THE CONSTITUTION OF PORTLAND CEMENT By F. M. LEA O.B.E. D.Sc. F.R.I.C. (DIRECTOR OF BUILDING RESEARCH DEPARTMENT OF SCIENTIFIC AND INDUSTRIAL RESEARCH) THE complex anhydrous silicate and aluminate products which are exempli- fied by the igneous rocks and certain industrial materials have been the subject of much investigation and none more intensively than Portland cement an industrial product of first importance. Studies in these fields of high-temperature chemistry demand specialised techniques and progress has run parallel with the development of modern physicochemical methods in general and of new techniques appropriate to the subject in particular. The study of the constitution of Portland cement has had as one of its objects the determination of the nature of the compounds or mineral species it contains and the origin of its cementing properties.Another purpose has been the relation of the constitution of the product to its composition and conditions of manufacture and to its mode of hydration and technical properties. It would be arbitrary to attempt to draw any distinction between those studies which have been pursued because of their scientific interest per se and those which have been followed because of their technical importance but the present Review will be confined largely to the chemical and mineralogical aspects of the subject. Portland cement is manufactured by burning a mixture of suitable finely ground calcareous and siliceous materials such as limestone chalk clay and shale. The large rotary kilns normally used for burning reach an extreme length of some 500 ft.and are slightly inclined to the horizontal so that as the kiln rotates the raw materials fed in a t the upper end travel slowly down to the lower end from which the kiln is fired and where the burnt product is discharged. The fuel used is most commonly pulverised coal blown in by an air blast but oil or gas is also used when economic. The raw materials as they progress down the kiln are first dried followed by dissociation of the calcium carbonate and the commencement of solid reactions between the lime and the clay. At about 1280" partial melting commences and the reactions accelerate and are finally completed in the hottest zone at a tem- perature of 1350-1500" when some 20-30% of the mix becomes liquid. This causes the materials to coalesce into small nodules mostly from 8 to 9 in.in diameter known as Portland cement clinker. The latter is then ground to a fine powder with a small addition of gypsum to control the rate at which the material sets on mixing with water. Though lime alumina ferric oxide and silica are the major components of Portland cement there are also present minor constituents which exercise a considerable influence on the constitution of the product. The com- position of different clinkers falls within the following percentage range CaO 60-67 ; SiO, 17-25 ; Al,O, 3-8 ; FezO3 0.5-6 ; MgO 0.5-5.5 ; Na,O + K,O 0-4-1.3 ; TiO, 0.1-0.4 ; SO, 0.1-0-5. Contents of 82 LEA TEtE CONSTITUTZON OF PORTLAND CEMENT 83 Mn,O varying from 0.5 to 3.0% are also found when certain types of blast- furnace slag are used as raw material for cement manufacture.The investigation of the constitution of Portland cement involves the determination of the manner in which these various oxides are combined a t the temperature of burning and the changes which occur during cooling. Despite the limited amount of melting which takes place there is good evidence that Portland cement mixes approach closely a state of chemical equilibrium a t the burning temperature apart from a small residue of uncombined lime which has not had sufficient time to react during the burning. On cooling however the equilibrium is not fully maintained owing to the sluggishness of some reactions and crystal inversiom and to the failure in varying degree of the liquid phase to crystallise. Essentially therefore Portland cement is to be regarded as an example of frozen equili- brium reproducing in its cooled state the equilibrium existing a t or near the burning temperature but modified in some degree by partial change towards the true equilibrium at room temperature.The more important compounds formed namely 3CaO,SiO, 2CaO,Si0, 3CaQ,Al,Q3 and 4Ca0,A1203,Fe20, are modified to a