NOVEMBER 8, 1889.]

دو

Precisely similar results have been obtained by reversing feeble agnetic forces. So long as the forces are very small, the comnsation for "reverse "is the same as for "make and for break," and the creeping of the magnetism in any given time cer make, break, or reverse is nearly proportional to the amount the preceding change of magnetising force.

537

In the following experiments the magnetising force was raised higher values, at which this proportionality no longer held good s before, the compensating coil was adjusted for each current balance the effect of "make," the iron being demagnetised by eversals immediately before the "make." When a stronger rrent was applied, the coil had to be pushed nearer the magnetoeter: but up to forces of 0.3 C.G.S. or so, it was practicable to ecure an instantaneous balance by doing so. Observations of the rift were taken at 5 and 10 second intervals during one minute.*

FIG. 2.

process was continued up the curve of magnetisation, it became apparent that the immediate effect was increasing; in other words, that there was under-compensation, and that the compensating coil would have to be moved a little forward if an exact balance was to be maintained. The results of this experiment are given below (Table II.), and are exhibited in fig. 4. The magnetising current was increased from one to another of the successive values shown in the table at intervals of one minute in each case, by moderately quick movements of the sliding block in the liquid rheostat. The changes of magnetic force were therefore not quite sudden; each of them took perhaps a quarter of a second to complete.

27

62

161

77

169

344

581

1098

79

171

347

586

1104

79

173

350

590

1109

80

175

354

595

1116

The

These are given on p. 273, the equivalent effect of the compensating coil being added in each case to the actual magnetometer readings. In fig. 2 curves are drawn to show the relation of the current to a) the immediate magnetisation; (b) the magnetism after ive seconds; and (c) the magnetism after one minute. gradient of the curve (a) at and near the origin is the same as that of the corresponding curve in fig. 1, when allowance is made for difference of scales. In the present instance one division of current is 0.0013 of, and one division of magnetism is 0.008 of J. The gradient begins to increase very sensibly when exceeds about 0:07.

Some of the results of Table I. are also shown in fig. 3, which gives time curves of the growth of magnetism for the first two stages (currents 27 and 62). Similar curves for the other stages may readily be constructed from the table. It should be noticed that the time rate of creeping is by no means excessively great in the first instants after contact is made; it is on this fact indeed that the practicability of the method depends.

Similar differences between the immediate and ultimate increments of magnetism present themselves when the magnetising force is increased step by step. In the following experiment the compensating coil was set so as to balance the immediate effect of a feeble magnetising current. Then such a current was applied, and the creeping up of the magnetism during one minute was observed. At the end of the minute the current was increased by a small step, and it was found that the compensation was still correct, or very nearly so; in other words, that the immediate effect of this small increase of magnetising force bore the same, or very nearly the same proportion to the increment of force as at

To make the drift large the top of the wire was this time only 4 cm. from the magnetometer.

Here the last step was too large for perfect compensation. One scale division of current corresponds to 0.0021 C.G.S. units of magnetising force, and one scale division of the magnetometer corresponds to 0.022 of 3. The immediate susceptibility to small increments of force, d /d, is now about 13. The magnetic viscosity is now so great that this immediate effect is less than one-fourth of the whole change which the magnetisation has suffered by the time 1 minute has elapsed.

To show clearly the region in the curve of magnetisation at which the experiment of Table III. and fig. 5 was made, a curve is drawn in fig. 6, showing, as the result of a separate experiment, the relation in absolute measure of the intensity of magnetism to the magnetising force produced by the solenoid. The region dealt with in Table III. is at the place marked (about 40 C.G.S.), and the dotted line drawn there shows the immediate value of d d after a 1-minute pause. The dotted line, P, shows the corresponding initial gradient, or immediate value of d/d,when there is no previous magnetisation.

Another step-by-step experiment of the same kind, made at a place higher up, where the magnetising force of the solenoid was about 4 C.G.S. and about 320, gave again about 13 for the immediate effect (d/d ); and this was followed by a creeping to the extent of six or seven times the immediate effect.

