Page images


λ =


h. m. S.

[ocr errors]


10 12

-24 8

[ocr errors]
[ocr errors]

--23 28
-23 12

[ocr errors]


amount of detail given concerning the careers of the notabili- 531 Perseids were seen, and the observations are recorded ties included, and something might be done with advantage in the order of the appearance of the objects, the time, the to reduce the lengths of some of the biographies, and thus

points of appearance and disappearance, and the general

characteristics of each meteor being given. to keep the volume of a convenient size. The “ Who's Who Yearbook " contains the tables which were formerly THE ORBIT OF SIRIUS.-In No. 3981 of the Astronomische included in “ Who's Who " itself. “ The English woman's

Nachrichten Prof. Doberck gives the results of a discussion Yearbook " will in its revised form continue to lighten the

of the observations of Sirius and its faint companion, and

includes a set of elements, an ephemeris for the period labours of women sharing in the useful work of the world.

1903-2–1917.2, and a table showing the differences between the observed and calculated values of position angle and

distance. Owing to the great difference between the magniOUR ASTRONOMICAL COLUMN.

tudes of the two components, the systematic errors of

observation are unusually large. RE-DISCOVERY OF TEMPEL's Second Comet.-A telegram

The following are the elements determined from the disfrom the Kiel Centralstelle announces that Tempel's second

cussion : comet was re-discovered by M. Gavelle at Nice November 30, and that the observation showed the daily

88 = 225° 49'

P-49-49 years ephemeris published in No. 3971 of the Astronomische

29° 54'

Nachrichten to be nearly correct.

= 43° 20'

a =7":513 The following is an extract from the above named

e =0.5871 ephemeris, which was published by M. J. Coniel :

The orbit is referred to the equinox of 1900. The motion 12h. M.T. Paris.

is retrograde, and the anomalies are considered as positive

before and negative after periastron.
a (app.)

8 (app.)
log A

The consideration of the errors of observation showy Dec. 8 20 7 38 - 24 19 0'29671 O'126 that they are inversely proportional to the aperture of the 20 15 4


object glass employed. 20 22 26

- 23 56

0'30161 O'I 22 HARVARD OBSERVATIONS OF VARIABLE STARS.- Part ii., 14 20 29 47 - 23 42 0*30414

vol. xlvi., of the Harvard College Observatory Annals is 16 20 37 4

0'30672 O'117 devoted to the observations, chiefly of variable stars, made 18 20 44 19


by Prof. E. C. Pickering with the meridian photometer 20 51 30

- 22 55

0*31206 0*113 during the years 1892-8. PARALLAX OF A Low METEOR.—Whilst exposing on the

The first chapter gives the results of the observations of Andromeda nebula with two Voigtlander objectives on

short-period variables, and then discusses the phases of the August 12 Herr P. Götz, of Heidelberg, photographed on

light-variations and the corrections to their ephemerides. each plate the trail of a remarkably low Perseid. From

Chapter ii. deals similarly with the observations of variables measurements of the trail on the two plates it was possible

of the Algol type, chapter iii. collates the observations of to determine the parallax of the meteor at definite points

various miscellaneous objects, and the fourth chapter gives, in its flight where the trail was considerably strengthened.

and discusses, the observations of planets and asteroids. The result showed a mean parallax of 28".12, whilst for

The early observations of variable stars, at Harvard, are six distinct points on the trail the following parallaxes were

collected into tables in the fifth chapter, whilst the last determined :

chapter discusses the observations of long-period variables,

and describes the eight light-curves given on the two plates 28":26, 37"-31, 27"-78, 25"-20, 17":14, 10"'o.

at the end of the volume. The base of the triangle Meteor-Voigtlander 1.-Voigtlander II. measured 68 cm., and it therefore follows that

CORRECTION OF THE LONGER TERM IN THE POLAR MOTION. the distance of the meteor at each of these points was

-In a previous communication to the Astronomische Nach498, 3:78, 505, 5.57, 8.27, 14.03 kilometres respectively,

richten Mr. Kimura, of the Mizusawa International Latithe coordinates of the meteor at each point being re

tude Station, showed that the cycle of the polar motion spectively :

might be approximately represented by two principal terms s=0h. 28.2m.,

oh. 1962m.,
oh. 16.8m., ob. 10°7m.,

of 365 and 438 days.

