The difficulty of breaking the high short-circuit powers of large modern systems has an important bearing on the security of electrical distribution. The oil switch, as usually constructed, is in some ways a fortuitous device and, as will be shown in this paper, the very act of breaking a circuit under oil, and thus bringing a high-temperature arc into contact with the oil, results in pressures being set up which are not necessarily dependent upon the actual circuit power being broken. Briefly, it can be shown that even with fixed pre-breaking conditions the formation of the arc and the oil pressure set up within the tank will vary. Moreover, such factors as the effect of phase angle, the influence of the layout of the system and, in another sense, the speed of breaking and the effect of oil head, have important effects upon the process of circuit rupture. The author has carried out an investigation in an attempt to assess the effect of these and other factors and to determine whether material improvement is possible without radical changes in design.The author's investigation has led him to conclude that the fundamental limit of breaking capacity, that is the short-circuit power which can be broken time after time by a given switch without breakdown, is for the largest examples of present designs of oil switches rather below 1 million kVA and that the present breaking-capacity rating of oil circuit-breakers is correct only if a certain factor of reliability is taken into account. For high-power breakers this factor expressed as a percentage appears to be about 96.5 per cent, and indicates that such circuit breakers can be relied upon in 96.5 operations out of 100, or, in other words, that there is a 3.5 per cent chance of breakdown. In order to eliminate this element of breakdown the breaking-capacity rating should be divided by a factor which the author's tests indicate to be 2.5. Moreover, the limitation of the short-circuit current by means of reactors may be responsible for troubles not hitherto identified with such a cause.A circuit breaker embodying entirely new principles has been evolved and tested under short-circuit conditions. Its performance is superior to that of the usual oil switch, and a unit of a given size and weight is capable of breaking many times the power broken by an oil switch of similar dimensions, while its performance appears to be wholly consistent.Throughout the paper, tests are quoted and illustrated by oscillograph records wherever possible, but it is, of course, quite impossible, while keeping the dimensions of the paper within reasonable bounds, to give anything more than a skeleton selection of the tests carried out. A few details of the tests and equipment used may, however, be of interest.The majority of the tests were carried out on a 6 000-volt, 6 000-kVA, 40-cycle alternator. Other tests were carried out at 11 000 volts with a 10 000-kVA, 25-cycle alternator, and at 6 000 volts with a 25 000-kVA alternator.The power used in the tests varied between 500 kVA and 270 000 kVA. In all, some 800 tests were carried out, in most of which such, factors as periodicity, kVA and power factor were kept reasonably constant. Factors which could not be so readily controlled were investigated by carrying out groups of rapidly repeated tests from which a selection was made in such a way that the results bore directly upon the point being investigated. This may appear to be rather a roundabout method, but it was found that groups of tests could be usefully co-ordinated and, for instance, results rejected from one group used in another. In addition to this group of 800 tests, at least an equal number was available which, although not directly forming a part of the work described in this paper, were sometimes used for comparison.The switching apparatus consisted of several model switches; 4 of these were of an experimental character, while the remainder, 5 in number, were of commercial or semi-commercial form. Compared with power-system conditions, the majority of the tests were carried out on a reduced scale, the switchgear being reduced in size so that it could be over-stressed if necessary with the power available. With a 50 000-kVA source of power, the switch models were regarded as being 1/64th of their commercial cubical content, with the object of obtaining a switch with a commercial maximum rupturing capacity of about 50 000 kVA, and comparing its performance with a full-scale switch on 1 500 000 kVA. Such a full-size switch, as the author has already suggested, would be about the largest commercially available, and thus one which would be likely to be used in practice on a short-circuit power of 1 500 000 kVA. With a larger source of testing power a correspondingly larger switch was employed. The largest used (excluding the quenched-arc switches) could break a maximum power of about 200 000 kVA.In no group of tests were fewer than 5 repeated tests carried out under similar conditions—a relatively small number, perhaps, for many conditions—but when there arose good reason to believe that closer investigation was necessary, as many as 25 repeated tests were sometimes made. The work described in this paper has been carried out—at intervals—during a period of about 8 years.