Analyst, March, 1971, Vol. 96, $@. 213-219 213 The Automated Determination of Silicon and Calcium in Portland Cement and Associated Raw Materials* BY J. A. FIFIELD AND R. G. BLEZARD (Tunnel Cement Ltd., West Thuwock, Gyays, Essex) The manufacture of Portland cement clinker is a continuous large scale chemical synthesis of specific compounds. The strength of the hydrated cement matrix in concrete is a function of the original clinker compound assemblage. Control of the production of the clinker compounds has to be related to the time of passage of materials through the rotary kiln and the conditions of processing. Continuous control of the major constituents is essential and knowledge of the effects of the variation of the constituents on the physical properties of the product is desirable.Methods are described for the automated determination of calcium and silicon in Portland cement and associated raw materials with respect to the demands, scale and nature of the process. THE analytical importance of the major constituents of a material can be a reflection of inherent specific properties as well as a guide to the completeness of the analysis summation. For example, in the analysis of cast iron or steels it is the minor alloying metals that are studied rather than the major constituent. However, there are other complex systems where there are several phase combinations between two elements and in these instances it may be important to know the elemental analyses with reasonable accuracy. This knowledge may help in deciding which phases are present, or in the continuous monitoring of a dynamic sytem where variables may affect the major constituents.The major oxide constituents of a Portland cement are lime and silica, which together account for 90 per cent. of the total content. Some typical results from the samples regularly analysed in a cement works laboratory are as follows- Lime, per cent. Silica, per cent. Clay . . .. .. 2.0 58-0 Chalk . . .. .. 50.0 2.0 Raw meal . . .. 44-0 15.0 Clinker . . .. .. 67-0 23.0 Cement . . ,. .. 65.0 22.0 On inspection of these results it would appear that methods having a confidence level . of 0.50 per cent. for one standard deviation should suffice for the control of the plant but, in practice, much better precision than this is required. In the cement manufacturing process raw meal is calcined in a kiln into a mixture comprising four main cementitious compounds- Molecular formula Mineralogical abbreviation 3CaO.Si0, .... .. c*s P-2CaO. SiO, .. .. P-C2S 3Ca0.A1,03. . .. .. C3A 4Ca0.A1,0,.Fe,03 . . . . C,AF These formulae are idealised; in practice, the silicate minerals contain interstitial aluminium, magnesium and titanium oxides, alkalis, etc. Early cement microscopists called C,S solid solution alite, C,S belite and the iron complex is now known as ferrite. Cement manufacture is basically a continuous large-scale synthesis of these compounds. The time of passage of the raw material through the rotary kiln may be of the order of 2 hours, and the analytical control system has to be devised with this in mind.* Presented at the symposium on “Accurate Methods of Analysis for Major Constituents,” organised jointly by the Society for Analytical Chemistry and the Analytical Section of the Royal Dutch Chemical Society, London, April 3rd to 4th, 1970. 0 SAC and the authors.214 FIFIELD AND BLEZARD : AUTOMATED DETERMINATION [Auzalyst, Vol. 96 The development of strength of a cement matrix is a function of the compound assem- blage, the relative influence of each of the four main phases being shown in Fig. 1. It must be stressed that strength calculations based on the hydraulic activity of chemi- cally pure cement compounds must be modified as a result of the inclusion of minor elements in solid solution. Compound composition is used as a guide to the potential strength of the finished product and analyses at a rate of at least one per hour are required.N I E E Time/days Fig. 1. Development of strength of pure cement compoundsf EFFECT OF ANALYTICAL ERRORS- Consider the introduction into a kiln of a carbonaceous raw meal having the chemical composition: silica 15-0 per cent., aluminium(II1) oxide 3.0 per cent., iron(II1) oxide 1.5 per cent. and lime 43.5 per cent. Then, under ideal conditions and assuming no ash absorption from the