Journal of Research of the National Bureau of Standards Vol. 49, No.6, December 1952 Research Paper 2375 Heat Generation in the Setting of Magnesium Oxychloride Cements Edwin S. Newman, John V. Gilfrich, and Lansing S. Wells

The hea ts of hardenin g of magnesium oxychloride pastes and fl ooring cements were determined by t hree methods. Two direct methods limited to 18 hours in durat ion gave heats of hardening of a bout 190 cal/g of MgO. T emperat ures up to 1470 C were obtained in the interior of t he samples. A heat-of-solut ion method, applicable only to the pastes, gave 60-day values for t he heat of hardening ranging from 135 to 320 cal!g of MgO, depending on the relative concent rations of MgO, MgCI2, and H 20 in t he pastes. A sample stored for 4 years gave a heat of hardening of 240 cal/g of MgO, as co mpared with the 60-day value of 21 5 cal /g of MgO for a d iff erent sample of similar co mposition. The low heats of hard­ ening, corresponding to high heats of solution, obtained for certain of t he pastes support t he conclusion that free MgO may remain in such pastes. 1. Introduction direct m ethods wer e used Lo deLermine the heat of har dening of the commer cial cemen ts. In one group Magnesium oxychloridc cem ent, also ealled mag­ of tests, t he heat generation was determined by nesite cemen t or Sorel eement, is used extensively for recording automatically the temperatm es attained floor sm faces. I t is formed b y the action of con­ when cemen t samples enclosed in vacuum fl asks were centr ated magnesium chloride solution on active allowed to harden. D etailed descriptions of the magnesium oxide. Numerous filler materials, such apparatus used in these tests have been published as asbestos fiber, hardwood sawdust, talc, and, and [4,5]. Weigh ed samples of the pastes were placed in powdered silica, are incorporated to impar t desired X-p t friction-top cans provided with t,hel'mocouple physical proper ties t o the finished floor. The com­ \vells. T hree-junction copp er-constan tan th ermo­ pound binding these materials together appears to piles were inser ted , and the assemblies wer e placed be t he magnesium oxychloride 3MgO.M gC12.11H20 in widemouthed I-p t vacuum fl aslcs. The fl asks [1 J.1 This compolmd is lUlstable in cont act with were corked and placed in an air bath m ain tain ed at solutions containing less than 11 p ercent of MgC l ~, 30.0° C. The temperatm es attained b y the pastes and consequently t he flo or must be protected from elm ing the ensuing 18 hl' were recorded. th e action of water . The condit ions of mixing an.d The vacuum fl asks were calibrated electrically. hardening are such that M gCl 2 solution is r etailled H eaters were built and embedded in plastic in cans in tbe interstices of t.he hardened cem ent . This identical with those used for the cement samples. solution preserves the oxychloride and r enders th e The assemblies were placed in the flasks, and the floor co nductive to electricity, a desirable proper ty power input was maintained at a constant level until in locations wher e sparks of static electricity may the t emperatm e within the cans became const ant. create a hazard. At this point the heat flow from eaeh flask was equal There appears to be little or no inform ation in th e to th e power input to that fl asl,;:. Suitable auxiliary litera tme on the amount of heat gener a ted by m ag­ resistances were arranged to enable the curren t and nesium oxychloride cemen t while hardening. Chas­ voltage across the h eat er coils to be recorded on the seven t [2] m easm ed the t emperature rise during sam e chart as the temperatures. M easurements sett ing of M gO-MgCl%-H20 mix tm es and fou nd t hat were made at two levels (besides zero) of power input the temperatm e rises of the pastes and tb eir 7-day to the flasks. compressive strengths increased together . However, The heat-transfer coeffi cients varied somewhat he gave no values for the quant ity of heat generated. with t emperat ure because of the large differen ces in The work reported in this p aper had for its primary temperature between the interior of the flasks and pmpose the determinat ion of the amount of hea t th e surrounding ail' bath. The tot al heat lost from generated by sever al commercial flooring cem ents. a cement sample was evaluated from th e time-tem­ In addit ion, similar information was obt ained con­ perature cm ve, using the m easm ed values of the cerning mixtm es of MgO and M gCl2 solution without heat-transfer coeffi cient. The sensible h eat remain­ th e fillers ordinarily used. ing in the sample, usually a sm all quant ity, was cal­ culated as the product of the final temperatm e dif­ 2. Apparatus and Procedure ference and the estimated heat capacity of the sample and container. The heat-of-solution m ethod [3] for determining A second series of tests were m ade with a conduc­ the heat of hardening is not easily applicable to com­ tion calorimeter similar to that described by Lerch mercial oxychloride cements b ecause of the large [6]. This apparatus was also calibrated electrically. proportion and inhomogen eous nature of tbe fill er Because of the more rapid flow of h eat from the materials used in their preparation. Consequently, specimen , the conduction calorimet er permitted the