The immediate magnetic effect of a small step is substantially the same whether the step is made quite suddenly by shortcircuiting a resistance coil in the circuit of the magnetising solenoid, or comparatively gradually by means of the liquid slide, so that the process occupies a sensible fraction of a second, or even as much as a whole second.

However small the step is it appears to be followed by a creeping up of magnetism. I have been able to discover nothing which would correspond with the limit of perfect elasticity in straining a solid (if there be any true limit of elasticity), either in the initial part of the process of magnetisation, or after the prolonged application of a constant magnetising force.

But the prolonged application of a constant magnetising force produces an effect which is a most interesting analogue of one effect of prolonged loading in a stretched wire. It is well known that when a load (sufficiently great to produce permanent set) is applied to a stretched iron wire, there ensues, with the lapse of time, not only a certain amount of supplementary viscous extension (the analogue of the magnetic creep) but also a quasi-hardening of the metal which becomes manifest when an addition is made to the load. One effect of this is that the wire responds with great sluggishness to the additional load, and this sluggishness is greater the longer has been the preceding interval during which the load has been maintained constant. To test whether, in like

Cf. Roy. Soc. Proc., No. 1205, 1880, or Encycl. Brit., art. Strength of Materials."

FIG. 5.

fig. 7, where the curve A shows the growth of magnetism durin 10 minutes when the step had been preceded by a 3-minute inter of constant force, and the curve B shows the growth of magnetis when a sensibly equal step was made, which had been preceded

REVIEW

hour interval of constant force. The times are in each case coned from the instant at which the step was made, and the ement of magnetism is in each case reckoned from the value hed just before the step was made. The immediate effect of 1 step (balanced by the coil) was equivalent to 51 scale divisions he magnetometer. The creeping up in 10 minutes was equal no less than 531 scale divisions in the case of curve A, gainst 320 in curve B. At the place marked with an asterisk urve A, it happened that the laboratory door was slammed,

In confirmation of the above, another experiment was made in which the magnetic force was increased by three successive small and very nearly equal steps. The first step was made after 5 minutes of constant force, the second after 1 hour of constant force, and the third again after 5 minutes of constant force. Time-curves of the growth of magnetism were drawn in all three cases. The first and third curves were not far from coincident; but the second curve lay very much below them, as B lies below A.

In the experiments to which figs. 4, 5, and 7 relate, the increment of magnetic force whose effects were measured was preceded by increasing magnetic forces; in other words, it was a step-up from a point on the up curve of magnetisation. I have also examined the effect of a small step-down from a point on the up curve -that is to say, a small decrement of previously increasing force -and find, as might perhaps be anticipated from what we knew about static hysteresis, that the immediate effect (d/d) of a step-down is decidedly less than the immediate effect of a step-up. When the compensating coil had been adjusted to balance the first effect of a step-up, it was found to give over-compensation for a step-down.

Another process has been examined, namely, the alternation of a step-up with step-down, many times repeated. After the magnetising current had been raised to a certain value, it was periodically altered through a definite narrow range by alternately putting in and pulling out the short-circuit plug of a small resistance coil in the main circuit, or by making and breaking a feeble circuit in a second solenoid wound over the first. It was only when this process had been repeated many times that the magnetic effects of the small changes of became approximately cyclic; the early cycles were associated with a progressive rise in the intensity of magnetism. But when a nearly cyclic state was reached, the compensating coil could be adjusted to balance the immediate effects ofor, and the same adjustment, of course, served to balance either.

Tested in this way, the gradient d /d (for the immediate effect of after many small + and steps), has, of course, a lower value than the gradient which is found when is first raised to + d. The latter, as we have seen, is greater when the magnetisation is moderately strong than when there is little or none. The former is nearly constant throughout a wide range of; its value is approximately the same as at the initial part of the magnetisation curve-namely, 10-until the region of saturation is approached, when it becomes distinctly less.*

The periodic changes of magnetism which are brought about by successive small increments and decrements of exhibit a lagging and creeping up and down precisely similar to that which has been illustrated in fig. 1. That figure may serve to show in a general way the relation of the change of a to the change of, when at any place in the curve a very small increment has been applied and removed often enough to establish a cyclic régime. I have not made any full examination of the variation which under these conditions the gradient d /d suffers when the magnetism on which the small cycle if superposed is gradually pushed up towards saturation, nor of the proportion which the subsequent creeping up or down bears to that part of the change of which occurs immediately on the application or removal of 8. The creeping which follows each repeated application and removal of is certainly much reduced when the iron approaches saturation; but the immediate effect is also reduced, and so far as may be judged by rather rough determinations, it appears that the proportion of creeping to immediate effect is much the same with high as with low magnetisation.

