i

l

The Microstructure of ConcreteKAREN L. Scnrvexen, Imperial College

The great tragedy of science: the slaying ol abeautiful hypothesis by an ugly lact.

Thomas Huxlev

The various aspects which contribute to the microstructure of concrete areconsidered and il lustrated, from the microstructure of the anhydrous cement to themicrostructure of mature concrete. At each strge the importance of quantitativecharacterization and methods of achieving this cre discussed.

Introduction

At some level the behavior of every material is related to its microstructure. Theunderstanding of these relationships between structure and properties forms the basisof materials science. Microstructure encompasses a wide range of structural levels,from the atomic scale to that of the engineering component, and includes alldiscontinuities inside and between phages, such as dislocations, grain boundaries,phase interfaces, pores, and cracks.' A complete characterization of the

microstructure of a multiphase material must also entail quantitative informationabout the relative proportions of the phases and their distribution in space. Therelative lack of success in developing rnigrostructure/property relationships.forconcrete is due, in no small part, to the lact of good microstructural characterization.

In microstructural terms, concrete is an extremely complex system of solidphases, pores, and water, with a high degree of heterogeneity. This heterogeneitlian be considered on several leveli. lt the simplest level, concrete consists of

aggregate particles, distributed in a matrix of cement paste. On a more detailed level,

thJ paste itself is a mixture of unreacted cement, hydration products, pores, andwater and at a still finer level these phases themselves have complex microstructures.

In this chapter, various aspeits of the microstructure of concrete will be

considered in anattimpt to provide a coherent picture. As an introduction, a brief

description of the chemistry and structure of ihe solid phases in cement pastes is

given, followed by a short survey of the physical methods used to study

microstructure. Initiatly, the components of the concrete microstructure are

considered separately: thi microstruciure of anhydrous cement; the development of

microstructurb auting the hydration of cemLnt paste; the microstructure of

Materials Science of Concrete t27

aggregares; ano the rntertace between cement paste and aggregates. To arrive at acomplete characterization of concrete microstructure, these components must beinteg,rated, but this is an area in which much work needs to be done and only a fewcomments can be made. For conciseness, it has been possible to cover only themicrostructural development of concretes at near ambient temperatures.

Solid Phases in Cement Paste

Anhydrous cement powder is a combination of oxides of calcium, sil icon,aluminum, and iron (plus small amounts of other oxides). The four principalminerals found in ordinary portland cement are alite, impure tricalcium sil icate (C3S);belite, impure dicalcium sil icate (CzS); and the aluminate and ferrite phases whichhave average compositions of c3A and c4AF, respectively. These minerals react withwater to give a variety of hydrates. The calcium sil icates react to give calciumhydroxide and calcium sil icate hydrate and the aluminate and ferrite phases reactwith the added calcium sulfate to give two groups of product referred to as AFt andAFm.

Calcium hydroxide (CH) is the only hydration product to have a well-definedstoichiometry and crystal structure and normally forms as massive, relatively purecrystals with a euhedral hexagonal habit.

The calcium sil icate hydrate in cement paste is a gel which shows no long-range crystall inity. Its composition is uncertain (and possibly variable) with a C/Sratio of about 1.7 to 2.0; consequently it is usually written as C-S-H. The short-range order of this phase is probably related to the layer structure of the cry5lall inecalc ium si l icate hydrates-1.4 nm tobermori te (C5S^Hn) 4n! jenni te (C'S6H1r).2.r Thisphase appears to adopt a wide range of morphologiris,*-o some based on ihin sheetswhich may give fibri l lar or honeycomb structures at early ages, others with a morecompact structure forming at later ages.

The term AFt denotes the phases related to ettringite-calcium aluminumtrisulfate (CaA . 3CS . 32HzO) with the F indicating the poss-ible substitution of ironfor aluminum in the structure. This phase forms when the concentration of sulfateions in solution is relatively high and has a morphology of hexagonal rods.

_ The AFm phases are isostructural with calcium aluminum monosulfate (CrA .CS. l2H2O). In addi t ion to possible subst i tut ion of a luminum by i ron, hydroxlde,carbonate, or chloride ions may replace the sulfate. These phases have a hexagonalplate morphology.

Techniques for Studying Microstructure

There are many techniques which have been used to study microstructure; theycan be divided broadly into two categories. Indirect or bulk techniques giveinformation on average features of the whole microstructure. Examples of indirecttechniques are thermogravimetry (TG) and X-ray diffraction (XRD), which can beused to determine the amounts of certain phases in the sample. Methods used toobtain information about the pore-size distribution, such as mercury intrusionporosimetry (MIP) and methanol absorption, are also indirect techniques as they giveno information as to how the pores are arranged in space. The other group oftechniques is direct or microscopical techniques which provide information aboutthe way in which the component phases are arranged in the microstructure.