small extent by solid solution with other components and the first two exhibit polymorphic transformations. The compounds formed by the minor components add their own complications. Methods of experimental investigation have been needed which would determine the equilibria existing a t the burning temperature enable the changes occurring on cooling to be traced and permit the identification of particular mineral species in the cooled clinker. The study of phase equilibria has been one essential line of attack using methods common to the wider sphere of mineralogy.Such investigations on silicate systems at high temperatures cannot be carried out by the method of cooling curves used in the study of meta,ls since crystallisation and inversion are too sluggish. The alternative method of heating curves can often be very useful for checking inversion temperatures of compounds exhibiting polymorphism and the points of initial and complete melting but taken alone the interpretation of the results is frequently uncertain. The primary tool and an elegant one is the quench met'hod. In this a small charge 0.1 g. or less of the desired oxide composition is placed in a small platinum bucket suspended by a fine platinum wire and heated in a vertical furnace at a constant temperature until equilibrium is attainecl.The supporting wire is then fused and the charge allowed to drop into water or mercury to quench the equilibrium. This freezes any liquid that was present a t the furnace temperature to a glass in which are embedded crystals of the solid-phase compounds. By examination under the petrographic microscope of powdered samples from a series of preparations of varying composition the temperatures of final melting and of the appearance of different crystalline phases can be determined and the phase equilibrium diagram a complete polytherm of the Bystem built up. For temperatures up to 1650" the furnaces used are normally wound with Pt-20% Rh wire and automatically controlled to maintain a temperature constant to & 1" over a small central section in which the charge is held.Syatems of four, 84 QUARTERLY REVIIEWS FIa. 1 System CaO-A1,O ,-SiO,. Abbreviated Notation ; C = CaO A == Al,O, S = SiO,. FIG. 2 System C'aO-2Ca0,Si0z-5Ca0,3A1z03-4Ca0,A1203,Fe,0,. Abbreviated Notation C = CaO A = A1,0, F = Fez03 S = SiO,. LEA THE CONSTITUTION OF PORTLAND CEMENT 85 and to a limited extent even five components have been examined by this means and much of the remainder of the nine-component system involved in Portland cement studied by selection of suitable “ cross-sections ” of the whole. The phase-equilibrium diagrams illustrated later (Figs. 1 and 2) are projections of the polytherm on to the composition base an equilateral triangle in the case of ternary systems and a regular tetrahedron in that of quaternary systems.The method of plotting is therefore different from the representation of solubility curves at some fixed temperature commonly used for aqueous systems. Instead the diagrams show the composition regions or primary phase fields in which any one solid is the first to separate when a completely liquid mix is allowed to cool. With the aid of tempera- ture contours the course of crystallisation can be followed and the proportion and composition of the solid and liquid phases present a t any desired tem- perature ascertained. The effect of deviations from the true equilibrium path during cooling can also be estimated. X-Ray methods have been considerably used for the determination of phase-equilibrium diagrams for nietallic systems but their scope is more limited in silicate systems owing to the low symmetry of many of the com- pounds that arise and the consequeiit lack of sensitivity in the detection of the first appearance of particular solid phases.There still remains however a wide field of use for the detection of individual compounds when present in sufficient quantity in mixes for the study of inversion phenomena with high temperature cameras and fGr structure analysis. Work of the last type has until recently been handicapped by the small size of the crystals obtainable from some of the most important compounds restricting the examination to powder preparations. Recently however R. W. Nurse at the Building Research Station has