A N N A L E N DER P H Y S I K 5.FOLGE, 1930, BANDY, H E F T 4 T h e Jena Wdederholung dee 2tllicheZsm96rsuch8 Vim G e o r g Joos With 11 Figures 1. Elinleitnng Soon after D. Miller') had made his sensational announcements about a positive hherwind effect in the Michelson experiment, as in other places%),so also in Jena a follow-up was started. The goal was to obtain continuous registrations of a light path of about the same size as Miller's, which would provide an unambiguous decision as documents that could be verified by anyone. The Carl Zei6 company made its great resources available in the roughest possible way, and in particular the difficult technical designs of the rotation apparatus were worked out by the "Astrod" design office (senior engineers D. Fr. Mey e r and Dip1.-Ing. For the construction of the apparatus and the photographs themselves, the author was assisted by Ing. Koppen and Mr. Ziege, the mechanic. It is our duty to thank them all. 2 Description of the Apparatus a) The Triiger of the Optics (Fig. 1) For the achievable accuracy, of course, everything depends on a stiffness-free mounting of the optics. Already the 1) D. C. Miller, Proc. Nat. Ac. Wash. 11. p. 306, 1925, further data of Miller in the discussiolr of the Michelson experiment. Astrophys. Journ. 68. 8. 341. 1928. 2) R. I. Kennedy, Proc. Nat. Ac. Wash. 18. p.621. 1926; A. Piccard and E. Stahel, h'aturw. 14. 5. 935. 1926; 16. p. 25. 1928; A. A. Michelson, F. G. Pease, and F. P e a r s o n , Nature 123. 8.88. 1929; Journ. Opt. SOC.Amer. 18. p. 181. 1929; K. K. I l l i n g w o r t h , Phys. Rev. 30. 8. 692. 1927. 23 Annalen der Physlk. 6th inst. 7. 386 G. Joos The question of materials required preliminary experiments, which were carried out by the author at the Physics Institute of the University of Jena. Initially, invar was considered, but an overlap calculation showed that magnetostriction in the earth's field can already produce effects of the order of magnitude of those expected, and also with double the period. The most suitable material would have been, of course, quartz glass, which, however, was out of question with the planned dimensions of the apparatus (arm lengths of 2 m!). Recently, the company S c h o t t und Gen. has been producing fused silica in almost any size according to the process of P f n n e n schmid t . Although the friction shows that the material is not homogeneous, but contains numerous small air inclusions, its coefficient of thermal expansion is not significantly different from that of clear quartz. For our purposes, however, it is important to know whether the material is "ax-working", i.e. whether spontaneous small variations in length occur. For this purpose, 1 m long plates were placed side by side and interferometrically checked for any relative length changes. It was found that at least in the times considered for a rotation of the apparatus (a few minutes to 1 hour) no longitude changes occurred. (It was not investigated, however, whether there were no long movements in the course of years). From this material, the S c h o t t company produced a large number of plates, for which I owe them and especially the head of this department, Mr. P f a n n e n s c h m i d t , my deepest thanks. The last form chosen is that of a rectangle of 193 x 41 cm, pointed on one side under 45O. A cross (Fig. 11) could then be assembled from four such plates. The plates have a thickness of 2 cm and are provided with a bent edge of 8 cm height for stiffening (see Fig. lb). The panels were assembled by neatly beveling the sides at the 45O ends and pressing them together without any intermediate layer by means of two crippling clamping screws. A plastic interlayer was not used, since it would result in permanent length variations under higher pressure. In the center of the cross thus formed, the The Jena rerun of the M ~ ~ l s m w r s u c h s 387 a square opening of 10 cm side width remained free, which served for the passage of the beam path. The plates were provided with apertures for the mirrors and glass plates. a The beam path Fig. 1 b) The rotating apparatus (Fig. 2) Once a satisfactory setup of the optics had been created, the second question arose: an exhaustive rotatable mounting of this cross. This question was solved in a way that was completely different from the previous constructions. Instead of letting the triiger of the optics float on mercury as before, we suspended the quartz cross on a very large number of springs F in such a way that the springs were first attached to the frame R 25 * * * * * * * * * *. 