The Early History of Seismometry (to 1900)

by
James Dewey and Perry Byerly

The Development of the Seismograph in Japan

Cecchi's seismograph notwithstanding, it seems clear to us that the credit for the introduction of the seismograph as an essential tool in seismology belongs to a group of British professors teaching in Japan in the late nineteenth century. These scientists obtained the first known records of ground motion as a function of time. Furthermore, they knew what such records could reveal about the nature of earthquake motion. They used their instruments to study the propagation of seismic waves, and they used them to study, for engineering purposes, the behavior of the ground in earthquakes.

The principal figure among the British seismologists in Japan was John Milne. It has been noted (Davison, 1927) that Milne's influence on seismology extends far beyond his own contributions, which were themselves considerable. The Seismological Society of Japan was founded at Milne's urging, in the spring of l880, following a sharp earthquake in Yokahama. The society was the first devoted to seismology; its founding marked the beginning of a period of rapid growth of seismology in Japan. Nearly two decades later, Milne was to be largely responsible for having similar seismographs set up at stations throughout the world, in order to collect data which could be evaluated at a central observatory. Milne's own investigations were rather more observational than theoretical. Although Milne devised several significant instruments, his primary importance to this paper is in the use to which he put both his own and others' instruments.

We will frequently cite the work of James Ewing and Thomas Gray, who designed the most important of the early instruments, in the years immediately following the founding of the society. Unfortunately, their stay in Japan was short; Gray left in 1881 and Ewing in 1883 (Davison, 1927). It was left largely to Milne to show the value to seismology of the Gray and Ewing seismographs. All three, Milne, Ewing and Gray, were visiting professors at the Imperial College of Engineering in Tokyo. The membership lists of the Seismological Society of Japan, in fact, reveal a greater number of foreign members than Japanese members (Seismological Society of Japan, 1880), and the instrumental innovations of the 1880's were due almost entirely to the visiting British professors.

In the decade preceding 1880, several instruments were constructed in Japan which deserve consideration although they did not function successfully as seismographs. Verbeck made the first pendulum observations of earthquakes in Japan in 1873 (Milne, 1880a). Wagener (1880) constructed a common-pendulum seismometer which successfully detected earthquakes, but apparently it did not write records.


Figure 7. Wagener's pendulum seismoscope (after Ehlert, 1897a).

Wagener's machine is shown in Figure 7. The motion of the pendulum bob M was magnified by a mechanical lever, T. The fulcrum B of the lever was attached to the ground, while the short arm of the lever was attached to the base of the pendulum bob. The displacement of the bob relative to the ground was magnified twenty-four times at the end C of the long arm of the lever. To this end of the lever was attached a piece of thread, which was, in turn, wrapped around a spool R. As the lever moved, the thread pulled off the spool, thereby turning the spool. An indicating needle, I, turned with the spool to indicate the relative size of the quake. The pulling of the thread also stopped a clock, thus giving the time of the earthquake. Wagener also built an apparatus which determined the azimuth of the earthquake motion and the amplitude of vertical motion.

In a seventeen-month period, in 1878, 1879 and 1880, in which Wagener's instrument was in operation, it detected twenty-seven earthquakes. The instrument seems to have been set up in the vicinity of Tokyo Bay (Wagener, 1880). Most of the recorded earthquakes were shown to have maximum ground-displacements of less than a millimeter, although two were indicated to have had a displacement greater than 2.5 millimeters. The observer reported that "the extent of the motion as given by the indicator has always been in perfect accordance with the violence of the shock, as far as this can be estimated by the feeling" (Wagener, 1880, p. 71). The vertical motion instrument detected vertical displacement in only eight of the twenty-seven events.

Wagener planned to record the motion of the indicator needle continuously on a translating, rotating drum. He does not appear to have carried out this part of his plan (Ewing, 1883a, p. 38). As it was constructed, then, Wagener's machine still did not constitute a seismograph. It did not write a record, but only indicated the maximum horizontal shaking.