developed a technique for growing single crystals from silicate melts and crystals of for example pure SCaO,SiO and its solid solutions as large as 0.5 mm.in diameter have been obtained. Various microscopic techriiques are used. Individual mineral con- stituents can be identified in powder preparations by measurement under the microscope of optical properties. Thin sections of synthetic silicate melts can be prepared for examination by transmitted light in the same way as in the petrographic study of rocks. The met,allurgical technique of polishing and etching surfaces for examination by reflected light is also applicable though rather more difficult than with metals and has proved an invaluable weapon for the study of the miiieralogy of cements. By the use of selective etches different mineral species can be brought into prominence identified and estimated. Other measurements such as microreflectivity have proved their usefulness in some cases.As a still further step in technique it has proved possible though more difficult to 21st Congress Ind. Chem. Brussels 1948. * See 33. H. Bogue “ Chemistry of Portland Cement ” Rheinhold Publishing Corporation U.S.A. 1947 ; F. M. Lea and C. €I. Desch ‘‘ Chemistry of Cement and Concrete ” Edward Arnold London 1935, 86 QUARTERLY REVIEWS polish and etch thin sections which can be examined by both reflected and transmitted light. In general it is difficult to identify interstitial material by examinahion under transmitted light but a very clear differentiation can often be obtained by reflected light with selective etching. A comparison of Figs. 3 and 4 will illmtrate this. The normal methods of mineral separ- ation by sedimentation in heavy liquids or centrifuging also find application.Electron microscopy has so far found its greatest use in the study of the hydration of cements where the grain size of the products is often extremely small. Special interest has attached to the presence of a glass phase in Portland cement since this has some very significant effects on the technical properties of the material in use. Its coinposition can be determined approximately from the phase-equilibrium data and its quantity estimated again approxi- mately by microscopic counts on polished surfaces or by a heat of solution method. In the latterY3 the heat of solution of the cement clinker before and after annealing is determined in a nitric-hydrofluoric acid solvent and the glass content estimated from data on the heat of crystallisation obtained in a similar manner.Two of the individual constituents of Portland cement clinker free calcium oxide and free magnesia,5 the mineral periclase can be determined by direct chemical methods. The quantity of the former is of interest since it indicates the extent to which complete equilibrium was not obtained during burning and both are important in relation to the volume stability of the set and hardened cement. The study of the constitution of Portland cement commenced with Le Chatelier who is regarded as the father of cement chemistry. His first contributiog appeared in 1883 and his thesis for the degree of Doctor of Science published in Paris in 1887 and in an English translation,* remains a classic. The basic system CaO-Al,O,-SiO necessary to the understanding of Portland cement was the first ternary silicate system studied at the Geophysical Laboratory Washington D.C.U.S.A. and its publication by R. W. Rankin and F. E. Wright 7 in 1915 is a landmark in the study of the natural and artificial silicates. It laid the method of approach and the basis of many of the high-temperature techniques from which later developments have been derived. The equilibrium diagram of the system Ca0-A1,0,-Si02 is reproduced in Fig. 1. The field of Portland cement compositions is such that a t equili- brium the compounds 3Ca0,Si02 2CaO,SiO, and 3CaO,Al,O are formed together with free CaO if more than a certain proportion of lime is present. A brief inspection of the diagram will show however that a t a temperature of say 1500" the liquid in equilibrium with 3CaO,SiO as solid phase or with a W.Lerch and L. T. Brownmiller J . Res. Nat. Bur. Stand. 1937 18 609. 4 W. Lerch and R. H. Bogue Ind. Eng. Chern. Anal. 1930,2,296 ; B. Bakewell and G. E. Bessey Building Research Spooial Report NO. 17 H.M.S.O. 1931. R. H. Bogue ref. (2) p. 75. 