388 G. Joos which support the crossbars on which the cross rests. Each individual load of the cross, such as mirror, dividing plate, etc., is relieved to prevent deformation by additional springs, which are attached to the crossbars at the 01% of the load& The frame is connected to the moving part of the apparatus. This consists mainly of a spherical central rod with four Section through the entire apparatus Fig. 2 and two vertical concentric cylinders extending downward. The horizontal extension tubes enclose the cross in an airtight manner. To dampen the oscillations of the suspended cross, soft-hair bursts are attached to the underside of all four ends, and these are counterbalanced by adjustable bursts attached to the tube walls. The steaming can be intensified by tightening these brushes. In the case of strong steaming, the start-up vibrations subside very quickly, but on the other hand vibrations are transmitted again, so that in the case of The Jena recovery of the 9 Michelson experiment 389 the damping was completely eliminated. Ball and extension tubes are cast from Lautal-Metal1 (an aluminum alloy). For assembly, the sphere and extension tubes had to be made of two halves with a horizontal separation surface, which could be pressed airtight onto each other. (The possibility of evacuation was left open from the outset, in case the disturbances caused by the trapped air became too great. However, no use was made of this, since these disturbances can be reduced to a minimum, while, on the other hand, complete freedom from gas - leaks through which some air continuously enters are, of course, very difficult to achieve). The iiu0er vertical cylinder 2, carries the drive pulley and four stiffeners V carrying the horizontal arms. The moving part of the apparatus runs on a ball bearing K l , which is placed on the upper end of the fixed hollow column S . The foot of this hollow column h n n be adjusted by three set screws Sch resting on concrete bases. Inside the fixed hollow column, with a small space in between, runs the second cylinder Z, which is firmly connected to the moving part, and which holds the camera K t r w . The fixing of the axis of rotation was initially attempted in such a way that between the fixed column and the iluSer cylinder 2, at a distance of about 1m, two ball bearings were fitted for guidance. This construction resulted in a smooth running, but had the disadvantage that in reality there were two not quite coinciding axes of rotation, one perpendicular to the plane of the support bearing K1, the other the line connecting the centers of the two bearings. The unavoidable inaccuracies meant that the direction of the axis, which could be controlled by adjusting an attached spirit level L, could only be fixed to an accuracy of about 2 0. Since this accuracy proved to be too low, the two guide bearings were completely removed and instead a ball was screwed onto the fixed sliule, on which four adjustable jaws, representing parts of a spherical surface, grind. The support bearing was then designed to simultaneously override the lateral guide at the upper end. By adjusting the jaws, the axis of rotation can be adjusted by less than 1". 390 G. Joos from the vertical. This advantage far outweighs the factual disadvantage of a somewhat heavier gear. The drive is effected by an electric motor M rermed with an astatic ZeiS-regulator Model1 I11 according to Mey e r by means of a belt transmission. This drive transmission proved to be superior to other experiments, such as a worm drive, with respect to freedom from vibration. The gear ratios are chosen in such a way that one revolution of the apparatus takes place in 10 minutes. c) The beam path (Figs. 1 and 2) In the first series of experiments, a quartz glass mercury lamp with a monochromatic filter for 5461 was set up outside the experimental chamber as the light source, and the light beam entering the experimental chamber horizontally through a h u n g was thrown in the direction of the axis by means of a mirror. With this arrangement, however, a ratsel-like phenomenon appeared : a simply periodically slight wandering of the interference fringe image together with the zero marks. Since with our arrangement the zero marks are at the place of the interferences (see below), no fallacies could arise from this phenomenon. The explanation is quite surprising: The image of the interference fringes (cf. below and Fig. 1a) is made through the two glass plates placed under 45O in the optical path. If now, due to insufficient centering of the light biindel during a rotation, these flats are penetrated by slightly differently inclined rays, a shift of the image must occur. Since it turned out that a sufficiently exact adjustment of the entering bundle was hardly to be achieved (with each tightening the direction was changed a little bit and apparatus axis and center beam were allowed to deviate from each other only about 10"), the disadvantage of a heat development in the observation room by the light source was taken into account and a Heraeus spot mercury lamp P (Fig. 2) of only 25 watts power consumption