I ne advantage ol the lndirect techniques rs that they provide information in aquantitative form so that different samples can be compared objectively. In contrast,information derived from direct techniques usually takes the form of images. Imagesare extremely valuable in conveying a vivid impression of the microstructure. butcomparisons between different samples are more subjective and rely on theexperience and interpretation of the observer. As most of the findings presentedhere have been obtained by microscopy techniques, their application to the study ofthe microstructure of cement and concrete wil l be discussed in more detail.

optical microscopy has been widely used in the study of cement and concrete.TThe use of suitable etchants makes it pQlsible to study the distribution of anhydrousphases in clinker and cement powder.6'v In hydrated pastes and concrete tiere isl itt le contrast between the phases in reflected l ight. However, in thin sections therefractive properties of different phases and the use of f luorescent resin can indicatecement and aggregate type, the presence of mineral admixtures, the w/c ratio. thequality of compaction, and the presence of alkali si l ica reaction (e.g., Refs. l0 andI l ) .

Ultimately, the resolution of optical microscopy is l imited by the wavelength oflight and for detailed microstructural studies electron microscopy must be used. Ascement is nonconducting, and due to the presence of water, the preparation ofcement pastes and concretes for electron microscopy poses problems. The possibleeffects of drying on the microstructure and the selectivity of different preparationtechniques must be considered when interpreting electron micrographs of cementpastes.

Fracture surfaces provide a relatively simple means to observe the three-dimensional arrangement of hydration products. These surfaces can be imaged bysecondary electrons in the scanning electron microscope with fairly good resolution(arl0 nm). While providing much useful information, the microstructure observedusing this technique is that of cement paste which has been dried and exposed to thehigh vacuum of the electron microscope. Given the amorphous nature of thehydration products, it is l ikely that some alteration in their morphology wil l occurduring specimen preparation. Also, only fracture paths are revealed, so that weakareas of the microstructure predominate. In young pastes the fracture path isinterparticular, showing only the outer surface of the hydration products. In olderpastes the fracture surface is dominated by areas of cleaved calcium hydroxide,which has grown in the originally water-fi l led space to engulf the other hydrationproducts. After hydration for longer than one day it becomes increasingly diff icultto assess the extent to which fracture surfaces are representative of the bulkmicrostructure.

Undried cement pastes can be examined if the area around the specimen can beisolated from the high vacuum in the electro_n microscope column. In the AEI EM7I MeV high-voltage electron microscope (HYEM) there is a comparatively largespace between the objective pole pieces, which permits the incorporation of anenvironmental cell. Two types of environmental cell have b_e^e1^ used to study thehydration of cement in tha HVEM. Double and coworke.sl2'I3 haue studiea thehydration of cement in a window-type cell, at a high water/cement ratio. With thistype of cell, the resolution is limited by the presence of windows and furt[gr by thelarge excess of water in the paste. f-he cell used by Jennings and Pratt" and byScrivenero and Scrivener and Prattr) has pairs of differentially pumped openapertures. Pressures of up to 300 r can be maintained in the cell with water-saturated gases and better resolution can be obtained than with the window-type cell.Using this technique, much information can be gained on the morphology of

Materials Science of Concrete Materials Science of Concreter28

_d|[,

t29

hydration products in the undried state and the way in which their morphologychanges on drying. However, due to the method of specimen preparation it is stillonly the outer surfaces of the hydrating grains which can be observed.

Ion-beam thinned sections allow t-hq s^t^ructure to be viewed in cross section inthe transmission electron microscope.6'1s-20 The brittleness of cement pastes andconcretes makes the preparation of electron-transparent sections by ion-beamthinning very difficult and young pastes must first be impregnated with Lpoxy resin.The use of scanning transmission electron microscopy (srEM) minimizes damage tothe cement by the electron beam. Fairly extensive electron-transparent areas can beobtained with careful thinning; nevertheless a certain amount of selectivity isinevitable.

Another technique uses backscattered electron images of polished sections. Withthick polished sections, large cross-sectional areas can be observed by SEM.However, they do not show good secondary electron contrast. Using a pair ofbackscattered electron (bse) detectors, images can be produced in which the intensityis dependent on the average atomic number of the scanned area. Thus, in polishedunetched sections the anhydrous material, massive calcium hydroxide, otherhydratio.n^products, porosity, and aggregate particles can be distinguished from oneanother.o'rr The contrast between these censtituents is sq-fficient to allow quantitativeimage analysis of their total area and size distribution.z The form and distributionof the hydration products observable is comparable with, though of lower resolutionthan, that seen in the STEM of thin foils. However, bse images may be taken overa large cross-sectional area, eliminating the selectivity of ion-beam thinning used toprepare STEM specimens.

Microstructure of Anhydrous Cement

An account of the microstructural development of concrete must start byconsidering the microstructure of the starting materials. The microstructtlre of themost important of these, anhydrous cement, is determined by the microstructure ofthe clinker produced in cement kiln and the way in which the clinker breaks upduring grinding. The resulting product is quite different, in both appearance andreactivity, from a simple mixture of monomineralic grains of the four mainanhydrous minerals.

An "ideal' clinker microstructure consists of crystals of alite and belite (whichare solid at the firing temperature) in a matrix of interstitial phases which solidifyfrom the melt during cooling. The alite crystals have a hexagonal habit, while thebelite crystals tQnd to-be lounqed. In the interstitial material, the alirminate andferrite phises ma! bt{ifte$ divided and intergrown. The details of themicrostructure, such as the shape and distribution of the phases, will depend on thephysical and chemical characteristics of the raw mix (fineness, homogeneity, andchemical composition) and the exact conditions of burning and cooling of the clinker.Nevertheless, the general features described above can be identified in mostcommercial clinkers, such as that shown in Fig. l. This clinker also shows extensivecracking, which commonly occurs due to thermal stresses during cooling.