6 " Experimental Rssssrches. on the Constitution of Hydraulic Mortars " tranalated 7 Anzer. J . Sci. 1915 33 1 ; see also J. W. Greig {bid. 1927 13 41. by J. L. Mack McGraw Publishing Co. New York 1905. FIG. 3 Portland cement clinker thin section by transmitted light ( x 350) FIG. 4 Portlarid cenieizt clinker etched polished surfciccl by reflected light ( s 600) LEA THE CONSTITUTION OF PORTLAND CEMENT 87 either CaO or 2CaO,SiO in addition has a compositioh failing on the low rime side of the composition line from 2CaO,SiO to 3CaO,Al,$.To maintain equilibrium on cooling therefore the liquid must react with solid 3Ca0,Si02 to remedy its deficiency in lime. This is one of the sluggish reactions referred to earlier which fails to occur completely a t the rate at which commercial Portland cement clinker is cooled. Evidence of its partial occurrence is however to be seen in Fig. 4 where the dark 3CaO,SiO crystals have jagged and roughened edges resulting from this reaction. Separation of a liquid magma at some stage of its crystallisation into solids which are more basic and a liquid which is more acidic or vice versa than corresponds to the final equilibrium state on cooling is not confined to the artificial silicates but is also characteristic of the igneous rocks.It is indeed typical of systems containing compounds which decompose below their melting point or which have an incongruent melting point such as the tricalcium silicate and tricalcium aluminate which occur in Portland cement. Studies on the system CaO-Al,O,-Fe,O * showed that a ternary com- pound 4CaO,A1,O3,Fe,O exists and more recently evidence has come forward for a further compound 6Ca0,2Al2O,,Fe,0 forming a complete solid solution series with it. The quaternary system Ca0-2Ca0,Si02- 5Ca0,3A1,0,*4Ca0,A1,0,,Fe203 which forms a complete system within the larger CaO-Al,O,-Fe,O,-SiO system contains within itself all the com- positions of lime alumina ferric oxide and silica which are of concern in the chemistry of Portland cement. The phase-equilibrium diagram l o of this system is shown in Fig.2. From it there can be derived the compositions of the liquids formed during the burning of Portland cement and the changes they should undergo if equilibrium is maintained on cooling. It also indicates that the compounds formed a t equilibrium from a mix of Port- land cement composition are 3CaO,SiO, 2CaO,SiO, 3CaO,Al20, and 4Ca0,A1,0,,Fe,03 (or its solid solution with 6Ca0,2A1,0,,Fe,03). A more detailed examination than is possible from the diagram further shows that with mixes of low Al,O Fe,O ratio the liquids contain excess of CaO and that if equilibrium is to be maintained on cooling reaction must occur between solid BCaO,SiO and liquid to form 3Ca0,Si02. This is the reverse of the case found in the Ca0-A120,-Si0 system but the Al,O Fe,O ratio of most Portland cements is such as t o make them behave similarly to the latter in this particular respect.Since however the liquid-solid reaction is too slow to occur to more than a limited extent during the cooling of the clinker the cooled cement represents a somewhat complex frozen equilibrium. The compounds formed by the minor components of Portland cement are still not entirely clear but much progress had been made by phase- equilibrium investigations and other means. In general they pass into the liquid phase during burning and if this fails to crystallise on cooling 8 W. C. Hansen L. T. Brownmiller and R. H. Bogue J . Amer. Chem. SOC. 1928 50 396; H. F. McMurdie J . Res. Nat. Bur. Stand. 1937 18 476. M. A. Swayze Arner. J . Sci. 1946 244 1 65. l*F. M. Lea and T.W. Parker Phil. Trans. 1934 A 234 1. * Reoent work indicates that the true composition of this compound is 12Ca0,7A1,08. 