1 Num bers in brackets indicate literature referenllCS at the end gf this paper. hardening of the cement to occm at lower tempera- 377 tures than did the vacuum flasks. The heat of Six samples of magnesia used for magnesium hardening was obtained from the calibration con­ oxychloride cement and for other purposes were stant and the integration of the time-temperature obtained from a producer. These samples were of difference curves. In both these methods the inte­ differing activities. In table 1 are shown some gration was extended over an 18-hr period. details of their physical and chemical properties. The heats of hardening of oxychloride pastes pre­ Magnesium chloride of analytical reagent purity was pared from MgO, MgC12.6H20, and water were de­ used to make solutions for use with these magnesia termined by a modification of the standard method samples. of determining the heat of hydration of portland cement [3]. The heats of solution of MgO and of 4. Results and Discussion the hardened pastes were determined in the vacuum­ flask calorimeter in 425 g of 2.00 N HOI. The heats 4 .1. Direct Measurements of dilution of MgC12 solutions were determined in the same quantity of acid to which 1.0 g of MgO had The compositions of the active portions of the nine previously been added. The MgO content of the commercial cements were calculated from the paste was determined by dissolving a sample in an magnesia contents and directions for mixing as given excess of standard hydrochloric acid and titrating the by the manufacturers. Figure 1 shows the composi­ excess HCl with N aOH solution. Methyl red was tions of the cement (half-shaded circles) in terms of used as an indicator. In order to prevent the pre­ MgO, H 20 , and MgC12 .6H20 . The compounds cipitation of Mg(OH)z, approximately 1 g of solid 5MgO.Mg012.12H20 and3MgO.MgC1 2.llH20, shown NH4Cl was added to the solution prior to the titra­ as closed circles, are believed to form in magnesium tion. The MgC1 2 content was calculated from the oxychloride cement. The two compounds chloride determined by titration with silver nitrate 2MgO.Mg012.6H 20 and 9MgO.MgC12.14H20 [8] are solution , using a second portion of the paste dissolved also shown. The composition of the pastes made in nitric acid solution [7]. The MgO content of the with the six samples of magnesia, 40Mg0:40MgClz,- active magnesia used in making the pastes was also 6HzO:20H20 by weight, was selected for study after determined by titration. The heats of solution were consideration of Chassevent's data [2] in the belief calculated on the basis of the MgO content. The that such pastes would develop the maximum heat heats of hardening, also expressed in calories per gram of hardening. The composition is very nearly that of MgO, were determined by subtracting the heats of the compound 5MgO.MgCI2.12H20 , the first to of solution of the hardened pastes from the heat of form in the hardening paste [2 , 9]. solution of the magnesia, corrected for the heat of The directly measured heats of hardening of the dilution of the MgC12 solution contained in the flooring cements (open circles) and of the MgO­ paste. MgC1 2-H20 pastes (closed circles) are shown in 3 . Materials figure 2 and given in table 2. Excluding magnesia sample 3, which had a very low surface area, the Nine samples of commercial magnesIUm oxy­ heats of hardening of the pastes and of the flooring chloride cement were obtained, representing the products of seven manufacturers. These samples MoO were chosen to represent the material made com­ mercially by the larger producers. Magnesium chloride solutions to be used with their product were supplied by the makers, either as solution or as flakes, MgClz.6H20, with directions for preparing solutions of the proper concentration.