Clinker is ground with gypsum to give a powder with a wide range of particlesizes from less than I pm to about 100 pm. Secondary electron images of cementpowder by SEM (Fig. 2) give some impression of this size distribution and show theagglomeration of many smaller grains on the surface of the larger grains. Suchimages also reveal fine particles of gypsum spread over the grain surfaces.

r30 Materials Science of Concrete l3rMaterials Science of Concrete

Ftg. 1. Mlcrostructure of ecommerclel clinker (bse lmege).Light-grry crystels of ellte rnddarker rounded crystels of belitecen bc seen in e metrix of theInterstitiel pheses-ferrlte (light)and eluminate (derker). Porosity(bleck) cen elso be seen.

Fig. 2. Scanning electronmlcrogreph of rnhydrous cementgnins showlng the wide reoge ofparticle slzes. Small lrthlikecrystals of calclum sulfate can beseen on the surfaces of the largergrains.

Flg. 3. Backscettered electronlmage of polishcd sectlon ofenhydrous cement greinsdlspersed ln resln. It cen be seenthat most of the gralns erepolyminerrlic.

The distribution of phases within the cement grains can be studied in polishedsections prepared from samples of cement dispersed in epoxy resin. Examinationof such sections with backscattered electrons by SEM (Fig. 3) indicates that fracturethrough the alite crystals dominates the grinding procels,.with l itt le evidence offracture along the interphase boundaries. Almost all grains larger than 2 to 3 pm arepolymineralic and on the surfaces of the larger grains only small areas of interstitialmaterials are exposed, between areas of alite. Relatively greater proportions ofaluminate and ferrite phases are exposed on the surface of smaller grains, as theirformation is more likely to involve fracture through the interstitial material.

At present, most cements are characterized solely by their surface area and oxidecomposition. The surface area is generally determined by permeabil ity methods.Such methods give a much lower value for the surface area than absorption methods(such as BET), because blocked channels are not accessible to the moving air stream.Surface area measurements give l itt le information about the particle-size distribution,which is of great importance in determining the packing of the cement grains inpaste and concrete. Laser diffraction methods can be used to measure the sizedistribution; the result for a typical cement is shown iUFig.4.

From the oxide composition, the Bogue calculationa may be used to estimate therelative proportions of the major phases. This calculation assumes that completereaction occurs in the kiln and takes no account of any solid solution between thephases or of the presence of minor oxides, and may give quite inaccurate results.Quanti-tptlye X-ray diffraction can be used to measure the proportions of the clinkerphases--^ and accuracies of l% are claimed for this method if a suitable range ofstandards is used.zo Analyses of clinker can also be obtained by optical microscopyusing a point-counting method on polished and etched sections of clinker.

As i l lustrated in Fig. 3, the distribution of phases within the cement grains canbe seen in bse images of polished sections. If these images are analyzed inconjunction with dot maps from energy dispersive X-ray analyses of the saqg areas,the relative amounts of sil icate and interstit ial phases.can be measured.4' Thistechnique can be extended to assess the relative proportions of the phases exposed onthe surfaces of the grains, which might be expected to determine the init ial reactivityof the cement. Preliminary studies have shown that these surface proportions maydiffer significantly from the bulk values. Other studies using secondary ion massspectroscopy (SIMS) have shown^that the surfaces of the cement grains may containhigh proportions ol impurit ies,4 which might also affect the rea-ctivity.

Proper characterization of anhydrous cement is an important f irst step inunderstanding differences in the reactivity of cements and the properties of theconcretes made from them.

Fig. 4. Particle-size distribution for r typical cement obtained by laser diffraction.

Development of Microstructure During Hydration or cement Paste

During hydration, cement paste changes, flom a fluid mixture of cement powder

ana *"t"r T" i iiglO tof iO. itt" r"actions ihich occur are exothermic and the overall

IrJn."ss of tne rEactionian Ue stuOiea by isothermal conduction calorimetry. From

i'ff;ilirat;Ji:rti"t-.*rution rurr" fo. an ordinary portland^cement, various

;di,-o|h)^i;.;;;;; ue iOentirieo, which will be used as a basis for describing the

,oliiort.u"iural development (ttre times indicated are intended only as a general guide

ffi-;tii U" "fi."t"O

Uy a variety of factors including temperature and admixtures):

I (about 0 to 3 hours). on mixing, a large amount of heat is evolved; the-

iate of heat evolution then decreases rapldty to a minimum after about 3

hours. This init ial period, during which the cement remains fluid and

workable, is often referred to as the induction period'

II (about 3 io 2a hours). During this period some 30% of hydration occurs'

which is reflected by the maior peak in the rate of heat evolution.

III (24 hours onward). After iUout Z+ hours the rate of heat evolution

declines, although hydration rnay continue indefinitely'

I (0 to 3 Hours)

It is particularly difficult to examine the microstructure of cement paste during

this ;;;J. Moit 'of the mixing water is present as free water. If this water is

ia."-""di" that specimens can bJexamined in the electron microscope, fundamental

;;;;s;i" the microstructure may occur. Even apart from an-y-.damage caused by

ii"ir"e. the time taken for specimen preparation makes it difficult to follow the

ir6"iEit "itfte

nvOiation accuiately. These problems make examination of specimens

i,i i f i" ' ., ipii" i;; i t";, in the environmental bell of the HVEM, an especiallv valuable

;";h;GG for stuOylng the early hydration- Using this technique, specimens can be

examined as l itt le as l0 minutes after mixing.