88 QUARTERLY REVIEWS most of the minor components remain in the glass formed. Study of the quaternary system l1 Ca0-Mg0-2Ca0,Si02-5Ca0,3A1203 and of part of the quinary system9 containing iron compounds in addition has shown that magnesia is soluble in the liquid phase formed during the burning of Portland cenoent to the extent of about 5%. Any excess of MgO is uncombined and present as the mineral periclase at the burning temperature and this is also formed from the liquid phase on crystallisation. The alkalis K20 and Na20 behave very differently from one another. The former combines preferentially with any sulphur trioxide present ,I2 and as molten potassium sulphate is immiscible with the clinker liquid it crystallises direct on cooling.Excess of K20 above that required for this reaction enters into solid solution in or forms a compound l3 with dicalcium silicate of the formula K20,23Ca0 12Si0,. The information regarding sodium oxide is rather less definite but it is probably capable of being present in several forms. It dissolves in the clinker liquid and the glass formed if this fails to crystallise can hold up to 8% of Na20.14 From investigations on portions of the quaternary system l4 Na20-CaO-Al,03-Si02 it seems that a compound Na20,8Ca0,3A1203 or a solid solution of this with 3Ca0,A1203 is probably formed on crystallisation of the liquid phase. Soda also enters into solid solution to a small extent in the dicalcium silicate and precipitates as inclu- sions if this inverts on cooling to the /?-form.Titania 15 may form 3Ca0,2Ti02 or CaO,TiO,. Manganic oxide l6 can substitute for Fe,03 in the compound 4Ca0,A120,,Fe203 and any Mn20 in Portland cement is probably present in it solid solution of this type. The two most important compounds of Portland cement are tri- and di-calcium silicate for they are primarily responsible for the cementing properties. Though the amounts of other components which these silicates can take up in solid solution are small they are nevertheless important for they influence the reactivity to water and also the polymorphic changes which the compounds can undergo. Tricalcium silicate is an unusual compound for it is stable l o only between 1250" and 1900" and at lower or higher temperatures decomposes to form the same products CaO and 2Ca0,Si02.Its rate of decomposition below 1250" is very slow and it remains in a metastable state indefinitely at ordinary temperatures. Only the upper decomposition point appears in the Ca0-A1203-Si02 diagram for the liquidus temperatures along the boundary between the fields of 3CaO,SiO and 3Ca0,A120 range from 1455" to 1470". By the addition of miiieralisers however the liquidus temperatures can be much reduced and the field of 3CaO,SiO then tapers to a point a t both ends. This is illustrated by the phase diagram obtained by W. Eitel17 for the system Ca0-2CaO7Si0,-CaF which is reproduced in Fig. 5. If the lower decomposition point of tricalcium silicate were another 100-200 O higher l 1 H. F. McMurdie and H. Insley J .Res. Nat. Bur. Xtand. 1936 16 467. 1 2 W. C. Taylor ibid. 1942 29 437 1 4 K. T. Green and R. H. Bogue ibid. 1946 36 187. l6 F. M. Lea and R. W. Nurse unpublished l6 T W. Parker private communication lS Idem ibid. 1941 27 311. 1' Zement 1938 27 455 469. LEA THE CONSTITUTION OF PORTLAND CEMENT 89 its formation in Portland cement would be difficult or impossible and Portland cements with the speed of hardening known today would not exist. There has been difficulty in determining the extent to which 3CaO,SiO can take other components into solid solution but as formed in Portland cement it may contain up to 3% of A1,03 Fe,03 CaO and MgO. Tricalcium silicate is also a constituent of open-hearth slags and crystals from this source have been found to contain some 3 % of FeO and MnO.Very recently Bernal and Jeffery have carried out an X-ray analysis on single crystals of pure tricalcium silicate grown at the Building Research Station by the technique developed by Nurse and on similar solid solution crystals. The pure compound is found to exist in two polymorphic forms a high-tempera- ture trigond form and a lower-temperature triclinic form. Th2 inversion CaF 2572" 1388 FIG. 5 Sysfern CaO-ZCaO,SiO ,-CaF,. temperature is still uncertain and may be above 1600" for the pure com- pound but lower for the solid solutions. The minerals found in Portland cement or basic open-hearth slags all have the trigonal form which is apparently transformed less readily when containing limited amounts of other oxides in solid solution. Dicalcium silicate has long been known to exist in three enantiotropic forms a $ and y of which the first two have cementing properties while the y-form is very inert to water and has no cementitious value.The ,8 + y