EXPERIMENTAL RESEARCHES ON THE CONSTITUTION OF HYDRAULIC MORTARS BY HENRI LE CHATELIER TRANSLATED FROM THE ORIGINAL BY NEW YORK McGRAW PUBLISHING COMPANY 114 Liberty Street 1905 AUTHOR'S PREFACE TO THE ENGLISH EDITION. My thesis for the degree of Doctor of Science, upon the Experimental Study of the Constitution of Hydraulic Mortars, has been, for fifteen years, the starting point of numerous studies; some of these, and especially those of Mr. Newberry, have notably confirmed my first results, in regard to the action of supersaturated solutions ; to the com- position of the calcium silicate, which is the active element in cements ; and to its manner of reacting in contact with water. These have been completed in other points, as by the studies of M. Candlot, to whom is due the discovery of two extremely important compounds which relate to the behavior of cements in the sea; the calcium chloro — and sulpho-aluminates. Lastly, my results have been cor- rected in certain points of detail, as, for example, the formula for hydrated calcium aluminate. When Mr. J. L. Mack offered to translate my memoir into English, I hesitated a little before accepting. Do studies so old as these possess still sufficient interest to merit reprinting ? It would have been necessary to complete- ly recast this work in order to incorporate into it the most recent researches. But I have not the time to do this, and I have taken a middle course. I have indicated in notes the most important advances in our theoretical knowledge of cements and have given the references to the papers in which these studies are recorded. [iii] _9 o £, 3 S iv PBEFACE I apologize for allowing so incomplete a work to be re- printed and I thank Mr. Mack for wishing to take the trouble to translate it. It is nevertheless a great satis- faction to me to think that in a country where such great advances have been realized in the cement industry, my researches still seem to have interest. H. Le Chatelier. Villers sur mer, August 1, 1903. * TRANSLATOR'S PREFACE. Although this thesis of Monsieur Le Chatelier appeared in 1887, it was ten years before any new work of impor- tance was published on the subject of the Constitution of Cements. In 1897 the matter again received attention, and the discussion and investigations started at that time have continued up to the present time, and have amplified and confirmed the results obtained by Monsieur Le Chatelier. His classic work therefore stands to-day as the first, the most complete and beautiful piece of work done upon the chemistry of Portland cement, and since the original is not easily obtained and all later work on this subject goes back to and rests upon this thesis, I have thought that it ought to be available to all who are inter- ested in the manufacture and use of Portland cement. In representing the chemical formulas, the rational form used by the author has been retained because of certain advantages which it possesses over the empirical form, al- though in some- cases the latter might be preferable. But it has been necessary to change some of the. formulas slightly to make them conform to the atomic weights which are at present used, for example, "HO" becomes H 2 0 and "CaCl" becomes CaCl 2 . In an appendix has been added some formulas for con- verting metric into English units. The table which ap- pears on page 122 has been calculated into percentages. vi translator's preface The author has very kindly read the corrected proof, and it is a pleasure to record my appreciation of the kind- ness and interest shown by him and others during the progress of the translation. Nor should I fail to mention the courtesy and patience of the publishers during prepara- tion of this book. J. L. M. Fordwick, January, 1905. CONTENTS. Introduction PAGE 1 PART I.— PLASTER. Historical • ? The Calcination of Plaster 3 The Set of Plaster 12 The Chemical Phenomena of Hydration 13 The Physical Phenomena of Crystallization 13 The Mechanical Phenomena of Hardening 21 PART II.— THE SILICATES OF BARIUM. The Synthetic Study of the Barium Silicates 24 Barium Cement 30 PART III.— CEMENTS AND HYDRAULIC LIMES. A. Air Mortars 32 The Calcination of Lime 33 B. Hydraulic Mortars 38 Classification 38 Historical Researches 39 Synthetic Study of the Calcium Salts 47 Lime 47 Calcium Hydrate 49 Calcium Silicates ^0 Calcium Aluminates 62 [vii] Vlll CONTENTS Calcium Ferrites G8 Micro-chemical Study of Cements 73 Anhydrous Cements 73 Hydrated Cements 81 The Manufacture and Use of Cement 86 The Composition and Calcination 87 The Set and Hardening 99 The Causes of Destruction of Hydraulic Mortars HI The Methods of Testing 120 Index 129 INTRODUCTION. The chief object of this work is the study of the chemical reactions which are produced in the limes and hydraulic cements, either during their calcination or during their hardening. Having found in these researches difficulties, the existence of which I had not at first suspected, I de- cided to take up first of all the study of more simple analogous bodies : i.e. plaster, then the silicates of barium, hoping that by understanding the phenomena found in these bodies, some light would be thrown upon the theory of calcareous cements. This work, therefore, includes three parts: Part I. — Plaster. Part II. — The Silicates of Barium. Part III. — Cements and Hydraulic Limes. [1] PART I. PLASTER. Histoeical. — The scientific study of the phenomena which accompany the calcination and hardening of plaster is due to Lavoisier. The details of his experiments are recorded in an extremely remarkable memoir for the time in which it appeared. Chemistry did not yet exist; its most important laws, that of definite proportions among others, had not even been suspected. But the young savant, then only twenty-one years of age, already pos- sessed the experimental method which he was destined to apply to all of his researches and from which he must justly be considered as the founder of modern chemistry. He has summed up his experiments in the following pas- sage, a brief note, inserted in the Comptes rendus de V Academie des Sciences:* "If gypsum, which has been deprived of its water of crystallization by means of heat, is again treated with water (to make what is commonly called mortar) it takes it up with avidity, a rapid and irregular crystallization occurs, and the small crystals which are formed are so en- tangled in each other that a very hard mass results." He had noted, further, that calcination at too high a temperature deprives plaster of the property of setting; finally he pointed out an important peculiarity in the * Comptes rendus, 17 fevrier, 1765. [2] PLASTER 3 calcination of plaster* which had escaped all authors who have studied this question after him. The dehydration of gypsum occurs in two stages: three-fourths of the com- bined water is much more easily driven out than the last fourth. Our present knowledge of plaster is limited, or very nearly so, to about the preceding results; it has scarcely been completed by any secondary points. Berthierf noticed that plaster such as is used in trade contains from 4 to 8 per cent, of water. This fact has since been confirmed by the recent researches of M. Lan- drin. £ Payen,§ about 1830, tried to determine the exact tem- perature of the calcination of plaster. He noticed that gypsum begins to lose its water at about 115° C. and loses it afterwards more and more rapidly in proportion as the temperature rises. My personal researches have been upon the calcination and the set of plaster. Calcination. — The daily experience of the manufac- turers of plaster has taught that the calcination of plaster occurs at a low temperature, well below a red heat. From the laboratory experiments due to Payen, complete de- hydration will occur between 115 °C. and 120° C. How- ever, the figures given by different authors are entirely dis- cordant, and vary from 110° C. to 300° C. It is interest- ing to take up again this study, which evidently remains incomplete. I have employed for this purpose the method by progres- Lavoisier CEuvres Completes, t. III., p. 122. Berthier, Ann. des Mines, 3e serie, t. XIX., p. 655, 1841. Landrin, Ann. de phys. et de Chim., 5e serie, t. III., p. 441. Payen, Chimie industrielle, 1851, p. 304. 4 CONSTITUTION OF HYDRAULIC MORTARS sive heating proposed by Regnault to study the allotropic modifications of molten sulphur. Pulverized gypsum was put into a glass tube in the middle of a paraffin bath, the temperature of which was raised in a progressive and very regular manner. A thermometer gave at each instant the temperature of the salt and the passage of the mercury column past each division of the stem was marked by a registering chronograph. The rise in temperature thus determined ought to remain regular unless some unique phenomenon occurred and should present, as it rose, at the moment of dehydration a period of rest in which it slack- ened, indicating a corresponding absorption of heat, and then should reassume its normal rate of increase. Con- trary to this expectation, instead of only slackening once, the rise in temperature has slackened twice, as will be seen from the table below: LAW OP HEATING OF GYPSUM. mature. Hours. Differences. C° min. sec. mm. sec. 100 0 0 105 20 20 110 45 25 115 1 20 35 120 2 0 40 125 2 50 50 130 18 0 15 10 135 21 30 3 30 140 23 30 2 0 145 24 40 1 10 150 25 40 1 0 155 26 50 1 0 160 27 50 1 10 165 30 40 2 50 170 31 40 1 0 175 32 40 1 0 180 33 30 50 185 34 15 45 190 34 55 40 195 35 30 35 200 36 0 30 8 S Q PLASTER 7 The temperature-curve very plainly shows two halting points, although not equally marked, the first, correspond- ing to a temperature of 128° C. is evidently produced by the dehydration of the gypsum ; the second, corresponding to 163° C, may be attributed either to the end of the de- hydration, on the assumption that these phenomena occur in two phases, or to an allotropic modification succeeding the dehydration. In order to decide this question, I have determined the loss in weight suffered by gypsum when it is heated to a temperature above one of these halting points only, and then above both of them. The experiment was made upon 5 grams of gypsum heated in a paraffin bath maintained at a constant temperature by means of a Schloesing regulator. Duration of Total loss Temperature. the heating. Loss percent. hours, min. on 5 grams. 155° C. 0 0.00 0.0 15 0.33 6.6 30 0.68 13.6 45 0.76 15.2 1 0 0.78 15.6 1 15 0.78 15.6 194° C. 0 0 0.78 15.6 15 0.89 17.8 30 0.99 19.8 1 0 0.99 19.8 1 0 1.04 20.8 270° C. 0 0 1.04 20.8 15 1.04 20.8 This experiment very clearly shows: First, the dehydration is incomplete at 155° and that it is complete at 194° • The quantity of water contained in pure gypsum, CaS0 4 .2H 2 0, is really 20.9 per cent. The two halting points of the temperature-curve correspond, therefore, to two distinct phases in the dehydration. Second, the quantity of water liberated during the first 8 CONSTITUTION OF HYDRAULIC MORTARS phase is perfectly definite and corresponds exactly to 1.5 molecules of water. The product obtained may there- fore be represented by the formula CaS0 4 >4H 2 0 and con- tains 6.2 per cent, of water. In this fact we discover the explanation of certain peculiarities which have been previously pointed out. The halting points in the heating of plaster noticed by Lavoisier result from the higher temperature necessary to drive out the last fourth of the water by hydration. The presence of a nearly constant quantity of water in calcined plaster is due to the fact that its industrial cal- cination is generally limited to the first phase. The pro- portion of 7 per cent, of water found by M. Landrin differs somewhat from the theoretical quantity, which is 0.2 per cent. The motive for this incomplete calcination seems to be entirely a question of economy of fuel, because the plaster when completely calcined sets equally well and has the advantage of solidifying with a larger quantity of water. It may be asked if the product of the incomplete calcina- tion of plaster is a chemical compound or a simple mixture of anhydrous calcium sulphate and of sulphate with two molecules of water. Calcined gypsum is an amorphous material which shows no guaranty of homogeneity. It was interesting to pre- pare this body in the crystallized state. I have obtained this by heating a saturated solution of calcium sulphate in a sealed tube between 135°-150° C. I thus obtained long, regular, exceedingly slender prisms crystallizing in the rhombic system. But the quantity of these crystals ob- tained in one experiment is very small, by reason of the slight solubility of gypsum. In order to obtain a quantity at all considerable, it is necessary to put an excess of cal- PLASTER 9 cium sulphate in the water ; the crystallization is then very- confused. It is indispensable to separate the crystals rapidly from the water in the midst of which they are formed because below 130° C. they again hydrate rapidly to become gypsum. They are obtained by breaking the warm tube, which causes the instantaneous evaporation of the larger part of the water, and throwing the moist ma- terial into absolute alcohol, then draining rapidly and dry- ing them in a drying oven at 100° C. The analysis of these crystals gave me these results : Observed. Calculated. Water 6.7 6.2 CaS0 4 (difference) 93.3 93.8 100.0 100.0 which shows that the body thus obtained is well represented by the formula CaS0 4 .^H 2 0 It is the same compound which constitutes the incrusta- tions of boilers using sea water. The following is the re- sult of an analysis of a similar sample coming from the boilers of the steamers of the Transatlantic company : Analysis. CaCo 3 0.3 FeA 2 -0 H 3 0 5.8 CaS0 4 (difference) 91.9 100.0 This hydrate obtained in the superheated water corre- sponds to calcined plaster, its temperature of rapid de- hydration lies between 160° and 170° C. When finely ground and gauged with water, it hydrates and hardens. 10 CONSTITUTION OF HYDRAULIC MORTARS The set, it is true, is less rapid and less complete than with plaster, which is probably due to the greater volume of the product. It therefore is fully demonstrated that pure, ordinary plaster of good quality, e.g. moulding plaster, is not an- hydrous calcium sulphate, as has been admitted hitherto, but a definite hydrate. CaS0 4 . y 2 H 2 0. We know since the experiments of M. Debray upon the saline hydrates that their decomposition is limited to a temperature given by a definite vapor tension and that the different hydrates of the same salt are characterized by different tensions at the same temperature. I have con- cluded directly from this that the temperatures of decom- position of the two hydrates of calcium sulphate are pre- cisely those for which their vapor tensions were equal to the atmospheric pressure, and that consequently the tem- perature of decomposition would be lowered by diminish- ing the pressure. The result has been entirely contrary to this expectation, as the following table shows : 1st temperature 2d temperature Pressure. stationary. . stationary. 760 mm. 128° C. 163° C. 316 128 280 ... 165° 200 125 164° The temperature of decomposition is therefore entirely independent of the pressure ; the phenomenon observed is not one of dissociation. I have discovered, however, in using the customary methods of experimentation that the dissociation temperature of calcium sulphate with two equivalents of water under the pressure of 760 millimeters is about 110°, that is to say, notably lower than its tem- perature of rapid decomposition. It is a very tedious PLASTER 11 operation to attain uniform pressure, which made precise experiments very difficult and forced me to give up the at- tempt.* The speed of decomposition, which is very slow at the normal temperature of dissociation, becomes more and more rapid as the temperature rises, becoming quite in- stantaneous or at least so rapid that it completely escapes measurement. It is this temperature of rapid decomposi- tion that Ave observe in the experiments of progressive heat- ing. The preceding experiments show that it is independ- ent of the pressure. I have noticed further that it differs little from that at which gypsum begins to dehydrate in the presence of liquid water. It can be foreseen that this temperature of decomposi- tion will vary with the rapidity of the heating. I have been able, indeed, to raise it to 133° C. by heating very rapidly the paraffin bath used in the experiments. For the same reason, the temperature of industrial calcination may be a little lower by reason of the greater time required for the heating of the relatively voluminous pieces of plaster stone. Experiments upon one hundred kilograms of material by M. Lacauchie, director of the plaster works at Argenteuil, have given 125° C, a figure scarcely differ- ing from 128° C. which I have found when working on ten grams of gypsum. It is therefore confirmed that the temperature of the calcination of plaster differs from its temperature of dis- *This question has been taken up again by Mr. van't Hofff and com- pletely solved in spite of its great difficulties. The dissociation-tem- perature of gypsum is still lower than I had supposed and differs little from the vapor tension of water. The salt with half a molecule of water has, contrary to what one might suppose, a greater tension than that of gypsum, and it can only exist in the metastable state. t Proceedings of the Academy of Sciences of Berlin, 1900-1901. 12 CONSTITUTION OF HYDRAULIC MORTARS sociation and that it is always higher than the latter, that it is independent of the pressure, that it is not theoretically susceptible of being precisely determined, but that in fact it varies within rather narrow limits, and that it is com- prised between 120° and 130° C. The Setting of Plaster. — Lavoisier, as I have said above, has shown that plaster during its setting takes up a quantity of water equal to that which had been liberated from it during its calcination, reproducing crystallized gypsum identical to that which constitutes the plaster stone. This analysis of the phenomenon of set is evi- dently incomplete ; it does not show how the hydrate which is produced crystallizes, nor how this crystallization leads to the hardening of the entire mass. In all the special memoirs or treatises of general chem- istry which hitherto have considered plaster, the physical phenomenon of crystallization has generally been passed over in silence and the hardening has been attributed, with- out any explanation, to the massing or felting of the crystals formed. However, we can scarcely understand how rectilinear crystals will be able to become entangled in the manner of twisted fibres which constitute a felt. The bending of these elastic threads gives a reactive force which presses them one against another and causes the development of rubbing stresses at the point of contact. This is the only cause of longitudinal resistance of similar systems. Rigid crystals are unable to cause such phe- nomena. A very simple experiment will show the insuf- ficiency of this explanation; by precipitating a solution of calcium sulphate by alcohol, a deposit of small crystals of gypsum is obtained, showing as complete a massing as can be wished, but nevertheless possessing no cohesion. I therefore think myself authorized to say that this question PLASTER 13 of the physical phenomena of the hardening of plaster has remained entirely untouched up to the moment when I began the study of it. The solution of the problem has a considerable importance because it was quite probable a priori, that the setting of all mortars occurs in an analo- gous manner, and that in each case we can distinguish between : The chemical phenomenon of hydration, The physical phenomenon of crystallization, and The mechanical phenomenon of hardening. Hydration. — Little can be added to what Lavoisier has said about the chemical reaction ; gypsum CaS0 4 .2H 2 0 identically the same as the raw stone re-forms. Only since plaster contains already half a molecule of water, it does not fix more than one and one-half molecules in contact with water. CaS0 4 .