main body has an inner diameter just slightly larger then the electrode outer diameter. This gives minimal total volume but allows visual adjustment of the height of the electrode above the liquid. The electrode-glass connection uses gum rubber tubing. A small stir bar in the chamber is driven magnetically. The liquid volume in the chamber is approximately 0.4 ml and the air gap is about 0.5 cm.
A liquid seal and level control in the chamber result from a small tee placed in the line downstream from the outlet. The exit pumping rate is set higher than the inlet rate to the chamber. The additional volume drawn by the outlet pump is made up by air entering through the tee. The outlet stream is a liquid stream between the chamber and the tee, but an air-liquid mixture after the tee. This fixes the level in the chamber and provides a sealed air-gap between the electrode and the sample stream. If the system were open to the atmosphere, sensitivity would be greatly reduced.
Based on advice from other groups that use this type of specific ion electrode, the sensing system was placed in a plastic box at constant temperature (32° C). This added only a small expense of a thermistor, and an optical relay, and a low-wattage heater. It was also determined that ammonium chloride standards at concentrations below 0.1mM were unstable at room temperature. For calibration purposes, all standards were chilled to 3-5*#176 C. To mimic the ionic strength of fermentation broths, standards below 1.0 mM NH4Cl were prepared in 3.0 g/l KCl.
Calibration using ammonium chloride standards shows that signal was not linear over the entire concentration regime, but linearity was assumed between calibration points in early runs or curve fit with later data. Both step up and down tests, performed using 10 mM and 1 mM NH4Cl standards respectively, show a delay of approximately 1 minute. The majority of this is transportation lag and is characteristic of our particular set-up. It could be improved by using narrower tubes and reducing the distances. The mixing chamber shows virtually no measurement delay within the time frame of our measurements. The system does show some non-linearity. If the response is assumed to be approximately first-order, the step up from 1 mM to 10 mM has a time constant of 35 seconds and reaches 95 percent response in 1.2 minutes. The step down from 10 mM to 1 mM however is slightly more sluggish and has a time constant of 54 seconds and reaches 95 percent response in 3.42 minutes. These time constants are an order of magnitude less than the response time of a typical fermentation system.
Both temperature and flow rate effects were examined, and only temperature was found to be important. A temperature change from 32° C to 28° C, caused a 23 percent change in the signal. An electrode originally at room temperature can take 3-5 hours to equilibrate at temperature setpoints above room temperature. The effect of flow rate through the chamber was investigated by raising the flow rate from the original 97 ml/hr to 168 ml/hr. This has the obvious effect of reducing the delay time, but no effect was seen on the steady state signal using ammonium chloride standards. The effect of flow rate on the dynamics was not investigated.
The signal is linear between 0.1M and 1.0 µ M and loses linearity between 1.0 µ M and 0.1 µ M. This is not necessarily a problem because computers are often used for measurement and control, but it does make calibration much more tedious. The linear range could be improved by the varying the sample-to-caustic ratio.
The ammonium controller use a digital proportional-integral (PI) algorithm. The larger time constant led to control action rate of 1 minute. The longer control action gave an effective increase in control action resolution. We were limited to using on/off cycles on the peristaltic pumps with a resolution of 1 second. By remaining with the 10 second implementation underneath the 1 minute controller calculation, fractions of 1/6 seconds were available. This is means a control action of 41.7 % full flow (i.e., 4.17 seconds) could be implemented by turning the pump on for 5 seconds out of 10 seconds for the first interval and 4 seconds for the remaining five, 10 second intervals.
Use "BACK" to go with the flow.