fuel, a clinker will be produced containing: silica 21.3 per cent., aluminium(II1) oxide 4.6 per cent., iron(II1) oxide 2.3 per cent. and lime 67.0 per cent., basing the transposing factor (slurry to clinker) on 100 (100 - ignition loss at 1400 "C) From this chemical analysis it is possible to calculate a theoretical compound composition of the clinker provided the percentage of unreacted or free lime present is known.Hence, the following compound formation can be postulated by using the method enunciated by Bogue2: alite 5843 per cent., belite 23-4 per cent., C,A 8.3 per cent., ferrite 7 per cent. and free lime 1.5 per cent. Assuming no error has occurred in the determination of aluminium and iron, then 0.5 per cent. error in the silica analysis gives rise to a 6 per cent. error in the alite estimate, and 0.5 per cent, error in the lime analysis leads to a 3 per cent. error. Thus the coupled effect of these two errors leads to underestimation of the alite by more than 9 per cent. and over- estimation of belite by the same amount. Alitel has a 7-day strength of 40-68 N mm-2 and a 28-day strength of 49-30 N mm-2 and belitel has corresponding strengths of only 0.69 and 6.90Nmnr2, other phases con- tributing very little to the strength of the cement. Although belite ultimately develops a strength equal to that of alite, it is early strengths that are of importance to the construction industry.The influence of variation of alite content upon compressive strength is shown in Table I. Thus, the 0.5 per cent. analysis errors can lead to strength estimations that may be about 4 N mm-2 low at 28 days. For control of the plant, calculations of compound composi- tion to +2 per cent. of alite are required and it is therefore necessary to use analytical techniques capable of determining silica to +0.07 per cent. and lime to +0*1 per cent.for one standard deviation.March, 19711 OF SILICON AND CALCIUM IN PORTLAND CEMENT 215 Assumptions made in the Bogue calculation are included in this study where a multi- component system (silica - aluminium(II1) oxide - iron(II1) oxide - lime) is being considered under specific conditions. It is assumed that all the iron, aluminium and silicon and the calcium not existing as uncombined lime or as calcium sulphate will be combined as C,S, C,S, C,A and C,AF. The particle-size distribution of the ground compound assemblage in the presence of gypsum is a further contributor to the strength of a hydrating matrix, mainly from the aspects of kinetics and efficiency of hydration. TABLE I INFLUENCE OF VARIATION OF ALITE CONTENT UPON COMPRESSIVE STRENGTH Alite content, * 7-day strength/ 28-day strength/ per cent.Nmm-2 Nmm-a 56.8 28.96 39-10 53.5 27-65 37-62 50.8 26.34 36.61 47.0 25.24 35.10 * The belite content will increase relatively as the alite content decreases. SAMPLE DISSOLUTION FOR AUTOMATED ANALYSIS- Fuse the sample3 with a strong alkali in a gold - palladium crucible and leach with hot water. Pour the alkaline suspension into a solution of sufficient hydrochloric acid to dissolve the calcium and iron hydroxides and, after cooling, dilute to a known volume (500 ml). The range of lime and silica to be measured together with the sample weight required for the various materials is given below in Table 11. TABLE I1 SAMPLE WEIGHT FOR MATERIALS ANALYSED Sample weight/g Lime, per cent. Silica, perlcent. Clay .. .. .. .. 0.250 0 to 6 48 to 64 Raw meal . . .. .. 1.000 42 to 54 12 to 16 Cement and clinker . . .. 0.600 63 to 68 20 to 27 Chalk.. .. .. .. 1.000 30 to 53 0.5 Sampler i - 7 Tube dia. I 1 Wash I I 0.030 mm I t I 10-045mm Sample ' I Hydrofluoric acid ! 10-073mm 1 Air i ;O-O65md i 10.073mm Boric acid i I v- W - Pulse suppressor + Lk;eb~ler drain I I ! I - Ammonium I T0.065mm molybdate I , Water ~ O - 1 0 0 m r n l W R e-sam ple I 10.030mrn I I 30" C 30" C Air i i0.073 mm 1 I 1 I I expander Fig. 2. Flow diagram for the automated determination of silica. Sampler speed 20 hour-l and wash ratio 1:2216 FIFIELD AND BLEZARD : AUTOMATED DETERMINATION [Analyst, Vol. 96 AUTOMATED DETERMINATION OF SILICA For the determination of silica an automated molybdenum blue method has been developed; the required reactions are given by the flow system (Fig.2). 