Several workers have repo-r1g{.the appearance of a gelatinous layer on the surface

of cement soon after mixing.6'13'la'2e Mictog.aphs of undried cement taken in the

environmental cell (Fig.5ishow an indistinct layer of produ_ct on.the surfacg

Similar early produci ttas been observed during the hydration 9f C-3A with 8yp6s1lm'--

In the hydration of C3S, the early formation of product has also been notecl'"'"^ but

in this c.se the product appeais to have a somewhat different morphology of

i ifofiating fifrt. ' Considering the polymineralic nature of cement grains,, the

gefatinoui layer observed in tie hydration -of

cement is probably. an amorphous

iottoiaat product, rich in alumina and silica, but also containing significant amounts

of calcium and sulfate, iftiE""i composition varying with the composition of the

;;J;;itG ii"in i"ri"d". Outsioe this ielatinous layer, small rods of AFt can alreadv

be seen afiei as rittr" .t lO .inutes hvd-ration (Fig. 5) and after abpP.t"83l,.b"tijh;;;

rods can be clearly seen on dried fracture surfaces (Fig' 6)'""''

environmental cell thev can be observed dispersed qn the carbon support film at some

distance from the rutf".i of the hydrating grain,$ indicating that they form by a

through-solution mechanism. when cement paste is- dried,. the gelatinous layer

cottapiis and shrinks il[; to the surface of ttre grai-ns. It is difficult to identify

tftii J.oOuct on fracture;rfaces, although indicqtipns of some gelatinou-s-product can

tre seen in ion-beamed thinned sections-(Fi8. 7).o't The AFt rods also fall back onto

the surface of the grains, although the denJity of rods is probably still related to the

comjosition of ttre'unOeilying giain. After some three hours hydration, the amount

of C-S-H formed h suiiiiirn-t ior this phase to be identifip{qp fracture surfaces as

small spiky outcrops on the surface of the grains (Fig' E)'"-"*

N

;o.a

Materials Science of Concrete Materials Science of Concrete

20

r32133

Fig. 5. (Left) cement paste hydrated for l0 min'in the environmental cell ofthe HVEM. The surfaces of the grains have e gelatinous appearance and seyeralshort rods of AFt can also be seen.

Tig.6_._ (Right) Dried fracture surface in the SEM of cement paste hydrated forI h. Many rods of AFt can be seen.

f!q.:. (Left) Ion-beemed thinned sectlon of cement peste hydrated for 2 h(srEM). Product with e gelatinous rpperrrnce can be seen with some AFt rods.

Iig, 8. (Right) Fracture surface of 3-h-old paste. Spiky outcrops of driedC-S-H can be seen.mong the AFt rods.

It (3 to 24 Hours)

The end of the induction period is marked by the rapid growth of both C-S-Hand CH. In the undried state, C-S-H has a filmy, foil-like morphology, (Fig. 9),which dries to a fibrillar morphology (diamond type I) where there is plenty of space,and to a honeycomb morphology (diamond type II) where space is more restricted.(The presence of foreign ions such as Cl- may also affect the morphology ofC-S-H.) Small grains of cement (which probably contain only alite) hydrate rapidlyto give rosettes or spherulites of C-S-H (Fig. l0).

In ion-beam thinned sections (Fig. l l), i t can be seen that the C-S-H shell isseparated from the unde;[ying grain. This separation can be observed after as littleas five hours hydrationo't' and is thought to occur because the C-S-H nucleatesoutside the gelatinous layer formed earlier, on a framework of AFt rods. The spacewhich is observed in dried sections is most likely filled with a highly concentratedor possibly colloidal solution in the wet state, although the discontinuity between thehydrate shell and core cag also be seen in undried samples examined in theenvironmental cell (Fig. l2)." The growth of C-S-H between the cement grains bondsthe paste together, causing the paste to set after about 3 to 4 hours of hydration.After about 12 to l8 hours, this bonding is strong enough for fracture to occurthrough the hydrate shells, such that the separation between them and the anhydrouscore can be seen on fracture surfaces (Fig. l3). Such separilted shells, which arecompletely hollow in some cases, were first noted by Hadleys on fracture surfaces

Flg. 9. (Left) One-day-old paste in the environmental cell of the HYEM. TheC-S-H which formed through solution during the major heat evolution peak hase morphology of crumpled foils; some rods of AFt can also be seen.

Ftg. f 0. (Rlght) Fracture surface of one-day-old paste in the SEM. 'Rosettes' ofdried ftbrillar C-S-H (diamond Type I) and long rods of AFt from the secondaryreaction of the aluminate phase can be seen.

'ij

'The w7c ratio of sll the pastes illustrated is 0.5 unless otherwise stated.

t34 Materials Science of concrete Materials science of concrete 135

l - ig. l l . Ion-besm thinnedsection of e l2-h-old oaste(STEM). Shells of hydrationproduct can be seen rround. butseparated from, the darkenhydrous cores. Rosettes fromthe hydration of small grains andsome hol low shel ls of C-S-H canalso be seen.