In No. 3981 of the same journal, however, he discusses $= +43° 13', +42° 1', +41° 28', +40° 58', +39° 47',

the latter term more fully, from observations made during + 38° 59'.

the period 1890-1904, and finds that it is probably a day

or two too long. Taking the two periods 1890-1896 and The path of the meteor was apparently rectilinear, but 1896–1902, he derived the value 437.1 days, whilst from the the observations indicated that it described a sharp curve periods 1892–1898 and 1898-1904 the value 436.6 days was in the third dimension with the convex side towards the obtained. The latter value, Mr. Kimura thinks, is likely observer.

to be the more correct, and consequently the cycle is not The path of the meteor extended from a=oh. 33.6m., exactly six years as was indicated by the former discussion. 6= + 44° 17' to a=23h. 52-2m., 8= +35° 28' (Astronomische The values given in the paper show that for the years Nachrichten, No. 3975).

1890 and 1891 the radius of the circular motion was DATE OF THE Most RECENT SUN-SPOT MINIMUM.–From a especially large, but from 1892 to last year it remained discussion of the observations of solar phenomena made at nearly constant. the Roman College Observatory during the period ARC SPECTRA OF THE ALKALI METALS.—In No. 9, vol. xl., November 25, 1900, to January 4, 1902, Signor E. Tringali of the Proceedings of the American Academy of Arts and deduces the date of the latest sun-spot minimum to have Sciences Mr. F. A. Saunders, of Syracuse University, gives been June 15, 1901, or 1901:45.

the results of a series of researches on the arc spectra of In Table i. of the communication the relative daily fre

lithium, sodium, potassium, rubidium, and cæsium. quencies of spots, &c., are given for the years 1878-9 and

The salts were vaporised on nearly pure carbon poles, and 1888–1903, and it is seen that the frequency of days without

the spectra were taken with a grating camera, special spots during 1901 was greater than obtained during the

arrangements being made to photograph the spectra well previous minimum (1889), but less than in the 1878

up into the red. minimum. The numbers given for 1878 and 1901 are 0.76 Several new lines, which fit into the respective series, and 073 respectively (Memorie della Società degli Spettro

were discovered, and in the lithium spectrum Mr. Saunders scopisti Italiani, No. 8, vol. xxxiii.).

believes that the dual character of the lines is real and not OBSERVATIONS OF Perseids, 1904.-In No. 9, vol. xxxiii., simply due to reversals as has been supposed by Hagenbach of the Memorie della Società degli spettroscopisti Italiani, and other spectroscopists. Prof. S. Zammarchi, director of the meteorological observ- A comparison of the arc spectra with spark spectra of atory at Brescia, gives in tabular form the results of the the same substances showed no relative enhancement of observations of Perseids made at that observatory during the any of the lines in passing from the conditions of the arc nights of August 9-14.

to those of the spark.

oh. 22m.,

oh. 77m.


Preliminary. DES ESCRIPTION of Phenomena.-A new material requires

a new name; that of " invar” has been adopted, on the suggestion of Prof. Thury, to avoid the periphrase “steel containing about 36 per cent. of nickel, which is characterised by possessing an extremely small coefficient of expansion or by the fact that its specific volume is practically invariable when considered as a function of the temperature. The name has been universally adopted, and the title of this article is thus justified.

The discovery of invar, as is the case with most discoveries, was preceded by observations indicating the direction of the researches from which it had its origin. As early as 1889 the late Dr. John Hopkinson noted the singular fact of the existence of a ferro-nickel containing about 25 per cent. of nickel, the density of which was found to have diminished by about 2 per cent. after cooling to the temperature of solid carbon dioxide ; and in 1895 M. J.-R. Benoît, director of the Bureau international des Poids et Mesures, having to determine the length of a metre scale composed of an alloy of iron with 22 per cent. of nickel and 2 per cent. of chromium, was extremely surprised to find that his measurements, made with an extreme range of temperature of about 2 degrees C., gave concordant results only on assuming for the alloy a totally abnormal coefficient of expansion, equal to that of brass, and consequently half as

later in a brief manner the common cause of the anomaly observed by Hopkinson and of the phenomenon which I have studied.