One may refer, in this connection, to the energy which is dissipated through hysteresis, in performing a small cycle by alternately applying and removing a very small force. The action is the same in kind whether there is or is not additional magnetisation. The energy dissipated in each cycle is a Id, and vanishes when the increment and decrement of go on pari passu with the increment and decrement of H.

Consider now fig. 1. When the repeated cyclic changes of are indefinitely rapid and go on without pause, so that creeping has not time to occur, a single straight (or sensibly straight) line such as o, A, represents the relation of the change of magnetism to the (very small) change of magnetising force, during both increment and decrement. The rapidity of the action prevents any loop from being formed, and there is consequently no sensible dissipation of energy through hysteresis. This state of things is perhaps nearly realised in the case of a vibrating telephone diaphragm, or, in regard to circumferential magnetisation, by an iron conducting wire in a telephone circuit. Again, let the cycle he performed indefinitely slowly. In that case the magnetism, at every stage of the cycle, creeps up or down to a steady value. A sensibly straight line, such as o, B, represents the relation of to

during both increment and decrement; and there is again no dissipation of energy. But with any frequency of alternation lying between these extremes of infinitely fast and infinitely slow, a loop will be formed, since the creeping will take effect most considerably at and near the ends of the range (the time rate of change of being least there), and there will be dissipation of energy. When the limits and mode of variation of are specified, there must be some particular frequency which will make the energy dissipated per cycle a maximum.

The phenomena described in the paper have been reproduced in several specimens of annealed iron wire, of course with quantitative differences. As to the amount of magnetic creeping much depends on the annealing of the specimen. Another piece of iron wire cut from the same bundle as the piece with which these experiments were made, and annealed at another time, showed almost exactly the same susceptibility to magnetism as the first piece, so far as immediate effect went; but in it the subsequent creeping up was decidedly less (in the proportion of about 4 to 5).

*Cf. Lord Rayleigh, loc. cit., on the approximate constancy of the static gradient di/dl.

540

When the iron is hardened by mechanical strain the phenomena of creeping vanish almost completely. A specimen from the same bundle was annealed, and showed much creeping. It was then put in the testing machine and pulled until it took a set of 1 or 2 mm. in a length of 40 cm. or so. It was then examined magnetically as before, and scarcely a trace of creeping could be observed when a feeble magnetising force was applied. When the compensating coil was properly adjusted the making or breaking of the magnetising current caused no more than a slight momentary quiver of the magnetometer needle, followed by no measurable drifting, although the whole magnetic effect (compensated by the coil) was equivalent to a hundred or more scale divisions. When a magnetising force of as much as 06 C.G.S. units was suddenly applied, the amount of creeping, if there was any, was certainly less than 1 per cent. of the immediate effect. With values of higher than this it became possible to detect creeping with certainty. The following notes relate to this wire :

These forces were in each case applied to this wire in a neutral state. Another trial of the same, with feebler forces, gave 5:3 as the value of d /d for the immediate effect of a very small force, applied when the iron was demagnetised. The same quantity in the annealed specimen was, as has been said, about 10. In fig. 6 the relation of (immediate) to as stated above, is represented by the curve, o, R; the creeping up at the last point is R, 8.

In speaking of soft iron it has been shown that the effects of creeping are most marked when a small addition, 8, is made to a previously increasing force, . In instances quoted above, the creeping up in 1 min. has under those conditions been many times greater than the immediate effect of 8.