^ H 2 0 +iy 2 H 2 0 = CaS0 4 .2H 2 0 The Crystallization of Plaster. — In all the known cases of crystallization of salts in contact with water, the formation of crystals is preceded by the solution of the salt, which gives to the molecules the freedom of motion necessary to allow them to dispose themselves according to a geometrical arrangement. It is therefore very probable, a priori, that the crystallization of plaster is produced by the same means, or if this is not the case, it is absolutely necessary that an intermediate state exists during which the molecules of plaster possess a freedom of motion analagous to that which solution will impart to them. There is no example of solid bodies taking a crystalline 14 CONSTITUTION OF HYDRAULIC MORTARS form without change of state.* The difficulty of admit- ting that the crystallization of plaster may be preceded by its solution arises from the fact that the quantity of uncombined water contained in mortar, which is about 20 per cent., is sufficient to dissolve only 1/2500 of the calcium sulphate with which it is mixed, and the reason is not seen why this salt when in solution should precipi- tate and thus allow the water to dissolve a new quan- tity. It seems to me that this difficulty may be very easily overcome by relying upon the following experiment of M. Marignac. This savant has noticed that if a solution of calcium sulphate is prepared by agitating calcined plaster with a certain quantity of water and filtering the liquid at the end of about five minutes, a solution about five times as concentrated is obtained as if it had been prepared from calcium sulphate hydrated with two molecules of water. But this solution quickly becomes cloudy, deposits crystals of gypsum, and at the end of a longer or shorter time re- turns to its normal concentration. The saturated solution of dehydrated calcium sulphate is therefore supersaturated by comparison with the hydrated sulphate, f *The false crystalline forms obtained by M. Spring by compressing solid bodies are only the sliding surfaces produced by the rubbing of the body upon the walls of the mould. The experiments of M. Fizeau upon the compression of precipitated silver oxide, made much pre- viously, however, to those of M. Spring, have been conclusive in this respect. t The intervention of the phenomenon of supersaturation has been used by M. Landrin (Ann. de phys. et v. Calcium fluoride, used as a flux instead of the chloride, has proved no more satisfactory. The cold mass has always pulverized in cooling, a characteristic which indicates the presence of A great number of attempts, pursued in the same man- ner, remained fruitless. I therefore find myself exactly Density Melting point, near 2.77 800° C. ra=110; ^ = 100; e^lOl. a :b :h=l : 0.726 : 0.287 Si0 2 .2CaO. CEMENTS AND HYDRAULIC LIMES 57 at the point where M. Fremy has stopped ; the impossibil- ity of obtaining in the laboratory a definite silicate of lime setting by the direct action of water.* I then had the idea of starting from the calcium chlor- silicate previously obtained and decomposing it by water vapor at a temperature above 450° C, the normal dis- sociation-temperature of calcium hydrate. I obtained in this way the reaction Si0 2 .2CaO.CaCl 2 +H 2 0=Si0 2 .3CaO+2HCl. The decomposition is very slow and remains superficial. The mass must be pulverized repeatedly in order to obtain a nearly complete elimination of chloride. The pur- est product thus prepared had the composition Observed. Calculated. Si0 3 (by difference) 26.7 26.3 CaO (tested alkalimetrically ) ... 72.1 73.7 CaCl a (tested by silver) 1.2 .... 100.0 100.0 This is a pulverulent mass, showing no evident trace of crystallization; it is not therefore possible to establish by fixed characteristics if we are really dealing with a definite compound, and consequently if it is identical to crystals of cement. But by finely pulverizing this mass, tempering it with water and allowing it to harden in boiling water, there is obtained, at the end of eight days, briquettes which are comparable in hardness to those of cement, and show- ing no trace of swelling or cracking. This absence of swelling is a certain indication of the absence of free lime (because, in fact, if we added only 1 per cent, of strongly calcined free lime to a cement of good quality which does * Mr. Newberry has succeeded in preparing this body directly by the fusion of its elements in the oxyhydrogen blowpipe, and he has obtained a material possessing all the qualities of a good Portland cement. 58 CONSTITUTION OF HYDRAULIC MORTARS not swell in cold water, we observe a considerable cracking and a swelling of volume of about 10 per cent.). This characteristic, together with the property of setting which none of the lower silicates possess, shows clearly that we have to do with a compound and not a simple mixture whose properties would simply be the sum of those of the mixed bodies. I pass by here the theoretical question whether it is a combination in definite proportions or a combination in variable proportions as the isomorphous mixtures are. Silicious Glasses. — The fusion of mixtures of silica and lime containing less than one molecule of lime for one of silica produces glass, that is to say, homogeneous mixtures or mutual solutions of silica and calcium silicate. The crystals of wollastonite become more and more rare and unformed, in proportion as the silica is increased. They have almost completely disappeared for the mixture Si0 2 . }4 CaO. These glasses are not attacked by water and are less and less attacked by acids in proportion as they become more silicious. The alkalies attack them slowly ; with lime they can act like puzzolanas, but rather moderately, however; they thus behave like all silicious materials containing free silica. The presence of a small quantity of the alkalies increases considerably their action as puzzolanas. They ought to be the puzzolanic silicates obtained by M. Fremy. Hydrated calcium silicate. 8i0 2 .Ca0.2}4HoO. — Hydrated calcium silicate cannot be obtained in a pure state by the hydration of the anhydrous silicates, inasmuch as some of them are entirely unchanged in water and the others only very slowly attacked, but the method indicated by Guy ton de Morveau at the beginning of this (nine- teenth) century can be followed, precipitating an alkali CEMENTS AND HYDRAULIC LIMES 59 silicate by a lime salt, or oy causing hydrated silica to react upon calcium hydrate. The precipitate thus obtained is amorphous and does not seem to have a definite composi- tion. According to the relative proportions of lime and silica, we can obtain formulae from Si0 2 . tVCaO up to nearly Si0 2 .2CaO. By working in the presence of an excess of lime, the composition of the precipitate which is less variable is still indefinite, inasmuch as the following differ- ent formulas could be assigned to it. Si0 3 .CaO or 3SiO a .2CaO Berthier Si0 3 .2CaO or 3Si0 2 .4CaO M. Landrin Si0 3 .3CaO or Si0 3 .2CaO Rivot In order to prepare this body in a pure state, I have precipitated a solution of colloidal silica by an excess of lime water; the precipitate formed is so voluminous that after complete deposition, one gram of this body still occu- pies a volume of two liters. The washings are thus made very tedious, but I have noticed further that the washings decompose it and take lime from it. By using a sufficient volume of water, we finally obtain a residue of nearly pure silica, but all the silica does not remain to the end of the operation, a part disappears in the washings, as had oc- curred with the barium silicate. It was interesting to study this decomposition of the silicate by water as much for the purpose of determining its composition as for the purpose of considering the con- sequences which may lead to the destruction of cements by water. A certain weight of the hydrated silicate was sus- pended in almost saturated lime water ; after settling, half of the liquor was removed, analyzed and replaced by pure water, and the same operation was repeated a certain num- 60 CONSTITUTION OP HYDRAULIC MORTARS ber of times. The table herewith gives the experimental results : Total quantities of Quantity of lime existing in solution lime removed. in one liter of the liquor. Grams. Grams. 0.000 1.00 0.50 0.51 0.755 0.27 0.89 0.14 0.955 0.085 0.99 0.065 1.03 0.053 1.07 0.052 It is seen that the strength of the solution decreased at first proportionately to the volume of water added, or very nearly so ; this indicates that the quantity of lime given to the wash water by the precipitate is sensibly zero. But when the lime content in the wash waters has fallen to 0.052 grams per liter, the phenomenon changes. The ad- dition of new quantities of water no longer changes the lime content of the liquor, which remains fixed at 0.052 grams per liter. The silicate gives up its lime to the water up to a constant strength, thus conforming to gen- eral laws of the decomposition of salts by water. In order to make an analysis of a definite calcium sili- cate it is therefore necessary not to carry the washing of it to the limit indicated above. Above this it would seem that we ought to find a precipitate of constant composition, but this is not the case. In the presence of saturated lime water and after a contact of six months, the composition of the precipitate approximates that indicated in the formula : Si0 2 .1.7CaO.Aq. whereas at the first moment of its precipitation in a CEMENTS AND HYDRAULIC LIMES 61 liquid impoverished of lime by the very fact of its precipi- tation, it approximates the formula : Si0 2 .1.3CaO.Aq. Lastly, by washing the precipitate and stopping at a lime content in the liquor a little higher than that for which the normal decomposition begins, the product Si0 2 .l.lCaO.Aq. is obtained. These facts lead me to assume that the normal composi- tion of hydrated calcium silicate is, as for the barium silicate Si0 2 .CaO.Aq* The lime in excess would be fixed by a phenomenon of superficial attraction well known for finely divided chem- ical precipitates, similar to retention of lime by the pre- cipitates of alumina and oxide of iron, of potassium sul- phate by barium sulphate, etc. The quantity of lime thus held necessarily increases with the concentration of the solution. *At the present day the formula Si0 2 2CaO.Aq. is quite generally assigned to the hydrated silicate formed during the set of cement. But this is simply a matter of interpretation. The experimental determinations made upon the composition of these bodies are com- pletely in accord with mine, that is to say, that always a little less than 2 molecular ratios of lime are found for 1 of silica. The number found being more nearly 2 than 1, this former number is chosen, but I persist- in thinking that the arguments drawn from the progressive dissociation of the silicate by water are still valid and prove that an excess of lime is fixed in some other way than in the state of definite combination. 62 CONSTITUTION OF HYDRAULIC MORTARS Analysis of the washed precipitate has given me as results : Analysis. , ' S SiO a 36.4 CaO 35.7 H 2 0 27.0 99.1 which leads, by deducting 1-15 of a molecule of Ca(OH) 2 to the formula Si0 2 .Ca0.2KH 2 0, which I consider as the only hydrated and definite cal- cium silicate able to be formed in the presence of water and of an excess of lime. The formation of these salts disengages for one molecule of silica : Colloidal Si0 2 + dissolved CaO = 6 cal. Calcium Aluminates. — The few definite aluminates which have been studied hitherto show very clearly that alumina is a poly-basic acid, several compounds are known of the form Al 2 0 3 .MO. The reaction of alumina upon sodium carbonate gives, as M. Mallard has shown, a com- pound of the same formula, Al 2 0 3 .Ea 2 0. In the pres- ence of water, alumina and baryta give, according to H. ■Sainte-Claire Deville, the compound, 2Al 2 0 3 .3BaO.Aq. Lastly, M. Fremy has shown that a solution of potassium aluminate deposits crystals having the formula Al 2 (k3K 2 O.Aq. It was therefore very probable a priori that several cal- cium aluminates ought to exist. Indeed, I have discov- ered at least three different ones. Molecular Ratios. 1.21 1.28 3.00 CEMENTS AND HYDRAULIC LIMES 03 Al 2 0 3 .CaO.—In order to study the anhydrous alumi- nates, I fused mixtures of alumina and lime in variable proportions; the melts obtained and cut into three sec- tions have been examined in polarized light by the micro- scope. Mono-calcium aluminate corresponds to a spinel and is difficultly fusible, has very great hardness and crystallizes in the cubic system. When reduced to a fine powder and tempered with water, it sets rapidly ; suspended in a large excess of water, the alumina and the lime dissolve. I shall return more in detail to this action of water when dis- cussing the following aluminate, which behaves in an analogous manner. 2Al 2 O s .3CaO. — Upon fusing a mixture containing iy 2 to 2 molecules of lime for one of alumina, a rather easily- fusible and very hard mass is obtained, which cut into thin sections shows crystals of very strong double refraction, belonging to the orthorhombic system (Fig. 5, facing page 86. There is therefore present a new definite aluminate dif- ferent from the preceding. Unfortunately, however, the fused masses are not completely crystallized, even after re- calcination, a vitrous part always remains which causes a little uncertainty to exist concerning the exact composition of the definite crystallized aluminate. The formula Al 2 0 3 .2CaO can be adopted very well. I have allowed myself to be guided in the choice of the formula by that of the barium aluminate. This calcium aluminate when finely pulverized and tempered with water sets with a rapidity comparable to that of plaster, but the hydrates thus formed have little stability because the mass heated to 100° C. in the pres- ence of water disintegrates and sometimes ends by being reduced to mud. 64 CONSTITUTION OF HYDRAULIC MORTARS When finely pulverized and agitated with a large ex- cess of water, this aluminate allowed alumina and lime to dissolve in variable proportions. I will cite as an example the following analyses of the filtered solutions : But this solution is supersaturated, it soon deposits crystals of the hydrated calcium aluminate and the larger part of the alumina is thus precipitated. Upon adding lime water to the liquor, the crystallization is immediate, and no more alumina remains in solution. Upon introducing calcium aluminate in large crystals into the water, the action remains superficial; a crust of hydrated calcium aluminate about half a millimeter thick is formed which absolutely protects the internal core. After remaining in the water three years the depth of the hydrated layer does not seem to have exceeded that which it had reached at the end of two or three months. If in- stead of immersing the fragments of the aluminate, they are left in moist air, they are seen to crack gradually and then to be reduced to powder. They then undergo a true slaking, and this action continues for some years without reaching its limit. Al 2 0^.3CaO. — Upon increasing the proportion of the lime in the mixture fused, the crystals with strong double refraction are soon seen to disappear. When the composi- tion Al 2 0 3 .3CaO is reached, the thin slabs cut from the fused mass cease to act upon polarized light. Neverthe- less, some very clear outlines are seen showing that the Materials dissolved after 10 minutes' agitation. 1 gram 2Al,0 3 .3CaO in liter of H 2 0 10 grams " " " A1 2 0 3 CaO grams. grams, 0.19 0.39 0.21 0.24 CEMENTS AND HYDRAULIC LIMES 65 whole mass is uniformly crystallized. Therefore a definite aluminate exists corresponding to the formula Al 2 0 3 .3CaO and crystallizing in the cubic system. It is the most fusible of the aluminates. When pulverized and tempered with water, it sets like the preceding aluminates. This aluminate is equally soluble in water. Upon filter- ing after 10 minutes' agitation the liquid obtained by put- ting 1 gram of the aluminate in a liter of water I have obtained the following results: A1 2 0 3 contained CaO contained in 1 liter. in 1 liter, grams. grams. Cold water at 15° 0.08 0.15 Cold water with 3% NaCl 0.12 0.40 Boiling water 0.03 Not determined. For higher proportions of lime a molten mass is still obtained, inactive on polarized light, and slaking rapidly in moist air. When pulverized and tempered with water mixtures below Al 2 0 3 .4CaO still set well; the more cal- careous mixtures swell more and more, and finally are slaked like lime without taking any set. All of these char- acteristics indicate that we are dealing with a mixture of lime and aluminate of lime having the properties of these two bodies. Hydrated Calcium Aluminates. — Several hydrated cal- cium aluminates exist. One of these which can be ob- tained by evaporating solutions of alumina and lime, very poor in lime, obtained by the action of water upon an anhydrous aluminate, has no interest from the standpoint of a study of cements, which after setting always con- tain an excess of free lime. I shall concern myself only with the most basic aluminate, the production of which in a state of purity is likewise the easiest. 66 CONSTITUTION OF HYDRAULIC MORTARS Al 2 0 s 4Ca0J2H 2 0. — The method of preparation which has given me the best results consists in starting with a filtered solution of anhydrous aluminate and in adding to it an equal volume of lime water. A white crystalline pre- cipitate is immediately formed, which settles rapidly. In order to obtain solutions containing as much alumina as possible, it is well to start with a very slightly calcareous aluminate and to conduct the operations of solution and filtration as rapidly as possible. It is necessary to work with a dozen liters of liquid to obtain one gram of the precipitate. When examined by the microscope, the precipitate ap- pears in the form of long needles extending in the direc- tion of their length and united around a central point in such a manner as to form spherical groups similar to those which are observed in the crystallization of all the super- saturated solutions. (Fig. 6, facing page 86). When the crystallization is so slow, compact spherolites are obtained, giving the black cross in parallel light similar to those which occur in the precipitation of calcium carbonate. This calcium aluminate, washed with water, is decom- posed by giving up lime and a small quantity of alumina ; the decomposition is stopped at a temperature of 15° with a content of lime in the solution equal to 0.225 grams per liter. It therefore is a phenomenon similar in all points to the decomposition of the barium and calcium silicates by water. The experiments have been carried out as with the silicate, by removing half of the supernatant solution and replacing it by pure water until the moment when the CEMENTS AND HYDRAULIC LIMES (IT alkatimetric strength of the liquid becomes constant. Here are the numbers obtained at a temperature of 17° : Total weight of the lime Weight of lime contained removed from the liquor. in 1 liter. grams. grams. 