was mounted on a rack high above the apparatus itself and let rotate. The current is supplied from the ceiling by two slip rings. Through a totally reflecting prism the light from The Jena recovery of the Michelson experiment 391 is thrown in the direction of the axis by a totally reflecting prism. Through an objective 0 removed from the light source at a distance equal to the focal length, the light is made parallel. A d the objective is also the monochromatic filter. A mirror placed above the center of the quartz cross and inclined below 45O first throws the light to another plane mirror 0 sitting at the end of one arm (cf. Fig. la). From here it reaches the dividing plate. Each of the two partial beams is now thrown back and forth three times in the plane of the cross. The resulting length of the two light paths arriving at the interference is 20.99 meters. After the reunion of the two bushes, the light hits the concave mirror imaging the interference, which first throws the beam onto a slightly higher mirror at the opposite end of the cross and from there onto a plane mirror mounted in the middle and inclined below 45O. This mirror brings the beams into the plane of the cross. This brings the rays down to the camera in the a c h e of the apparatus. In front of the mirror where the interferences are located - that is, the one where the beam passing through the dividing plate is thrown back on itself - three diinne driihts are stretched in a FGihmchen to produce zero mkken. It proved necessary to apply several zero marks, since very often one mark was so out of line with the interference fringes that the measuring accuracy became insufficient. On the other hand, sometimes a strip became unusable for measurement because a mark was placed just at the maximum. The marking of a certain position of the apparatus is done by the fact that a narrow strip of metal sheet fixed to the ceiling of the experimental room covers the beam path shortly behind the spot lamp during the rotation. In this position, the arm carrying the electromagnetically movable mirror forms an angle of 20° to the west with the north direction. On the photographs reproduced in Fig. 9, it can be seen that these marks, which appear vertical there, are not exactly perpendicular to the interference fringes. This is not because the fringes are not quite perpendicular to the slit at the image location G. Joos 392 lay. Rather, the reason was that the sheet metal strip producing the mark first clears one half of the strongly diverging bifurcation, which corresponds to one plate half, and that by the time the other half is cleared, the plate has already been transported one more step. d) Details of the optics (Figs. 3 and 4) The mirrors are square surface-silvered glass plates of 7 x 7 cm, the half-silvered dividing and compensating plates have a size of 7 x 11 cm. b a hiontization of the mirrors Fig. 3 To avoid excessive reflections, these two plates are ground to a slightly wedge-shaped form and positioned so that together they act again as a plane plate. Fig. 3 shows how the mirrors are mounted. Coarse adjustment is made by three simple screws, fine adjustment by h d e r i n g the spring tension in the counter bearings by means of three further screws. The holders are in turn fastened to the quartz plates by a screw, two cutting edges and a ball serving as a support. l) It has been shown that with manual j u s t i e m g the mirrors can be adjusted by one of the plates embedded in the tube wall. 1) In the working drawing shown in Fig. 3 b, 2 balls are indicated, but this has been omitted in the execution. The Jena recovery of the Michelson experiment 393 In the case of the Michelson experiment, such thermal stijrungen arose in the closable hand holes, &I3 it took hours until the interference fringes had calmed down again, whereby the position was by all means not always the desired one. For this reason, and for adjustment when the apparatus was evacuated, a fine electromagnetic movement was attached to one mirror (Fig. 4). Three adjusting screws - one for parallel displacement of the entire mirror, two for inclination - are each connected to two toothed riders, which are each moved forward or backward one tooth by an electromagnet when contact is made. Six contact buttons for these three screws - one each for the electromagnetic forward and backward movement - are mounted on a switchboard attached to the wall of the experimental room next to the telescope used for ocular observation. Fig. 4 The cables are led inside the apparatus to a plug-in contact sitting in the bore wall, into which the cables coming from the contacts are plugged during the adjustment. After the adjustment, these cables must of course be removed again for the rotation. m e) The moving camera. (Fig.5 ) At the lower end of the inner rotating cylinder 2, in Fig. 