inversion which occurs at 675" is accompanied by a volume increase of some loyo resulting in the break-up of the solid into a fine powder. This is the cause of the " dusting " during cooling or subsequently of some types of blast-furnace slag which contain dicalcium silicate. A similar behaviour is sometimes found to occur with Portland cement clinker stored for a 90 QUARTERLY REVIEWS considerable time. With the pure compound the inversion occurs readily and indeed it is diacult to cool it sufficiently rapidly to stabilise the ,&form but with quite small amounts of other components in solid solution the inversion becomes very sluggish and the p-form can then be preserved indefinitely.Both a- and P-dicalcium silicate can take small amounts of numerous other oxides such as SiO, Al,O, MgO Na,O B,O, or P,O into solid solution with a lowering of the inversion temperature indicating a greater solubility in the a- than in the /3-form. With CaO the inversion temperature is raised so that the solubility must be greater in the /I-fGrm. Because of the influence of SiO and CaO in respectively lowering and raising the inversion temperature its precise value is in doubt but it appears to be either 1438" or 1456". The a-form has long been supposed t o be monoclinic or triclinic and the /?-form orthorhombic but in 1942 M. A. Bredig l9 published the first of a series of papers on the structure of compounds of the A2XQ type.These compounds e.g. K,SQ4 Na2SO4 CaNaP04 all occur in more than one modification but in their high-temperature (a) form they are isomorphous and have hexagonal symmetry. Bredig therefore suggested that a-Ca,SiO was hexagonal and he was also led to postulate the existence of an ='-form because the compounds or solid solutions K20,23Ca0 12Si0 and 27Ca0,P,05,1 2Si0 have a structure corresponding to D-K,SO with a symmetry lower than a but different from @-Ca,SiO,. This "-form was considered to have a range of stability between the a- and the ,%form. Recently C. E. Tilley and H. C. G. Vincent 2o have found the a'-form in metamorphised limestone from Scawb Hill and in spiegeleisen slag and have designated this new phase bredigite. Confirmation of the hexagonal spmmetry of a-Ca,,Si04 has been obtained by A.V. Van Valkenburg and H. F. McMurdie 21 from an X-ray powder diffraction pattern taken in a high- temperature camera a t 1500" using a Geiger counter in place of a photo- graphic film. A solid solution of Cs,,SiO containing 10% of CaNaPO prepared at the Building Research Station has also recently been found to have the a-structure at 1350". From other recent evidence obtained by Nurse a t the Building Research Station the a'-form is probably stable only a t the highest temperatures inverting to a as the temperature is lowered. For the pure mineral the inversion temperature probably lies between 1650" and 1750" but is lowered by solid solution. The attention that has been paid to the crystal form of tri- and di-calcium silicates and also of other cement compounds reflects the interest in the origin of their reactivity and cementing properties.Both E. Branden- berger 22 and Bredig l9 have endeavoured to relate with quite different results the variation in reactivity to differences in the co-ordination number 18 E. S. Newman and L. S. Wells J . Res. Nat. Bur. Stand. 1946 36 137 ; K. T. Greene ibid. 1944 32 1. 19 J . Physical Chem. 1942,46,747 801 ; 1943 47 587 ; 1945 49 537 ; Amer. Min. 1943 28 594. ao Min. Mag. 1948 28 255. 2% Schweiz. Arch. Angew. Wiss. Tech. 1936 2 45 ; Symposium on the Chemistry of J . Res. Nat. Bur. Stand. 1947 38 415. Cements Stockholm 1938 pp. 122 174. LEA THE CONSTITUTION OF PORTLAND CEMENT 91 of the calcium atoms and the part these play in the crystal structure ; in the absence of complete structural analysis no generally accepted conclusions have yet been reached.With the recent development of a technique for growing single crystals large enough for X-ray analysis further progress now becomes possible. Data on the heats of formation of the cement compounds from the oxides CaO Al,O, SiO, and Fe203 are available and also the heats of hydration. The data are shown in the table but their degree of precision varies. Thus the heats of formation of the silicates are in doubt to & 0-7 kcal. per mol. and those of the alumina-containing compounds to & 1 kcal. per mol. owing to uncertainties in the heat of crystallisation of amorphous silica and of the heat of the reaction 2Al(s) + SO2(& = Al,O,(s) the values of which enter into the respective computations.The formation of SCaO,SiO from #?