W. S. Chaplin constructed a horizontal-pendulum seismometer in Japan in 1878 (Ewing, 1883a, p. 21). James Ewing described it as a "wooden rod, free to turn about a vertical axis, and carrying at its end a rigidly attached block". The instrument did not work, a failure which Ewing attributed to either low magnification or excessive friction at the pivot points. Chaplin's instrument is important as the first attempt to build a seismometer with a horizontal pendulum as a sensing element.

A most interesting and important paper, entitled "On a neglected Principle that may be employed in Earthquake Measurements" was presented in 1877 to the Asiatic Society of Japan by J. Perry and W. E. Ayrton (1879). The authors analyzed the motion of damped and undamped mass-spring oscillators subjected to a periodic force. The relative displacement of the oscillator mass and the ground is expressible as a sum of steady-state terms and transient terms. To minimize the effect of the transient terms, Perry and Ayrton recommended using an oscillator with a period much shorter than the periods expected in an earthquake. That is, they recommended an accelerometer for the most accurate recording of earthquakes. They recognized that the relative displacement of the mass and the earth, with such an instrument, would be less than the actual displacement of the Earth, but they don't seem to have thought that this would prevent the apparatus from obtaining useable records. Perry and Ayrton also considered oscillators with periods greater than or equal to the periods of earthquake motion. They pointed out that these oscillators, when damped with viscous material, could also be used to obtain reasonably accurate records of earthquake motion.


Figure 8. Ewing's common-pendulum seismograph (after Ewing, 1883a). Shown in cross-section are the ring-shaped bob A spanned by the cross-bar C. Attached to the middle of the cross-bar is a brass plate, L, which has a slot in it. The lever M rests loosely in the slot. Motion perpendicular to the slot is transferred to M and writes on the recording surface S. For motion parallel to the slot, M slides in the slot and is not affected. Another lever, not shown, records the component of horizontal motion parallel to the slot. In this way, the horizontal ground motion is resolved into two perpendicular components.

Perry and Ayrton designed an instrument to have their recommended short period. It was to consist of a heavy spherical weight, free to move horizontally and vertically, and held in equilibrium by five springs. The motion of the ball was to be resolved into three rectilinear components and written on a moving band of paper. This seismograph was never built. James Ewing (1883a, p. 69-70) criticized the design of the instrument on account of the probable difficulty of its construction. Ewing concluded, "... it is not remarkable that the principle on which it is based should, even after attention was directed to it, have remained neglected."

In rejecting Perry and Ayrton's proposed short-period instrument, contemporary seismologists seem also to have rejected the suggestion that the records of longer-period seismometers would be more accurate representations of the Earth's motions if these seismometers were damped with a viscous fluid. The authors had recommended the use of material of high viscosity, such as a mixture of tar and pitch. Perhaps this suggestion hindered acceptance of their proposal. We shall see that artificial solid friction was tried as a means of damping a seismometer. Not until the turn of the century was viscous damping introduced into seismology, with much success, by Wiechert.

It is time now to consider the work of John Milne, James Ewing, and Thomas Gray in developing the first successful seismographs in Japan. [An interesting series of letters by Milne, Gray, and Ewing in Nature (1886-1887) discusses the question of priority for the invention of certain of these seismographs. It would appear that very few of the instruments can be considered the invention of one of the men exclusively.]

Ewing's first seismograph (Figure 8) used a twenty-one foot long common pendulum as a sensing element (Ewing, 1880a). Ewing anticipated that the pendulum's five-second period would be sufficiently long that the bob would remain stationary under the short pulse-like motions of the ground which he believed occurred in earthquakes. He took care to build a rigid frame for the pendulum, so that the motion of the frame would not contribute spurious oscillations on the record.

The relative motion of the pendulum bob and the ground was magnified six times and recorded on a continuously-revolving, circular smoked-glass plate. The magnification was achieved with a simple lever whose fulcrum was attached to the ground. The short arm of the lever was attached to the pendulum bob and the long arm was the indicating lever, scratching a trace on the smoked-glass plate. Expecting that force due to the friction of the indicating lever would be transmitted back to the bob, Ewing attached the lever to the center of mass of the bob, so that this frictional force would not exert a torque on the bob and cause the bob to rotate or wobble. The motion of the pendulum bob could be resolved into two perpendicular components by the recording apparatus.