TABLE 1. Properties oj magnesia samples

Sample

J 2 3 4 5 6 ------Loss at 110° C .... percent-. 0. 05 0.04 (a) 0. 62 (0) 0.12 Water in Mg(OH), percent b __ 1. 29 . 08 4. 13 6. 02 1.33 1. 44 Loss on ignition... percent.. 2. 04 . 16 6.32 8.43 1. 73 2. 42 CO, ...... percent' .. 0.70 . 04 2.19 1. 79 0.40 0.86 Bulk density...... g/mL . 41 1.38 0.55 0.39 .79 . 30 Surface ...... m'/gd .. 28. 3 0.93 83.5 86.6 28.3 31. 3 H ydration in 24 hr percente __ 80 0 58 77 46 78 Expansion. __ .. __ percent ' .. 0.19 11. 0 0.09 0. 12 0. 17 0.26 FIGUR E 1. Composition oj magnesium oxychloride pastes and cements . • Gained weigbt at 110° C. b From loss at 450° C. O. Weight·percent compositions of tbe pastes prepared for heat-of-solution , By difference. determinations; ct. composition of the cementitious material in the commercial d By nitrogen adsorption . flooring cements; • • composition of compounds tbat bave been identifi ed in the • By ignition loss of 2·g sample mixed with 200 ml of H20 and allowed to stand system MgO-MgCb-H,O. 'rhe numbers attached represent the molar propor­ 'Mixed in the weight ratio 1:19 witb portland cement. allowed to harden for tions of MgO, MgCb, and H20 that bave been assigned to the compounds. 24 hr, and autoclaved 3 br at 300 Ib/in 2. Letters near 0 identify t be samples for easy reference in the text and fi gure 6. 378 L cements were of the same order. Both sets of samples figure 2, are of the sam e brand. Presumabl.v similar showed a wider range of 18-hr heat of hardening in magnesias were used in their preparation. The four the vacuum flasks than in the condu ction calorimeter. points representing these two cements are in excellent Magnesia samples 1, 5, and 6 all had about the same agreement with the dotted lines of figure 2. This surface area (table 1), and samples 5 and 6 were agreement is to be expected because these are point~ known to be magnesias used for oxychloride cement. far enough removcd from the rest to be import.ant These three each developed 180 caljg of MgO in the in giving direction to the curves. Taken by them­ vacuum flask compared with an average of 193 selves, however, the two cements show a similar caljg of MgO for the flooring cements, and an trend. This agreement may be taken as additional average of 203 caljg of MgO in the conduction evidence for th e correlation indicated by the dotted calorimeter compared with an average of 177 caljg lines. The variation of heat of hardening with MgO of MgO for the cements. There is some indication, content will be demonstrated more fully in the see dotted lines in figure 2, that the heat generation section concerned with h eat-of-solution measure­ is greater with smaller percentages of MgO in the ments. cementitious material. However, as th e cements were products of different manufacturers, the T ABLE :2 . H eals of hardening of magnesium oxychlo,·icle magnesias used in their preparation were probably cements and pastes different, and, as the closed circles of figure 2 show, the accidental variat ion of the h eat of hardening Composition of H oat Maximum Time to reach I I cementit ious generated in temperature maximum I among different samples of . oxychloride magnesia material a 18 hours reached- temperature could account for this apparent correlation. The Sample two cements showing the higher heats of hardening, Durn - In con- Con - ber In d uc- d ue- MgCI,. VacH· Calo- tion Vacu- tio n M gO IT,O um ri m- vacu- um 6H, O um calo- fla sk calo- fla sk eter fl ask ' ri m- lim- etcr d ctcr • 20 0 • M gO-MgCh-H,O mixtures without fill er --°---- 0--_ 0 • -- o ...Q. __ callv callv °C °C hr hr % % % MgO M gO o 00 _ C> 1 40 40 20 130 188 120 54 )4 1 :::!! 2 40 40 20 280 178 11 4 60 ~ 1}1 LL o 3 40 40 20 44

~ o ~-__~ _ _ _ L____ L_ ___ ~ --~ ...J 1 3S 28 34 180 164 97 29 3 ~ 7}1

I __ J 150 r------~------~

I 50 I 1.50 I I .., Q Yo "'C " :E. >-- 100 4 0 1.00 I- U .(,) 5i J-> W ~ , a: ....J W ::J W a:: ti lI- :::> a: 0 w (f) I- a. ::J « :::E ....J a:: w ::Ja W I- 0 :::E a... 30 0.50 ~ (,) W ~ I- ~ B z 50 A b