-2Ca0,Si02 and CaO it will be noted is endothermic. Compound. 3Ca0,Si02 . . . . /3-2Ca0,Si02 . . . y-BCaO,SiO . . . 3Ca0,A120 . . . 4Ca0,A1,0,,Fe20 . CaO. . . . . . MgO . . . . . Heat of format,ion from oxides at 20". Kcal per mol. 29.4 29.8 30.9 2.0 10.0 - - Cal. per g. 129 173 179 7 21 - - Heat of hydration at 20'. Kcal. per mol. 27.4 10.7 57.8 48.6 15.7 8.2 - Cal. per g. 120 62 214 100 279 203 I From the detailed information now available on the constitution of Portland cement it is possible to calculate its approximate compound composition on certain assumptions to relate this to the properties of cements and to check it by microscopic measurement on etched polished surfaces of the cemBnt clinker. A simple calculation of the main compounds can be made fiom the oxide analysis assuming that the cement has crystallised completely and is in a state of equilibrium.For different types of Portland cement the content of 3CaO,SiO varies from 10 to 65% of 2Ca0,Si02 from 5 to GOYo of 3CaO,&O3 from 3 to 15y0 and of 4Ca0,Al,03,Fe,03 from 5 to 15% except for white Portland cements where it falls to below 2%. The content of uncombined lime varies from almost zero to a few units yo depending on the hardness of burning. Alternatively other assumptions may be made as to the degree of crystallisation of the clinker liquid and this can be checked by determination of the glass content of the clinker and the subsequent calcidation of cornpoilnd content modified in various ways. Portland cement clinkers are found to vary very considerably in their contents of glass the values ranging from almost zero up to 20% or more.The contents of alumina and iron compounds decrease as the glass increases since they are found during the crystallisation of the clinker liquid. The properties of Portland cements are closely related to the compound content but are affected also by the fineness of grinding and other factors such as the proportion of water used to mix them. I n general tricalcium 92 QUARTERLY REVIEWS silicate makes the major contribution to the strength developed over the first 28 days after mixing with water while the contribution made by dicalcium silicate becomes important between 7 and 28 days and practically equals that of tricalcium silicate wit,hin 90 days. Perhaps the major problem remaining here is the influence of solid solution of minor com- ponents and of crystal form on the reactivity and rate of strength develop- ment.There are also other technical properties such as the heat of hydration which becomes of major importance when large masses of concrete are placed as in dams and the volume changes which the set cement under- goes with change in moisture content that can be rela,ted to the mineralogica.1 make-up of the cement and the nature of its hydration products. The hydrated compounds formed from Portland cement present a still more complicated problem 2$ 23 than that of its constitution and only the broadest outline can be given here. The aqueous systems involved in the determination of final equilibria are far from easy to study and demand all the resources which modern physicochemical techniques can bring to bear for solubilities are low and metastable equilibria common.The systems CaO-Al,O,-H,O 24 and CaO-Si0,-H,O 2 5 are now reasonably well though not entirely known and also the quaternary system 26 Ca0-A1,0,-S03-H,0 which is implicated because of the addition of gypsum to Portland cement to control its setting time. The quinary systems 27 formed by the addition of Na,O or K,O to CaO-A1,0,-SO3-H,O and the quaternary 28 CaO-Na,O- Si0,-H,O have also been studied and to some extent the quaternary system Ca0-A1,0,-Si0,-H20 and the systems involving Fe203. It is perhaps characteristic of the relative difficulty of their study that the information on the equilibria in many of these systems at temperatures above 100" under high steam pressure is more precise than that at normal temperatures.With the former well-crystallised products are