Figure 9. Ewing's horizontal-pendulum seismometer (after Ewing, 1880b).

Ewing's common-pendulum seismograph was constructed in 1879. This seismograph was not operated continuously from its birth. It therefore did not record its first earthquake until over a year later. At that time, a more sophisticated instrument, Ewing's horizontal-pendulum seismograph, was also in operation. The long common pendulum thus did not have a major influence in the development of seismology by the British in Japan. As we shall see, however, very long common pendulums were later built by Italian seismologists, and these played a significant role in the history of seismology.

Ewing's was the first successful attempt to use a horizontal pendulum to detect earthquakes (Ewing, 1880b, 1881c). Ewing's suspension is illustrated in Figure 9. A light rigid frame F is pivoted at points A and B so as to swing like a garden gate around the axis of rotation AB. At the axis of percussion of the frame is pivoted the cylindrical mass M. [The mass was pivoted on a light frame in order to obtain the maximum inertia for a given total mass and a given distance of the axis of percussion from the axis of rotation (Ewing, 1883a, p. 17). Milne (1898a, p. 138) also pivoted the masses of some of his horizontal pendulums.]


Figure 10. A record, obtained with Ewing's horizontal-pendulum seismograph, of a strong local earthquake on March 8, 1881 (reproduced from Nature, 30, 1884, p. 174). Two of the pendulums write on the same surface. The recording plate revolves continuously with one revolution every fifty seconds. The "beginning" of the earthquake is marked as a, a' respectively on the EW and NS traces. In order that the indicator pens not interfere with each other, the EW trace is put on the record at a point approximately ninety degrees clockwise from where the NS trace is put on the plate. In the center, the traces have been aligned on a common time scale.

The motion of the horizontal pendulum was magnified by means of a long indicator fastened to the frame of the pendulum. A record was inscribed on a revolving smoked-glass plate (Figure 10). There was no way to separate the signal written in one revolution from the signal written in another revolution. It was expected that the indicator would trace and retrace over the same circle on the revolving record during periods of quiet and, during an earthquake, would yield a signal superimposed on the circle. In practice, because the zero-line drifted, later versions of the instrument used a seismoscope, such as the one invented by Palmieri, to start the recording surface into motion at the time of an earthquake. In some instruments, a short pendulum was released in an earthquake to put time marks on the record. Otherwise, the relative arrival times of different phases were determined from the known speed of the recording surface.


Figure 11. "Shake table" test of Ewing's horizontal seismograph (reproduced from Memoires of the Science Dept., Univ. of Tokyo, no. 9, plate XXII). a represents the motion of the "table"; b the motion of the table as it was recorded by the seismograph. A time scale is not given.

With few exceptions, the seismographs built by the British in Japan were intended to have periods longer than the periods of the waves occurring in earthquakes, so that the seismographs would function as displacement recorders. In order to test the accuracy of his horizontal-pendulum seismograph as a displacement recorder, Ewing subjected it to a "shake-table" test (Ewing 1881c, 1883a, p. 86). Two identical instruments were placed side by side on a table, the bob of one being clamped to a nearby wall, so that it would not move, and the bob of the other being left free. The table was then shaken. It was found that the records written by both instruments were nearly identical (Figure 11), showing that the bob of the unclamped seismograph had remained nearly stationary during the shaking. The "frequency" of the shaking motion was not reported.

Both of Ewing's instruments, the common-pendulum and horizontal-pendulum seismographs, recorded a small earthquake on November 3, 1880, giving the first lengthy seismograph records of earthquake motion as a function of time. [Milne (1880b) obtained a short seismogram, three seconds in length, of the Yokohama earthquake of February 22, 1880. He recorded the motion of a long common pendulum on a moving plate.] Assuming that the seismometer functioned as a displacement meter, and assuming harmonic motion, Ewing calculated the maximum displacement (0.29 mm.) and maximum acceleration (1.6 cm/sec-sec) of the ground (Ewing, 1881a, 1883a, p. 54-55).

