A Method for Determining the Amount of Oxygen Dissolved in I C.C. of Water

IN the well-known Winkler method for determining the amount of oxygen dissolved in water, an accuracy of about 0.1 c.c. O₂ per litre may be reached in the case of pure¹ water, saturated with air. With very careful manipulation the accuracy may even be increased to 0.01 c.c. O₂ per litre. If, however, the water to be analysed has a very low oxygen pressure, difficulties arise ; if, during sampling, no particular precautions are taken to avoid contact between water and air, the water will take up oxygen and the figures found will be too high². Various means are used to eliminate this error. Usually an excess of water is allowed to flow through the bottle; while, for limnological investigations, various special types of sampling apparatus have been designed (cf. Maucha, 1932). Bjerrum (1904) derived a correction formula, and Powers (1918) used special sampling bottles which allow the addition of reagents without exposing the water to air; Weinland (1918) collected the samples over mercury, Hogben and Zoond (1930) protected the water from contact with air by means of paraffin oil and Ellis (1934) described a technique which allows both sampling and addition of reagents without exposing the sample to air.

Some authors seem not to be aware of the fact that an inadequate sampling method may introduce serious errors. In the work of Helff (1928) and Helff and Stubblefield (1931), for example, it is nowhere indicated that precautions were taken to warrant accurate sampling. Hence the conclusion reached by these authors, that in the case of water of low oxygen pressure oxygen was sometimes given off by the animals under investigation, is probably erroneous³; in my opinion the irregularity of their results is a strong indication that their sampling technique was defective.

Several authors have tried to adapt the original Winkler method (sample of 250-100 c.c.) to smaller quantities of water¹. Winkler himself (1924) modified his method for samples of 25 c.c., Allee (1929) analysed samples of 13 c.c., Lund (1921), Thompson and Miller (1928) and Nicloux (1930) worked out micromethods in which samples of only 5.10 c.c. are required. Concerning these micromethods the following general statement may be made : their accuracy has always been determined by analysis of water saturated or nearly saturated with air². And here we meet with a serious drawback to all these micromethods : when analysing water of a very low oxygen pressure, the error introduced by sampling is much more serious than when the macromethod is used³. Of course, here too this error may be eliminated by allowing a large excess of water to flow through the sampling vessel⁴ ; but since samples much larger than 5 c.c. are then needed, the typical advantage of a micromethod is thus more or less lost. In the present paper a simple micromethod is described which requires samples of only $12$⁠, and which has the further advantage of avoiding all contact between water and air both during sampling and during the addition of reagents ; this means that it gives reliable results even in the case of water with a very low (or very high) oxygen pressure.

The sample is drawn by means of a modified Krogh and Keys syringe pipette and the ordinary Winkler reagents are added without exposing the sample to air. The titration of the liberated iodine is performed by means of Rehberg’s microburette.

The original syringe pipette as described by Krogh and Keys (1931) was designed to deliver small quantities of fluid with a very high degree of accuracy.

It consists of a glass tube with a closely ground glass plunger ; to the tube, which is fitted in a metal frame, an injection needle is cemented. The deliverance volume of the syringe can be adjusted to any desired fraction of the maximum capacity by means of a screw, which is released or locked in position by a little set-screw.

In order to adapt this syringe to micro-analysis of dissolved oxygen, I have modified it in the following way (cf. Fig. 1) :

• (1) In order to avoid the use of cement, which would, sooner or later, be attacked by the reagents, the injection needle was replaced by a strong, heavy-walled glass capillary (1) about 5 cm. long, which has a very small bore (inner diameter 0.15-0.20 mm.) and which was welded to the tube.

• (2) A spiral spring (2) was introduced, which pulls the head of the plunger (3) up against the screw (4).

