THE ON LINE MEASUREMENT OF LOW LEVEL DISSOLVED OXYGEN

THE ON LINE MEASUREMENT OF LOW LEVEL DISSOLVED OXYGEN MR. EBBE HÖFFNER, OxyGuard International A/S This paper does not deal with common and already well-known details such as installation, sample preparation, temperature and pressure compensation etc. Instead some basic theory and the major differences between two of the most commonly used types of sensors will be discussed. It is also the aim to try to correct a few common misunderstandings. OxyGuard International A/S was founded by Mr. Ebbe Höffner in 1987. The company specializes in the design and manufacture of oxygen and ph measurement and monitoring equipment and has become a leader in its field. OxyGuard International A/S, Blokken 59, 3460 Birkerød, Denmark. Phone +45 4582 2094 Fax +45 4582 1994 E-mail oxyguard@oxyguard.dk web site www.oxyguard.dk

TERMINOLOGY More than 99% of the instruments used today have sensors of the so-called membrane covered polarographic type. These sensors are amperometric devices as they generate an electrical current when they measure oxygen. Two different sensor types make up this 99%: The Clark cell type and the Mackereth cell type. Some people call the Clark type for polarographic and the Mackereth type galvanic. This is wrong; both types are polarographic as they both contain a pair of polarized electrodes. The difference lies only in how this polarization is created. As it will later be shown the term galvanic is a good name for the Mackereth type of cell, but please never again reserve the term polarographic for the Clark cells only! A membrane covered polarographic oxygen sensor (MCPOS) consists of a cathode and an anode "submerged" in some kind of electrolyte behind a gas permeable membrane, with the cathode touching the membrane. The anode and the cathode are polarized to a potential of some 800 mv, cathode negative. Any oxygen on the cathode is quickly reduced leaving the cathode and the liquid layer between the membrane and cathode virtually oxygen free. As long as there is the slightest amount of oxygen in the sample the oxygen pressure in the sample will be higher than inside the sensor. Oxygen will automatically diffuse from the sample, through the membrane, onto the cathode. In other words it is the partial pressure of oxygen in the sample that is the driving mechanism. DIFFERENCES BETWEEN THE CLARK SENSOR AND THE MACKERETH SENSOR. The Clark sensor, patented in 1956, has an anode made of silver and a cathode made of gold. The electrolyte is usually KCl or KBr and in a few cases KOH. The necessary potential of some 800 mv is applied from an external source, i.e. comes from the transmitter. It is correct to name the Clark Sensor the Externally Polarized Sensor. The Mackereth sensor, described first time in 1964, is nowadays constructed with an anode made of a very electropositive metal such as Zn, Cd, Pb or Amalgam. Even Al and Cu have been tried. The cathode is either a semi-noble or noble metal; e.g. Ag, Au or Pt. Cu works fine, too. The electrolyte is a whole chapter in itself. Depending on the choice of electrode materials many different electrolytes can be used, ranging from NaCl and KCl over Na 2 CO 3 to very strong basic stuff like concentrated KOH. KI is also used despite the fact that it oxidizes quite easily. The electrolyte plays an important role in creating the potential by "dissolving" the anode. Through an internal circuit the electrons released are transferred to the cathode that becomes negatively charged and polarization takes place. The potential is electrochemical - or galvanic if you like. It is correct to name the Mackereth Sensor the Internally Polarized Sensor or The Galvanic Sensor. 2