0.00 1.00 0.50 0.50 0.75 0.26 0.88 0.230 1.01 0.220 1.12 0.225 1.23 0.225 The analysis of different samples of the calcium alumi- nate, drained and pressed between filter paper has given me: I. II. III. IV. V. Weight of material employed (grams) 0.1 1.0 0.5 0.664 Water 47.8 (diff) 39.2 40.8 38.4 Alumina 17.2 19.2 16.8 17.3 19.6 Lime 35.0 40.5 33.2 39.6 40.0 100.0% 98.9% 97.7% 98.0% These results lead to the following rough formulas : I. ' Al 2 0 3 .3.75 Ca0.15H 2 0. II. AL03.3.7 Ca0.12H 2 0. III. Al 2 0 3 .3.6 Ca0.14H 2 0. IV. Al 2 0 3 .4.2 Ca0.13.5H 2 0. V. Al 2 0 3 .3.7 Ca0.11H 2 0. I have adopted the formula Al 2 0 3 .4Ca0.12H 2 0 as the most probable one, assuming a priori that the number of molecules of alumina and lime are in a simple ratio. This is the only hydrated aluminate stable in the pres- ence of an excess- of lime ; therefore, it is the only one which we can find' in cements after their set,* *Some later researches by M. Candlot have shown that the exact formula of this aluminate was Al 2 0 3 .3CaO.Aq. and not Al 2 0 3 -40aO.Aq., 68 CONSTITUTION OF HYDRAULIC MORTAES Calcaro-Magnesium Aluminates. — The substitution of a certain quantity of magnesia for lime in the mixtures sub- mitted to the fusion gives crystallized products, the fusibil- ity of which increases at first with the proportion of mag- nesia, passes a maximum and afterwards decreases, so that the pure magnesia aluminates are infusible. The alterability of aluminates by water decreases very rapidly by the substitution of magnesia for lime. Double salts seem to be formed, liberating a notable quantity of heat by their combination, and in consequence becoming much less sensible to the action of chemical reagents. This is an analogous fact to that observed with dolomite which is slightly attacked by acids, whereas the two carbonates which compose it taken separately are very easily at- tacked. Calcium ferrites. — Iron sesquioxide combines with lime in the same manner as alumina and ought therefore to give as I had assumed. This body, therefore, is formed by direct hydration of tri-calcium aluminate and not with addition of lime. It is almost certain that calcium ferrite has the same composition. Calcium aluminate gives, with calcium chloride and calcium sul- phate, two extremely important compounds, the existence of which has been noted for the first time by M. Candlot. Calcium chlor-aluminate, the composition of which has been de- termined by Mr. G. Friedel, corresponds to the formula Al 2 O 3 .3CaO.CaCl 2 .10H 2 O. It is decomposed immediately upon contact with water, but seems to be capable of being formed momentarily by the action of saturated solutions of calcium chloride upon the anhydrous aluminates which explains the changeable solubility of these compounds. Calcium sulpho-aluminate, the exact composition of which has been determined by Mr. Deval, corresponds to the formula Al 3 0 3 .3Ca0.3 ( CaO.S0 3 ) .30H 2 O. It appears to be the principal, perhaps the exclusive, cause of the chemical decomposition of cements in sea water. But the method of its destructive action and of the swelling which it causes is yet un- known. (Bulletin de la Societie' d' encouragement, juillet, 1890, and fevrier, 1900.) CEMENTS AND HYDRAULIC LIMES 69 as varied products, but I have not succeeded in obtaining them. When it is attempted to fuse a mixture of oxide of iron and lime in the proportion of a molecule of each, the very high temperature necessary to effect this fusion leads to the partial reduction of the sesquioxide to the state of the magnetic oxide of iron, even on making the flame as oxidizing as possible. On adding two or three molecules of lime, the mixture melts easily without reduction of the oxide. The melt is of such a dark color that cut into this section it is not sufficiently transparent to enable it to be studied by the microscope. All of these calcium ferrites, when treated with water, swell and slake more or less rapidly and none of them take any set. A hydrated calcium ferrite exists, to which Pelouze has assigned the formula Fe 2 0 3 .4CaO.Aq. It is a white sub- stance which is very rapidly altered by carbonic acid, which colors it brown by liberating the hydrated sesqui- oxide of iron. It is decomposed by water until the lime content of the liquor is about 0.60 grams per liter. But the limit is much less clearly defined than with the cal- cium aluminate and silicate. I have been able to obtain this compound by the prolonged contact (about one month), of moist hydrated iron oxide 2Fe 2 0 3 .3H 2 0 and calcium hydrate, but I have never been able to prepare it pure enough to establish its formula and determine its water of hydration. Calcium Alumino-ferrites. — The mixture of alumina and iron sesquioxide considerably augments the fusibility of the corresponding lime salts. Thus I have obtained the double salt Al 2 0 3 .Fe 2 0 3 .3CaO 70 CONSTITUTION OF HYDRAULIC MORTARS very clearly crystallized in long needles of a beautiful red color. It is truly a double salt and not an isomorphous mixture, because if more than one molecule of alumina is used, for one of the oxide of iron, colorless crystals of cal- cium aluminate are seen to be formed in addition to the red crystals. I have tried, as for the calcium silicate, to cause the aluminates and ferrites to crystallize in a bath of molten calcium chloride ; these attempts have been entirely fruit- less. With the aluminates a vitreous very fusible material is obtained, which is alterable by water and dissolves cal- cium chloride, and leaves an insoluble gelatinous residue containing alumina. With the ferrites, on the contrary, very beautiful brown crystals are obtained, which are un- changed by water and dilute acetic acid and permit the easy separation of calcium chloride and of the excess of lime. But this compound is a chloro-ferrite : Fe 2 0 3 .CaO.CaCl 2 Chemical Analysis. Observed. Calculated. FeA 51 49.0 CaO 18 17.2 CaCl 3 32 33.8 101 100.0 This body is very well crystallized, but it has such an easy cleavage that we can hardly gather anything but cleavage plates. I have found 129° 20' and 109° 10' for the angle of the cleavage plane with the two adjacent faces, which gives 121° 30' for the prism angle. These plates are terminated by an obtuse systemmetrical pointing, the plane angle of which is 139°. CEMENTS AND HYDRAULIC LIMES 71 Multiple Silicates of Alumina, Iron and Lime. — The number of double silicates at present known is very con- siderable and still increases. I soon gave up pursuing in this direction the researches which I had begun. It is the least complete side of my work. But I will show later, that from the standpoint of the theory of cements these com- pounds only play a secondary role. Among the double silicates the only interesting ones are those which contain the most lime; the acid silicates like the feldspars, which by fusion give glasses, cannot exist in cement. I have assured myself that mixtures of silica alumina and lime containing twice as much oxygen with the silica and alumina as with the lime, give upon fusion glasses which are unalterable in water. The mixtures containing an equal number of atoms of oxygen with the silica and alumina on the one hand, as with the lime on the other, give crystallized melts which, examined in thin sections, seem to be made up of two different kinds of crystals, the relative proportions of which vary with the proportions of alumina and silica. Some little cells with rounded contours and having weak double refraction, are essentially silicious, the others, elongated crystals, extin- guishing in the direction of their length are essentially aluminous. I have not succeeded in determining the com- position of either of them. Lastly, I will recall that the most basic calcium double silicates which we know are formed in blast furnace slags. These are Melilite SiO s .y 5 Al 2 0 3 .1.3CaO ' Idocrase SiO a . % A1 S 0 3 - CaO Gehlenite Si0 2 .y 2 AI 2 0.,.1.5CaO all of which contain less than two molecules of lime to one of silica. 72 CONSTITUTION OF HYDRAULIC MORTARS I have tried to obtain some hydrated double silicates; we know that the alkali silicates and ahiminates combine with the liberation of some alkali, to form an insoluble silico-aluminate, the composition of Avhich, according to H. Sainte-Claire Deville, may be represented by the formula 3Si0 2 .AL0 3 .^a 2 0+Aq. It was possible that the corresponding lime compounds might behave in the same manner. I have mixed pre- cipitates of calcium aluminate and calcium silicate with lime water diluted with an equal volume of water, that is, of such a concentration that neither the aluminate nor the silicate can be decomposed by the sole action of the water. The mutual reaction of these two compounds ought to cause the strength of the lime solution to vary. I have observed nothing of the kind, which shows that either no reaction has occurred, or, if one did, the quantity of the combined lime did not vary, contrary to what happens for the alkali silico-aluminates. To sum up, three different anhydrous calcium silicates exist, only one of which, the tri-calcium silicate, Si0 2 -SCaO, is attached by water and capable of taking set; three calcium aluminates, all of which take set very rapidly in water; some calcium ferrites, all of which slake and swell like quick lime, and lastly, numerous multiple silicates, none of which, amongst those so far studied, is alterable by water. The only corresponding hydrated salts able to exist in the presence of an excess of lime are the silicate Si0 2 .Ca0.2l4H 2 0. the aluminate Al 2 0 % .Wa0.12E 2 0. the ferrite Fe 2 0 3 JfCaO.Aq. CEMENTS AND HYDRAULIC LIMES 73 These salts are decomposed in the presence of an excess of water, liberating lime. MICROCHEMICAL STUDY OF CEMENTS. The cements and hydraulic limes contain variable pro- portions of their constituent elements : silica, alumina and lime ; they therefore are not definite compounds, but mix- tures of definite compounds. Chemical analysis can teach nothing about the nature of the compounds thus mixed; in order to solve this question, I have proposed to employ the microscopic method which has allowed us to make such great progress in the study of the rocks of the earth's sur- face. But this method can only be used for crystallized bodies; among the cements and limes, the cements known as Portland, or slow setting, alone fulfill this essential con- dition ; it is with these that I have concerned myself. Anhydrous Cements. — The examination by the micro- scope of a thin section cut from a clinker of Portland cement shows immediately two predominating constituents which recur without exception in all the samples. (Fig. 2, facing page 86). 1° Colorless crystals, with weak double refraction, with square or hexagonal cross sections and very clear borders much resembling those of the cube. It is by far the most abundant constituent. 2° In the space between these crystals, a ground mass, the color of which is always dark and varies from a yel- lowish red to a greenish brown. Its double refraction is stronger than that of the preceding material, but it does not possess any clear crystalline contours. 3° Beside these two essential elements, accessory ele- ments are frequently found, varying in different samples: 74: CONSTITUTION OF HYDRAULIC MOKTAKS a. Crystalline sections of forms and dimensions analo- gous to those first given, but which are distinguished from them by a light brownish, slightly yellowish color, a com- plete absence of transparency, and by very fine striae in- clined to each other about 60°. This constituent, al- though scarcely plentiful, is found, however, in almost all samples of cement of good quality. b. Very small crystals of sufficiently energetic double refraction to give polarization colors. This constituent is always in small quantity and is sometimes entirely absent. It is found especially in underburned cements. c. Certain forms without action upon polarized light and of negative character which do not give any distin- guishing test. This microscopic study, which of itself is insufficient to make known the nature of the crystallized compounds ob- served, notwithstanding, when it is taken in connection with the absence of fusion of cements during their calcina- tion, reveals this very important fact : The pseudo-cubic crystals, which are constituents of the first consolidation are not fused, but have formed by chem- ical precipitation in the midst of the fusible brown ma- terial, the constituents of the second consolidation, which after having acted as a flux and made the chemical re- actions possible, has solidified by cooling, filling up all the spaces ivhich then remained open. The action of chemical reagents upon a thin plate of cement placed under the microscope can be studied. I shall successively pass in review the action of the acids, the salts of ammonia, etc. The acids even when very dilute and weak, like acetic acid, attack all the constituents of cement very rapidly, the destruction can be recognized in polarized light by the CEMENTS AND HYDRAULIC LIMES 75 total extinction of the plate. From this it can be concluded that all the visible compounds contain lime because silica and the silicates of alumina and iron, the presence of which in a free state in cement is assumed by some authors, are entirely unaffected by weak acids. Upon allowing the plate to dry after being attacked, a white skeleton of silica is seen, not showing any important break in continuity, the essential elements of cements are therefore all silicious. The ammoniacal salts destroy in about a quarter of an hour the pseudo-cubic crystals which constitute the major part of cements, whereas the other constituents of cements active under polarized light only disappear after the end of several hours. If account is taken of the fact that the ammoniacal salts exercise a more energetic action than water attacking, for example, di-calcium silicate, which is not altered in water, we may conclude from this test that the pseudo-cubic crystals, are the only bodies, among those acted upon by polarized light, which play any part during the hardening of cement. Farther on, we shall see the con- firmation of this fact in studying hydrated cement. Ferro- and ferri-eyanides added to hydrochloric acid enable us to recognize the distribution and the degree of oxidation of the iron. The sesquioxide is found in abun- dance in the colored flux which surrounds crystals of the first consolidation and only there. The protoxide of iron is found in isolated and very rare opaque, slowly attacked grains which are magnetic oxide, arising generally if not exclusively from the emery used to grind the plates. More- over, this absence of protoxide of iron is confirmed By directly attacking cements by sulphuric acid, and adding permanganate, which in the greater number of cases is not decolorized. The presence of protoxide of iron in cements is only accidental. 70 CONSTITUTION OF HYDRAULIC MORTARS There are no characteristics which enable us to recog- nize the distribution of the alumina, but it is not rash to assume that it occurs with the sesquioxide of iron. This being admitted, one may conclude from this chemical study. The pseudocubic crystals, the essential constituent of cements, as well as the opaque striated cells, and possibly also the crystals of strong double refraction are composed of silica and lime. The colored flux which fills the voids left by all these crystals is a multiple silicate of alumina and iron and of lime. Of these compounds, only the first seems sufficiently alterable to enable it to take any important part during the hardening. This study teaches nothing of the nature or even the existence of compounds which are not abundant and which would be without action upon polarized light. It remains to determine the chemical formulas of the crystallized compounds brought out by polarized light. Numerous attempts to make the proximate analysis of cement clinkers either by reducing the ferruginous parts by hydrogen and separating them by the magnet, or by making use of a mercuric iodide solution, have been un- successful. I have thought that by seeking among the cements those which contain the least alumina and iron, one might get an approximate idea of the composition of tri-calcium silicate, the essential constituent of cement. I have succeeded beyond my hopes in this direction. A cement similar to Portland cement is made by grinding ''grappiers" or the residue from the slaking of first-class hydraulic limes, which are generally very silicious limes and containing scarcely any alumina. CEMENTS AND HYDRAULIC LIMES 77 These grappiers, examined before grinding, are a mix- ture of very various products, of which half at most is really cement. By a microscopic and chemical examina- tion, I have discovered the following materials: 1° Fragments of limestone, or unburned fragments; white, hard grains, effervescing with acids, showing in thin plates the strong double refraction characteristic of cal- cium carbonate. 2° Unslaked lime, white, friable, porous, grains, slak- ing by a long continuance in the air, it is unslaked lime. 3° Hydrated and hardened lime, white friable, porous grains, losing considerable water upon calcination; these are fragments of lime which, instead of slaking, have set because they were too wet. 4° Wollastonite, very hard, translucent, almost trans- parent, grains, coloration active in polarized light; their analysis leads almost exactly to the formula Si0 2 .CaO after subtracting from 3 to 4 per cent, of alumina and iron and a corresponding quantity of lime. The analysis of divers similar samples of different origin have indeed given me the following results : Molecules of CaO for 1 molecule Si0 2 CaO A1 2 0 3 Fe 2 0 3 H 2 0 & C0 2 MgO &c. of Si0 2 Teil 43.0 46 5.5 1.0 0.0 1.0 1.15 Senonches.43.0 45 3.3 1.2 0.7 1.2 1.14 Paviers ..45.5 48 4.4 1.4 0.3 0.7 1.14 47.0 49 2.0 1.5 0.3 1.0 1.13 The excess of 0.15 of a molecule of lime for one of silica corresponds to the proportion of alumina and iron, which is, on the average, 0.1 molecule. It seems singu- lar at first view that these losses arising from the irregular action of the silicious walls of the furnace and the fuel ashes should show so regular a composition. Samples 78 CONSTITUTION OF HYDRAULIC MORTARS ought to be met with containing a greater proportion of lime ; if this is not the case, it is because the silicate in the form of di-calcium silicate pulverizes spontaneously and consequently passes totally into the bolted lime. 5° Grappiers of Cements. — These are hard grains of saccharoidal fracture, green or grey, showing in their sec- tions crystals identical to those of cements. They are al- most exactly joined one to the other, the colored flux being almost absent. (Fig. 1, facing page 86). Their analysis ought therefore to give very exactly that of the crystals in question, and consequently to be identical whatever may be the origin of the grappiers. The following table of analysis shows that this composition differs little from the formula Si0 2 .3CaO. The grappiers analysed come from the lime of Teil (Ardeche) Senonches (Eure) and Paviers (Indre-et- Loire). Analyses. Molecules of CaO for 1 H 2 0& MgO molecule Si0 2 CaO AL,0 3 Fe 2 0 3 C0 2 S0 2 &c. of SiO a GreyGrappiers.Paviers 26.0 G6.0 3.0 1.2 1.1 1.0 2.75 Teil 26.0 66.0 3.5 0.8 1.0 1.0 2.75 Green Grappiers, — . . 