Reversible Alloys.-The alloys of iron and nickel which contain more than 25 per cent. of the latter metal may or may not be magnetic, according to the temperature at which they are studied. The passage from one state to another is gradual, the magnetism declining continuously as the temperature is raised, whilst on lowering the temperature the reappearance of the magnetism follows the same curve. The temperature at which the magnetism totally disappears depends on the composition of the alloy. For alloys containing from 26 per cent. to 27 per cent. of nickel it is little above o° C.; as the proportion of nickel increases it rises very rapidly until a maximum, corresponding with 70 per cent. of nickel, is reached at a temperature fixed by M. Osmond at 550° C., when the curve falls to the transformation point of nickel at 340°. This curve of variation is, so to speak, an indicatrix of the properties of the alloys; above the curve the expansion is abnormally great, but at the moment of crossing it with descending temperature the rate of the contraction diminishes, and a region is soon reached in which the anomalous negative expansion exists. Subsequently at a much lower temperature the normal state is reached. The curve given in Fig. I shows the general character of the variation for alloys of this class ; its phases are more or less elongated, the different regions more or less inclined, but the curve always consists of a region of negative abnormality with two confluent curves, one side being characterised by large expansions at high temperatures, the other by a normal expansion. The abnormal region covers generally several hundred degrees.

[ocr errors]
[merged small][ocr errors][merged small][merged small][ocr errors][merged small][ocr errors][ocr errors][ocr errors][merged small][merged small][merged small][merged small][ocr errors][merged small][merged small][merged small][merged small][merged small]

0 Fig. 1.-General form of the expansion curve for a reversible nickel-steel.

great again as that required by the law of mixtures generally applicable to such cases. This alloy was not magnetic, and thus resembled Hopkinson's alloy before cooling, although the latter after exposure to a low temperature became endowed with magnetism.

It was natural to coordinate these two anomalies and to consider the non-magnetic iron of the second alloy as being very expansible. At the time I considered that the alloy, after being rendered magnetic by cooling, would possess a normal coefficient of expansion; but as the alloy studied by M. Benoît did not become magnetic either in carbon dioxide or in liquid air, I was forced provisionally to renounce this hypothesis. For the liquid air I was indebted to the kindness of Sir James Dewar at a time when liquid air was not obtainable in Paris. I did not, however, abandon this research, and it was in seeking for alloys capable of a transformation similar to that observed by Hopkinson that I was led to examine alloys possessing a negatively abnormal coefficient of expansion. I may add that I was able later perfectly to reproduce Hopkinson's discoveries and to extend them in various directions, but I am unwilling to linger over the details in an article of a practical character, these discoveries having hitherto been fruitless of industrial applications. It will be sufficient to consider

The temperature indicated by the abscissa of the point A corresponds sensibly with the ordinate of the indicatrix in question at the point belonging to the same alloy; in other words, it is at this point that the magnetism finally disappears as the temperature rises.

Curve i shows that it is impossible to assign a general value to the expansion of a particular nickel-steel; the value chosen must always apply to a definite region and to a more or less extensive range of temperature. If we consider, for instance, the temperatures 0° and 50° C., the two curves of Fig. 2 can be traced, representing at these two temperatures the inclination of the tangent to curve i for all the reversible alloys of iron and nickel. It is the minimum of this curve which corresponds with invar, strictly so-called. This minimum will be displaced toward the left for alloys considered at lower temperatures and conversely,

It should be noted that beyond the minimum the curves cross; we are then in the region corresponding to the lefthand side of curve i, where the true expansion diminishes with rising temperature. This result of the measurements is of interest because, independently of its being observed for the first time, it has given rise to an interesting application.

Theoretical Views.-Without entering into the details of a theory for the development of which I may refer to an article in the Revue générale des Sciences (July 15 and 30, 1903), I will indicate at least the source of the phenomena which have been described.

In the two transformations which take place successively

in iron in passing from the a condition to the B and y con- It is necessary to mention, however, a retardation in a ditions of Osmond, the metal undergoes different apparent minor part of the change which follows very slowly the changes, of which the most characteristic are the transitions, principal instantaneous phenomenon. This retardation, in two distinct stages, into the non-magnetic state and a due perhaps to a migration of some of the molecules engaged

in the change, is rendered visible in the case of invar, strictly so-called, by a gradual elongation with time. It is enormously accelerated by heating the alloy, for example, at 100° C. Nevertheless, when a bar of invar has been heated thus it still increases in length very slightly after several years at the ordinary temperature. At the end of five or six years the total elongation is nearly 1/100 mm. per metre, but the subsequent lengthening each year does not exceed a fraction of a micron.