Fig. 12. Undried, one-day-oldpaste (environmental cell,HVEM). Although this is not ecross-section, there is e region oflow-density material visiblebetween the hydrete shell andcore.

Fig. 13. Part ial ly reacled grainof anhydrous cement with eseparated shel l of hydretionproduct revealed on the frecturesurface of a one-day-old paste.Rods of AFt and plates of AFmcan be seen between the shel l andcore.

Fig. 14. Backscattered electron image of e polished section of a l2-h-old paste.A large mass of calcium hydroxide (l ight gray) crn be seen. This has grownaround many of the hydrating grains. In this paste rlmost sll of the cementgrains have separated hydrate shells.

The calcium hydroxide which is produced by the reaction of the sil icate phasesis deposited as massive hexagonal crystals in the space originally occupied by thewater. In bse images this phase can be seen as extensive l ight-gray areas (Fig. l4).The number of nucleation sites appears to be small and the crystals may engulf someof the smaller cement grains as they grow.

After about l6 hours, a renewed growth of AFt occurs with long rods growingthrough the C-S-H layer (Fig. l0). These microstructural changes correspond to theshoulder on the main heat evolution peak which is sometimes observed incalorimetric .u.lres.6'27'35 The chemical changes responsible for this secondaryreaction of the aluminate phase are probably related to the reaction of the alite. Thisreaction of the aluminate phase may lead to larger local separations between the coreand its hydrate shell.

At the end of the main heat evolution peak, all grains smaller than about 5 pmwill have completely hydrated. Many of these which originally contained somealuminate phase wil l leave hollow shells of hydration product.

m (24 Hours Onward)

As the hydrate shell thickens, it becomes less permeable and continuing reactionof alite leads to deposition of C-S-H on the inside of the shell. As the volume ofC-S-H formed is greater than that of the alite which reacts, this process decreasesthe separation between shell and core. If the grains are large enough, this separationdisappears after about seven days hydration, by which time the shell is about E pm

thick (Fig. l5). Further hydration of alite is very slow, with change being observableonly on the scale of years. There is evidence to suggest that th-ig period of hydrationoccurs by a topochemical mechanism, as suggested by Taylor." In old pastes, three

regigrs of C-S-H product can be identified in the relics of fully hydrated grains (Fig.

t6).37 The outermost of these is a thin layer (sl pm) of nouter" C-S-H which formed

through solution, in the originally water-filled space, during the major heat evolutionpeak.-Inside this is a layer, about E pm thick, which formed through solution while

and such phenomena are often referred to as'Hadley grains.' During fracture, largeanhydrous cores may fall out, to leave large hollow ihetts. Howevlr, on poiistrJosections, while it can be seen that practically all grains have separated hydraie shells(Fig. l4), the separation between shell and core is seldom moie than about I pm.

The separation between the reacting grain and the C-S-H product indicates thatthe formation of the latter occurs by a through-solution mechanism. At this stagethe shell is very open and porous and ions migrate through it, leading to the growihof C-S-H on the outer surface of the shell and increasing the sepaiation b6tweenshell and core.

r36 Materials Science of Concrete Materials Science of Concrete r37

:

1t

Fig' 15. Partiallv hvdrated grein in a 69-dav-ord paste (bse image of porishedsection)' Ar rhe ourer edge or rne rriorate;'L"ii;ffi;r;;of'.i l;r,c-s-Hformed during the main tti"q.";"!ril9-1-;;;k;"" be seen."i;ri;. thi; is e rayer ofc-s-H formed throueh sorurion iniii l i ' i-},. 'r.p"ration between sheil and core.within this leyer, arong the bottom ed-ge-oiirr" grain, is e rounded crystat ofuureacted belite. At thc top right edge"is

" i.gi6'o oi'uor"""tlJ-r-"-..itu.(Courtesy of K. D. Baldie.)

Fig. 17. Fracture surfac,e of e r4-day-ord paste. The edge of a hydrate sherlruns diegoneilv across the microgreph-.--oo'i i" inner_surrece (top"reft) manypreres of AFm can be scen. .ours-ide'i-lie in"ii iri-,itil;ililiril ,r. por"space. (Courtesy A. Ghose.)

Fig. 18. Fracture surface of_a one-day-ord white cement (63% arite, 27o/oberite,5olo aluminate bv oXDA). rh;;ir i l i" i"-prr".., unresrricred by ferrite, has readto the formatioi oi targ6 efm pt"t. i l-- '- '" '

Inside the shelr. the concentration of surfa.te in sorution drops rapidry as thealuminate phase reacts. This results in ttre .eaction or iuitr,ii.iuriin"t, phase withanv AFt present on the inside of the shelr tJ giui ;i;;;;;;n;^io tr,. reaction(ignoring any substitution of iron):

Fig' 16. Relic of fuilv reacted grain In e 23-year-ord paste (bse image). Inadditlon to the two regions or E-s--ri- i i . i i ir i"a in Fig. 15, rhere is an innerregion formed by e tof,ocheri"al ;;h;is;.

the hydrate shell was separat!d from the anhydrous core. This is almost all "innernproduct, occupying spaie.originallv.tiiriJ6-i the anhydious-ir"i"l"rr," innermostregion is also "inner' oroductl in ttris casi i6rmeo by a topochemical or sorid statemechanism.