By way of putting the specimen of hardened iron to the same test, I have applied a magnetic force of 1:46 and raised it by a small step to 1:49. The immediate effect of this step (which was balanced by the compensating coil) was equivalent to 22 scale divisions of the magnetometer, and this was followed during one minute by a creeping equal to six scale divisions. In itself this creeping is considerable, but compared with the corresponding creeping in soft iron it is extremely small.

Pieces of steel (containing a good deal of carbon) have also been examined, with the result that whether the steel be annealed or in its commercial temper, the phenomenon of creeping is even less visible than in hardened iron. With annealed steel, a force which produced an immediate (compensated) magnetic effect equal to 124 scale divisions caused barely a single scale division of creeping. With a stronger current, giving an immediate magnetism of 340, the subsequent creeping was 3. In steel and in hard iron the creeping seemed to be completed in a few seconds after the institution of the magnetising current. The steel specimen, like the iron, had a diameter of rather more than 4 mm. Its susceptibility (annealed) was considerably less than that of the iron in the hard state.

It is scarcely necessary to observe that the protracted and extensive creeping or magnetic "nachwirkung" in soft iron which these experiments illustrate cannot be ascribed to the subsidence of the circumferential currents which are generated by the imposition of logitudinal magnetic force. The creeping is equally conspicuous whether the magnetic force is suddenly or gradually imposed. Lord Rayleigh has shown that circumferential currents started and left to themselves will subside to e-1 of their initial magnitude in the time.

=

=

where a is the radius of the cylinder, μ its permeability, and p its specific resistance.* In the present instance, taking the case of the annealed iron rod, a = 0.202, μ 125, p 9,827 (Everett), and is less than 1th of a second. The subsidence would be practically complete in a small fraction of a second; but the creeping persists during many seconds, and even minutes, with no excessive change of rate. Again, comparing soft iron with hard iron, in which μ is less and p is greater, the values of r will differ, but not by any means so much as to correspond with the very wide difference in magnetic lag.

In view of this it is puzzling to find that the diameter of the rod experimented upon has a most important influence on the magnetic lag.

In testing various samples of soft iron wire, most of which were of less diameter than the piece used in the above experiments, I noticed that the phenomena of creeping were less marked in the smaller rods. I then tried a bundle of nine very soft annealed iron wires, which were bound together with fine copper wire, and formed a core of about the same length and aggregate diameter as that of the solid rod formerly used. With this bundle there was some creeping, but very little in comparison with what was observed in the solid rod, as the following notes show :

Brit. Assoc. Report, 1882, p. 446.

[NOVEMBER 8, 1889

Finally, another bundle was built up, consisting of a m larger number of fine annealed iron wires. With this the cre was almost insensible.

It may be that the comparative absence of magnetic cry or "nachwirkung," in these last experiments is to be an the quickness with which the process of creeping completes in a finely divided mass of iron; in other words, that the pr is practically completed in a time much shorter than the pen the magnetometer needle. The marked difference in effect her a solid core (a single thick wire) of soft iron and a laminated (a bundle of fine wires) of the same material, suggests that former much more than in the latter the process of creepe retarded by the eddy currents which are set up by those mokė movements in which the process itself consists.

[July 11th.-In seeking an explanation of the differe behaviour it may be worth while to bear in mind that the probably a considerable difference in molecular structure be a solid core and a laminated core of iron. If we accept the that the magnetically neutral state is due to the molecular :. nets forming closed rings, these rings will for the most par closed within the limits of the separate constituent pieces d laminated core, whereas in the solid core they may be much b their dimensions being limited only by those of the core itse I have received very valuable help in these experimente two students, Mr. David Low and Mr. William Frew, who prosecuted a troublesome research with much patience and

A NEW THERMOMETRIC SCALE.

By GEORGE FORBES and WILLIAM HENRY PREECE THE recent Electrical Congress at Paris has adopted the Je. the unit of work, and the Watt as the unit of power. The fo ing definitions were unanimously accepted: The practical of work is the Joule. It is equal to 107 C.G.S. units of work is the energy expended during one second by an ampère 3 ohm."

"The practical unit of power is the Watt. It is equ 107 C.G.S. units of power. The Watt is equal to a Jodie second."