In his report on this and four other small earthquakes recorded in the same month, Ewing noted the most striking features of these early seismograms. They were: "(1) The very gradual beginning and ending of the disturbance. In none of the observations did the maximum motion occur until after several complete oscillations had taken place. (2) The irregularity of the motion. The successive undulations are widely different both in extent and in periodic time. (3) The large number of undulations in a single earthquake, and the continuous character of the shock. (4) The extreme minuteness of the motion at the Earth's surface" (Ewing, 1883a, p. 55). The importance of these observations can be imagined. For the first time, scientists had a representation of earthquake motion, and this representation revealed a much different manner of shaking than that which had been previously thought probable. Specifically, Robert Mallet's widely accepted view that an earthquake consisted primarily of a longitudinal pulse was shown to be incorrect. Ewing (1881b) realized that his observations were fatal to methods of determining the "velocity of the earthquake" from the times shown on stopped clocks (as had been attempted by von Lasaulx, cited above). For the first time, also, seismologists could design their instruments with some knowledge of the phenomena the instruments were to record.

A seismograph for recording vertical motion was still to be desired. Thomas Gray (1882) introduced a method of increasing the period of a mass-spring system by attaching the spiral spring to the short arm of a lever and attaching the mass to the long arm of the lever. This increased the period by increasing the effective mass of the pendulum bob. It was, however, desirable to diminish further the tendency of the mass to return to its zero position. Accordingly, Gray attached to the bob a container which was connected by a siphon to an external reservoir of mercury. When the bob was deflected downward, mercury would flow from the reservoir into the container on the bob, thus increasing the weight of the bob and neutralizing the increased restoring force of the spring. When the bob was deflected upward, mercury flowed from the container to the external reservoir, lightening the bob and, again, neutralizing the tendency of the bob to return to its equilibrium position.

James Ewing (1882a) suggested that the period of Gray's seismometer could also be increased by attaching the spring to the lever arm at a point below the line connecting the center of mass of the pendulum and the pivot of the lever (Figure 12). This gives the moment arm, to which the spring is applied, a first-order dependence on the boom deflection, with the moment arm always changing so as to decrease the restoring moment of the pendulum. This particular suspension was adopted in Japan, and was to see widespread use in other vertical-motion instruments before the introduction of the zero-length spring. Gray (1887b) and Grablovitz (1891, 1896a) introduced other suspensions which also had the effect of decreasing the restoring moment acting on the lever, when the lever was displaced.

The changing length of the spring in the Gray-Ewing vertical seismometer caused considerable drift. In the 1882 instrument, the seismometer recorded on a revolving glass plate which, because of the drift, was not run continuously. A Palmieri-type seismoscope was used to put the recording surface into motion in an earthquake.

James Ewing, in 1882, joined a common pendulum with an inverted pendulum, so that the two would move horizontally together (Ewing, 1883b, 1886). He obtained thus a system in nearly neutral equilibrium, with the unstable tendency of the inverted pendulum lessening the stability of the common pendulum to which it was attached. This was Ewing's "duplex-pendulum" seismometer (Figure 13).

The duplex-pendulum seismometer is of particular interest to us because in 1887 and 1888 this type of seismometer was placed at ten sites in Northern California and Nevada (Louderback, 1942). The first seismographic observatories in the Western Hemisphere, at Berkeley and Mount Hamilton, were equipped with the duplex instruments as well as Ewing horizontal-pendulum seismometers and Gray-Ewing vertical seismometers. The duplex seismometers in use here wrote two-dimensional particle-motion diagrams on a stationary, smoked glass plate.


Figure 12. Ewing's suspension for detecting vertical motion (after Ewing, 1882a). The spring is attached below the line between the pendulum's center of mass and the pivot point, B.

We must introduce here yet another form of the horizontal pendulum, the so-called "conical pendulum", invented by Gerard in 1851 (Gerard, 1853; Davison, 1896), and introduced to seismology by Thomas Gray (1881). In this suspension, the nearly horizontal boom is held up by a flexible wire and pivots about the axis of rotation on a bearing joint. In order to lessen the friction at the pivot point, James Ewing (1883a, p. 27) forked the boom and extended it beyond the axis of rotation (Figure 14). From the end of the fork, Ewing had a flat steel spring coming back to the axis of rotation and serving as the joint. The spring was held in tension by the boom; the only friction occurred internally within the flat spring. This type of joint has been widely used ever since.