This phenomenon is of theoretical interest. Practically B D

it restricts the use of invar, and although, by systematic heating, a much smaller limit of variation can be reached than that above indicated, such a change prevents the alloy

from being employed in the preparation of standards of the с

first order. It is necessary to point this out before proceeding to consider the apparatus in which invar has introduced decided elements of progress. For a consideration of other qualities which may render it valuable I will refer to information already given in this Journal. I can describe here only a few of the uses of invar, and will choose three of the most typical.'

Applications. Standards of Length.-If the slight defect of stability referred to above prevents the employment of invar in the

preparation of fundamental standards, the requirements of А


which are infinite, a wide field of application still remains

in the construction of standards which can be referred from Fig. 3.—Expansion of iron.

time to time to fundamental units, and during these intervals

are employed at temperatures which are not readily ascersudden diminution of the specific volume of the iron at the tained, as is the case with the majority of measuring instrumoment it reaches the higher condition. The expansion of ments which cannot be maintained in a liquid bath. With iron up to high temperatures is indicated by a curve such a brass scale, for instance, an uncertainty of o.1 degree C. as ABCD, Fig. 3. "The addition of a little carbon modifies this curve considerably, as was observed especially by M. Le Chatelier and MM. Charpy and Grenet. The addition of nickel begins to separate the change more and more into two inverse transformations, which

very different temperatures (Hopkinson's phenomenon); as the proportion of nickel increases, the change again becomes simple, but instead of being sudden, as with pure iron, it is spread out over a wide interval of temperatures, at each of which the reciprocal solution of iron in its two extreme states and of nickel strives to attain a stable equilibrium. For the greater part the attainment of equilibrium is practically instantaneous; it is much more rapid, for example, than that which is observed in an aqueous solution in which large crystals are placed, and resembles rather that which would occur in a saturated

Fig. 5.-Rolling of a 2 km. wire on an aluminium drum. solution containing


infinite number of crystalline nuclei of the same density as in the temperature introduces an error little less than 2u per the solution. In a medium thus constituted equilibrium metre of length. But a rod of invar, thoroughly annealed is reached almost instantaneously. The perfect dissemin- and aged, will not change to the same extent in an interval

of three years. The interpolation of definite values up to five or six years can be made with even less uncertainty. Measurements in which the instability of invar will introduce an unacceptable error are very rare; in the case of standards prepared with the usual metals they would correspond with errors of temperature which ar exceeded in nearly all ordinary measurements.

But the greatest claim that invar can make to utility is in

its application to geodesy; working in the open air under FIG. 4. -Scale at the end of a wire (the divisions are millimeties).

extremely variable atmospheric conditions makes the deteration of iron throughout the nickel or the converse is 1 The variation of the rapidity of the change with temperature seems to evidently a very important factor of the phenomenon. For

follow van 't Hoff's law of geometrical progression. Hopkinson's phenomenon the same transformation is still

2 NATURE, No. 1822, September 29, vol. Ixx. P: 527.

3 A more complete description will be found in my recent work, "Les produced, but with an enormous thermal hysteresis.

Applications des Aciers au Nickel."





As a

mination of temperature very uncertain, and, on the other plied by factors of a variable nature, but all greater than hand, a control on returning, by means of a standard of unity. reference in a geodesical or metrological establishment, is These uncertainties disappear completely with a wire the request of General Bassot, have designed for the use the manufacture of comparatively small quantities of the of the Geographical Service of the French Army a scale of alloy when it is required for particular purposes in which 4 metres which is made of invar, and has been found so the price, between certain limits, is a secondary considerpractical by the surveyors that four other scales of the same ation ; samples may be chosen so as finally to descend below type have been constructed for other countries.

the minimum of the curve in Fig. 2 and cut the axis of This scale has an H-section with a side of 40 mm. ; its the abscissæ. Zero and even negative expansions have thus direction lies in the plane of the neutral fibres, and it has been realised. The specimens having a minimum ex such rigidity that the flexure is quite admissible in an pansion are strictly reserved for geodetic purposes, and accurate standard supported at only two points.