16-0 REAGENTS- fusion (as for samples). Wash reagent-An aqueous solution of 0.22 g 1-1 of pure precipitated silica prepared by Hydrojuoric acid-A solution of 5 ml of 40 per cent. hydrofluoric acid in 1 litre of water. Boric acid-A saturated aqueous solution (approximately 50 g 1-l) . Ammonium molybdate-A solution containing 2-5 g in 1 litre of 0.3 per cent. v/v sulphuric Ascorbic acid-An aqueous solution containing 10 g 1-1 of ascorbic acid. acid. 30 METHOD- The addition of hydrofluoric acid to the flowing sample stream complexes the iron, decomposes any polymerised silicic acids and converts them to a silicofluoride complex. Excess of hydrofluoric acid is then complexed by excess of boric acid, which also decomposes the silicofluoride ion to give fluoroborate and silicate ions.A yellow silicomolybdic complex is formed by adding acidified ammonium molybdate and the product is finally reduced to the molybdenum blue complex by ascorbic acid. Phosphate will interfere but in practice (con- sidering a sample sequence at an individual plant) phosphate content is almost constant and is very low; in the United Kingdom it does not exceed 0.05 per cent. If the phosphate is known to have a detectable variation, any interference that it may cause can be suppressed by the addition of citric acid.4 Initially the analyser was set up to give an optimum performance over the range 0 to 16 per cent. of silica and the charts obtained looked very promising, but a 1 per cent increment of silica concentration was represented by only 2.5 per cent.transmission and interpretation gave errors outside the control limit. For raw meal analysis the range measured is 12 to 16 per cent., so washing with water was replaced by washing with a solution containing silica equivalent to 11 per cent. in the sample. The base-line was set at 98 with the full colour development from 11 per cent. of silica passing through the flowcell and peaks were obtained - 15.0 14.0 Continuous sampling 14.0 Fig. 3. Determination of silica: recorder traceMarch, 19711 OF SILICON AND CALCIUM IN PORTLAND CEMENT 217 for the standards. In this way the resolution improved to 3.5 per cent. transmission per 1 per cent. increment. A x 4 range expander was then employed to boost the resolution to 14 chart transmission units per 1 per cent. of silica.By using this technique the peaks shown in Fig. 3 are obtainable, the base-line varying from peak to peak by less than 1 chart division (+0.035 per cent. of silica). If peaks are allowed to reach their maximum then a precision of +0-035 per cent. of silica can be obtained but, in practice, the sampling time is insufficient to allow a state of equilibrium in the system. However, *0*04 per cent. can be achieved for one standard deviation and this precision meets the requirements of the method; Fig. 4 shows the calibration graph. Having developed this system it was extended to the analysis of clays, chalks and clinkers. 1 I I 1 12.0 13.0 14.0 15.0 11 Silica, per cent.0 Fig.. 4. Determination of silica: calibration graph For clay analysis 0.25 g of sample is used instead of 1 g (see Table 11) so that the 12 to 16 per cent. standards represent a range from 48 to 64 per cent. in the clay, and similarly 0-6 g of cement and clinker is taken to measure the range 20 to 27 per cent. For chalk samples a method of standard addition can be used by adding the equivalent of 11 per cent. of silica to the solution before diluting to volume. In this way the standards range from 1 to 5 per cent. Cement samples have been analysed by the AutoAnalyzer and by the reference B.S.I. m e t h ~ d . ~ Each sample was analysed three times by each technique and thirty samples in all were used. The average deviation between the two methods for the ninety results was 0.03 per cent.and the standard error between the mean of the triplicates 0.06 per cent. AUTOMATED DETERMINATION OF LIME For the determination of lime up to the 70 per cent. level normal colorimetric procedures were discounted because a 1 per cent. increment of lime would be represented by less than 0-5 per cent. transmission. An EDTA titration would seem to be better for this analysis, but in order to achieve a faster output of results, a system that is a compromise between titrimetry and colorimetry has been developed. In Fig. 5 (see p. 218), a manifold for the analysis of lime over the range 0 to 4 per cent. is shown. REAGENTS- Triethanolamiute-A 