obtained for example 3Ca0,Si0,,2H20 from tricalcium silicate and several dicalcium silicate hydrates from BCaO,SiO, but with the latter amorphous or sub-micro- crystalline materials are frequently encountered. It is in this field that the introduction of the electron microscope promises to be of considerable help. The reaction of SCaO,SiO with water produces a supersaturated solution of the same stoicheiometric proportions from which Ca(OH) and a hydrated calcium silicate separate while with P-ZCaO,SiO in a limited amount of water only traces of calcium hydroxide appear. The hydrated silicate is an apparently amorphous product but G. E. Bessey 23 has obtained micro- scopically visible crystals from ,f?-2Ca0,Si02 after hydration for a year.The formula ascribed to the compound by R. Hedin 29 is 2Ca0,Si0,,4H20. Other evidence 2 from phase-equilibria studies indicates that at. a pH value 2 3 Symposium on the Chemistry of Cements Stockholm 1938. 2 4 L. S. Wells W. F. Clarke and H. F. McMurdie J . Res. Nut. Bur. Stand. 1943 2 6 F. E. Jones Trans. Faruday Soc. 1939 35 1484 ; J . Physical Ghem. 1944 48 27 Idem ibid. 1945 48 356 379 ; G. L. Kalousek ibid. 1945 49 405. 2s Idem J . Res. Nut. Bur. Stund. 1944 32 285. 08 Proc. Swedish Cement and Concrete Institute 1945 No. 3 30 367. 311. 2 6 H. H. Steinour Chemical Reviews 1947 40 391. LEA THE CONSTITUTION OF PORTLAND CEMENT 93 which is very close to and may be slightly above that of a saturated lime solution this hydrated disilicate hydrolyses to form 3Ca0,2Si02,aq.as another apparently amorphous phase. There is still therefore some uncertainty as to which of these silicates is formed from Portland cement. But this does not appear to be the final equilibrium product since with alumina present as well the formation of hydrated calcium aluminosilicates is possible. After hydration for three years H. P. Flint and L. S. Wells 30 have obtained a crystalline compound 3Ca0,A1,0,,3Ca0,Si0,,3Q-32H20 from pure cement compounds immersed in a saturated lime solution. The hydrated calcium aluminates and ferrites form a complex series of but the manner of the hydration of 3CaO,A1,0 and 4Ca0,A1,0,,Fe20 in Portland cement is controlled by the presence of the gypsum added. Calcium sulphate forms two double salts with tricalcium aluminate having hexagonal or pseudo-hexagonal symmetry and represented by the formulze 3Ca0,A1,0,,3CaS0,,32H20 and 3CaO,A1,O3,CaS0,,12H,0.Corresponding sulphoferrites are also known to exist. The compound 3Ca0,A1,0,,3CaS0,,32H20 is probably formed initially in the hydration of Portland cement but this is not the final end-product for it may be trans- formed 31 into a solid solution of 3CaO,A1,O3,CaS0,,12H,O with the com- pound 4Ca0,A1,0,,13H20. The latter arises from the hydration of 3CaO,A1,0 in the cement surplus to that with which the gypsum can combine. The close analogy between the calcium sulphoaluminates and aluminosilicates also suggests the possible existence of a calcium sulpho- aluminosilicate or of solid solutions. The subject is however even more complex for there also exists a compound 3Ca0,A1,0,,6H20 which belongs to the cubic system and is related in structure to the garnet series.The sulphate-containing solid solution mentioned above and possibly also the compound 4CaO,~,0,,13H2O are metastable with respect t o it at room temperature. A complete series of hydrogarnet solid solutions 32 of the general formula.? 3Ca0,A1,03 6H,0-3Ca0,A1,0,,3Si0 3Ca0,Fe,03,6H,0-3Ca0,Fo,0, 3Si0 I I are formed at high temperatures and steam pressures and there is some evidence to suggest that they may eventually be formed at normal temperatures. It will be evident that although the initial hydration products of Portland cement at normal temperatures are probably Ca(OH),,2Ca0,Si02,aq. or 3CaO,2SiO2,aq. 3CaO,A1,O3,3CaS0,,32H,O 4CaO,Al2O3,13K,O and an uncertain iron compound yet these do not represent the final chemical equilibrium in the multi-component system involved. In such systems however the final equilibria may not be attained for many years if indeed it is ever reached within the life of man. 30 J . Res. Na,t. Bur. Stand. 1944 33 471. 31F. E. Jones J . Physical Chem. 1944 48 311; 1945 49 344. H. P. Flint H. F. McMurdie and L. S . Wells J. Res. V a t . BUT. Stand. 1941 26 13.