considerable quantities of wire have thus been obtained of consequence, the scale can be placed on a light support which it is unnecessary to know the temperature within which is subjected to no especial conditions of rigidity, since about 10 degrees even for the most precise measurements of it has not, as in most of the older apparatus, to assure the base lines. Commonly, a knowledge of the temperature rigidity of the standard. The support which we have within 5 degrees is sufficient; an error of this magnitude adopted is an aluminium box that completely envelops hardly makes a difference of 1 part in 1,000,000. the scale and protects it from shocks, dust, and accidents These advantages could not escape surveyors. As early of all kinds, as well as from rapid changes of temperature. as 1898, M. Jäderin himself requested me to obtain for him The complete apparatus weighs 56 kg., whilst the old wires made of invar for the purpose of perfecting his form of Brunner, consisting of two scales and a rigid method, at a time when M. Benoit and myself were undersupport, weighs 72 kg., and affords no protection for the taking, at the Bureau international des Poids et Mesures, standards.

experiments to ascertain their suitability for such a pur. For direct employment in the field, especially when the pose. The trials were so encouraging that the following apparatus has to be carried to great distances (the scale will, year it was decided to equip the Swedish-Russian expedi

tion to Spitsbergen with similar wires, by means of which all its base lines were measured. At this time, however, the experiments were not sufficiently advanced to obviate the need of taking many precautions, and the expedition acted very wisely in not considering the wires as standards of length. The true standards were two iron bars, previously verified at the Bureau international, which served to measure the short bases (the Swedish base was 96 metres long) on which were standardised the wires of 24 metres, which subsequently served to measure the true bases of several kilometres in length. This was the first practical trial of invar in the field, and. according to the reports which I have received from several members of the expedition, notably from M. Jäderin, the success exceeded every hope. Two independent

of the

Swedish base showed a difference of Fig. 6. - Reading the position of the end scale of the wire against a movable mark.

19 mm. per 10 kilometres, that is, of 1/500,000 without introducing any

correction for the temperature. in the near future, be used in the Andes), the facilities intro- The same sense of safety in the employment of these duced, compared with those existing in older apparatus, are wires is felt after reading the report by M. Backlund, of considerable, and if they constituted the sole progress in the Russian expedition, and of Commandant Bourgeois, on geodesy they would deserve serious consideration. But the the measurements of the French Survey in the territory use of invar has permitted a more complete transformation of the Republic of Ecuador. The difference in the measurein the measurements of bases. Twenty years ago M. Edw. ments of a base made in 1901 with a bimetallic scale and Jäderin made trial of a method which consisted of the use with a wire of invar was 1/3,300,000 ; the agreement is 30 of long wires stretched under a constant load and serving good that it must be attributed partly to chance, but such the purpose of fixing between two limits of the base the chances are rare when the systematic elimination of errors distance of a series of movable bench-marks, ranged has not been pushed to extremes. between these limits. The advantage of this method, the In any case a more complete study of the wires of invar rapidity of measurement, lightness of material, and facility became necessary, and, on the ground of the studies already in the choice of ground, will be readily appreciated, but it commenced by M. Benoît and myself, the International will also be recognised that the uncertainty of the tempera- Committee of Weights and Measures entrusted to us, at ture of the wires made the method doubtful in cases where the end of 1900, on the request of the International greater accuracy was required than that usual for the Geodetic Association, a detailed investigation of this ordinary requirements of topography or land-surveying. question. M. Jäderin has diminished these uncertainties by employing We therefore erected against a thick basement wall, protwo wires of brass and steel respectively, by means of which tected by the building of the laboratory of the bureau, a each of the ranges was successively measured. The differ- series of bench-marks spacing out a length of 24 metres at ence observed for the two wires was taken as an indication intervals of 4 metres, measured by means of an invar of their common temperature, whence the temperature of standard. On the outside of the last uprights are two the steel wire, considered as the principal standard, was pulleys on ball bearings over which pass two cords that deduced. Without going into the details of the calculations carry weights of 10 kilograms and are attached to the wire necessary to the method, it is easy to see that small in- on which observations are to be made at the distance of evitable errors influence the result; the real difference of temperature of the two wires at the time of the measure

1 These wires were manufactured at the steel works of Imphy belonging

to the Société de Commentry-Fourchambault and Decazeville, by whose ments and errors of reading reappear in the result, multi- collaboration 1 was enabled to carry out the work described in this article.