The heat produced by this reaction is sometimes viqi.bre as a row, broad peak in the3t9 9f heat evolution, as noted by prar and Gt ose.35- it e Ji.""t'.-on"""rsion of AFtt-o AFm in the absence of aluminate phase does not appear to occur in cement. onfracture surfaces, AFm_may bepgserveo insiae tne hydration shelrs and AFt outsidein-the same specimen Gii.'tii.e'n

-iro*"r.ilir rhe concentration of sulfate ions in

solution drops before th; hydrate J"rt r,"r. thickened and partially isolated theanhydrous grains, much rarg-er prates of ,tFm-may be formed irrilugirout the paste,as has been observed in somJ tuhite cements *t ict'contai" ;;?u;;i;;;il; ifits:ls).d- _ .

The be.lite phase appe.ars to take rittre part in the earry rrvci"ir"" *action. Smailregions of belite within alite may be left unreacted rrie. rsi. -Aiier'auout

l4 days,

2CrA + caA . 3cS . H3z * 4H * caA . cS . Hlz (t)

r3EMaterials Science of Concrete

-5--

Materials Science of Concrete t39

the syrfa_ce! of_be.[ite particles revealed on fracture surfaces show signs of localreaction (Fig. l9),rr although there is still relatively little hydration pro?uct visibtearound the particle after several months (Fig. 20). This reaction appears to occurpreferentially along the twin boundaries. ln a 23-year-old paste, t-h; belite grainshave completely hydr,qted, although the original limellar oi the twin structure ispreserved (Fig. 2l)." This strongly sugge!-ts that hydration has occurredtopochemically, as originally suggested by Funk$ and McConnell.3e

_ The ferrite phase appears virtually unreacted in bse images of polished surfacqs(Fig. 22), even in pastes many years old. close examination of s 23-year-old paste3T

Fig. 19. Bel i te crystal reveeledon lhe fracture surface of e l4-dry-old paste, showing signs oflocel rercl ion. (Courtesy A.Ghose.)

Fig. 20. Backscattered electronimage of bel i te crystal in e 3-month-old paste showingpreferential reaction along thetwin planes.

Fig. 21. Relic of fully reacledbelite crystal in 23-year-oldpaste. The original twin lamellascen sti l l be identif ied (bse image).

Flg.22. Rel ic of large grains in e one-year-old pasle ln whlch the ferr l te phase(whlte) ls apparent ly unreacted (bse image).

, !9rq ,

Flg. 23. Summary of microstructural development for a grain of cement. (a)

Unhydrated sect ion of polymineral ic grain (scale of Interst i t ia l phase ls s l ight lyexaggerated). ( r ) -10 min. Some C3A (and/or Fss) reacts wl th calc lum sul fateIn solul ion. Amorphous, aluminate-r lch get forms on the surface end short AFtrods nucleate at edge of gel l and in solut ion. (c) -10 h. Reaction of C3S toproduce nouler" product C-S-H on AFt rod nelwork leaving - l trrm belween graln

surface and hydrated shel l . (d) -18 h. Secondary hydration of C3A (rnd/or Fss)produclng long rods of AFt. C-S-H inner product starls to form on Inside ofshel l from continulng hydration of C3S. (e) 1-3 days. C3A reacts with any AFt

lnslde shel l forming hexagonal p lates of AFm. Cont inulng format ion of " lnner"product reduces separation of anhydrous grain and hydrated shel l . ( f) -14 days.

Suff lclent " lnner" C-S-H has formed to f i l l In the space between grain and shel l .The "outer" C-S-H has become more f ibrous. (8) -years. The remalnlnganhydrous mater ia l reacis by a s low sol id state mechanism, to form addl t ional" lnnern product C-S-lI . The ferr l te phase rppesrs lo remaln unreacted.

{\

' , . : . ;

* t ' t

Materials Science of Concrete Materials Science of Concrete140

L

t4 l

has revealed two distinct regions in the iron-rich areas. It is possible that there issome leaching of calcium and aluminum from the outer layer to leave amorphous,hydrous ferric oxide. If the interstitial phases are intimately mixed, the reaciion ofaluminate phase will probably be restricted.

. Th9 development of microstructure for a large grain of cement is surnmarizedin Fig. 23. After the first day or so, the net movehent of ions during the hydrationprocess is very small. This fact was noted by Taylor and Newburyao from theirelectron microprobe study of a cement paste which had been cured for 23 years. Inbse.images,of.this (Fig. 24) and other mature pastes, the relics of originai cementgrains can be identified. Ne.vertheless, the continued growth of calciuir hydroxideand some c-s-H in the originally water-filled space has gradually reduced thevolume and connectivity of the large pores.

Fig.24. Microstructure of a 23-year-old paste (bse image). This paste is fullyreacted apart from the bright regions of ferrite phase; these help to identify t irerelics of the original grains which now mainly consist of dense inner' c-s-H.Among the relics can be seen tortuous regions of calcium hydroxide and amixture of other hydration products (dark gray) and porosity (black).