24.0 69.0 2.7 0.3 1.0 1.0 3.08 Senonches 25.5 68.0 3.6 0.7 0.7 1.3 2.85 The pseudocubic crystals, which are the most abundant and at the same time one of the most alterable constituents of cements, are therefore composed of a calcium silicate differing little from the formula : Si0 2 .3CaO* * The hydraulic limestone of Teil which on the average corresponds almost exactly to this composition, gives by a suitable calcination a cement of the very first quality equal, if not superior, to the best Portland cements. It is an industrial proof of the exactness of the formula which I have assigned to the active element of cements. CEMENTS AND HYDRAULIC LIMES 79 I have not been able to establish in even an approximate manner the composition of the multiple silicate which serves as a flux ; but I ought to add that the solution of this question only has a secondary interest from the standpoint of the theory of cements, since this slightly alterable com- pound does not seem to take any part in their hardening. Among the non-essential elements, I have cited at the head of the list, the opaque, light yellow, finely striated, crystals. However, they do not occur in very calcareous cements ; their abundance increases with the proportion of silica, they are especially observed towards the surface of clinkers at points contaminated by the fuel ashes. Their presence has an intimate relation with the spontaneous pulverization of cements. These opaque crystals are only cracked, unorganized crystals, pulverizing spontaneously, a property pertaining to the silicate Si0 2 .2CaO. All of these characteristics are so concordant that one is tempted to assume the formation of this compound in a pure state, but on the other hand, from microscopic examination the general appearance of these crystals is so similar to those to which I have assigned the formula Si0 2 .3CaO that it is very difficult not to assume the continuous passage of one into the other. It might be supposed that the two silicates Si0 2 .2CaO and Si0 2 .3CaO form isomorphous mixtures, the pulveriza- tion of which is more complete and easy as the proportion of the first of these silicates is greater. The crystals of sufficiently strong double refraction to give colors in polarized light remind us of wollastonite, SiOo.CaO. They are too rare and moreover resist re- agents too strongly to take any part whatever in the hard- ening of cements. The crystals which do not act upon polarized light are so CONSTITUTION OF HYDRAULIC MORTARS those whose study presents the greatest difficulties, since nothing warns us of their presence and no characteristic permits us to recognize their destruction under the influ- ence of reagents. It was natural to seek among these bodies for the presence of lime which crystallizes in the cubic system. Since Rivot, the presence of free lime in cements has been admitted and it often happens that it is attempted to determine it by analysis. In fact, Rivot had recognized the presence of lime only in the hydrated cements, and he had assumed as evident that if the hydrated cements contained lime, it ought to be the same for the calcined cements, an entirely erroneous hypothesis. Quick lime can be very easily recognized, because of its property of slaking, the effect of which can be further ex- aggerated by using water heated to 100° C. The pres- ence of traces of lime in a cement suffices to cause a very evident swelling and cracking. The addition of 1 per cent, of lime from the strongly calcined nitrate, to Portland cement of first quality, after pulverization has produced the following results : The pure cement was made into mortar and then mixed with the lime, using the quantity of water strictly neces- sary ; the briquettes were fashioned in the form of cylinders two centimeters high and two centimeters in diameter, which were, 24 hours after taking set, heated in a water bath at 100° C. during 24 hours. Strength. Increase in bulk. Pure cement 320 Kg. None Pure cement 1% CaO 112 10% These figures allow cements of good quality to be con- sidered rigorously exempt from free lime. It is necessary to class certain aluminates among the cubic compounds, and as a consequence, without action CEMENTS AND HYDRAULIC LIMES 81 upon polarized light. I have tried to recognize them by using their solubility in water. Finely ground Portland cements agitated with a large excess of water always allows small quantities of alumina to dissolve. This fact, con- sidering the inalterability of the double silicates of alumina and lime seems therefore to establish the existence in the calcined cements, of small quantities of a calcium alumi- nate. This solution of the aluminate is much more important in Roman cements. The following figures give the quan- tity of alumina dissolved from 10 grams of cement from Vassy by ten minutes' agitation in 1 liter of water. The 3 samples came from 3 different factories : A1 2 0 3 in 1 liter, grams. Cement A 0.04 Cement B 0.045 Cement C 0.035 To sum up, the absence of free lime in Portland cements of good quality is very certain; the existence of the alumi- nate is only probable. This chemical study of the calcined Portland cements shows therefore that they are formed essentially of a cal- cium silicate differing little from the formula 8i0 2 .3CaO , which is the active element of hardening, and that the com- pound is produced by chemical precipitation in the midst of a molten double silicate, which has acted as a vehicle for the silica and lime to allow them to combine, but which re- mains sensibly neutral during their hardening. Hydrated Cements. — A briquette of Portland cement which has hardened under water for several months shows a clearly crystalline structure under the microscope ; small crystals in the form of hexagonal plates are recognized in all the cavities. The surface itself bristles with these 82 CONSTITUTION OF HYDRAULIC MORTARS crystals when the hardening has taken place protected from the carbonic acid of the air. Cut into thin sections and examined by polarized light, these cements have sharply illuminated borders showing brilliant colors which very clearly constitute the prolongation of the hexagonal crystals that cover the briquettes. Outside of these isolated borders only a whitish, slightly translucent mass without action on polarized light is seen which shows no apparent indication of crystallization. This negative characteristic has no value ; we have seen, indeed, that hydrated plaster formed by the massing of long needles of crystallized gypsum has the same appearance. It is because these crystals are very slender and within the thickness of a thin plate, a great number, oriented in all directions, are super- posed. We will presently see that it is the same with cements. Finally in the midst of this mass, we perceive from place to place outlines of larger grains of cement which are recog- nizable in the ferruginous flux which preserved them al- most unaltered, and which still gives the contour of the crystals of the calcined silicate, which has been entirely transformed. The hexagonal plates with which the briquettes of cement bristle sometimes attain a size of several millimeters; it is then easy to detach them for examination. The angles of these plates measured by the turning stage of the micro- scope are very exactly 60°. Two consecutive angles of a similar plate thus measured have been found equal to 59° 52' and 59° 36', numbers which differ from 60° by less than 1°, the error possible with the method of measuring used. The double refraction along the hexagonal axis is very weak and irregular, no direction of extinction ex- ists, but perpendicular to its axis the double refrac- CEMENTS AND HYDRAULIC LIMES 83 tion is very strong. Tn converging light the black cross is very clearly observed. These are unaxial negative crystals; their chemical analysis has given the following results : Weight of material, grams.. 0.062 0.088 0.243 0.183 CaO 71.0% 0.3% 00.5% 73.6% HX> 27-4 23.0 18.5 22.6 SiO a (by difference) 1.0 0.7 17.0 3.8 100.0 100.0 100.0 100.0 It is seen, therefore, that they are calcium hydrate con- taminated by a little silicate, the presence of which tends to give, along the axis, a weak double refraction which is entirely absent in pure calcium hydrate. Even the manner in which the crystalline shapes are distributed in the cement mass, their absence of crystalline boundaries shows indeed that the calcium hydrate crystallizes by surround- ing and joining all the foreign substances which it con- tains, just as the calcium carbonate in the calcite of Fon- tainebleau surrounds the sand in the midst of which it has crystallized. The relatively considerable dimensions of these crystals when taken in connection with the slight solubility of cal- cium hydrate, makes it evident that the production of this body is the result of a very slow reaction which could not be the simple hydration of the quick lime. By following the growth of these crystals, it is easily ascertained that it is prolonged through several months. The amorphous mass which constitutes the most impor- tant part of the hardened cements can be studied, as was done with plaster, by following with the microscope the progress of the hardening. Some special precautions are necessary to avoid the decomposition of the hydrated cal- cium salts, as much by excess of water, as by the carbonic 84 CONSTITUTION OF HYDRAULIC MORTARS acid of the air. I take a thin section of cement glued upon an object glass and attach a cover glass without sticking it to it. Then I immerse all in lime-water diluted with its own volume of water, taking care to put the plate of cement underneath. By taking this plate out from time to time and examining it by the microscope, the progress of the hydration is easily followed. At the end of two or three days, wide crystals with strong double refraction are first noticed ; they are calcium hydrate. Besides, the cement is often bristling all over with extremely fine needles being scarcely 1-100 of a millimeter long, the subsequent massing of which finally forms the body of cement. (Fig. 4, facing page 86). More frequently only fibrous masses are seen to form, resulting from the union of all of these small crystals. The ammoniacal salts instantly destroy these crystals*, leaving gelatinous flocks of silica. They are coin- posed of a calcium silicate to which I have assigned the formula of the only hydrated calcium silicate which I have been able to produce synthetically: Si0 2 .Ca0.2^H 2 0. Lastly, there is formed about the plate at variable dis- tances, small spherolites giving the black cross in paralleled light. The distance at which they are formed is the indica- tion of a certain solubility which reminds one of the aluminate, but this is only a simple hypothesis. Tlie formula should be Al 2 0,.4Ca0.12H,0* The only aluminate which may be able to exist in tie presence of an excess of lime. The formation of hydrated aluminates during the hardening of cements is confirmel, at least in the case of Roman cements, by the study of * See footnote, p. 67. CEMENTS AND HYDRAULIC LIMES 85 their progressive decomposition in the presence of water. Upon shaking some cement from Vassy in the presence of water until complete hydration, and removing half of the water every day, we notice that the strength of the liquor in lime is at first maintained constant and nearly 1.2 grams (per liter) which confirms the existence of free lime in the hydrated cement. Then the strength decreases al- most proportionately to the quantity of water added, until it stops at a new stationary strength 0.22 grams (per liter), which corresponds exactly to the decomposition of the aluminate. In one experiment I have observed the fol- lowing results : Lime per liter, grams. 3rd day °- 2 % 4th day °- 22 5th day 0.22 6th day °- 18 If now we compare the results to which the study of calcined and hydrated cements has led us, we will he led to conclude that the fundamental reaction which brings about the hardening is the splitting up of a basic calcium silicate into mono-calcium silicate and calcium hydrate: Si0 2 .3CaO + Aq. = SiO,.Ca0.2#H,0 + 2Ca(OH) 2 a reaction analogous to that which accompanies the setting of di-barium silicate. There is formed accessorily, a basic calcium aluminate, the rapidity of hydration of which intervenes in the more or less rapid set of different cements : AL0 3 .3CaO + Ca(OH 2 ) + Aq. = Al 2 0 3 .4Ca0.12H 2 0. this second conclusion being stated only with reserve. Lastly, iron certainly takes no part in the set of Port- land cements; there is no calcium ferrite formed; we 86 CONSTITUTION OF HYDRAULIC MORTARS would recognize it immediately by the brown coloration under the influence of carbonic acid, as a result of the liberation of hydrated oxides of iron. It, on the contrary, intervenes very clearly in Roman cements. I have spoken above repeatedly of the crystallized sili- cates and aluminates, which are formed either during the calcination. or during the set of cements; the figures op- posite show the structures observed in their sections, namely : Figure 1. — A thin section of grey grappier from Teil showing very feebly double refractive crystals of Si0 2 . 3CaO separated by a small quantity of a ferruginous matrix. Figure 2. — A thin section of Portland cement from Boulogne showing the same crystals as the grey grappier from Teil, and a much more abundant ferruginous matrix. Figure 3. — A cement similar to the preceding in which some crystals have become developed in an exceptional degree. Figure 4. — The beginning of hydration of a Portland cement, showing the elongated crystals of the hydrated silicate, Si0 2 .Ca0.2£H 2 0. The dimensions of these last have been considerably exaggerated, nearly doubled, in order to give more clearness to the design. Figure 5. — Crystals of the aluminate 2Al 2 0 3 .3CaO pro- duced synthetically. Fi gure 6. — A crystallized precipitate of the hydrated aluminate Al 2 0 3 .4Ca0.12H 2 0 * MANUFACTURE AND INDUSTRIAL USE. From the preceding investigations we are able to deduce certain conclusions relative to the manufacture and the in- * See footnote, p. 67. CEMENTS AND HYDRAULIC LIMES 87 dustrial use of cements. I shall successively treat of the different kinds of hydraulic products : Portland cements, Roman cements, hydraulic limes and puzzolanas. I shall study their composition and their cal- cination, their hardening, the causes of their destruction and the methods of testing. COMPOSITION' AND CALCINATION. Portland Cements. — It is possible to define rigorously the extreme limits of composition which completely cal- cined cements, such as Portland cements generally are, can possess. It suffices for this purpose to recall that they must not, on the one hand, contain any free lime which fixes a maximum for the proportion of lime, and that, on the other hand, the presence of too great a proportion of di-calcium silicate leads to spontaneous pulverization upon coming out of the kiln, which fixes the lower limit of the lime content. 1° Upper limit of the quantity of lime. In the pres- ence of increasing quantities of lime, the compounds which tend to form are : Si0 2 .3CaO and Al 2 0 3 .3CaO. The silico-aluminates tend to disappear completely; we therefore ought to have as the upper limit of the lime CaO+MgO Si0 2 +Al 2 0 3 = The quantities of each body being expressed in this for- mula not in weights but in molecules, I have not included iron sesquioxide in the calculation, since calcium ferrites swell in slaking. It therefore is not necessary to combine the oxide of iron with lime. 88 CONSTITUTION OF HYDRAULIC MORTARS 2° Lower limit for the lime content. Upon diminish- ing the lime the proportion of silico-aluminate increases, and when that is completely formed, di-calcium silicate Si0 2 .2CaO begins to be produced. The formula of the silico-aluminate is not known. We know only that it is at least as calcareous as the most calcareous of those which are known at present, gehlenite 2Si0 2 . Al 2 0 3 .3CaO. I have assigned this formula to it for the purpose of calcula- tion. We then find as the lower limit of the molecular ratios : (2) CaO+MgO SiO s _(Al 3 0 3 +FeA Here it is not necessary to separate the iron and alumina which act in the same manner with respect to the forma- tion of multiple silicates.* It is easy to ascertain that the two conditions stated above are satisfied for all the cements of good quality made in France, as the following table shows : Formulas. Origin of the Cement. , * > (1) (2) Boulogne 2.22 3.6 Desvres 2.28 3.8 Frangey 2.55 4.05 Grenoble 2.4 3.9 * The question whether magnesia should be added to the lime in these formulas and the oxide of iron added to the alumina has been the subject of numerous and tedious discussions, without decisive proofs resulting one way or the other. I must, however, recognize that the rather summary arguments which I have invoked are in- deed quite insufficient to settle this question which has not perhaps, after all, had the importance which has often been attached to it. The permissible variation in the composition of cements is, on the one hand, sufficiently large and, on the other hand, the quantity of these two bodies is never great. CEMENTS AND HYDRAULIC LIMES 89 On the contrary, for mixtures giving pulverulent cement, I have found : And, lastly, for a cement with an excess of lime made at Boulogne for the purpose of experiment and which swelled enormously : It is seen by these figures that the theoretical predic- tions are fully confirmed. But it may be asked why the composition of the cement is never represented by formula (1 corresponding to the exclusive production of calcium and aluminum silicates \ The lime is always too low. The ratio of bases to acids which ought to be equal to 3 very rarely exceed 2.5. The reason of it is that in the industrial calcination of cements the reactions are never entirely com- plete in consequence of the lack of fineness and homogeneity in the slurry, and the duration and temperature of the cal- cination are insufficient. In order to obtain the whole re- action, an increase in cost entirely out of proportion to the improvement of the quality of the product would be involved. It is preferred, in order to avoid the presence of free lime, to increase a little the proportion of the acids, silica and alumina. There is no inconvenience from this excess of acids, except in the formation of a little of the inert silico-aluminate. In each factory the conditions of homogeneity of the slurry and of the calcination remain sensibly the same. Thence it results that the slurry giving the best product should have a perfectly definite composition for each factory, but will vary from one factory to another with the Boulogne Desvres , 2.7 2.6 Boulogne 3.2 90 CONSTITUTION OF HYDRAULIC MORTARS conditions of manufacture. This is the explanation for the well known fact which at first seems paradoxical, that in one factory the variations of the proportion of clay in the slurry must not reach 1 per cent., whereas, as between one factory and another the difference may be more consider- able. These narrow limits do not allow the use of natural mixtures of limestone and shale designated under the name of marls or marly limestones. In order to obtain a composition of slurry of suitable homogeneity, it is neces- sary to make artificial mixtures, the preparation of which becomes an important item of the cost of manufacture. Moreover, the mixture can be made from pure limestones and clay, with each other or with marly limestones ; these mixtures may be effected in the dry or wet way, etc. The desired result is equally well obtained by very varied methods, the choice of which depends upon the conditions which prevail at each factory, and with the materials which they have at their disposal. The mixtures, although intimate, are not and cannot be absolutely homogeneous ; the grains of each material, how- ever small, still have finite dimensions. Those of calcium carbonate may reach 1 millimeter. The much finer particles of silica and clay are only several thousandths of a millimeter in diameter, but generally they are grouped in more voluminous bunches. It is difficult, in spite of all precautions, to avoid the presence of grains of limestone a little too voluminous, which give free lime on burning. It is necessary to slake this lime as much as possible before the use of the cement; for this purpose, the ground cement is left for a certain time in the air, or sometimes materials which will give up water vapor when they are heated are added in the mills CEMENTS AND HYDRAULIC LIMES 91 during the grinding, for example, gypsum, or already hydrated cements.