A little care is necessary in comparing the color of the tubes. If parallel rays of light are passed through them they act as lenses, con centrating the light to a line and making comparison difficult. If ground (sand-blasted) glass is used to disperse the light, the granular appearance prevents the most exact estimation of the color. We found the only satisfactory background to be a piece of very thin opal glass placed close behind the tubes. The middle portions of the tubes then appear uniform. In order to increase the accuracy of the observa tions, we removed the lenses from a stereoscope, placed their inner edges together, and fastened them at such a distance (8.5 to 10 cm.) in front of the tubes that when one eye was closed and the other held about 6 inches from the lenses, the centers of the tubes were brought together in a sharp line. Two colorimeters of this nature were made and proved satisfactory. The only serious difficulty we had was in making the illumination symmetrical. This is easily done by allowing direct sunlight to fall on the opal glass of the colorimeter and adjusting a pair of gun-sights (that have been placed for the purpose on the top of the instrument) in line toward the sun, but the eye is made unfit for colorimetric work for some minutes unless a dark glass is used. Owing to the small part of the time during which we had sunlight, or even sufficient daylight, we tried mazda lights and found that we could compare o-cresolsulfophthalein (with some loss in accuracy) by means of a 250-watt stereopticon lamp, but this would not do at all for thymol sulfophthalein. The best artificial light we have tried is a 100-watt "daylite" mazda (blue bulb) with aluminium reflector, but we have great difficulty in making the light symmetrical, probably owing to irregu larity in the thickness of the blue glass. This adjustment is made easier by a further dispersion of the light by means of a sand-blasted glass placed several inches behind the opal glass, but in every case it is necessary to verify a color-match by interchanging the tubes. The lamp was placed as close as possible to the tubes, but it might be supposed that the use of a larger lamp or thinner opal glass, thus enabling the lamp to be placed at a greater distance, would more nearly approximate the conditions with sunlight. We did not ascer tain that the opal glass is colorless, although it appears pure white. The sand-blasted glass had a slight greenish hue noticeable on the edge, but we were unable to obtain colorless sand-blasted glass before the close of the experiments. The first attempt at using the color imeter may lead to the supposition that the tubes may be matched more accurately by simply holding them before a piece of opal glass in the window, but this is due to failure of symmetrical lighting in the colorimeter. With a little experience it is possible to distinguish tubes in the colorimeter, which appear alike without it.
The indicators may be used in aqueous solution, since dilution of the sea-water has little effect on the In our experiments, however, it was necessary to use very concentrated solutions of the indicators, because we used the same sample of sea-water for the hydrogen electrode, the colorimeter, and the analysis of total and dilution has an enormous effect on the content at constant tension.
We made up 0.1 per cent solutions of the indicators in ethyl alcohol redistilled over sodium. The o-cresolsulfophthalein dissolved on heating, but the thymolsulfophthalein had to be assisted by a little NaOH. Enough NaOH to bring the color of the indicator to the middle of its range makes it more sensitive, but unless it is pro tected from it will absorb quantities of this gas and become correspondingly insensitive. If NaOH is used to assist solution, it is safer to add an equivalent of HC1, as the small amount of this salt formed will not be appreciable when mixed with sea-water. Since we used only 0.1 c.c. of 0.1 per cent alcoholic solution of the indicator for 10 c.c. of sea-water, we do not believe the alcohol influenced the indicator method appreciably, but we suspect that both alcohol and indicator tended to depolarize the hydrogen electrode and increase the time necessary for equilibrium.
It is very difficult to completely prevent the concentration of the indicator solutions by evaporation while in actual use, and we did not attempt to do it. We kept the solutions in small Erlenmeyer flasks, each having a pipette, graduated in hundredths of a cubic centimeter, passed through the stopper into the solution. Perhaps the most convenient way of correcting the concentration of these solutions is to make up one tube of distilled water containing the same concentration of the indicator as the standard buffer mixtures, at the time the standard tubes are made. There is enough in distilled water to turn these indicators to their maximum yellow color, yet not enough to carry them near their secondary color range (red in strong acid). The yellow color is constant over a very wide P. range. It is then only necessary to determine the quantity of indicator solution neces sary to color distilled water to the same degree of yellow as the sealed distilled-water tube, in order to have the amount necessary to put in the same volume of sea-water in determining the P. Even without this special tube, it is possible to determine colorimetrically (by repeated trials) the amount of indicator to add to seawater; the color of each standard tube is the resultant of two colors (red and yellow or blue and yellow). It is impossible to match it with these colors in any other proportion, no matter how the concentration is changed. If the proper concentration of indicator is unknown, it may be found by repeated trials to match the color, using sea-water of any P. within the range of the tubes. It was this consideration which led us to discard phenol phthalein and a-naphtholphthalein, since each showed only one color.