Effect of Mineral Replacements on Microstructure

Various minerals are often used in concrete as partial replacement for cement.These minerals have some capacity to hydrate, usually reacting with calciumhydroxide, to give space-filling product and thus refining the porosity of the paste.The most widely used mineral replacements are fly ash, blast-furnace slag, and silicafume. The fly ash is generally used at replacement levels of up to about 30%. Forslag replacements, levels at around 40% and around 70% are used for differentapplications. With these two minerals the development of microstructure of the pasteduring hydration is only slightly modified from the pattern previously described.The replacement levels for sitica fume are much lower, typically around i096, but theeffects on the microstructural development are much more drlmatic.

Fly ash is a by-product of coal burning in power stations and consists of smallspherical particles with a similar particle-size distribution to anhydrous cement.About 80% of a fly ash is a reactive silica glass containing calcium, aluminum, iron,and other minor oxides. The crystalline content is typically mullite, ferrite, andquart?. In cement paste, the glassy component of the fly ash undergoes a pozzolanicreaction with the calcium hydroxide produced by the cement to give C-S-H.

In terms of microstructural development, fly ash starts to react only after severaldays hydration, although impurities, such as alkali sulfates, in the fly ash may affectthe hydration of the cement earlier. Nevertheless, deposition of C-S-H and AFtfrom the reaction of the cement may be observed on the surflce of the fly ashparticles, almost immediately,.in the environmental cell (Fig. 25)al and after severalhours on fracture surfaces.'o''" After several days the dissolution of the glassy phase

can be seen on fracture surfaces, where the crystalline material in some of the largerparticles has been exposed (Fig. 26).* Rims of hydration product can also be seenaround the fly ash particles on fracture surfaces and in bse images of polishedsections (Fig. 27).

Fig. 25. Cement/f ly esh paste(70/30\, hydreted for l0 min(TEM environmentel cel l) .Product cen bc seen on thecemeot greins and deposited onthe surfaces of the spherical f lyash part icles.

Fig.26. Large part icle of f ly ashon the frecture surface of a 28-

day cement/f ly ash paste.

Reaction of the glassy phase in

the f ly ash has exposed somecrystal l ine material. (Courtesy ofY. Halse.)

Fig. 27. Backscattered electronimage of a 69-day-old cement/flYash paste. Several particles of flyesh ere visible in a massive regionof calcium hydroxide, some ofwhlch show signs of reaction.

t42 Materials Science of concrete Materials Science of concrete r43

Fig- 28. (Left) Backscattered electron image of cement/slag paste (30/20)hydrated for one day. A large mass of calcium hydroxide hii grown around rcollection of angular slag grains. (Courtesy of M. p. Cook.)

Ftg. 29. (Right) Partially hydrated cement grain (top) and sleg grain (bottom) inan 8-month-old cement/slag paste (60/40). The different grey level of thehydration rims indicates their different composition (bse image).

Blast-furnace slags have a glass content of about 90% glass of typicalcomposition: 35% SiOz, 45% CaO, l0% Al2O3, 8% MgO, and small amounrs of-otheroxides. Although slags have some inherent hydraulicity, an activator (such as calciumhydroxide) is needed and the slag plays l itt le active part in the early microstructuraldevelopment. However, it does appear to have some effect on the distribution ofcalcium hydroxide. In bse images of slag/cement pastes (Fig. 28), it is noticeablethat large masses of cH have grown around the slag particles. over a period ofseveral months the slag slowly reacts and the amount of calcium hydroxide decreases.In old pastes, reaction rims can be seen around the slag particles (Fig. 29); these aredarker (with a lower average atomic number) than the hydration product around thecement grains and conlgin significant quantit ies of Mgo. Such layers have also beennoted by Tanaka et al."' Analytical work by Harrison et al.6 sug,gests that these rimsare composed of C-S-H mixed on an atomic scale with another product, possibly ahydrated magnesium aluminate of the hydrotalcite group.

Silica fume consists of very small particles (s0.1 pm) of amorphous silica. Thehigh surface area of the fume means that it reacts much more rapidly than fly ashor slag. As it is very difficult to disperse the silica fume, superplasticizers are ahvaysadded to such mixes in practice. Several workers have noted that the microstructureof pastes contgiq[ng silica fume is very different from that developed in ordinarycement pastes,'"* having a dense amorphous appearance. The bse image of a maturecement silica fume paste (Fig. 30)4v illustrates this dense microstructure. Despite theuse of superplasticizer, some clumps of fume are still present (these are marked Si inthe micrograph). Some large cement grains appear to have hydrated to leave hollowpores containing unreacted ferrite phase. The reasons for these differences inmicrostructural development are not fully understood at present. Silica fume alsoappears to have a profound influence on the microstructure of the interface betweencement past and aggregate, which will be discussed subsequently.

Fig. 30. Polished section of a mature cement/sil ica fume paste (85/15). Clumpsof sil ice fume cen be identif ied (Si). No massive regions of celcium hydroxidecan be seen. Some of the large pores ere associated with the hydration of cementgrains to leave hollow shells of hydration product.

Characterization of the Overall Paste Microstructure

In the previous sections the development of microstructure was described mainlyin terms of the local changes which occur around individual grains of cement duringhydration. The resulting microstructure for a given paste wil l depend on:

l. The phases present and their distribution within the grains.2. The grain-size distribution.3. The amount of reaction (related to age, moisture state, and temperature).4. The water/cement ratio.