* During the calcination, the first effect of the heat is to decompose the clays and dehydrate them at a temperature in the neighborhood of 600°. This decomposition simul- taneous with the dehydration is evidenced by the action of a potash solution and of sulphuric acid, the first of which dissolves silica and the second, alumina more easily than before the calcination. Between 800° and 900° the lime- stone is decomposed, liberating its carbonic acid and being transformed into quick-lime. From this moment, the ele- ments of the clay begin to react upon the lime, and this reaction becomes more complete as the temperature is higher and its action more prolonged. At the points of contact of the grains of lime and the particles of clay, fusible products are formed which are diffused in opposite directions, becoming more basic on the one hand and more acid on the other. If we break up a nodule of clay, we will have in the centre the elements of clay, infusible silica and alumina. Then the slightly calcareous fused glasses, afterwards a fused mixture of double silicates analagous to slags, with the mono- and di-calcium silicates, all fusible at the temperature of calcination of cements, lastly, the most basic salts, the active constituent of cement, infusible tri-calcium silicate, and fusible calcium aluminates, and in the last place of all, quick-lime. The proportion of these diverse elements varies in a continuous manner with the degree of advancement of the calcination, tending towards a limit dependent only upon the relative proportions of the elements present. * The usual temperature of a Griffin mill being about 75° C, and of a Davidsen tube mill being about 83°, it is doubtful if gypsum loses any moisture. — Translator. 92 CONSTITUTION OF HYDRAULIC MORTARS With a large excess of lime, the final products will be quick-lime, tri-calcium silicate and tri-calcium aluminate. By diminishing the quantity of lime, we would have these two salts and no quick-lime. Afterward the calcium alumi- nate will disappear and will be replaced by a multiple silicate of a composition analagous to that of the basic slags from blast furnaces. This will be followed in turn by the disappearance of tri-calcium silicate, which will be replaced by di-calcium silicate with spontaneous pulverization ; then by the mono- silicate. Finally, glasses analogous to the acid slags of blast furnaces will be produced. In order to obtain calcination, the temperature must be the higher, as the multiple silicate which serves as a flux and allows the diffusion of silica and of lime in opposite directions is itself less fusible. A silicate containing only either alumina or sesquioxide of iron will be less fusible than if these two bases are both present at the same time, and it is possible to conclude from analysis that the maximum fusibility will be obtained with equal equiva- alents of these two bodies, say in round numbers, half as much again sesquioxide of iron as alumina. It is seen from this how slurries of variable composi- tion will be able to give similar products by different cal- cinations ; a moderate calcination applied to a slurry low in lime will be able to give, by reason of an incomplete calcina- tion, free lime and calcium aluminate just as a complete calcination applied to a mixture richer in lime will do. For example, in order to increase the rapidity of the set, an indispensable quality in certain works, the calcina- tion will be produced at a lower temperature to make the reaction less complete and augment the proportion of the CEMENTS AND HYDRAULIC LIMES 93 aluminates, but at the same time, the proportion of the lime will be diminished by several per cent, to avoid the possibility of any of it remaining uncombined. In addition to the normal Portland cement studied above, which comes from the kiln in greenish clinkers with a scorcacious appearance and very hard, there will be taken from the kiln about 25 per cent, of refuse,* or at least mediocre cement, which for reasons of economy has habitually been made to pass with the rest. These are, first, pulverized materials designated under the name blue powders arising from the disintegration of well calcined clinker, but containing too large a proportion of di-calcium silicate. The presence of this silicate is due to excessive burning in the case of a too argillaceous slurry, and likewise to the superficial action of fuel ashes, or of the siliceous walls of the kilns upon (the surfaces of) the clinker. These powders harden slowly, but in the long run may take set with great hardness, like the hydraulic limes. On the other hand, they have the great merit of being certainly exempt from free lime, and consequently are not subject to swelling. Secondly, unburned slurry is found, brown or grey, and porous, which little by little slakes from the moisture of the air, giving a brown powder. These unburned slurries contain free lime and calcium ferrites and aluminates. After a gradual slaking they give a cement which takes set rapidly, but of slight and very irregular strength. Lastly are found near the furnace walls, silicious scoriae, glasses and often perfectly crystallized masses of wollas- tonite Si0 2 .CaO, inert material which often has been- erroneously considered as overburned clinker. Roman or Quick Setting Cements. — The necessity of * This refers to the old shaft kilns. — Translator. 94 CONSTITUTION OF HYDRAULIC MORTARS preparing the slurry for Portland cement by artificial mix- tures considerably increases the cost of production of the latter. The calcination of the natural mixtures constitut- ing the marly limestones, which is much more economical, may be used for the manufacture of cements of ordinary quality, but these differ entirely in their composition, cal- cination and use, from the Portland cements. Experiment has demonstrated that the best results are obtained by using limestones rich in clay, by burning the rock slightly, and, lastly, by letting the cement slake in the air before using it.* These cements after calcination appear in the form of brown porous, light, friable masses, having no external appearance of fusion. They seem to contain besides a small quantity of tri-calcium silicate, some free lime and calcium aluminates and ferrites, and at the same time acid silicates, all products resulting from an incomplete re- action. A partial slaking before use is advantageous in order to eliminate the free lime, but too long standing in moist air deadens them completely. This slaking is facili- tated by the porosity of slightly calcined cement. Calcium ferrite is easily recognized by the discoloration of mortars during their set, due to the formation of white hydrated calcium ferrite, soon followed by a brown superficial color- ation upon contact with the carbonic acid in the air, by reason of the liberation of the hydrated sesquioxide of iron. These cements of inferior quality have only in their favor their low cost of production, due to the use of natural *A11 the Roman cements contain a rather high percentage of sul- phuric acid. Mr. Candlot has shown the necessity of the presence of this body, without which those cements which have a strong propor- tion of clay will fall to powder upon cooling. (Comptes rendus de la reunion des membres francais de l'association internationale des methodes d'essais, mai, 1903.) CEMENTS AND HYDRAULIC LIMES 95 limestones, to the small quantity of fuel necessary for their calcination, and to the small expenditure of me- chanical work necessary for their pulverization. Hydraulic Limes. — The hydraulic limes are obtained by the calcination of argillaceous, or better, silicious lime- stones containing less silica and alumina than cement rocks. The proportion of free lime remaining after cal- cination ought to be sufficient to bring about the complete disintegration of the mass by slaking without employing any mechanical process. For these reasons, the cost of production of limes is much less than that of cements; but, on the contrary, they have the disadvantage of harden- ing less rapidly and less completely, part of the active con- stituents having been destroyed during the slaking. A good limestone for hydraulic lime ought to be constituted almost exclusively of silica and calcium carbonate, all the other materials giving compounds which remain inert dur- ing the set, either because they are not attacked by water or, on the contrary, being too easily altered like the aluminates, they hydrate during the slaking. Iron and aluminia, however, facilitate the calcination, as they do for cements, making it more economical, by allowing the use of a lower temperature and a less prolonged calcina- tion. From this standpoint only, the presence of small quantities of these materials may be advantageous. In limes of good quality observation shows that the proportion of these does not exceed 3 per cent. The infusibility of silica and lime is an obstacle to their mutual reaction, which is always very slow and easily re- mains incomplete. Even in the presence of an excess of lime it is difficult to avoid the formation of insufficiently basic, and inert, silicates. We can hope to obtain a suitable hydraulic lime by the calcination of a silicious limestone 96 CONSTITUTION OF HYDRAULIC MORTARS only when the silica present is in a very fine state of division. In the limestone of Teil the silica appears in small spherical grains of less than ^ of a millimeter in diameter. Limestones containing silica in the state of quartz sand, the grains of which always have an appreciable diameter, cannot be advantageously used, but it is only the size of the grains, and not their crystalline state, as has been maintained, which makes it unsuitable to yield hydraulic products. The most suitable relative proportions of silica and cal- cium carbonate in a limestone, taking account of the for- mula of only the active silicate Si0 2 .3CaO, will be : SiO a 16.6 CaO.CO, 83.4 100.0 But, indeed, on the one hand, more limestone will be needed to furnish the free lime which by its slaking will permit the spontaneous pulverization of the material, and, on the other, more silica will be necessary, a part of which always escapes from the reaction which remains incom- plete. These two requisites practically compensate each other, and observation shows that the limes having the best reputation approach the composition indicated above. The intensity of the calcination has a great influence upon the quality of lime, a strong calcination rendering the reaction of the silica with the lime more complete, in- creases the proportion of tri-calcium silicate, and in conse- quence improves the degree of hydraulicity of the product, but, on the contrary, it diminishes the proportion of free lime, which makes the slaking more difficult. With the slightly silicious limestones containing more than four molecules of lime for one of silica, the calcination can CEMENTS AND HYDRAULIC LIMES 97 never be too intense ; sufficient free lime will always remain to insure the slaking. It will be quite the contrary with limestones not con- taining more than three molecules of lime for one of silica ; a very intense calcination gives what is called black grap- piers of cement, or the limit limes of Vicat. This product is not pulverized by slaking, and can begin to set only when it is finely ground, without previous slaking, but then it relaxes at the end of a certain time, swells, and ends by disintegrating under the influence of the small quantity of free lime which it contains. If the grinding is caused to be followed by a sufficiently prolonged air slaking, in order to make it complete, a hydraulic product of very good quality will be obtained. As the burning is less complete, a lime which is more easy to slake will result and no grinding will be necessary, but, at the same time, the degree of hydraulicity will con- tinue to decrease. The most suitable degree of calcination will depend upon the means of slaking which is used. The more a factory aims at a careful slaking, the more it will be able to push the calcination toward the production of black grappiers, and the more it will augment the quality of the product. This is the point which sharply differ- entiates factories calcining similar limestones. The improvement in quality resulting from the addition of ground grappiers to lime is often contested. This dif- ference of opinion comes from the fact that very dis- similar materials are known under the name grappiers. The word grappier designates all those things which es- cape pulverization by slaking; it is a mixture containing, on the one hand, unburned limestone, badly slaked lime, lime soaked and already set, and wollastonite, all white or yellowish materials and nearly inert, and, on the other 98 CONSTITUTION OF HYDRAULIC MORTARS hand, greyish black very hard grains, which are true grains of cement. The proportion of these last vary, according to the nature of the lime and its calcination, from more than 50 per cent, to less than 1 per cent. If in the first case the addition of grappiers is advantageous, it can only be detrimental in the second case. The slaking has at least as considerable an influence as the calcination has upon the quality of the lime ; generally this is the part of the manufacture of limes which leaves the most to be desired. The slaking, or pulverization, under the influence of a chemical reaction, is a very common phenomenon, pro- duced especially by the action of gaseous bodies upon solid bodies (the spontaneous oxidation of pyrites, the rusting of iron, the hydration of lime by atmospheric moisture). This slaking is the indispensable condition of all reaction between gaseous and solid bodies. If the solid body does not disintegrate, a superficial reaction only is produced, the new compound forms a coating, which prevents the contact, and terminates the reaction (the oxidation of sheets of zinc in the air) . The slaking upon contact with liquids is, on the contrary, exceptional, only three examples of it are known : the hydration of lime, of baryta and of strontia ; and even in these cases the final division obtained is less perfect. The elevation of tem- perature which accelerates all chemical reactions makes the slaking more rapid and thus seems to make it more com- plete, by causing water to react in a vaporous state. In the slaking of a hydraulic lime, it is necessary to obtain the complete hydration of the lime while altering the silicate as little as possible. This result can only be obtained by working at a very high temperature, which keeps the water employed in the state of vapor. In this CEMENTS AND HYDRAULIC LIMES 99 state water has not, so to say, action upon the silicates, which it does not cause to slake, whereas it does cause this rapidly with lime, the aluminates and the ferrites. The required elevation of temperature is furnished by the considerable heat of hydration of lime ; all losses of heat ought to be avoided with the greatest care, by giving as great a volume as possible to the pile of lime and pro- tecting it against all causes of external cooling. Even under these conditions the slaking is very slow; for the limes which are strongly hydraulic, it normally lasts a week. But this duration is very variable with the nature of the lime. Those which are slightly hydraulic, either as a result of a lack of insufficient calcination, or as a result of a deficiency of silica, may slake in forty-eight hours : for grappiers, on the contrary, the duration of air slaking, whether before or after grinding, ought to be counted by months. An insufficiently prolonged slaking gives limes the set of which is very rapid, but in time they swell up and disintegrate. Slaking at too low a temperature gives soaked limes, which, having partially hardened during slaking, are no longer able to harden in use, and which, on the other hand, are frequently insufficiently slaked in the work, as a result of the lowering of the temperature, and swell with time. The slaking of hydraulic limes is therefore the most delicate part of their manufacture. It is impossible to effect it on the ground at the moment of their use ; it always ought to be done at the factory. Hardening. — The hardening of hydraulic materials is divided into two phases, the setting and the hardening properly speaking ; the distinction of which, from a chemi- cal standpoint, is not very clear, but from the practical 100 CONSTITUTION OF HYDRAULIC MORTARS standpoint, these two phases have a special importance suf- ficiently great to merit a separate study. The setting is perhaps only the beginning of the hard- ening; it is characterized by a progressive diminution in the fluidity of the mortars : it begins to manifest itself by the persistency of breaks of continuity ; the voids produced artificially in the paste, which do not close up again under the influence of gravity or capillarity. From this moment the mortar is no longer good for use. The more quick the set, the more difficult is the workman's task. The set is considered as ended when the mortar paste does not yield under the influence of moderate pressure, such as that of the thumb nail or that of a steel rod loaded with a weight (Vicat needle). The chemical reactions which accompany the setting may be different from those which continue to occur during the hardening, but it is not always so, in the hydraulic limes, for example, it is impossible to establish such a dis- tinction. With the cements, on the contrary, which have undergone only a little or no slaking, there exists, besides tri-calcium silicate, certain compounds, the hydrating of which becomes complete at the end of a few moments : this is the case with the calcium aluminates and ferrites and quick lime. The thinnest parts of the tri-calcium silicate, the surfaces of the grains, undergo thus a rather prompt hydration. These different chemical reactions may at the beginning give a very rapid course to the hardening and constitute the period of set. It is easy to foresee the circumstances which may cause the rapidity of the set to vary. They are so numerous, that there is scarcely any of the elements of quality in cements, which may be susceptible to such great irregulari- ties. The fineness of the cement, depending upon the CEMENTS AND HYDRAULIC LIMES 101 processes of grinding and bolting, the porosity of the grains of cement calcined rapidly at a low temperature, the nota- ble proportions of aluminates and ferrites, giving quicker and quicker cements. Lack of slaking in the hydraulic limes produce the same effect. The circumstances of their use, the external temperature, and the chemical composition of the waters, are not without considerable influence. When the progress of the set is followed by means of a thermometer, a more precise method of observation than those which are based upon the study of the increase in hardness, at first a period of inactivity is often noticed during which the temperature does not rise : it seems that the chemical reactions do not begin immediately. Then at the end of some moments the temperature begins to rise abruptly and the set becomes rapid. This peculiarity is very simply called the phenomenon of supersaturation, the existence of which I have noted. So long as solution only occurs the disengagement of heat is very slight or none at all, and the setting does not commence. It is only at the moment when supersaturation by crystallization be- gins that the setting occurs manifestedly itself by a rapid disengagement of heat. The hardening,, properly so called, is occasioned almost exclusively by the slow and progressive hydration of the calcium silicate, which, after having begun during the setting, continues entirely alone. The quickness with which hardening is completed is not of great interest, only the final hardness is of conse- quence. It is necessary to study the circumstances which may influence this final strength. This supposes that cements reach a definite hardness which they afterwards keep indefinitely. Often this is not the case : after passing a maximum the strength decreases. There is, therefore, a 102 CONSTITUTION OF HYDRAULIC MORTARS third phase in the hardening of cements during which they undergo more or less complete retrogression. For the present, I shall leave to one side this part of the question, and return to it later in the study of the causes of the destruction of mortars. I have shown at the beginning of this study that the hardness of a hydraulic mortar depended upon two factors, the cohesion of the crystallized compounds, which are formed, and their adherence, either to each other or to the sand with which they are mixed. There is nothing to be said of cohesion, but it may be interesting to return to adhesion and to study more in detail the considerable modifications which they often undergo by reason of the variations in external circumstances which apparently are most insignificant. Adhesion depends chiefly: 1st. Upon the nature of the bodies in contact and the relative direction of the faces in contact, in the case of crystallized bodies. 