The Therm, as the unit of heat which was proposed by British Association Committee at Bath last year, did not comm itself to the French members. They preferred, for the pres to retain the term Calorie, notwithstanding the confusion there being two units of that name. It is said there is only Calorie in the C.G.S. centrigrade system.

But the question arose is there any need for either the T: or the Calorie? Cannot the Joule be made a thermal unit of for the latter is only a unit of work. The heat generated P seconds by c ampères flowing through R ohms (or driven by 1 v is c2 RT ECT Joules. If we take the mechanical equiva heat as approximately 42,140,000 ergs, or 42 Joules, it that 4-2 Joules will raise one gramme of pure water at 4 degree.

We have here the scientific mode of forming a C.G.S. th metric scale. If we take the fiducial points of freezing and t points of pure water at normal pressure on a column of r as usual, and divide the distance into 420 divisions, ther division is a true C.G.S. unit of temperature, and will be pa by the expenditure of one Joule per gramme of work.

Instead of employing degrees of temperature we can use t units of temperature, and we can replace the angular sy x by t or 0. Thus, to raise one gramme of water from freema point to boiling point requires 420 Joules, or to raise it to b point from 62 units it requires 358 Joules. There would necessity for coefficients, and calculations would be simp Degrees centigrade would be simply converted into these by multiplying by 4-2 and degrees Fahrenheit by deducting and multiplying the remainder by 2:33.

An objection to such a scheme has been raised by Prof. P of Paris, and it is that the mechanical equivalent of heat be yet been determined with sufficient accuracy to justify its adg as final. The pr

He thinks it wrong by nearly 1 per cent. measurements are:

NOVEMBER 8, 1889.]

PROCEEDINGS OF SOCIETIES.

Physical Society, November 1st, 1889.

Prof. REINOLD, F.R.S., President, in the chair. following communications were read ::-"On a New Electriciation Meter," by Mr. W. G. GREGORY. The meter consists of a fine platinum wire attached to a delicate magnifying spring e Ayrton and Perry type, and stretched within a compound of glass and brass. At the junction between the wire and ng a small mirror is fixed. When the tube is placed parallel Hertz's oscillator in action, the mirror is turned in a direction cating an extension of the wire. The arrangement is so sensithat an elongation of 000 of a mm. can be detected, and a placed at a distance of a metre from the oscillator, the appaextension is such as would correspond to a change of tempere of 0·003° C. By its aid the author has roughly verified tz's statements that at considerable distances the intensity of ation varies as the inverse distance; but before he can profurther it is necessary to greatly increase the sensibility of apparatns; and with a view of obtaining some suggestions in direction, he exhibited it before the Society.

rof. PERRY asked if the E.M.F. required to produce the rved results had been calculated; he also believed that the ibility might be increased by using copper instead of plam, and replacing the spring by a twisted strip.

r. BLAKESLEY enquired whether the effect of increasing the city of the ends of the wire had been tried.

r. Boys said that if the observed effect was due to rise of perature, he would like to see it measured thermally. He also ight the effect might be due to extension caused by rapid tric oscillations in some such way as the elongation of an iron caused by magnetisation.

answer to this, Prof. S. P. THOMPSON said the matter had 1 investigated experimentally, but with negative results. rof. HERSCHEL suggested the use of a compound spring, such re used in Breguet's metallic thermometers.

reply, Mr. GREGORY said he had estimated the E.M.F. by rving that a Leclanché cell through 50 ohms produced about same result. No improvement in sensitiveness was obtained ising copper wire or by increasing its capacity, and attempts to sure the rise of temperature by an air thermometer had been n up as hopeless.

he PRESIDENT, in thanking the author for his paper, congratud him on the ingenuity and courage displayed in producing an iratus to measure such microscopic quantities as are here lved.

On a Method of Driving Tuning Forks Electrically." by Mr. GORY. In order to give the impulses about the middle of the ke, the fork is arranged to make and break the primary circuit small transformer, the secondary circuit of which is comed through the electromagnet actuating the fork. The prongs he fork are magnetised and receive two impulses in each

od.