2, is the camera attachment, which can be moved to the correct distance by means of a worm. In front of a 0.2 mm wide gap, which cuts out a small piece of the interference image perpendicular to the stripes, the cassette slowly passes during rotation. In order to achieve any desired transport speed, the following arrangement was made: the movement of the cassette K s 304 G. Joos is effected by a toothed wheel which engages in a toothed rack connected to the cassette. The gear wheel itself is on the same axis with a grinding wheel S, which grinds on an inclinable spherical cap K, which is rotated by the motor. Depending on the zone on which the grinding wheel rests, a higher or lower transport speed can be set. The drive axle A, which serves to rotate the spherical cap, consists of the following components The camera and the drive Fig. 5 consists of two square tubes that can be telescoped into each other. This is necessary so that the length of the drive axis can be varied continuously when focusing the camera. Transmission from the drive motor is by means of a toothed gear. For a more elastic connection, the lens is attached to both the drive wheel and the calotte with a cardan joint. The used plate format is 3 x 12 em. For ocular observation, a prism can be inserted into the cassette, which after removal of the slit diaphragm The Jena repetition of the Michelson experiment 395 which, after removal of the slit diaphragm, allows observation of the entire interference image through a telescope placed to the side. Of course, this is possible only in a certain position of the apparatus; during the rotation an ocular observation is not possible. 3. xonti ng and jueti ng of the apparatus I n a basement room of the Zeiss factory, the apparatus was assembled. First, the lower parts of the rotary apparatus were assembled and the lower hips of the tubes were bolted to the frames supporting the cross. Then the quartz cross itself was first assembled floating on a pulley above the apparatus and finally lowered into its resilient mounting, whereupon the mirrors were attached. The rough balancing of the light paths was done by means of Prazisioosmat3stilben, which ergttben a length of the interfering light beams of 20.99 m. The mutual inclination of the mirrors was done with the help of the "Prazisioosmat3stilben". The mutual inclination of the mirrors was first adjusted by blanking out a fine beam through a pinhole at the location of the dividing plate and always directing it to the center of the mirror. A finer adjustment was then achieved by placing the observation telescope on the images of a needle held in front of the graduation plate and collapsing it: After putting on the covers, interferences could already be seen in the telescope. The more precise alignment of the light paths was then performed by means of the colors appearing in the non-filtered Hg light. An exact equality on light wavelength is not necessary. The last adjustment was made with the apparatus completely closed by means of the electromagnetic fine movement of one mirror. The main difficulty was to adjust the fringes perpendicular to the slit for a longer time, because only then the movement of the plate can produce an image of the interference fringes pulled apart with any periodic shifts. It turned out that it was best to give the wedge generating the curves of equal thickness a wagrecht edge, because then a h d e r u n g of the temperature stratification of the enclosed air only a wandering, but not a many1 396 G. Joos Dead view of the interferometer Fig. 6 The bearing of the optics Fig. 7 The Jena repetition of the Michelson experiment 397 The result is that the rotation of the interference fringes could be storendered. By means of the set screw attached to the Fu6 and a level placed on the apparatus, the axis of rotation was then set vertically to within 1". Fig. 6 shows the fully assembled apparatus, Fig. 7 the quartz cross with the optics made visible by removing the upper tube halves. 4. course of the exposures Both with the first version, in which the light source was fixed outside the experimental space, and with the final version, in which the light source rotated with it, countless photographs were taken in order to study the influence of the axis position, temperature and other sources of error. An indication of a periodic displacement corresponding to an xtherwind effect with the l l b e n revolution could only be obtained when, with extreme skewing of the axis (inclination of approximately 1'), the crenz was periodically deformed by sitting on the damping surfaces produced by unilateral heating. The photometrically detectable strip shifts (Fig. 8), which were periodic with a full rotation, occurred even with relatively small temperature differences (a few tenths of a degree difference between the inner and outer walls of the test chamber). To check whether the cause was really local temperature differences, the inside was heated by an electric oven, resulting in the curves shown in Fig. 8. After removal of the oven and temperature equalization had occurred, the strips became straight again. (The trembling appearance of the strips is due to the fact that, in order to save time, a relatively strong dampening by the inoculation brushes was set for these exposures, which caused a rapid decay of coarse distortions, but also a transfer of shuttering). Apparently 398 G. Joos -M it348 -M 5 N 17m 7 1950 -M M 8 21°b -M 9 2lU M M 10 M 2258 -M -M 11 2338 Registrations on 10. 