It is not a bad plan to get the glass exceedingly fluid round the wire--even if the lump has to be blown out a little afterwards--as it cools. The seal should finally be well annealed in asbestos, but first by gradually moving it into the hot air in front of the flame. It was observed by Professor J. J. Thomson and the author some years ago (Proc. Roy. Soc. 40. 331. 1886) that when very violent discharges are taken through lightly sealed-in electrodes in lead-glass tubes--say from a large battery of Leyden jars--gas appears to be carried into the tube over and above that naturally given off by the platinum, and this without there being any apparent want of perfection in the seal. This observation has since been confirmed by others. Consequently in experiments on violent discharges in vacuo where certainty is required as to the exclusion of air, the seals should be protected by a guard tube or cap containing mercury; this must, of course, be put in hot and clean, on hot and clean glass, and in special cases should be boiled in situ. A well-known German physicist (Warburg, I think) recommends putting the seals under water, but I cannot think that this is a good plan, for if air can get in, why not water? which has its surface tension in its favour. The same reasoning prevents my recommending a layer of sulphuric acid above the mercury-a method used for securing air-tightness in "mercury joints" by Mr. Gimingham, Proc. R. S. 1874. Further protection may be attained for many purposes by coating the platinum wire with a sheath of glass, say half an inch long, fused to the platinum wire to a depth of one-twentieth of an inch all round. In some cases the electrodes must be expected to get very hot, for instance, when it is desired to platinise mirrors by the device of Professor Wright of Yale. In this and similar cases I have met with great success by using "barometer" tubes of about one-twelfth of an inch bore, and with walls, say, one-tenth of an inch thick. [Footnote: "Barometer" tube is merely very thick-walled glass tubing, and makes particularly bad barometers, which are sold as weather glasses.] This tube is drawn down to a long point--say an inch long by one-eighth of an inch external diameter, and the wire is fused in for a length, say, of three-quarters of an inch, but only in the narrow drawn--down part of the tube. At different times I have tried four such seals, and though the electrodes were red hot for hours, I have never had an accident--of course they were well annealed. Fig. 37. For directions as to the making of high vacuum tubes, see the section dealing with that matter.