/ I I " 2Q o 0 10 20 lYEAR

ELAPSED TIME, HOURS

FIGURE 4. Temperature-time relations of a magnesium oxy­ chloride cement and a paste in the Lerch conduction calorim­ eter, and modul1ts of elasticity-time relation of the pasie. o~------L------~ o 10 20 Curve B represents the behavior of a commercial flooriug cement (sample 1, table 2), and curve A, that of the magnesia (sample 5, table 2) from which it was ELAPSED TIME , HOURS made. Curve C represents the dynamic modulus of elasticity of a 1- by 1· by ll ~" -in . bar prepared from paste made with magnesia sample 5 and magnesium FIGURE 3. T emperat'ure-time relations of a magnesium oxy­ chloride solution. Temperature of the surrouudiugs, 25° C. chloride cement and a paste enclosed in a vacuum flask.

Curve B represents the behavior Gf a commercial flooring cement (sample 1, table 2), and curve A, that of the magnesia (sample 5, table 2) from which it was made. Temperature of the surroundings, 30° C. Similar curves obtained with the conduction calorim­ eter are shown in figure 4. The maximum tempera­ ture was reached from 3 to 5 hI' later in the conduc­ tion calorimeter. This del2,y indicates that the much higher temperatures reached in the vacuum flask accelerate the hardening of the materials. The compressive strength of magnesite cement is reduced by high temperatures during hardening, such as would occur in the interior of mass concrete. In figure 5 are shown the relative compressive strengths of 2- by 4-in. cylinders that were allowed to harden for 24 hI' enclosed in vacuum flasks . The cylinders were then removed and stored at 25 ° C in o L-----______~ ______L ______"'-~ ___1~ the laboratory. Companion specimens were allowed o 50 100 150 to harden without protection from loss of heat and were similarly stored. The 7-day compressive MAXIMUM TEMPERATURE REACHED,OC strengths of the protected specimens, calculated as FIGURE 5. Effect of hardening temperature on the compressive the percentages of the strengths attained by the stl ength of magnesium oxychloride flooring cements. unprotected specimens, are plotted against the The strengtbs of 2- by 4-iu. cylinders allowed to harden iu a vacuum flask, calcu­ maximum temperatures reached by the protected lated as the percentages of tbe strengtb of companion cylinders unprotected agaiust loss of heat, are plotted agaiust the maximum temperature attaiued in the specimens. The cements are those listed in table 2, vacuum fl ask. Storage conditions; Protected cyliuders, 24 hours iu vacuum flask, 6 days at 25° C and 50-percent relative bumidity; unprotected cyliuders, except for sample 4. R eduction in strength occurred 7 clays at 25° C and 50-percent relative humidity. Tbe position of tbe dotted liue for seven out of the eigh t cements and amounted in was calculated by the metbod of least squares. Point A, representing the oxide from wbich cement B (sample I, table 2) was made, was omitted from the calcula­ one instance to 65 percent. The dotted line in tions. 380