2d. Upon the extent of the surfaces in contact. The Nature of the Bodies in Contact. — Crystals of the same nature meeting each other as a result of their de- velopment are joined to each other more or less completely. If their relative directions are such that the surface of contact determines two reticular superposable planes, they will be able to unite in such a manner as to form only a single individual. In this case, their adhesion will become equal to their internal cohesion, but, in general, this is not the case, and it is observed that the adherence of two individuals united in any manner whatever is considerably lower than their cohesion. This fact can be shown by allowing to crystallize by cooling, the solution of a salt chosen from among those which do not have easy cleavages, CEMENTS AND HYDRAULIC LIMES 103 and by breaking by hand the groups of crystals obtained. It is observed that the force necessary is less than that re- quired to break an isolated crystal and, further, that the fractured surface shows plane faces, indicating that the fracture is produced along the faces of contact of the crystals and not across them, the rupture of which would have produced an irregular surface. When a foreign inert, body is added to one which sets by itself, sand, for example, in order to increase the vol- ume of mortar obtained, account should be taken of the adherence of the crystallized salt with the foreign body. This adherence depends upon the chemical nature of the bodies present, and the condition of the surfaces. Daily laboratory experience shows that salts crystalliz- ing in a glass flask adhere more or less strongly to the walls. Some allow themselves to be broken rather than be de- tached. I will cite, for example, the crystals of barium silicate, which are formed in a flask of baryta water; others, on the contrary, are detached under the influence of their own weight ; this is the case with calcium sulphate. I have ascertained that this last salt had no greater adherence to carbonate of lime and quartz, that is, to the constituents of ordinary sand. The addition of sand can, therefore, only diminish the tensile strength of plaster mortars. The grains not joining with the calcium sulphate produce simply surfaces of contact, diminishing the full section almost as air bubbles would. This conclusion is fully in accord with practical results : plaster is always used without the addition of sand. The state of the surfaces also plays a very great part in the adherence. We know that in order to make lime mortar, the sharpest sands give by far the best results. The influence of the sharp- ness results, first, from the increase of the surfaces of 104 CONSTITUTION OF HYDRAULIC MORTARS contact; it is indeed evident that the development of the real surface of a piece of roughened glass is much greater than that of a piece which has been polished, but the prin- cipal cause of this influence of roughness, it seems to me, should be searched for elsewhere. We observe, when we attempt to stick any object whatever upon a plate of polished glass with gutta percha or Canada balsam, bodies which adhere strongly to glass, that the peeling off, when once begun at a point extends afterward under the influence of a very slight force, and that this does not occur if the glass is roughened. This phenomenon, it seems to me, is related to the facility with which fissures are propagated in a homogeneous and very strong body, such as the glasses, which, once cracked, split completely under a very mod- erate force, which does not happen with bodies the fracture of which is rough. Extent of the Surfaces of Contact. — I have said that the second factor upon which the strength of a mortar depends is the extent of the surfaces of contact of these diverse elements. This extent varies with the form of the crystals, with the volume of the voids remaining in the mortar after the complete set and with the manner of distribution of the voids. Form of the Crystals. — A very weak adherence, per unit of surface, if it is multiplied by a considerable extent of surface, will give as the result a very great total force of adherence, which may even equal the internal cohesion of the crystals, that is to say, the maximum strength which the mortar may be able to show. The surface of the crystals increase, moreover, in proportion as they are more elongated, as they differ more from a spherical form, the surface of which is a minimum. The form of elongated prisms which I have discovered in the crystallization of CEMENTS AND HYDRAULIC LIMES 105 plaster and all similar products is, therefore, eminently favorable to the development of adherence. Volume of Voids. — The quantity of water which it is necessary to use for mixing up a mortar to suitable con- sistency is generally greater than that which may become fixed by the setting. In order to mix up the plaster as stiff as possible, the quantity of water necessary is about 30 per cent, of its weight, whereas the quantity of water which it fixes is only 15 per cent. Therefore, at least twice as much water is used, and generally still more water is put in in order to slow the set and facilitate the use of the mortar. This water in excess remains imprisoned and forms breaks of continuity which diminish the extent of the surface in contact. If these voids did not exist the crystals would touch each other on all their surfaces and the adherence would be a maximum. In proportion as the volume of the voids increases, the extent of the surfaces of contact diminishes up to the point when the crystals no longer touch, then there is no setting at all : there is simply a mass of mud formed by the more or less complete suspension of a precipitate in the midst of the excess of water. The volume of voids sufficing to pre- vent the contact of the crystals, and, as a result, the quan- tity of water which a mortar can stand without ceasing to set, depends essentially upon the form of the crystals. Elongated prisms will be able to stand a much greater excess of water without failing to touch each other than would be the case with cubes. We will assume, cubes which may be regularly dis- tributed in space, and the centers of which, for example, form a cubic network, but which may be oriented around their centers in any manner whatever. The requisite condi- tion that they may no longer touch two and two is that their 106 CONSTITUTION OF HYDRAULIC MORTARS greatest length, that is, their diagonal, shall be at most equal to the side of the cubic network. The volume of voids, which is easily calculated, will be then a little less than twice the volume of the crystals. With elongated crystals whose length would be, for example, equal to ten times their width, which, like the cubes preceding, would have their centers upon a cubic network and which would be united in any manner what- ever, the contacts would become impossible when their greatest length, that is, their axis, is equal to the side of the cubic network. The volume of voids will then be 99 times that of the crystals, that is to say, that the prismatic crystals ten times as long as they are wide, will be able to stand an excess of water 50 times greater than cubic crys- tals before completely failing to set. The evident conclusion of these considerations is that the set of mortar, which always contains an excess of water, can only result from the formation of much elongated crys- tals : this, indeed, the result to which the direct observation of the facts leads, as I have shown above, aside from all theoretical ideas. Distribution of the Voids. — In what precedes, I have assumed the crystals uniformly distributed and conse- quently the voids also. In reality, this cannot be the case : in some places the crystals will be more compacted, else- where they will be spaced further apart. This unequal division may have a very great importance in regard to the positive strength. We will take the preceding example of cubes, regularly distributed in space and not touching each other: that is, for which the ratio of voids to filled spaces is equal to two and which, consequently, forms a system of no strength. If we shall modify their distribu- tion in such a way as to place some besides the others CEMENTS AND HYDRAULIC LIMES 107 along certain directions, for example, following the faces of a cube bounding a certain volume, we will be able to form a solid collection. It is easily ascertained that the tensile strength of such a system will be half of that which it would have if there had been no voids, that is, half of its maximum value. I have assumed here that the spacing of the crystals is according to a definite law, an hypothesis which at first view may seem very far from the reality, in fact, similar spacings are produced spontaneously in a very great num- ber of cases, and contribute to notably increase the strength of mortars. We will take a mortar formed of a mixture of finely pulverized cement, as is the custom, and ordinary sand, that is, rather coarse. The grains of sand, more or less rounded, touch each other in certain points, around which the interval which separates these grains increases progressively. It is conceivable that if all the cement can be concentrated in a restricted zone which surrounds the points of contact, the useful effect of a given quantity of cement will reach its maximum. If, on the contrary, the cement is uniformly distributed in all the voids left by the grains of sand, it is very evident that the crystals formed drowned in a great excess of water and for the most part removed some distance from the grains of sand, they will be in much less favorable a condition for strength. Now this concentration of cement at desired points may be very simply obtained by the capillarity of the water. It is sufficient to let the mortar drain, the most voluminous voids are emptied and the water lodges in the narrower spaces, carrying with it the pulverized cement which col- lects at the desired points. We know, indeed, that the strength of test briquettes of cement is doubled when they 108 CONSTITUTION OF HYDRAULIC MORTARS are put to drain on a porous surface of dry plaster as soon as moulded. To an analogous grouping of crystals we must, without doubt, assign the increase of strength which plaster takes upon desiccation. The increase of strength thus produced is notable, and it is the greater as the excess of water used for mixing has been greater. Cylinders of dry plaster, afterwards soaked in a saturated solution of calcium sul- phate, have given me the following figures on breaking : Strength per square centimeter. Kilograms. Dry cylinders f56 J G2 Wet cylinders | 28 l 30 The decrease in strength has therefore been exactly half. It might be supposed that the process of the phenomenon is analogous to that of the hardening of the clay by desic- cation, but in the case of plaster the effect is less marked because the crystals being already joined to each other do not have the entire mobility of the particles of clay. They are only able under the influence of the capillary forces developed by the evaporation of the water to be de- flected in the interval between their points of contact in such a way as to come into touch with each other and to adhere to each other at new points. This movement will always be produced in the direction in which the crystals are closer and consequently will diminish the number of breaks of continuity of the mass by increasing, on the other hand, the volume of those which remain ; these are the most favorable conditions for the increase of strength. The shrinking of the mass is very slight because it is of the order of magnitude of the approach of two ends of a straight line which is slightly bent: it cannot be com- CEMENTS AND HYDRAULIC LIMES 109 parable to that of clay in which every particle approaches the other. We see, therefore, in summing up, that the magnitude of the adhesive force, and consequently the strength of mor- tar, are intimately bound up with the elongated form of crystals and their manner of distribution, conditions which themselves depend upon phenomenon of supersaturation. The elongation of crystals is the greater as they are precipitated in a more strongly supersaturated solution: fineness of grinding, which increases the extent surfaces of solution will therefore be favorable to hardening. This influence of the rapidity of chemical reactions makes itself particularly felt with plaster. We know that the latter when strongly calcined only gives a very mediocre mor- tar; natural anhydrite does not harden at all. This is not, as is commonly believed, a result of lack of hydration. The water is active, its action is only very slow and the hydrate formed is found in large short crystals which ad- here to each other very poorly. This explanation is con- firmed by the fact that the crystallized hydrate S0 3 .CaO,- J/2H2O, which does not differ from calcined plaster, sets very poorly. The absence of porosity and the extent of surface of contact with the water suffice to modify the con- ditions of the crystallization. Roman cements, calcined at very low temperatures owe, doubtless, to their porosity, and consequently to the rapidity of their hydration, their ability of acquiring a certain hardness, which would not properly be expected from incompletely calcined prod- ucts, in which the proportion of active material ought to be relatively small. Raising the temperature, which accelerates chemical re- actions, ought to be favorable to the hardening of cements, but experience seems to indicate that between 0° C. and 110 CONSTITUTION OF HYDRAULIC MORTARS 100° C. the final hardness of products of good quality changes little. Merely the limit is reached much more quickly. It is possible that the temperature might have a special influence upon the elongation of the crystals in a contrary direction to that which produces an increase from the supersaturation. It will be understood that I leave out of consideration the products of poor quality insufficiently slaked, the destruction of which is thus increased, as I shall treat it farther on. But these are not the only causes which influence the development of crystals: the presence of small quantities of foreign substances in the mixing water may have a very considerable action. I shall recall, in connection with this subject, the experiments of Le Blanc upon the crystalliza- tion of alum; but our knowledge upon this subject is too incomplete to enable any application of it to an actual case to be attempted. The existence of such influences seems to be well shown by the difference of results obtained by mixing up a cement with rain water, river water, or selenitic water. In regard to the way the last circumstance may be able to influence the development of crystals, I will point out the number of centres of crystallization produced in a supersaturated liquid ; the fewer of them it has, the more the crystals will be elongated, the more favorable the con- ditions will be. But it is rather difficult to foresee the circumstances which determine the development of the first crystals. Doubtless in this direction must be sought the cause of the bad results obtained with dead cements. Hydration by the air of only several per cent, of the active elements deprives a cement of all of its qualities. There seems to be no proportionality between cause and effect. The final strength ought to diminish several per cent, if CEMENTS AND HYDRAULIC LIMES 111 the hydrated parts behave like inert materials. By creating an infinite number of centres of crystallization on the surface of all the grains, they ought to oppose themselves to a suitable supersaturation of the solution. CAUSES OF DESTRUCTION OF HYDRAULIC MORTARS. After having studied the causes of hardening of hy- draulic mortars, it is interesting to review the circum- stances which may cause their destruction. These mor- tars may be used in three essentially different conditions : In free air, Under ordinary water, Under sea water.