CHEMISTRY OF THE EXTERNALLY POLARIZED (CLARK) SENSOR Seen from outside the two sensors may look identical but there are major differences when it comes to the chemistry. Both types are corrosion cells in which the corrosion rate is controlled by the amount of oxygen in the sample. When corrosion takes place two different reactions are going on. One is the anode reaction (where the oxidation takes place) and the other is the cathode reaction (in which the oxygen is being reduced). The reaction mechanism of the cathode process is actually very complicated but the result can be expressed quite simply as in the following: Anode reaction: U + 4Ag + 4Cl - 4AgCl + 4e - Cathode reaction: Au O 2 + 2H 2 O + 4e - 4OH - Net reaction: 4Ag + 4Cl - + O 2 + 2H 2 O + (4e - ) 4AgCl + 4OH - + (4e - ) (The electrons are in brackets because they do, of course, not exist in solution). It is beautiful chemistry; oxygen is being reduced and the anode is being oxidized and for each oxygen molecule 4 electrons flow from the anode to the cathode delivering the sensor's output signal. The linearity is close to perfect so why worry and look for anything else? Because there are several problems not so easy to overcome. Electrolyte is consumed; i.e. chloride (bromide) and water disappear. Hydroxyl ions are formed and electro neutrality is maintained, but the electrolyte composition changes continuously and after a period of time depending on the amount of oxygen, the amount of electrolyte and the temperature, all of the electrolyte has undergone chemical change and the sensor stops working. Shortly before this happens the sensor output becomes very unstable or noisy. In order to extend the life-time of such a sensor some of them have fairly large electrolyte reservoirs but even these stop working sooner or later. The problem with the latter is that the anode gets coated or blocked by AgCl. Initially this just increases the response time of the sensor. At some point in time the response time becomes unacceptably long, and finally the anode becomes completely coated and the sensor stops working. Both the change in the electrolyte composition and the coating of the anode lead to changes of the sensor's slope and linearity. This can to some extent be overcome by frequent calibration, but a Clark sensor can only be really accurate just after it has been calibrated. One more problem exists - the residual current generated from the sensor under zero oxygen conditions. As seen from the reaction scheme the current should be zero when no oxygen is present but this is not the case. There is always a residual signal and it is worth mentioning that both sensor types suffer from this problem. However, the externally polarized type normally has a higher residual output than the internally polarized type. The reason is that the externally polarized sensor has one more source of residual current than the internally polarized type. This problem will be dealt with in detail later in this paper. 3

CHEMISTRY OF THE GALVANIC (MACKERETH) SENSOR As several different metals can be used as anode we have chosen to use the symbol Me. Me can be Zn, Cd, Hg (as Amalgam), Pb, Al, Cu and others depending on the choice of electrolyte and cathode material. Most of the metals listed react in exactly the same way when they are being oxidized. The anode reaction therefore is as follows: H 2 O 2Me 2Me 2+ + 4e - The cathode reaction is the same as for the externally polarized sensor: 4e - + O 2 + 2H 2 O 4OH - Net reaction: Ag, Au, Pt (4e - ) + O 2 + 2H 2 O + 2Me 2Me 2+ + 4OH - + (4e - ) If we concentrate on the right side of the equilibrium sign it is quite evident that metal hydroxide is formed: 2Me 2+ + 4OH - 2Me(OH) 2 The hydroxides of the listed metals are all heavily soluble but most of the corresponding oxides are even less soluble. This means that the hydroxides are converted to oxides: 2Me(OH) 2 2MeO + 2H 2 O The net reaction has now this elegant form: H 2 O 2Me + O 2 2MeO Indeed a simple expression; oxygen hits the cathode and is being reduced as the anode is getting oxidized. No consumption of anything, neither electrolyte nor water. The electrolyte no longer takes part in the chemical reaction. The reaction can go on and on until there is no more anode material left as long as you can prevent the anode from being coated and blocked. The idea of using an alloy instead of the pure metal is because pure metals passivate easily whilst alloys can be composed in such a way that passivation does not take place. If the crystal structure of the oxide layer is "messy" enough the layer is porous and has poor adhesive properties to the underlying metal. Hydroxile ions are therefore not impeded in their passage to the metal of the anode. There are even alloys of some metals available specially made to be non-passivating and fast corroding. With regard to the small miracles some vendors perform with the electrolyte - most of this is kept relatively secret and considered proprietary. The trick is, however, to mix the electrolyte in such a way that the oxide is dissolved as soon as it is formed. OxyGuard have researched thoroughly into both the above solutions. For the measurement of low level dissolved oxygen we have chosen to perform miracles with the electrolyte and the result is quite remarkable. The OxyGuard sensor has an operating time of 5 to 10 years or even more between overhaul. In addition it keeps its calibration exceptionally well simply because the internal chemical conditions are constant. A well designed galvanic type of sensor can therefore maintain its accuracy and calibration for a much longer time than a "traditional" Clark sensor. Our experiences support this. If you keep the membrane reasonably clean calibration accuracy is seen to be maintained for many months. It has been claimed by some sworn Clark Cell Purists that the galvanic type of sensor changes with time to the detriment of measurement accuracy and calibration. It must be clear from the above that such claims are completely incorrect, both theoretically and practically. And it is, in fact, relatively easy to keep the anode fully active. The solution is either to use an alloy instead of the pure metal for the anode or to perform small miracles with the electrolyte. 4