• (3) The little set-screw was functionally replaced by a metal collar (Fig. 1 (5) and Fig. 2 (2)). This collar can be taken off (Fig. 2), so that the plunger can be pressed in wholly by means of the screw ; the sample may then be drawn by means of the same screw. A small excess of water is drawn in, the collar is replaced in its original position, and the excess water is driven out. This method must be used if the sampling has to be carried out very carefully, e.g. if a sample must be drawn from the siphons of a Lamellibranch (cf. van Dam (1935)). If special care is not required, the sample may be drawn more rapidly in the following way : the collar is left in position and the plunger is pressed down by hand ; the spiral spring is then allowed to pull up the plunger with a moderate velocity. If this velocity is too great, gas bubbles may appear in the water, owing to the diminished pressure (high resistance in the very narrow capillary’)-When the collar meets the metal tube (Fig. 1 (6)), the syringe has a very’ definite deliverance volume of about⁠, to be measured by weighing the quantity of water delivered by pressing down the plunger of the filled syringe (cf. Krogh and Keys, 1931, p. 2438).

The dead space (Fig. 1 (7)) and also, if the syringe has been kept dry, the space between plunger and tube is filled with ordinary water. The dead space is rinsed three or four times with a small quantity of the water to be analysed, and is then finally filled.

Since the dead space is very’ small (at most 0.01 c.c.), only a small quantity of water (at most 0.2 c.c.) is required for a thorough rinsing. The required quantity of manganese chloride solution is then added by turning back the screw through a certain angle. Example: From the MnCl₂ solution about 0.5 per cent, of the sample volume must be added ; the thread of the screw being about 0.8 mm., the length of the filled part of the syringe tube about 40 mm., the screw has to be turned over approx. 90 °. Since considerable latitude is allowed in the amount of reagents added it is not necessary to measure them very accurately. The MnCl₂ solution which remained in the dead space is screwed into the water of the syringe, care being taken that no air enters. The water and MnCl₂ solution are then mixed by gently rotating the syringe backwards and forwards on an axis rectangular to its length. The capillary is again entirely filled by turning the screw, the potassium hydroxide-potassium iodide solution is added (in the same way as described for the MnCl₂ solution), and the rotating movements are repeated in order to disperse the precipitated manganous hydroxide throughout the sample.

The mixing movements must not be performed with too great vigour, and attention must also be paid to contraction of the fluid, in order to prevent any air entering the syringe ; by means of the screw the meniscus can easily be kept in the capillary.

After a few minutes thorough mixing¹ the meniscus is adjusted to the end of the capillary and sulphuric acid is added by turning the screw through an angle, about two and a half times greater than in the case of the first reagents. Again the fluids are mixed until the precipitate has dissolved completely, i.e. until all the iodine is set free.

The iodine is then carefully delivered into a small titration vessel (Fig. 3) by pressing down the plunger by hand. Since the reagents contain only traces of oxygen and are added in relatively very small amounts, the quantity of the delivered iodine solution practically equals the deliverance volume of the syringe (cf. p. 82).

This vessel is then placed in the slot of the arm of the microburette described by Rehberg (1925) and the iodine titrated with sodium thiosulphate 0.02 N. During titration the fluid is stirred by means of a moderate stream of air bubbles (cf. Rehberg, 1925) ; no more air than is necessary should be used in order to prevent loss of iodine by volatilisation. For the same purpose a large part of the required quantity of sodium thiosulphate may first be put into the titration vessel; the amount required may be determined by a preliminary test, or, after some practice, roughly estimated by observing the colour of the iodine solution.

Starch in a suitable, invariable quantity (e.g. 0.025 c.c of a 1 per cent, solution) is used as an indicator, a piece of filter paper serving as a background. For comparison a similar vessel, filled with the fluid decolorised in a former titration, is hung beside the titration vessel. After each determination the syringe is rinsed thoroughly with water.

The accuracy of the method described was compared with that of the ordinary Winkler method and of the micromethod of Nicloux (Tables I andII).

Table I shows an accuracy amply sufficient for most purposes¹.

Table II shows that the macromethod and (especially) the micromethod of Nicloux give too high figures in the case of water with a very low oxygen pressure, if no water is allowed to overflow the sample apparatus.

That the water of the filled syringe is practically shut off from the air by the water in the long and very narrow capillary was proved by the following test: a syringe was filled with oxygen-free water and analysed 24 hours later, after the water in the top of the capillary had been screwed out (some excess of water had been sampled); it was found that no detectable amount of oxygen had dissolved into the water.