5. 30 Fig. 9a (The marks extinguished one of the interfering biindels, they appear dark at the places of coarse brightness and bright at the places of lowest brightness). The Jena repetition of the Michelson experiment 399 -M 15 3sb M -M J[ 21 94* -111 22 10" =M -df 23 11"' -M M 400 G. Joos the mirror, which is equipped with the electromagnetic fine adjustment, responds particularly easily to temperature changes. After such control tests had achieved a certain mastery of the stiirations, the final recordings were made. Since with exact measurement of the registrations an even if very small displacement effect is to be expected - the zero is for the physicist as asymptotic a point as the infinity for the mathematician - it was aimed from the outset to receive series of recordings, which extend over 24 hours, so that from the course of the fluctuations a judgement can be won whether it concerns a genuine atherwind effect or coincidental scatterings. In the first series of this kind, in which measurements were made continuously, there were still very many small shocks, especially at the beginning of each plate. The reason for this was that during the plate change, which was necessary after two revolutions of the apparatus, shocks occurred which had not yet completely subsided at the beginning of the exposure. (The plate change itself is somewhat difficult, so that small shakeouts are unavoidable. The observer must lie under the apparatus on the back and insert the cassette into the camera during the rotation in the dark. The plate transport is switched on electrically). In the case of the recordings of May 10, 1930, which were evaluated, these problems were avoided by making only 2 registrations on one plate in one hour at a rotation time of 10 minutes. After the plate change, a half hour was waited and the plate transport was switched on electrically by the observer located outside the experimental room. Of course, the rotation of the apparatus was not interrupted. On the contrary, it had already been running empty for two days to equalize the temperature at the beginning of the recordings. The external shocks were reduced by the fact that the recording day was set to the time from Saturday noon to Sunday noon, thus to the time of the rest. The temperature conditions were also the best achievable. The difference between indoor and outdoor The Jena repetition of the Michelson experiment 401 The absolute value of the temperature, which is not so important, remained constant for a few tenths of a degree ( 15,5O C.) Only during the night did differences of several tenths of a degree occur, due to cooling of the outer wall and heating of the room by the light source. They are the cause of the gradual widening of the stripes. Fig. 9 shows copies of the original plates. The first picture still shows a certain unevenness of the stripes, because the distortions caused by switching on the light source were not yet completely compensated. On the 5th recording, there is a rough stuttering caused by the slamming of a door in the factory. In general, the apparatus responds relatively strongly to acoustic disturbances: If one claps the hands in the test room, one gets a clear jag in the roger curves. In addition, since this plate 6 has some areas with friction haze, it was excluded from the evaluation. 5 Evaluation of the plates No trace of periodic displacement is visible to the eye on any of the images. However, in order to be able to indicate the limit of a possible effect, a rather complicated evaluation procedure was applied: The distance corresponding to one revolution and made 4 cm long in the final photographs was divided into eight equal parts and each of these parts was photometrically measured with a G. H a n s e n c h e r recording microphotometer perpendicular to the direction of transport. Since 48 revolutions of the apparatus were recorded during 24 hours, this means 384 registrations for one series of recordings. In the previous designs of recording photometers, the shadow of the electrometer filament is imaged and thus the game recording plate is exposed down to the curve describing the shadow image when it is run through once. Thus, in our case, a tremendous consumption of plates cannot be avoided. In the latest construction of the Z e i s ssche Photometer, however, Annalen der Physlk. 