RESIDUAL CURRENT As mentioned earlier the externally polarized sensor has one more source of residual current output than the internally polarized type. It is unfortunately not possible to be certain what exactly causes the residual current that is common to both sensor types. but much indicates that the fitting of the cathode into the isolator is of major importance. Even with the best fit there is a risk that crevice corrosion takes place along the sides of the cathode and it is believed that this type of corrosion can create the zero current. Personally, I am not convinced that it is so simple. It is, however, evident that extreme care and accuracy in the production are essential for a good result, especially when dealing with sensors designed for low level measurement. The production of a ppb DO-sensor is much more difficult than you can imagine; in short, it is a thousand times more difficult than it is to produce a ppm DO sensor. The other source of residual current that the externally polarized type has is a lot easier to describe, although it is seldom mentioned by the manufacturers of such probes. The silver anodes of these sensors are coated by silver chloride (or silver bromide or silver oxide). These silver salts are all heavily soluble but they are not completely insoluble. There is therefore a small amount of silver ions in the electrolyte, and as soon as an ion comes near to the cathode it immediately plates itself onto it: Ag + + e - Ag There are in fact two problems associated with this; one can change the slope of the sensor, the other produces a residual current. After some the cathode turns gray - it is no longer a gold cathode but a silver plated cathode. This can lead to changes in the slope of the sensor. Frequent calibration can compensate for this. The other problem is perhaps worse. A current flows when this plating takes place, and the transmitter has no chance of knowing if the electrons flowing are a result of the normal function of the probe or as a result of such a plating action making a significant contribution to the residual output. How big the problem is depends much on the construction and dimensions of the sensor. Theoretically the problem is smallest when the electrolyte is KBr and biggest when it is KOH. This problem only applies to the externally polarized types. The galvanic sensors do not suffer from this problem. They dissolve like this: AgX Ag + + X - 5