L figure 5 represents the equation calculated from the groups of samples were prepared by adding MgO data by the method of least squares. Point A, to stock solutions of approximately these concentra­ which was not included in the calculation, represents tions, and deviations of the points represen ting these magnesia sample 5, and point B represents flooring mixtures from straight lines passing tlu'ough the cement 1. These are the samples represented by MgO ver tex of the diagram (fig. 1) can be due to curves A and B of fi gures 3 and 4. The eylinders analytical error only. The first group of samples represented by point A disintegrated in 5 days; was prepared by mixing MgO with the supernatant the 7-day compressive strength, therefore, was liquid formed by adding water to a l-lb bottle of plotted as zero. MgC1 2.61-I20 crystals. The concentration of this The sonic method of determining the dynamic solution undoubtedly varied from time to time. modulus of elasticity, E, was used to follow the The heats of solution of the pastes having compo­ progress of the hardening of 1- by 1- by l1%-in. bars sitions indicated by the points in figure 1 were made of paste prepared with magnesia sample 5, determined at various times after mixing. The heat table 1. Curve C, figure 4, shows part of the results of solution of the MgO was also measured. The heats of one such series of measurements. The bar of dilution of the various MgCJ2 solutions in 2.00 hardened sufficiently in 2 lu· to be removed from the N HCl containing 1 g of added MgO per 425-g mold, and measurements were begun at once. The charge of acid were also determined. Duplicate modulus of elasticity at 2.2 lu' was 260,000 Ib/in. 2 determinations were made using 2- 5- and 10-ml At 6.4 hr it had increased to 920,000 Ib/in.,2 and at quantities of each of the four solutions. These data 20 hI' to 1,240,000 Ib/in.2, about three-fourths of the were plotted and interpolated values of the heats of value at 1 yr. The most rapid increase in E occurred dilution appropriate to each heat-of-solution experi­ during the period of rapid increase of heat evolution ment ,vere obtained. The heat of solution of the indicated by curve A of figure 4, although suffi cient hardened paste was subtractcd from the sum of the strength had developed to enable the removal of the heat of solution of the MgO and the heat of dilution specimen from the mold about 2 hI' before the rapid of the magnesium chloride solution. The resulting evolu tion became evident. The early and rapid quantity is the heat of hardening of the paste, and development of a modulus, that is a substantial the values so obtained are plotted in fi gure 6. fraction of the final value, is charaeteristie of mag­ In this figure the pastes in a given column were nesium oxychloride pastes of widely differing prepared with solution of the approximate concen­ compositions. tration indicated at the top of the column. The Insufficien t d aLa are available in this work with columns, therefore, correspond to the lines drawn to the commercial cements and active magnesias to the MgO vertex in figure 1. All the samples in a separate the effects of temperature of hardening given row contained approximately the same per­ and of variations of the properties of the magnesia. centage of MgO. The rows of figure 6, therefore, The work does show, however, that the heat of hard­ correspond approximately to lines drawn in figure 1 ening of ordinary magnesium oxychloride pastes and parallel to the base and at 10-percent intervals. cements is about 190 cal/g of MgO (about 50 calories The heats of solution are subject to analytical per gram of cement) at 18 lu', and that temperatures errors from two sources. The first is the inherent near the boiling point of water may be reached in the error in the determination of the MgO in the paste interior of a mass of magnesite cement. This qml,ll­ sample used for analysis. Judging from the differ­ tity of heat is perhaps fi ve or SLX times as much as ences between the results for duplicate samples for would be developed in the same period in an equal the same paste, the error in the determination of weight of portland cement concrete. Because the MgO was small in mo t instances. The second oxychloride cement is usually less dense than con­ source of elTor was the difficulty in obtaining and crete, from three to four times as much heat would transferring certain samples. Paste containing less be developed in a given volume of oxychloride than 30 percent of MgO were of such consistency cement as in the same volume of concrete. that many could not be weighed accurately and quickly and introduced rapidly into the calorimeter. 4 .2. Heat·of-Solution Measurements Consequently, these pastes were exposed to the r- A systematic study was made of the heats of soll.!l­ atmosphere longer than desirable. No entirely tion of mixtures of magnesia and magnesium chlo­ satisfactory technique was devised for handling ride solutions. Magnesium oxide sample 6, table 1, pastes of soapy or greasy consistency. As a result, was used with solutions of MgClz of analytical reagent there was som e uncertain ty in the weight of the purity. The compositions studied are shown as calorimeter sample and in the conespondence of the lettered open circles in fi gUTe 1. These points rep­ calorimeter and analytical samples. resent the average compositions calculated from the It will be seen from figure 6 that the greatest part analyses of the paste samples made when the heats of the heat of hardening was developed in a short of solution were determined. For convenience in time. Figure 6, N, shows a sample, 30-percent of reference, samples A to E are considered as made MgO- 21-percent of MgC12 solution, that at 1 day with a solution containing 35 percent of MgClz; had developed nearly 80 percent of the 60-day heat samples F to Ie, with a 27-percent solution; L to P, of hardening. Sample 0 developed nearly 95 per­ with a 21-percent solution (220 Baume); and samples cent of the 60-day heat of hardening in 2 days. A Q to V, with a 15-percent solution. The last three sample of approximately the same composition as 3S1 CONCENTRATION OF ~C12 300 27 'Yo Mg CI2 2 1%MgC12 300 15% Mg el2 SOLUTION