6. Folge. 7. 26 402 G. Joos can also be used with dark field illumination of the electrometer filament. By a small adjustment of the optics several registrations can be made on the same plate. Fig. 10 shows the registration of the first half of plate 9. Fig. 10 shows the registration of the first half of plate 9: All 8 registration curves belonging to one revolution find place on one plate. The sinusoidal blackening curves with the three zero marks can be seen. The distance of the maxima and minima from the zero marks is decisive, not, as one could think at the sight of the figure, from a vertical. It was not possible to start photometry exactly at the same place when continuing to tick the plate to be registered, so that the near perpendicular interposition of the corresponding points has nothing more to mean than that a clear separation of the individual curves is achieved. The registration plates thus obtained were first measured by an assistant, for whose grant I would like to thank the Notgemeinschaft Deutscher Wissenschaft very much. The series of May 10, 1930, which contains almost no errors, was re-measured by the author in the most exact way, whereby the differences between the two measurements resulted in a maximum of 6/1000 strip widths (the previous recordings contained, for reasons which are discussed in detail above, too many errors of error to be considered equivalent to these recordings). The plates were measured by placing them on graph paper and observing them in transmitted light. The horizontal straight line in Fig. 10, produced by switching off the illumination of the photometer, served as a straight line. The position of each extreme was determined by reading the intersections of the curves with horizontal straight lines at two or three points about 3 mm above or below the ectremum to an accuracy of tenths of a millimeter and taking the average of these two readings. In the majority of all plates, however, three to four extrema could be taken. The Jena repetition &s Michelson experiment 403 could be used for the measurement. In the same way the exact position of the zero marks was determined. Finally, for each registration, the difference between the mean value of the extremes and the mean value of the marks was measured in thousandths of a strip. In the case of recordings 1-10, there are quite irregular fluctuations around the mean value hernm. From about the 11th recording onwards, a clear continuous qang yon about stripes per revolution can be seen, which is due to the deterioration of the temperature conditions. This Recording curves of the interference fringes of Plstte 9 Fig. 10 gear, however, can be easily corrected: The difference between two successive equal positions of the apparatus was determined and the eighth part of this difference was added each time from registration curve to registration curve. Of course, the percussive part of the displacement is not influenced by this correction of the rate. Also hints of a full-periodic fringe shift could be recognized in these later images. Their elimination is done by taking the average of the 1st and 5th, the 2nd and 6th, the 3rd and 7th, the 4th and 8th partial ticks, whereby the 26 * G. Joos 404 half-periodic shifts again remain uninfluenced. In addition, the mean of the two revolutions contained on one plate was still taken. Only in the case of recording 1 the first revolution had to be omitted because of the irregular strokes. Likewise, shot 5 was omitted completely (cf. above). The final result is shown in Fig. 11. If we take these curves as real &herwind effects, the deviation zero means that the arm carrying the electromagnetically moved mirror is under 45O or 135O to the direction of the horizontal component of the atherwind. A maximum means that this arm $%of0 i in that direction, a she@n minimum that it is perpendicular to it. (The sign is of course 21 undecidable because of the quadratic nature of the effect). N W S N W S N W S N W S One can now kbnn fringe displacements on rotation dee apparatus. (The indicatede Himhthinking, by meler directions refere to the righ- Fourier analysis the tion of thee highlighted arm) fundamental vibration of these Fig. 11 curves and thus to determine the amplitude and direction of the "&herwind". However, this method does not make much sense for the following reason: If a monotonic component of the displacement is present due to insufficient gear correction, this component, which certainly has nothing to do with the atherwind, also contributes to the fundamental frequency, because the Fourier analysis requires a periodic repetition of this gear and not a uniform continuation. This means, that depending on the more 5