A complete summary and discussion of the instruments invented in Japan prior to 1883 was given by Ewing (1883a). At that time, unfortunately, the period of an instrument was not considered too important, except that it was usually intended to be greater then the period of the waves the seismograph was to record. Ewing writes, "When registering on a continuously moving plate, a well-constructed and well-adjusted seismograph should make one complete oscillation in about 5 seconds, and its decrement of amplitude should not exceed about 1 mm. per oscillation" (Ewing, 1883a, p. 78). Although Ewing's damping specification is vague, it suggests that an undamped system was considered most desireable. In fact, we have seen that Perry and Ayrton suggested the use of viscous damping, from theoretical considerations. Gray (1881) and Milne (1881a) successfully used pendulum seismometers which were intentionally heavily damped with solid friction. Ewing (1883a) pointed out that the use of solid friction would mean that there would be a minimum recordable acceleration, below which earthquakes would not record at all.


Figure 13. The Ewing duplex-pendulum seismometer (after Ewing, 1883a). The bob, B1 is connected at A to the bob B2 of the inverted pendulum by means of a "ball and tube joint". The indicator G is attached to the bob of the inverted pendulum.

The use of viscous damping does not seem to have been attempted by the British in Japan. In fields other than seismology, viscous damping had been used to suppress high frequency tremors of pendulums so that the longer period motions could be studied (Rood, 1875; Darwin, 1882). In particular, the attempts of the Darwins (1882) to measure, with a damped pendulum apparatus, the lunar disturbance of gravity, were well known to the British in Japan. But viscous damping was not introduced into seismology until much later.

With the seismometers of Gray and Ewing available, seismologists in Japan were able to make important advances in the understanding of earthquake motion. John Milne, especially, made extensive use of the new instruments.


Figure 14. Ewing's modification of Gray's "conical-pendulum" seismometer (after Ewing, 1883a). The pendulum rotates about P. The member A is forked so that it does not touch P. From the left end of A a flat steel spring comes back to P.

Milne and his colleagues conducted experiments on the propagation of elastic waves from artificial sources (Milne and Gray, 1883; Milne, 1885b). The most revealing records are those of dynamite explosions recorded on moving glass plates by Ewing horizontal-pendulum seismometers and Gray-Ewing vertical seismometers (Figure 15). The moving plates were started into motion shortly before an explosion. Time signals were put on the plates. The records showed the normal motion outracing the transverse, and a slight vertical motion outracing them both. The apparent velocity of the wave of vertical motion was 500 feet per second. The maximum frequency of a disturbance was found to decrease with distance. In evaluating these observations today, it must be understood that the static magnification of the instrument was very low (four or six); early arrivals may not have been recorded. Also, the stations were close to the source (the furthest was 400 feet away), so that inaccurate timing would grossly affect transit velocities. Milne clearly believed that the normal and transverse waves were compressional and distortional body waves, respectively. He thought the vertical displacement was due to "free surface waves" (Milne, 1885b, p. 81). [A number of investigators had written of seismic "surface waves" before Rayleigh (1885) demonstrated mathematically the possibility of surface waves on an elastic medium. The idea of surface waves must have been suggested to earlier scientists by the observation of "waves" on the surface of the ground in an earthquake. J. Le Conte (1882) spoke of "surface waves", but meant the surface projection of body waves. Mallet (1862b, vol. 2, p. 300) envisioned true surface waves, set up by a compressional wave impinging on the surface above the focus of the earthquake, and propagating with velocities which were probably greater than the compressional wave velocity. Mallet's waves were transverse in a vertical plane. It seems likely that Milne thought he was observing the surface waves hypothesized by Mallet.]