* Their destruction in these different conditions may be connected with four chief causes: The destruction of cer- tain hydrates by efflorescence under the influence of heat * The study of the decomposition of cements by the sea has made very great progress in these latter years. Vicat has shown that of the salts of sea water magnesium sulphate alone exerted a really harmful action. M. Candlot has shown that magnesium sulphate only exerts its action after having been transformed into calcium sulphate and that the destructive action of sea water ought to be charged principally to this body and to its combining with calcium aluminate. Mr. Maynard made evident the very important fact that contrary to what would be supposed the penetration of the salts of sea water does not take place to the interior of mortars, if I may say so. An- alyses of mortars long since immersed do not indicate the presence in the mass of notable quantities of sulphuric acid nor of magnesia, as long as the decomposition is not very far advanced. Nevertheless, the mortar constantly loses lime until all hardness disappears, even when, because of its small content of alumina, it is not the seat of any swelling. I have verilied the accuracy of these two apparently contradictory facts, of the elimination of lime without the precipitation of mag- nesia. It is very easy to follow the progress of the phenomenon by using, instead of magnesium salts, salts of colored oxides; of silver, mercury, cobalt and copper. At the surface a less and less per- 112 CONSTITUTION OF HYDRAULIC MORTARS and dryness ; hydration accompanied by slaking of certain ultrabasic products ; the decomposition of the lime salts by solution of their lime, and, lastly, the decomposition of these same salts by salts of magnesia. These different influences will be made unequally sen- sitive, depending upon compactness of the mortar; con- sequently, their action will vary with the proportions of meable crust of colored oxide, is formed, producing a semi-permeable partition. Lime is diffused from the interior of the mortar toward this surface, where it meets the soluble salt, and decomposes a new quantity of it, which reinforces the superficial crust. On breaking briquettes, after several months' immersion in like solutions, it is recognized without difficulty from the absence of color that no pene- tration of the colored oxide had taken place. When the salt is a chloride or nitrate, the reaction with the lime solution, but without penetration of the oxide, is continued indefinitely without any ap- parent alteration of the briquettes. When, on the contrary, the salt is a sulphate of copper or of cobalt (used as a concentration of 6 grams per liter), the effects are exactly the same as with magnesium sulphate used at the same concentration. That is to say, that at first a colored crust is formed at the surface, as with the chloride, but this crust swells up, cracks and reforms anew, and at the end of a certain time the briquettes are seen to split, to swell and to break to pieces completely. It seems that this expansive crust may have a force sufficient to break the cement to which it adheres. This expansive action of calcium sulphate is difficult to explain. I have discovered, in the first place, that in the formation of calcium sulpho-aluminate there is no increase of absolute volume. The com- bination of the bodies present is, on the contrary, accompanied by a contraction. Moreover, exactly the same thing occurs in the slaking of lime. The volume of the calcium hydrate is less than that of the quick lime and of the water which enters into its composition. The swelling is the result of an increase of apparent volume. The solid elements repel each other by an unknown means. However it may be, it suffices that these forces of unknown origin may have sufficient magnitude to give the explanation of the destruction of mortars. This is indeed the case from the slaking of lime which develops enormous force. We know that a small grain of lime in the middle of a brick suffices to cause it to crack. But it is entirely different with calcium sulpho-aluminate. The forces which accompany its formation are extraordinarily weak. They are not sufficient to crack a thin glass or test tube, and yet with time they succeed in dislocating extremely strong masonries. The reason for this paradoxical fact is a consequence of the general CEMENTS AND HYDRAULIC LIMES 113 sand and water employed for the mixing. A mortar with- out voids can be attacked only superficially by fresh or salt water: it will ''resist much longer than a permeable mortar. The difference will be the same as that which is observed in the solution of a crystal of sugar, or of a piece of lump sugar. The porosity, on the other hand, will re- tard destruction by slaking, by allowing the swelling to fill up the internal voids before making its action felt out- wardly. I do not think it useful to dwell longer on these facts, which to-day are well known. These being assumed, we will rapidly review the different causes of the destruction of mortars stated above. Dehydration by Dry Air. — A certain number of hy- drated salts, e.g., sodium carbonate, lose their water in dry air, and at the same time disintegrate and are finally reduced to a white powder having no cohesion. Among the theory of the rule which I have given. All force acting upon a solid hody in contact with an uncompressed liquid increases the fusibility or the solubility of the body considered, which melts and gives a supersaturated liquid in comparison with the same solid uncompressed body. The liquid will then allow the uncompressed crystals to crystallize in the voids and will thus cease to be saturated with respect to the compressed body, which will be dissolved anew, and thus repeatedly, as long as the forces act upon the solid parts of the mass. A porous moist mortar submitted to a continuous stress will be deformed indefinitely until rupture occurs, no matter how weak the stress may be. The quickness of the deformation will change only with the magnitude of this force. The accuracy of this theory is very easily verified by taking a stick of plaster 5 mm. thick, and 250 mm. long, carried on two supports and loaded at its middle point with a weight equal to one-fourth of the breaking stress. It is done comparatively on two similar rods, one dry and the other immersed in a saturated solution of calcium sulphate. The first is not bent; the second bends progressively until it is bent several centimeters and finally breaks. The continued expansive action of calcium sulpho-aluminate suffices to produce like effects, but this action often takes many years to reach its limit, because of the weakness of the stress put into play. 114 CONSTITUTION OF HYDRAULIC MORTARS hydrated lime salts produced during the set of cement the calcium aluminates are the only ones which to me hare seemed susceptible of meeting with a similar alteration. This cause of destruction is very clear with Eomari cements: briquettes left for six months to set under water and afterwards placed in a dry atmosphere are cracked by shrinking. After drying out, heating them for some min- utes at 100° is sufficient to cause them to lose half of their strength. Nothing similar is observed with Port- land cements of good quality. The hydration by moist air or even by water followed by swelling is by far the most frequent cause of the de- struction of mortars. The presence of foreign substances susceptible to slacking; ultra basic calcium aluminate, magnesian lime or even pure lime, may result from an ex- cess of calcium carbonate in the mixtures submitted to cal- cination, from an insufficient calcination, or, lastly, from a lack of slaking in the hydraulic limes. This effect is more marked as the grains of cement are more voluminous : the interior escapes the action of the water during the hardening, as I have explained in connection with the action of water upon the aluminates, and can ultimately be slaked upon contact with atmospheric moisture. This cause of destruction is more formidable in the air than in water : certain compounds susceptible of taking set upon contact with liquid water are slaked, on the contrary, under the action of moist air. But if the final swelling is greater in the air, it is, on the contrary, much more slowly pro- duced. I will cite, as an example, a cement rather rich in lime coming from one of the most noted French factories, which, after six months of setting under water, had given per- fectly intact briquettes and showing good strength. These CEMENTS AND HYDRAULIC LIMES 115 briquettes, removed from the water, and left in the air, be- gan to crack at the end of three years, and the sixth year were completely disintegrated and reduced to the state of a sandy mass. The finest powder, separated by bolting is shown to be richer in lime than the larger fragments. The disintegration, therefore, is very certainly due to the slak- ing of a small quantity of free lime. This different action of water in the liquid or in the vaporous state can be made evident with the greatest ease by experimenting upon salts which are rapidly hydrated, as sodium carbonate, when fused and finely ground and then mixed up with liquid water they take set quickly without slaking ; on the contrary, when left in the air, they hydrate slowly and swell enormously. Upon enclosing them in tubes, or simply in paper bags, the envelope is soon seen to burst on all sides, and the salt seen to come out through the crevices. In order that the slaking of lime may produce such dis- astrous effects, it is necessary that its slaking shall be very slow, and not become complete until after the hardening of the mortar. The lime obtained by the calcination of pure limestone is an extremely porous mass, which is slaked in a few seconds : in this state it is scarcely harm- ful, because its hydration is ended before the beginning- of hardening, properly speaking. This is what seems to happen with Roman cements calcined at very low tem- peratures and very porous ; but if the lime is compact and does not have pores possessing an extremely considerable surface of attack by the water, the hydration becomes very slow. This compactness is obtained every time that the lime is produced in contact with molten bodies, which dis- solve it and allow it to crystallize. Thus this occurs in the decomposition of calcium nitrate, in the calcination of 116 CONSTITUTION OF HYDRAULIC MORTARS lime from the carbonate in contact with fusible substances, such as calcium chloride, the calcium aluminates, and cal- cium sulphate. The retardation is still more considerable when the lime crystallizes in the presence of magnesia, with which it is intimately mixed, either by chemical com- bination or as an isomorphous mixture. In these cases, as I have shown above, the slaking will take years. There is no other cause to be sought for in the bad results obtained with magnesia cements, it is useless to attribute a char- acteristic action to magnesia. The substitution of magnesia for lime, molecule for molecule in a cement, has only one result, that of weakening the action of the water, if indeed for a certain magnesia content it would not have taken set at all. A cement capable of taking set can only be obtained by raising considerably the proportions of the bases. Then lime and magnesia remain in excess, which may take set like calcined dolomite, but likewise are disintegrated in time. The proportion of free lime necessary to alter in a notable manner the qualities of a cement is extremely slight. I have shown above that the addition of 1 per cent, of lime from the nitrate to a cement of good quality suffices to reduce its strength by half. This cause of destruction seems to me to deserve much greater attention than is usually given to it. I would be induced to believe that at least nine times in ten it is the only cause of the disintegration of hydraulic mortars used in the air or in fresh water. If it escapes observation fre- quently, it is because its action often does not become sen- sible until the end of several months, that is, long after the end of the tests which are usually made at present. The dissolving action of water would cause, with time, the total decomposition of hydraulic mortars, if the car- CEMENTS AND HYDRAULIC LIMES 117 bonation which is produced simultaneously did not protect them against this influence. This carbon ation is more or less efficient, according to the slowness with which it has occurred. When produced rapidly with the mortars of hydraulic limes, which contain pulverized slaked lime, the precipitate of calcium carbonate will be equally pulverized, without adherence, and will be carried along where water has a passage through the fissures of the masonry. These, instead of choking up, continue to increase. On the con- trary, with the cements, in which the free lime is less abundant and especially less finely divided, the slower re- action will cause the adherent crystallization of calcium carbonate, and this will end by filling up the fissures in the masonry, which will thus be produced by the dissolving action of water. A laboratory experiment which I have cited above makes very clearly evident these two different modes of carbonation. If calcium silicate or aluminate is left under water, we see forming gradually, under the in- fluence of the carbonic acid of the air, crystals of calcite adhering to the walls of the flask. If the aluminate or silicate is replaced by slaked lime, there will be formed only at the surface of the liquids crusts of calcium car- bonate, which gradually fall to the bottom of the liquid without acquiring any adherence with each other or with the walls of the flask. It, therefore, is only by virtue of the slow carbonation of the lime that the hydraulic mortars can escape the de- structive action of water and resist it in those places where plaster, calcium sulphate, which cannot carbonate itself, is finally entirely dissolved. The magnesian salts of sea water produce an action analagous to that of pure water, but much more energetic. They dissolve lime, as Vicat has shown, by allowing hy- 118 CONSTITUTION OF HYDRAULIC MORTARS drated flocculent magnesia to precipitate. A liter of sea water which contains 2 grams of magnesia can dissolve 2.8 grams of lime, whereas pure water can liberate only 0.052 grams per liter from the silicate. It seems that the mag- nesia, on being precipitated, ought to crystallize, and con- sequently to play the same part as calcium carbonate. This doubtless would happen if the precipitation were in- finitely slow, but having regard to the slight solubility of magnesia, to the concentration of sea waters in salts of magnesia, and to the concentration in lime of the waters of saturation of the mortar, the reaction is infinitely too rapid to give this result. With so great a dissolving power of sea water upon cement, the attack would be as rapid and complete as it is for a mortar of plaster placed in running water, if the pro- gressive carbonation of the lime did not come in to protect it against the action of the magnesian salts. It is not neces- sary to search elsewhere for the cause of the more or less prolonged resistance of cements to the destructive action of the sea. It is, therefore, interesting to study all the circumstances which may favor the formation and the crystallization of carbonates. The crystallization is the more perfect as the carbon- ation is effected upon the less soluble lime compounds. It, therefore, seems that the more or less complete elimination of calcium hydrate would be favorable : the addition of a certain quantity of puzzolana to the cement would allow this result to be reached. The use, like sand, of calcium silicates, inert to the action of water, but capable of being attacked by carbonic acid, as the slags of blast furnace are, it also seems ought to be recommended. I have shown that these compounds set perfectly in water charged with car- bonic acid. Doubtless they would not be altered by salts CEMENTS AND HYDRAULIC LIMES 119 of magnesia. But the best way to assist the carbonation would be to furnish the cement with carbonic acid in greater proportion than sea water can, the content of car- bonic acid of which is only 50 liters per cubic meter. The use of alkali carbonates is very costly, it may be that it would be possible to utilize the decomposition of certain organic substances by the lime of the cement, wood saw- dust, for example, either natural or transformed into hydrocellulose to make it more easily decomposed. We know, indeed, that all organic substances heated with lime break up in such a manner that all their oxygen passes off in the condition of carbonic acid: at ordinary tem- peratures the reaction is still produced, but much more slowly. A similar proceeding would allow carbonation to the center of a piece of masonry to occur, and would protect it against all infiltration. Such a method was em- ployed, intentionally or otherwise, by the Romans in the preparation of coatings of fat lime for their frescoes. I have recognized, upon samples brought from Pompeii, that they mixed with the lower layer of the coating a great quantity of fragments of chaff, of which only the casts re- main to-day. The organic matter has been completely de- stroyed in giving the carbonic acid which has aided the carbonation of the lime and which is now complete. I do not express these ideas as theoretical consequences of the chemical researches mentioned above; they should not be accepted until they have been submitted to the direct control of experiment. I have not spoken so far of the carbonation of the mag- nesia precipitated in the pores of a cement ; it ought to be produced in the same time as that of lime, and even give place to a more complete crystallization, because of the greater solubility of the bicarbonate. Moreover, mag- 120 CONSTITUTION OF HYDRAULIC MORTARS nesium carbonate crystallizes with a certain quantity of water, a favorable condition for the obstruction of the pores of masonry, but the hydrates of this carbonate stable at different temperatures are not the same, and it might be that the crystals at first formed may be afterwards de- stroyed by disintegrating as a result of variations of tem- perature: this is a question which would deserve to be studied. I have not spoken of the hydration of calcium carbonate, because this salt is ordinarily produced anhydrous, and, once formed, it cannot combine with water at any tem- perature: it is indefinitely stable. However, a hydrate of this salt exists, CaO.C0 2 .5H 2 0, which can be produced by the carbonation of lime below 5° C. ; but it is very un- stable, and the least rise of temperature transforms it into pulverulent anhydrous calcium carbonate, devoid, conse- quently, of all strength. It may be useful to take account of this fact, in the hardening of cements at low tempera- tures, especially for experimental briquettes, exposed to the air or placed under a thin layer of water. TESTING OF THE HYDRAULIC PRODUCTS. In concluding this study, I will broach a question which has great interest from a practical point of view, that of the testing of hydraulic materials. It may be asked at the very first if their quality depends upon certain very definite conditions, and, in the second place, if characteristics ex- ist which will allow us to discover, before all, in what meas- ure these conditions are fulfilled. The reply to the first question is evidently in the affirmative ; I have shown that a cement of good quality has a perfectly definite composi- tion, and the reciprocal ought to be assumed: all cement CEMENTS AND HYDRAULIC LIMES 121 showing this composition will be able to give a cement of good quality when ground and afterwards mixed under suitable conditions, but the means of controlling this con- dition of composition are absolutely lacking. Rough chem- ical analysis teaches nothing of the nature of the compounds formed; it does not distinguish between a mixture of silica and lime, a silicate of lime, a cement simply decar- bonated which cannot set, and a true cement properly calcined. The only useful information which may be drawn from the chemical analysis is whether the cement does not con- tain an excess of bases, that is, more than three molecules of protoxides (CaO.MgO), for one molecule of the acids (Al 2 0 3 .Si0 2 ) ; the theory that I have given above shows, indeed, that with this proportion of base, it will not contain free lime if the mixture was perfectly homogeneous, but only compounds capable of taking set Si0 2 .3CaO and Al 2 0 3 .3CaO. However, the mixture and the calcination never being perfect, it is preferable to take it a little below this limit. In the calcination, it is not necessary to allow for oxide of iron, since the ferrites are slaked like lime. It is neces- sary, on the other hand, to deduct from the protoxides the quantity required to saturate the sulphuric acid, and the chlorine, that is, to bring into the analyses, not sulphuric acid, but calcium sulphate, and likewise for the chlorine. These conclusions are fully in accord with the practical re- sults of the cement industry. I will give below the analysis of two cement clinkers: the first of good quality, the sec- ond setting with difficulty and soon disintegrating by swelling, as the result of the presence of an excess of lime. The calcination in both cases was complete. I will express the results of these analyses in molecular ratios, silica 122 CONSTITUTION OF HYDRAULIC MORTARS being taken as unity, as I have previously done, which makes the comparisons much more easy : Normal Cement. Cement with excess of iime. Losses and sundries 0.94% 1.82% Si0 2 1.00 1 1 01 1.001 A1 2 0 3 0.21 J 1J1 0.17 J 1.21 o!l7 1 1 17 Fe 2 0 3 0.04' 0.03 CaO 3.29 1 0 _ 3.71 1„„, MgO 0.08 J 3 37 0.05 J 3 /0 CaO.SOa 0.015 0.01 Acids „ = 2.78 = 3.2 Bases This table, calculated