RESPONSE TIME Generally speaking the response time of a ppb DO sensor should be as fast as possible. In normal use it is hardly very important if it takes a few seconds more to reach an end value, but it is very important that the sensor recovers quickly after calibration. Only sensors that have ultra fast response time measure correctly at low levels within reasonable time. It should not take more than 3-4 minutes after calibration in air to get back to 10 ppb. The less time you are off line the safer you operate. Usually the time it takes to get down after calibration depends on how long time the probe has been subjected to high oxygen concentrations and, consequently, the shorter the calibration time is, the faster the sensor can measure low values again. During calibration the sensor gets more or less soaked with oxygen. Depending on the mechanical as well as the chemical design it takes some time before all oxygen has been rinsed off the sensor again. It is here the response time really means a lot. If you take a sensor that will calibrate within a minute then you are actually back on line within 5 minutes. Compare that to a sensor that needs 20 minutes in air to calibrate. Such a sensor requires several hours to get down again. Fast response time is even more important if the temperature is low, since lowering the temperature will always increase response time. A sensor that takes 20 minutes to respond at, say, 25 C will probably take several hours if the temperature is lowered to 5 C. After several hours in air such a sensor may need a day or two before it again measures correctly at low ppb levels. There are many things that influence the response time; too many to list here, but mechanical and chemical design, membrane material and thickness as well as thermal mass are some of the most important. The good advice here is to go for systems with as short a response time as possible. Such instruments are a lot easier to work with; if you are in doubt of anything you can pull the sensor out into the air, check the calibration, put the sensor back and a few minutes later you are on line again. And you can forget all the fancy ideas about calibrating in special gas mixtures (they are expensive, too!), Farady cells, automatic air injection, etc. etc. A fast response galvanic cell with its excellent zero characteristics can accurately, easily and inexpensively be air-calibrated and return to normal measurement within a very short space of time. The OxyGuard sensor is one of the fastest responding sensors available. OPERATION AND MAINTENANCE The costs of operation and maintenance are, in fact, much more important than most people realize. Equipment specifications and initial cost are things that you can easily find out before you buy the instrument, but the true cost of operation and maintenance can often only be found when the instrument is taken into use in the actual application. In fact, the cost of spares, man-hours used and the need in some cases for back-up equipment to avoid the loss of a measurement can all add up to more than the initial purchase price. This is specially true for nuclear power plants - handling contaminated items is expensive! Since modern electronics should not need attention the cost of operating and maintaining a DO measurement system are centered around the sensor. There are three things that can be done to DO sensors: 1) calibration; 2) membrane cleaning; 3) sensor renovation 1) Calibration Generally speaking the Clark type cell needs much more frequent calibration than the galvanic cell. As shown above, the chemical changes that take place in the Clark cell affect its performance both with regard to slope and linearity, so it is necessary to calibrate to compensate for this. The galvanic cell is inherently superior to the Clark cell in this respect, and can be made so that it does not suffer from any such changes. Permanent changes in the oxygen permeability of the membrane affect both types of sensor - this is in fact the only factor affecting the sensitivity of a well designed and well manufactured galvanic sensor. The increased response time of a Clark cell when it no longer is completely new make calibration much more time consuming. 6

2) Membrane cleaning Deposits on the membrane of all types of sensor will affect its permeability to oxygen. How often cleaning is necessary depends on the actual use and on the design of the sensor more than the type of sensor. At least one commercially available sensor has come a long way towards having very little need for membrane cleaning. The trick is to use a membrane material with a very high diffusion impedance. Normally a high impedance will affect the response time negatively, but by choosing the right materials it is possible to obtain a high diffusion impedance with an ultra-fast response time! 3) Sensor renovation. There is an immense difference between sensors both with respect to the need for renovation and to the difficulty or ease of doing it. A Clark cell will always need renovation - or even replacement - after a period of time. The galvanic cell - at least when made optimally - only really needs attention if the membrane gets damaged. Just look at the chemical processes that take place. In the Clark cell the electrolyte is depleted, the anode gets blocked by AgCl and the cathode gets silver-plated. In the galvanic cell the anode is slowly consumed - but if you calculate how long it takes to consume the entire anode when measuring at 10 ppb you will arrive at a figure of tens of thousands of years! Renovating some oxygen sensors can be a complicated process involving pages and pages of instructions, expensive replacement parts, special tools, dipping items in concentrated chemicals etc. etc. Special training can be necessary. At least one manufacturer has made a video showing how to do it. We at OxyGuard are proud that our sensor is so easy to handle. We call the process membrane replacement - but it renovates the sensor completely: Unscrew the cap, discard the electrolyte, screw the ring from the cap, remove the old membrane and O-ring, rinse and dry the parts, put a new O-ring and membrane in the cap, screw the ring into place, fill the cap with electrolyte and screw it up onto the probe body. That's it and that's that. Admittedly - we have stretched the instructions to about half a page, and have added some drawings, but it really is so easy that anyone can do it in just a couple of minutes. And the parts you need to replace the membrane five times come with the unit. Experience so far indicates that you may need to do this once every five or ten years. The reference list for the OxyGuard PPB analyzer includes some interesting users. It can be seen on our web site, www.oxyguard.dk. References: Hitchman, Michael A.; Measurement of Dissolved Oxygen; John Wiley and Sons 1978 Gnaiger E. and Forstner H. (editors); Polarographic Oxygen Sensors; Springer-Verlag 1983. ISBN 3-540-11654-0 and 0-387-11654-0 eh-paper 0401 7