Milne placed a number of similar seismometers at different locations to test the effect of topography and geology on earthquake motion (Ewing, 1882b, Milne, 1886a, 1887a). It had long been observed that damage to structures seemed to depend on the "ground" on which they were built (see, for example, Dolomieu (1784) on the Calabrian earthquakes of 1783). Milne found that the amplitude of seismic waves recorded on similar seismometers also depended on the nature of the ground in their immediate vicinities. In "moderately strong" earthquakes, the recorded amplitudes were generally "very much greater" on "soft ground" than on "hard ground" (Milne, 1887a, p. 30-31). For "very small" earthquakes, this was not true. "Soft ground" in Milne's experiment was marshy land where the water-table was a few feet from the surface. "Hard ground" was higher land where the water-table was at least ten feet from the surface. The recording sites used by Milne were separated by several hundred to a thousand feet horizontally, with a maximum difference in elevation of twenty feet. It is worth noting that all the records were started into motion simultaneously by a seismoscope, so that the seismographs should have recorded the same phases. On the other hand, none of the instruments recorded the earliest phases, and there is no way of knowing when in an earthquake they did start recording. In one of the early experiments of this sort, Milne tested the similarity of the seismographs he used by placing them side by side and recording with them simultaneously. He observed that "they practically gave similar diagrams" (Milne, 1884b).


Figure 15. Elastic waves from dynamite explosions (reproduced from Trans. Seis. Soc. Japan, 8). The distance of each station from an explosion of three pounds of dynamite is given. Time marks are put on every half second. The experiments were carried out in 1881.

Milne performed a similar experiment in different types of building construction and on different floors of the same building (Milne, 1888d). He compared the motion in a brick two-story house and a wood two-story house, situated sixty feet apart on "similar ground". Seismographs were placed on the two floors "vertically above each other and as nearly as possible in similar positions" (Milne, 1888d, p. 67). Milne found that the wood house moved more than the brick house, and the upstairs of each moved more than the downstairs.

In 1882, Milne became concerned with the possibility that neighboring points of ground might move relative to each other in an earthquake. The presence of large strains would pose engineering problems, as well as render seismograph recordings less significant. He constructed an instrument to measure the relative motion of two neighboring points of ground, i.e., a strain seismometer (Milne, 1885b, 1888c). The instrument is shown in Figure 16. Two posts A and O were driven into the ground three feet apart. A rigid beam R was extended from one of the posts almost to the other post. The motion of the end of the beam with respect to the other post, hopefully the same as the relative motion of the two posts, was magnified six times by means of a simple lever B and the record written on a smoked glass plate C. In the earliest form of this seismometer, used in 1882, unusually large amplitudes were obtained (Milne, 1885b), which suggested to Milne that the beam was moving independently of the post to which it was attached. The apparatus was apparently then improved so that Milne felt such motion of the beam would be unlikely. Relative motion of several millimeters was obtained for some local earthquakes recorded by the system.


Figure 16. Milne's instrument for recording the relative motion of neighboring points of ground (after Milne, 1888c).


Figure 17. Oddone's strain-measuring seismometer (adapted from Oddone, 1900). A recording device, which was intended to record the rise and fall of the water in the tube C, is not shown.

Before leaving the strain seismometer, we mention another such instrument, built in 1900 by E. Oddone (Oddone, 1900). Oddone was attempting to construct a seismological apparatus which did not depend on a pendulum for a sensing element, because of the inaccuracies introduced by pendulum eigen-oscillations. His instrument, shown in Figure 17, was similar to Milne's strain seismometer, with a long, rigid beam attached to one pier at A and extending almost to a second, neighboring pier. A "piston" was attached to the end of the beam so as to slide in the "large cylinder", B, of a hydrostatic press which was mounted on the second pier. A small displacement of the piston in the large cylinder, equal to the relative motion between the two piers, caused a 3600 times larger displacement of the water in the "small cylinder" of the press, C. Oddone encountered difficulty in recording the displacement of the water in the small cylinder, however. He mentioned only one earthquake recorded with his instrument.

Milne stayed in Japan until 1895. In 1892, the Seismological Society of Japan ceased to exist due to a lack of interested members (Milne, 1893a). The "Transactions" of the society was continued as the Seismological Journal of Japan, which was edited by Milne and was published until 1895, the year Milne left Japan. Seismology in Japan was carried on by Japanese seismologists, of whom Fusakichi Omori made particularly important contributions in the decade after Milne left. When Milne returned to England, he began important work in the systematic registration of large distant earthquakes. We will meet him again later.

From the Bulletin of the Seismological Society of America. Vol. 59, No. 1, pp. 183-227. February, 1969.