Another device was suggested, where the prongs retively operate contacts which successively charge and harge a condenser through the coils of the actuating magnet. rof. S. P. THOMPSON said the methods, if perfect, would be of it service, and suggested that a fork so driven be tested cally by comparison with a freely vibrating one. He regarded mercury contacts used as objectionable, for their capillarity and esion would probably cause the impulses to lay behind the ointed epochs.

rof. M'LEOD remarked that Lissajou's figures gave a satisfacmethod of testing the constancy of period and could be readily rved without using lenses, and in reference to liquid consers suggested by the author for his second device, said that inum plates in sulphuric acid were found to disintegrate n used for this purpose. He thought lead plates would prove able.

rof. JONES, who read a paper on a similar subject in March , said he now used bowed forks with which to synchronise the ed of the disc there described, and the frequency is determined causing the disc to complete the circuit of his Morse receiver e each revolution.

On a Physical Basis for the Theory of Errors," by Mr. C. V. TON, D.Sc. After pointing out that the law of error for any ticular measurement depends on the nature of the conditions erning such measurement, the author considers several simple es, and deduces their curves of error. A kinematic method of bining two or more independent errors, each following known 8, is then described and applied, and the general formula ained leads to Laplace's law of error in the case of an infinite mber of similar errors. Referring to most advantageous comtions of measures, it is shown that the method of least squares nly a particular solution of the general equation, and is derived assuming the individual errors to conform to Laplace's law. jective errors are next considered, and in conclusion the author 8 that "the law of error in a set of observations depends on nature of each special case, and what may be called the prole law of error is determined by our knowledge of the condiThe combination of three or more sources of error of parable importance gives in general a law not seriously ering from that of Laplace, so that the method of least squares 1 be practically the most advantageous, except where a single rce of error with a very different law is predominant above all

rest."

541

"A Note on the Behaviour of Twisted Strips," by Prof. J. PERRY, F.R.S., had been prematurely announced by mistake, and he accordingly gave only a brief outline of the paper. In a previous communication Prof. Ayrton and the author enunciated a working hypothesis in which the strips were imagined to be split up into pairs of filaments, each pair acting as a bifilar suspension. The resulting formula for the rotation produced by a given load did not agree with experiment, and quite recently the author had recognised why the formula was incorrect. The bifilar law they had assumed was only true for small twists, but he now saw another method of treatment by which he hoped to verify the formula derived from experiment before the next meeting.

Prof. FITZGERALD reminded Prof. Perry of a method of attacking the problem suggested by the speaker some time ago, in which each filament was supposed to be wrapped round a smooth cylinder, and said that on working it out the formula was found to be very complicated.

Mr. TROTTER thought the pairs of strips might be regarded as twisted ladders, and Mr. Gregory said this suggestion reduced the problem to a series of bifilar suspensions which had already been worked out.

"On Electrifications due to Contact of Gases and Liquids," by Mr. J. ENRIGHT. For some time past the author has been studying the electrical phenomena attending solution, by connecting an insulated vessel in which the solution takes place with an electrometer. As a general rule no effect is observed if nothing leaves the vessel, but when gases are produced and allowed to escape the vessel becomes charged with + or electricity, depending on the nature of the liquid from which the gas passes into the air. As an example, when zinc is placed in hydrochloric acid, the deflection of the electrometer is in one direction whilst the liquid is chiefly acid, but decreases and reverses as more and more zinc chloride is produced. From such observations the author hopes to obtain some information relating to atomic charge. Owing to the lateness of the hour the latter portion of the paper and the discussion on it were postponed until next meeting.

16880. Improvements in electric bells." L. W. WINNALL. Dated October 25.

16887. "Means for ascertaining the state of the vacuum in incandescent electric lamps." F. H. W. HIGGINS. Dated October 25. 16890. "Improvements in elements for secondary batteries." E. FRANKLAND, A. H. HOUGH, and D. G. FITZGERALD. Dated October 25.

16917. "Improved winding of armatures of dynamos." J. FARQUHARSON and F. V. ANDERSEN. Dated October 26. 16947. "Improvements in treating and rectifying or ageing alcohol or alcoholic liquors or the like by electricity." MERITENS. Dated October 26. (Complete.)

A. DE