che extreme marks. These wires carry at their extremities scales of invar, having the form represented in Fig. 4, with their edges in the same line as the axis of the wire." This arrangement, somewhat complicated in appearance, is necessary to ensure constancy of length, whatever be the inclination of the scale in a transverse direction.



the needs of a surveyor ; the relative error of the base has fallen below that of the angles; bases can be measured across broken ground, cultivated land, streams and rivers. Above all these advantages, the complete staff, including auxiliaries, need not exceed ten men for a rate of progress of 5 kilometres per day. This arrangement, compared with that by which ten years ago fifty men using rules and microscopes could advance 500 metres a day, exhibits an economy of 98 per cent.! To-day the measurement of a base with all the accuracy required in geodesy costs little more than chaining, and the proof has been so thorough that the French Survey finds its advantageous to measure all its bases by the new method.

The advantages of measurements by wires have been quickly recognised by surveyors. Several departments of survey have requested

the Bureau international standardise wires suitable for base measurements; we have thus had the satisfaction of examining the apparatus for use by the Argentine Republic, Australia, Cape Colony, France, Germany, Japan, Mexico, Roumania, Russia, Servia, and Switzerland.

This simplification in the fundamental measurements of the survey will lead to a reversal in the future of the respective positions of the base and angular measurements. In the old method of surveying measurements of bases were reduced as much as possible and angles multiplied indefinitely; in the new geodesy angles will be controlled by frequent measurement of numerous long bases. This general plan has already been introduced in the United States in the fine work carried out during the determination of the length of the 98° meridian.

Horology and Chronometry.—The possibility of constructing a compensated pendulum with its rod of invar is so obvious that it is hardly necessary to emphasise it. It will be sufficient to observe that the slight change which invar undergoes is not for this purpose a serious defect. As it is necessary to determine the rate of a clock at frequent intervals, variations in the daily rate of the order


0 FIG. 7.-General for.n of the curve of change of Young's modulus for a

reversible nickel-steel.

[merged small][merged small][ocr errors][merged small]


[ocr errors][ocr errors][merged small]

During four years measurements have been made weekly with a great number of wires which have been submitted to different treatment. Owing to the complexity of the subject more than a hundred thousand comparisons between the wires and the base were necessary to elucidate all the questions relating to the stability of the wires and the precision that they guarantee. After four years, and after the method of treatment of the wires has been gradually modified so as to ensure the greatest possible degree of stability, we can emphatically assert the excellence of the method of measurement by wires constructed of invar, When a wire of the usual diameter of 1.65 mm. is stretched by loads varying from an insignificant weight to that of 20 kilograms, the permanent elongation which it undergoes is not measurable; moreover, it can be rolled as often as desired on a drum (Fig. 5) of sufficient diameter (at least 50 cm.), or kept rolled for months without showing on subsequent measurement a variation greater than that due to errors of observation. Several wires which measured at the bureau were returned after use in the field; in the beginning, variations in the length of the order i in 200,000 were observed in several instances, but recently the constancy of the length has become much more decided. Whilst reserving the results obtained by long trials in severe climates, it may be concluded from the results obtained in the laboratory that a surveying expedition equipped with several wires constructed of invar and subject to mutual control will be able to measure several long' bases without fearing a departure from accuracy in the wires greater than that permissible in such measurements, assuming, of course, that the wires are always handled with due care.

The considerable increase in the accuracy of geodetic measurements, caused by the substitution of wires of invar for those of steel or brass, necessitated a corresponding improvement in the apparatus. We have therefore proposed certain new principles which have been realised in instruments constructed with the aid of M. Carpentier, of which a provisional model has been already mentioned in NATURE." A description of the final types which have been adopted would carry me too far; Fig. 6, which indicates one of the measures, may take its place. It will be sufficient to add that, thanks to the new material which has been discovered, the measurement of a base by means of wires answers all

1 June 2, 1904, vol. Ixx. p. 104.



Fig. 8.-Diagram of the compensation of a chronometer with ia steel-brass


of a few hundredths of a second in a year will be merged in the variation of the longer period, and will give rise to an error hardly to be feared; but other applications will need some explanation.

In order not to prolong the preliminary part of this article, I omitted to mention a singular property of the nickel-steels,

« PreviousContinue »