100 ml 1-1 aqueous solution. Bufer, $H 13-An aqueous solution containing 10 g 1-1 of sodium tetraborate plus 20 g 1-1 Colour reagent-A methanolic solution containing 0.75 g 1-1 of glyoxal bis-2-hydroxyanil.Standard EDTA-An aqueous solution containing 8.35 g 1-1 of EDTA disodium salt of sodium hydroxide. (dihydrate) standardised against B.C.S. 372.218 FIFIELD AND BLEZARD AUTOMATED DETERMINATION [ArtabSt, VOl. 96 - Colour reagent I !O-OSl mm Colori- 570 nm I meter a I I O - O ~ O mm V Recorder Fig. 5. Flow diagram for the automated determination of lime. Sampler speed 20 hour-1 and wash ratio 1 : 2. Colour reagent glyoxalbis-2-hydroxyanil, which must pass only through Solvaflex tubing 60 70 80 90 METHOD- Triethanolamine solution is added to the sample to complex iron and the pH is adjusted to 12.6 by a buffer. A 1 per cent. glyoxal bis-2-hydroxyanil (bis-(2-hydroxyphenylimino)- ethane) solution in methanol3 is then added to produce a red 1 : 1 complex with calcium and the absorbance measured at a wavelength of 570 nm.Fig. 6 shows a typical recorder trace for lime and Fig. 7 illustrates a calibration chart. - 1.6 - 0.8 f - - 0 4-0 3.2 40 'I- 2.4 50 t Continuous sampling 2 4 100' Fig. 6. Determination of lime: recorder trace 2o I I I I 1 0 0.8 1.6 2.4 3.2 1 Lime, per cent. Fig. 7. Determination of lime : calibration graph A calculated amount of EDTA is added to test solutions so that not more than 4 per cent. but greater than zero lime is left uncomplexed in solution. The mixture is then analysed by the AutoAnalyzer for excess of lime content. The EDTA solution is made such that 1 ml is equivalent to 1 per cent. of lime in 50 ml of test solution.So, for raw meal analyses, 42 ml of EDTA (43 to 45 per cent. of lime being determined) are added to 50 ml of test solution, the volume is diluted to 100 ml with water and the mixture is analysed for excess of lime.March, 19711 OF SILICON AND CALCIUM IN PORTLAND CEMENT 219 The percentage of lime in the sample is then 42 @Zus the chart reading. From Table 11, cements and clinkers range from 63 to 68 per cent. of lime, but because 0.6g of sample is used to prepare a test solution and not the l-g standard, the measuring range becomes 37-8 to 40.8 per cent. of lime. To 50 ml of cement solution 38 ml of EDTA solution are added before dilution to 100 ml with water. The excess of lime is determined and 38 + chart reading 0.6 Percentage of lime = Chalk samples range from 30 to 53 per cent.lime content but at a factory site the samples fall into two categories called high and low chalks. Low chalks contain 30 to 34 per cent. of lime and high chalks 49 to 53 per cent. To 50 ml of chalk solution 30 or 49 ml of EDTA, whichever is appropriate, is added before dilution to 100ml with water and measurement of excess of lime. The lime content is then the volume of added EDTA @Zus the chart reading. Clays contain less than 6 per cent. of lime so in this case the neat test solution is analysed directly. As 0.25 g is used and the solution is not diluted further, the results are obtained by multiplying the chart reading by two. Results obtained for cements agree well with results from the B.S.I. rneth~d.~ Triplicate analyses of 30 samples showed an average deviation between methods of 0.05 per cent. with a standard error of 0.13 per cent. CONCLUSION The two methods outlined and also methods395 for aluminium(II1) oxide, iron(II1) oxide, sulphur trioxide and free lime5ps have been used for plant control of two factories for the past 3 years. Samples are analysed at hourly intervals and a considerable improvement in quality stability has been achieved. Some thought has been given to an automated digestion system so that the analyser can be operated in the plant itself, and when this problem has been overcome the system may then find a place in many more factories. REFERENCES 1 . 2. 3. 4. 5. 6. Bogue, R. H., and Lerch, W., Ind. Engng Chem., 1934,26, 837. Bogue, R. H., Ind. Engng Chem., Analyt. Edn, 1929, 1, 1929. Fifield, J. A., and Blezard, R. G., Chem. & Ind., 1969, 1286. British Standard 4550, Part 2 : 1970. Blezard, R. G., and Fifield, J . A., “Advances in Automated Analysis,” Volume 2, Mediad Inc., Fifield, J. A., and Blezard, R. G., Analyst, 1969, 94, 503. Received June 24th. 1970 Accepted October 12th, 1970 Chicago, 1970, p. 283.