KEY 200 200 0

0 SAMPLE % MilO 100 100 NO. IN PASTE K 60% MOO V ,6O",(,t.4qO 0 0

300 35% Mg CI2

0- 200 W f- (J) 100 ro ~ E 50% MOO 50% MOO P 50". MoO U 50'1'. MoO W 0 I f- 300 ~ 0 0 200 0> :E 100 ~ D 40% MOO 40'. MoO 0 40% MqO T 40\ t.40 0 :E 0 r t C2 (9 300 0- 0:: 0 W a.. 200 (J) W 100 a: C 30% MO O H 30% t.400 N 30'1'.1<400 S 30% MoO g 0 r S 300 (9~ 0 z Z 200 -<>- ~ 100 0:: 20% MoO M 20%M~0

T he compositions of t he samples, determiQed by analys is, are indicated by the corres ponding letters in fi gure 1. These pas tes were prepared using M gO No. 6, table 1.

0 ~'" 300 35% ~ 27% '0 u - i3 100 ~

0 0 Kl 20 60 MIlO IN PASTE, PERCENT 382 sample 5 developed 240 callg of MgO in 4 yr. The as apparently occurred in the work cited. X-ray average 60-day heat of hardening of the sample patterns of certain of these pastes also indicated that represented in figure 6 is 225 callg of MgO. These MgO was present. The presence of unreacted mag­ figures and the curves in figure 6 indicate that after nesia in the pastes made with the more concentrated the first few days h eat generation in magnesium magnesium chloride solutions could account for the oxychloride cements is very slow. Attention is low heats of hardening found with the larger amounts invited to the curves representing the behavior of of MgO. samples A, B, R , and S. The values obtained for the heats of solution of these samples indicate that the heat of hardening first increased and then de­ The authors acknowledge with gratitude their creased with the elapse of time. The authors can indebtedness to E. D . West, who made most of the at present offer no convincing explanation of this direct measurements, to R . B. Peppler, who devised unexpected behavior, if the apparent trend is real. the analytical methods and made many preliminary Although three of these samples were in the composi­ heat-of-solution measurements, and to L . P . McGin­ tion range where it was difficult to transfer the paste nis, who prepared many of the pastes and made quantitatively because of their consistency, errors in numerous heat-of-solution determinations. analysis from this cause are not believed to be large enough to cause deviation from the expected trend 5 . References of the magnitude shown here. The 50-day heats of harden ing from fi gure 5 arc [1] C. R. Bury and E. R. H . Davies, J. Chern. Soc. (London) shown in fi gure 7, plotted against the percen tage of p . 2008 (1932). [2] L. Chassevent, Chimie & industrie 30, 1020 (1933). MgO in the pastes. Here is definite evidence of a [3] F ederal Speci fi cation SS- C- 158b, May 20, 1946. change in the heat of hardening of those pastes as a [4] P. H . Bates, Proc. Am. Concrete Inst. 25, 88 (1930) . function of the MgO content. Three of the curves [5] C. H. Jumper and G. L. Kalousek, Rock Products 45, 54 of figure 7 show maxima in the neighborhood of 20 (April 1942). [6] William I-erch , Am. Soc. T esting Materials Proc. "6 percent of MgO. These curves show that the 50-day 1252 (1946). heat of hardening of mixtures containing large [7] W. C. Pierce and E . L. Haenisch , Quantitative analysis, amounts of MgO is less than the heat of hydration 2d ed., p . 247 (John Wiley & Son, New York, N. Y., of MgO to Mg(OH)2, 224 callg of MgO. It bas 1940). been shown, [9] that unrcacted magnesia may re­ [8] L. Walter-Levy and Y. Bianco, Compt. rend. 232, 730 (1951). main in magnesium oxychloride pastes after harden­ [9) W. Feitknecht and F . Held, H elv. Chimica Act a 27, 1480 ing. The pastes studied here were in sealed con­ (1944) . tainers, so that no hydration subsequent to harden­ ing could be caused by moisture from the atmosphere, W ASHING1'ON , August 1952.

383