into percentages,* gives : Normal Cement with an Cement. excess of lime. Losses and sundries 0.94% 1.820% Silica, SiO, 21.43 20.085 Alumina, Al 2 O s 7.66 5.814 Ferric oxide, Fe 2 O s 2 - 2 9 1-607 Magnesia, MgO 1.15 0.677 Lime, CaO 65.80 69.545 Calcium sulphate, CaS0 4 0.73 0.456 Total 100.00 100.004 The ultimate analysis of a cement can thus only indicate a limit of composition, beyond which all the products ob- tained are certainly bad. The proximate analysis, if it were possible, would be able to throw some light upon the qualities of a cement. Rivot had proposed, in this direction, to estimate the free lime by the dissolving action of water ; this test is still made in many laboratories. In reality, it has no value, since calcined cements, as I have shown, do not contain free lime. That which is thus obtained comes from the decomposition of the basic silicate, a very incom- plete decomposition in forty-eight hours, since the com- * Using Stas' numbers, as given in Tomassi's Electro- Chemistry, 1889. CEMENTS AND HYDRAULIC LIMES 123 plete hydration requires months, so that in this method very little is measured but the rapidity of attack, that is, the degree of fineness of the cement. I have tried, without success, different methods of proxi- mate analysis, more or less indirect: the action of ammonium salts ; the calorimetric measurement of the heat of solution in acids; the microscopic examination of thin sections, but by none of these methods have I succeeded in obtaining conclusive results. If it is impossible to verify the conditions of composi- tion, it can be attempted without being biased by the com- position, to discover whether the cement after hardening fulfills the desired conditions of solidity. These me- chanical tests to-day are in great favor, although, indeed, they may be scarcely more instructive than the chemical analysis. The analogous tests made upon the metals are, it is true, excellent, because they measure directly the conditions of elongation and of strength which define the quality of the metal. This quality, after use, will remain constant as long as a foreign cause does not act to destroy or alter the metal. For cements, the problem is entirely different; indeed, from the moment of their taking set, they experience a slow evolution which may require years before reaching its limit. This limit only is it interesting to know, and the tests which may require to be made on short notice can teach us nothing of our subject. They give only the tangent at the origin of a curve of which we would want to know the asymptote at infinity, and there is no relation between these two extremities of the curve, which are generally separated by a maximum of variable importance. Certain very basic cements, magnesian or not, which 124 CONSTITUTION OF HYDRAULIC MORTARS give very good strengths during the tests, end after several years by being completely disintegrated; the asymptote then blends with the axis of the coordinates. I will refer here to some examples borrowed from the interesting work of M. Candlot upon Portland cement. The numbers of the table give the tensile strength in kilograms of bri- quettes of 16 square centimeters cross-section: 7 13 6 Days. Month. Months. Months. (1) Cement with excess of lime 350 650 606 506 (2) Cement incompletely calcined. .. 568 712 609 412 (3) Lumpy powders of cement.... 295 410 510 560 We see that the tests at one month, the only ones really practicable, give an enormous superiority to cements con- taining free lime, which, however, are so dangerous in use : the lumpy powders, coming from the spontaneous pulveri- zation of slightly calcareous cements, the set of which is slow, but which with time acquire very great hardness, would be placed far behind. In my opinion, the mechanical tests ought to be man- aged differently. It would be necessary to conduct them in such a way as to bring into relief the causes of destruction, by exaggerating these latter by every possible means. The raising of the temperature allows the desired results to be easily obtained. The strengths obtained at higher tem- peratures are certainly not equal to those which have been obtained at ordinary temperatures by a more prolonged hardening, but, nevertheless, they seem to classify the different products in an order comparable to that which will be obtained by the tests at normal temperature, which are prolonged a very long time. In every case, this method of testing of hydraulic materials seems to give results more in accordance with the real facts than that which is CEMENTS AND HYDRAULIC LIMES 125 habitually employed. I will give, as an example, a series of experiments which I have made comparatively upon the lime of Teil, the best known lime of France, the lime of Pavier, known as of good quality, and upon limes of second grade coming from different factories of a particular province : Normal Test. Test at 80°. 1st 3rd 6th 1st week. week. week. week. Swelling. Strength. Swellii Kilograms. Kg. Lime of Teil. .12.5 19.5 39 None 69.0 None Lime of Pavier 3.6 6.5 17 None 48.0 None 1 19.5 33.0 51 Slight 30.0 15% 2 .. 8.0 13.0 33 Slight 7.5 30% 3 16.5 21.0 27 Slight 15.0 30% These numbers express the crushing strength per square centimeter of small cylinders 2 centimeters high and 2 centimeters in diameter. We see from this table that the tests made in the cold give a great superiority to limes of inferior quality. This initial excess of strength results wholly from lack of slak- ing, that is, from careless manufacture. The hot tests, on the contrary, immediately place these different products in their respective places, and show, by a considerable swelling, the lack of slaking in the limes of second quality. In order to make these tests, the briquettes are left to set for twenty-four hours at the ordinary temperature and are then immersed in a water bath heated to a temperature near to 100° C. It is necessary to avoid placing them in boiling water, which, by its agitation, would disintegrate those briquettes whose set is slow. The result obtained in this case would be more complex, being a function both of the quickness of the set and the final hardness. Slow set- ting limes, like those of Teil, would be completely disin- tegrated, whereas, very mediocre Roman cements would 126 CONSTITUTION OF HYDRAULIC MORTARS resist this action. Forty-eight hours' heating in hot water seems to suffice for obtaining the larger part of the effects of swelling, and consequently, the integral disintegration which it is purposed to make evident. At the end of seven days the maximum strength seems to be reached. These hot water tests might be completed by a desicca- tion at 100° C. in dry air, which brings about the destruc- tion of the slightly stable hydrates. By this mode of com- parison, we assign again to their true place certain products whose cold tests made at normal temperatures, and after too short a time after taking set, we might assign a different rank. Certain Roman cements, such as those of Vassy, the inferiority of which, as compared with Portland cement, no one doubts, may, however, occasionally give higher figures for tensile strengths in the tests made at one month. If, on the contrary, before breaking the briquettes they are put after desiccation into air heated at 100°, we see that the Roman cements lose half of their strength, whereas, the Portland cements are unchanged. Test of the strength after seven days of setting : Test of the strength after seven days of setting. Briquettes. Not heated. Heated to 100°. Roman cement from Vassy 630,650,750 400,325,340 Cement from Teil 480 460,430 Portland cement 625 640 It seems to me that similar tests combined with measur- ing the quickness of the set, a fact equally useful to know, would allow the value of a cement to be judged much more exactly than the actual methods of testing do, but I am obliged to add immediately that my experiments upon this subject are not sufficiently numerous to allow, at present, the absolute accuracy of this mode of classification of hy- CEMENTS AND HYDRAULIC LIMES 127 draulic products to be affirmed. Numerous experiments ought still to be made before it could be accepted in prac- tice. Therefore, no manner of testing exists meriting any confidence, the only means of judging of the quality of hydraulic materials is of assuring one's self of the identity of the product considered with a similar product having the same origin, and used successfully many years before.* From this particular point of view, the object of the tests ought to be merely that of establishing this identity : the re- sults will be the more certain, as the identity of a greater number of different characteristics have been established: chemical composition, heat of solution, density, strength, &c. If it is desired to make a choice, it will be necessary * This way of looking upon the tests by reducing them to a simple control of identification, is extremely hurtful to the progress of the industry of the hydraulic products : there is more and more a ten- dency at present, and with reason, to abandon this point of view. If still we have not succeeded in denning the tests of quality sufficient for an exact classification of all the hydraulic products, in view of all their applications, still some progress has been made in this direction. The hot water or vapor tests give an absolute security against the presence of the expansives, lime or magnesia. (H. Le Chatelier. Sur I'essai a chaud des produits hydrauliques, Bulletin de la Societe d' encouragement pour Vindustrie nationale, novembre, 1898). The mechanical 28-day tests give at least for Portland cements sufficient indications of quality, because with the fine grinding used at the present day, these cements have, after a month's set, almost the maximum strength. The same is true for slag cements. On the contrary, they are wholly insufficient for Roman cements, the cements from coarse powders and the hydraulic limes. For these two last classes of products the tests of hardening in boiling water would doubtless be employed very advantageously. Still, the satisfactory test for controlling the resistance of a cement to the decomposing action of sea water does not exist. It seems, however, that a satisfactory solution will be reached shortly. Possibly it will be sufficient to mix the cement tested with a given quantity of gypsum, from 20 per cent, to 50 per cent., and to follow the decomposition of the briquettes, or their expansion measured with a rather precise apparatus, or yet their mechanical strength. 128 CONSTITUTION OF HYDRAULIC MORTARS to try for characteristics which allow the greatest differen- tiations between similar products to be established, that is to say, to those which are susceptible of the greatest varia- tions, and, at the same time, of the most precise measure- ment. It is not necessary from this particular point of view to apply ourself to the characters having any relation whatever with the necessary qualities for the use of a cement. APPENDIX. 1 gram (<7)=15.4322 grains. 1 kilogram (kg.) = 1000 grams = 2.2046 pounds. To convert kilograms per square centimeter into pounds per square inch, multiply by 14.223. To convert degrees Centigrade into degrees Fahrenheit, %C° + 32 = F°. INDEX. A PAGE Adherence 22, 23, 102 Aluminates (see under barium and calcium) Analyses, barium hydrate 30 boiler scale — from sea water . . 9 clinkers, good and bad 122 "hexagonal plates" 83 hydrous mono-barium silicate . 27 hydrous mono-calcium alumi- nate 62, 67 plaster, crystallized 9 Analysis, proximate 42, 122 proximate methods 76, 123 ultimate 122 B Barium aluminate 62 cement 24 cements 30 hydrate 30 silicate, di-barium, properties of 25 decomposition of 29 set of 30 mono - barium, properties of 24 decomposition of 29 mono -barium, hydrous, analysis of 27 crystalline characteristics and reactions 28 r61e in setting 31 silicates, anhydrous 24, 25, 30 hydrous 25, 31 Bergmann — Cause of hydraulicity. 39 Berthelot — Law of the decomposi- tion of basic salts by water 29 Berthier water content of plaster . . 3 confirms Vicat's researches ... 42 formula for silicates 59 C Calcaro-magnesian aluminate 68 Calcium aluminate 62,66,67,72,91,92,93 conclusions on 81 expansive action with calcium sulphate Ill hydration of 100 as flux. 116 di- calcium, preparations and properties 63 hydrated 65 84 [129] PAGE dehydration of 113 micro-photograph of 86 mono - calcium, preparations and properties 63 quadri-calcium 65 tri-calcium 67, 68, 91, 121 preparations and properties . 64 sesqui-calcium 63 micro-photograph of 86 alumino-ferrite 69 carbonate, decomposition of . 33, 34 dissociation of 36 hydrous 120 chlor-aluminate 68 chloride, as a flux 55 chlor-ferrite 70 ferrite, conclusions on 72 hydrated 69 slaking of 100 ferrites 68, 69, 93 in Portland cement 86 fluoride, as a flux 56 hydrate, properties 49 analysis 83 contraction on tempering . 113 oxide 47 peridote 50, 52 salts, synthetic study of 47 silicate, chlor- 55, 57 di-calcium 56, 75, 91, 92 dusting of 78, 79, 93 preparation and proper- ties of 50 r61e in hardening 55 hydrated 84 decomposition of, by water 59 preparation and proper- ties of 58 micro-photograph of . . 86 mono-calcium 79, 91, 92 preparation and proper- ties of 50 multiple 71, 79, 81, 86 tri- calcium 72, 78, 79, 91, 92, 96, 121 conclusions on 55 decomposition of by water 85 hydration of 100 micro-photograph of . 86 silicates, anhydrous 50 basic 71 130 INDEX simple 50 sulphate 116 .expansive action of. .111-113 in Roman cements 94 solubility in water 14 (see under plaster and gypsum.) sulpho-aluminate .68, 113, 114 Candlot's experiments 68, 69, 94, 111, 124 Carbonation 117, 118 Cement, Barium 24, 30, 31 Cements 38 anhydrous 73 conclusions on 81 hydrated 81 micro-chemical study of 73 Portland 38 quick setting 38, 93 Roman 38, 93, 113, 118, 128 slag. . . 128 slow setting a° Collet-Descotils', experiments 40 Conclusions on barium cements ... 31 cements . . 81 compounds in cement 72 crystallizations and supersatu- ration 20, 21 Fremy's work 45 hardening 21,85, 90 heating of gypsum 7 microscopic study 74, 76 r61e of iron in setting of Port- land cements. 85 supersaturated solutions 20 temperature of calcination of plaster • • 1 1 Constitution of mortars, historical resume' of theory of 39 Couple — graduation of thermo-elec- tric 35 Crystallization, conclusions on.. 20, 21 of plaster 13 Crystallized potassium sulphate. . . 53 Crystallographic determination of calcium chlor-silicate 56 chlor-calcium ferrite. . . 70 hydrous mono-barium silicate . 28 D Debray, experiments of 33 on saline hydrates 10 de Coppet, on supersaturated solu- tions 16 Dehydration by dry air 113 Deval, experiments of 68 Deville, experiments on aluminates, 62, 72 Dolomite 48 Dimorphic lead oxide 54 mercuric iodide 37 Dimorphism and disintegration ... 53 Dissociation of calcium carbonate . 36, 37, 38 temperature of calcium hydrate 57 magnesium carbonate 34 PAGE Ditte, experiments on supersatur- ated solutions 18 E Expansives, calcium sulphate . .111, 113 Lime • • • 127 F Fremy, experiments of . . : . . . .44, 57, 62 Friedel, experiments with chlor- calcium silicate 68 Fizeau experiments in false crystal forms 14 G Gehlenite 71, 88 Gernez, experiments on supersatu- ration 16 Glasses 91,92,93 r61e of silicious 58 Grappiers 55, 76 analysis of 77 black ' 97 constituents of ... 77 of cements, analysis of 78 Gypsum, calcination of 3,4 conclusions on heating 7 loss of water at different tem- peratures 7 Lacauchie's experiments on .. . 11 law of heating of 4 Le Chatelier's experiments . . . 3, 4, 7 temperature of calcination .... 3 Van't Hoff's experiments on dissociation temperature of . 11 Guyton de Morveau, experiments of 40,58 H Hardening 97 conclusion on, cement 85 plaster 21 different modes of .• • • • ^2 of air mortars by carbonation . 33 dessication. . 32 of hydraulic materials 99, 101 of plaster 13, 21 Heat of dissociation of calcium car- bonate 37 of formation of calcium car- bonate 37 hydrous mono-barium silicate 28 of solution of calcium chlo- silicate in HC1 56 Hexagonal plates 82 analysis of 83 Hot water test 126 Hydration and swelling 113 of plaster • ■ • 13 Hydrated salts (see under calcium barium, etc.) Hydrates, destruction of by efflor- gscghcg * ■ 111 Hydraulic limes . 38, 39, 94, 95 slaking of 98 INDEX 131 PAGE Hydraulic mortars 32, 38 constitution of 39 products, testing of 120 I Idocrase 71 Iron, r61e of in the setting of port- land cements 85 L Lacauchie's experiments on gyp- sum. 16 Landrin's experiments on cements, formula 42, 59 theory of set 45, 50 plaster, theory of set 14 water content 3 Lavoisier, experiments on plaster . . 9,12,13 Le Chatelier's limiting equations. . . 88 temperature of calcination of plaster 3 thermo-electric couple 33 Lime 47 calcination of 33 compounds, (see under calcium.) free, action of 57 , determination of 42, 122 in portland cement 81 mortars 38 Limes (see hydraulic limes.) Limit limes 48, 97 Litharge. 54 M Magnesia as a flux 116 in calculating mixtures 88 Magnesian aluminate, calcaro- .... 68 Magnesium carbonate dissociation, temperature of 34 determination of crystal forms 26, 56, 62 Mallard dimorphic transformations. 53 Marignac, experiments of 14, 19 Massicot 54 Maynard Ill Mechanical tests 123, 124 at high temperatures. . . 124, 125 Mellilite 71 Merceron, theory of the set of plas- ter 45 Mercuric iodide 26 Micro-chemical study of cements . . 73 Microscopic examination of reac- tions 84 Monticellite 52 Mortars, causes of destruction of . . . Ill destruction of by efflorescence . Ill magnesium salts Ill solution Ill historical resume" of theories of constitution of 39 air 32 fat lime 38 hydraulic 32, 38 researches in constitution of . . 39 N PAGE Newberry, experiments 58 P ...... Payen, temperature of calcination of plaster 3 Pelouze on hydrated calcium ferrite 69 Pisani on hydrated barium silicate . 29 Plaster, calcination of 9 calcined, a definite hydrate ... 9 conclusions on temperature of calcination of 11 crystallization of 8, 13, 15 dissociation, temperature of . . 11 Lavoisier's experiments 2 Landrin's theory of set 14 the setting of 12, 13 temperature of calcination of. 3 water content of 3 Portland cement, calcination of . . . 90 manufacture of 86 Proximate analysis 122 methods 123 limits of composition of 87 uses of 86 Pseudo-cubic crystals 74, 76, 78 Pseudo-crystal forms 14 Puzzolana mortars 39 Puzzo-Portland of Landrin 45, 52 Puzzo-silicates 58 Q Quick setting cements 93 Quick-lime 80, 91, 93 R Regnault cited 4 Rivot, formula of calcium silicate . . 42, 59, 33 free lime in cements 80, 122 Roman cements 115 free lime not dangerous 93 calcium aluminate swell 113 strength 126 S Saussure, experiment on Swiss limestones ' 40 Sea water, decomposition of cement by 68 Set of hydraulic material 99, 100 Silicate, plaster, multiple 12, 13, 86 Silicate acid, in cements 71 double 91. 92 multiples of aluminum and lime 71,79,81 (See under calcium barium.) Silicious glasses, role of 58 Silico-aluminates 72 Slag cements 128 Smeaton, quoted 39 Sorel, on supersaturated solutions . 17 Solubility of gypsum and plaster, relative 19 Spontaneous pulverization . . .51, 52, 79 Spring, on false crystalline forms . . 14 INDEX 132 PAGE Supersaturated solutions 17, 18 conclusions on 20 Supersaturation, role of 26 T Theory of the constitution of mor- tars, historical resume 1 of 39 Hardening of mortar Landrin's 45 Thermo-electric couple of Le Cha- telier 33 Test, hot water 126 Testing of hydraulic products 120 Tests, mechanical 123, 124 at high temperatures 124, 125 U PAGE Ultimate analysis 122 V Van't Hoff, on dissociation temper- ature of gypsum 11 Vicat, experiments on air mortars . 32 hydraulic limes 39, 41 sulphates 48, 97 magnesia 111,117 W WoUastonite, 50, 52, 54, 58, 77, 79, 93, 97 ^ Date Due