PRODUCTION AND PROPERTIES OF GLASS CONTAINERS 29 as Na20, may be due to mixed alkalis, lime and magnesia. There is no doubt that the results obtained by a complete analysis of the solution are un- exceptionable, but such a determination is laborious and the additional information gained from it, rarely worth the labour. In addition, the bottle glass is assumed to contain not more than 1% boric oxide. If this were not in fact the case, the titration procedure would have to be modified. In other variations of the boiling •nethod, glasses are often evaluated according to loss in weight, electrical conductivity of the extract or determi- nation of the total solids extracted, by evaporation of the solution. In this connection it must be noted that the formation of flakes during the tests bears no relation to the alkalinity or conductivity of the solution, and the flakes should, in fact, be removed and evaluated separately. However, the question still remains whether the resistance offered by glass to the attack of water is a satisfactory criterion of durability. The enormous difference between the attack upon optical glass, by buffer solutions near neutrality, and that produced by water indicates that data obtained for water must not be used too freely in predicting resistance to attack by aqueous solutions. Autoclave methods, in which samples are heated with water under pres- sure, have been frequently used, but they give drastic treatment which often rates the chemical durability of various glasses in an order at variance with actual service. Convenience, however, is an important consideration when deciding which durability test to employ, and for pharmaceutical chemists the use of autoclaves is convenient. The autoclave method has therefore, been adopted by many official bodies, such as the British and Swiss Pharmocopoeias, and the American National Formulary and Society for Testing Materials, especially for the testing of containers for injections. The British Pharmocopoeia 1958 Test for the Limit of Alkalinity of Glass prescribes that six containers be filled with acid methyl red solution (corresponding to pH 3.8) and sealed by fusion, or with close-fitting inert closures. After heating at 121øC (15 p.s.i.g.) for half an hour in an autoclave, the solution in none of the containers shall have changed from pink to the full yellow colour of •nethyl red (pH 6-3 or over). Containers are not washed out before testing, unless they have deteriorated in storage after once having passed the test. In that event, washing is carried out internally with 5% v/v solution of glacial acetic acid, and the test repeated. Where possible, the test should be carried out not more than fourteen days before the containers are to be used. Various criticisms have been levelled at the B.P. ampoule test, mainly on account of the uncertainty of definition of some of the conditions. For example, no indication is given of the time factor involved in heating and •cooling the autodave, neither are instructions given as to the orientation of 'ihe ampoules. If the acid test solution gets into the necks it can become

30 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS yellow in circumstances where it would remain pink were the samples kept upright. The volume of the container is also a factor not taken into con- sideration, so that a small container is subjected to a more severe test. The same glass may fail in a small bottle and pass in a large one. This appears illogical at first sight, but it is a test of service conditions. The Swiss Pharmocopoeia, on the other hand, directs that the interior surface area of the container be calculated and a quantity of standard acid and indicator per unit area of surface added. Thereby, the quality of the glass itself is tested and the effect of container size eliminated. The International Pharmocopoeia 1959 "Rapid Test for Glass Containers. for Injection" attempts to overcome these objections by going into greater detail regarding the specification of the autoclave, of the distilled water and of the method of raising and lowering temperature, etc. Containers filled with distilled water are heated for one hour at top temperature, and supported in a rack in an upright position above the water level. Not less than three containers are treated from each batch, so that the total volume of water is not less than 250 ml. 100 ml from each container (or group of containers) is transferred, after the test, to a conical flask and titrated with sulphuric acid, using methyl red indicator. Two limits of alkalinity are given, depend- ing on whether the containers are of capacity up to or above 100 mi. Tests for the absence of arsenic and lead can also be carded out on the aqueous extract. The use of powders as a means of accelerating the attack on glass has much to recommend it, and powder methods have been adopted as standard tests by the English and German Societies of Glass Technology, and the American Society for Testing Materials and the National Formulary. Earlier editions of the British Pharmocopoeia, from 1932 onwards, included both the "surface test" previously described and a "crushed test", with the latter as final. (This was the result of the Report of a Sub-Committee of the Pharmocopoeia Commission, which had considered and tested the various methods used in foreign countries.) Later editions of the B.P. do not include the "crushed test", however, presumably owing to the introduction of surface treated glass. This has a resistant skin of silica or other material which enables the container to pass the surface test (thereby giving an indication of its behaviour under service conditions) but not necessarily the "crushed test". In the powder test, the preparation of the sample requires close attention to details of procedure, if results obtained in different laboratories are to be comparable. As with other methods, the attacking medium may be water, acid or alkali, and the amount of action by the solvent may be estimated by complete analysis of the solution, by evaporation of the solution to dryness and weighing the residue, by titrating the solution with acid thus estimatinõ