The first two are determined by the anhydrous cement powder, whosecharacterization has already been discussed. The grain size, in conjunction with thedegree of reaction, is particularly important in determining the distribution of theC-S-H product in the microstructure. As previously indicated, grains smaller thanabout 5 pm (about 6 wt% of the total) will hydrate completely during the major heatevolution peak to give "outern C-S-H in the originally water-filled space. Grainsbetween about 5 and l5 pm (about 30 wt%) will have reacted after about seven daysto give "outer'and 'through solution, inner" C-S-H. Only the cores of grains largerthan about l5 pm will then eventually hydrate to give "solid state, inner'C-S-H.

The water/cement ratio determines the amount of water-filled space in the freshcement pastes; for tw/c of 0.5 this is about 60%. At early ages the effect of the w/cratio on the appearance of fracture surfaces is quite dramatic (Figs. 3l and 32).After l8 hours, the fracture of the cement paste with w/c = 0.5 is almost completelyinterparticular (Fig.3l), whereas, in the paste with w/c = 0.3, the closerspacing ofthe cement grains means that more bonding has occurred and there is more fracturethrough the shells, revealing more anhydrous cores on the fracture surface (Fig. 32).At later ages the effect of w/c ratio on the microstructure is less obvious.

t44 Materials Science of Concrete Materials Science of Concrete t45

Fig. 31. Fracture surfrce of cement paste with w/c = 0.5, hydratedThe fracture hes been predominsnlly interparticular.

for 18 h.

o w/COll

o Yrcoat

ANr.o

-"* '

o

o lY/co{t

r w/C Oll

0.a

0.2

Iz

0.1

0.q

f "

dot

0.1

0.0 +0.0 o;2 o:3

P,*"

f ig. 33, Analysis of microstructural constituents in cement paste. (a) Bse imagefrom 28-day-paste; (D) gray-level histogram, showing peaks for: anhydrousmaterial (AN), calcium hydroxide (CH), and other hydration products; (c)anhydrous material; (d) calcium hydroxide; (e) porosity; (f) comparison ofanhydrous material measured by image analysis (ia) and by loss on ignition(LoI); (g) comparison of porosity from image analysis and methanot absorption.

the bse-images at the magnification used (x 400). Nevertheless, there was goodproportionality between the two sets of measurements for pastes at a wide range ofages and w/c ratios (Fig. 33(g)). It should also be possible to determine the ambuntof calcium hydroxide by this method, but, because of the uneven distribution of thisphase, measurements must be made over a larger area than was done in this study.

The importance of the analysis of bse images lies in the potential it offers forobtaining quantitative descriptions of the distributions ol the microstructuralconstituents. The choice of suitable parameters to characterize a distribution mustbe considered carefully and the best strategy will depend on the nature of thedistribution and the end purpose of the characterization. It is also important tounderstand the stereological problems involved in trying to derive information aboutthree-dimensional distributions from the study of two-dimensional sections.

The anhydrous material exists as discrete particles. In this case the distributionof particle sizes might be suitable for some applications, while for others the total

O:l0.20: l

Fig. 32. Fraclure surface of cement paste with w/c = 0.3, hydrated for lg h. Inthis case there has been more fracturi through the hydreie ih"l lr, revealing theseparation between these shells and the anhydious cores.

Given these various. factors, quantitative descriptions are necEssary forrnicrostructural characterization. The simplest meani of quantification'is todetermine the relative proportions of the Constituents p.esent. The gray-levelcontrast between the microstructural constituents in bse images makes it p6ssi'bte forthem to be discriminated by an automatic image analyzer, is illustrated in Fig. 33.This method can be used to measure the quantlties of anhydrous material, missivecalcium hydroxide, other hydration products, and pores. If a sufficiently large areais-analyzed, the quantities determined will be rlpresentative of the

-uutt-paste.

Although the same quantities can be measured by indirect techniques, sivera!techniques must be used to measure all the constituenis. A study by Scrivener et a1.22showed that measurement of the amount of unreacted anhydrous material by analysisof .b..se iqrages gives comparable results to values measured by loss on igniiion (rig.33(/). The measurements of porosity by the image analysis rnethod were lower iha-nthose obtained by methanol replacement, owing to ttre timitea resolution (0.5 pm) of

t46 Materiats Science of Concrete Materials Science of Concrete t47

surl'ace area of the grains or the uniformity of their distribution in space might bea more appropriate description of the microstructure.

In contrast, the pores in cement paste are not discrete and the connectivity of thepore space has an important influence on such properties as permeability. In thiscase the pore-size distribution and connectivity measured in two-dimensional sectionsmay bear little or no relation to the three-dimensional pore structure. For somepurposes, characterization of the.pore surf?^cermay be more appropriate and hereconcepts such zrs fractals may be important.ru'

The distribution of calcium hydroxide poses similar problems due to the tortuousshape of the crystalline masses. Here the density of nucleation sites might be animportant parameter, which could be estimated by analysis of nearest-neighbordistances.

It is clear that there is much to be done in this area, but with appropriate basicwork it should be possible to establish ways of characterizing the microstructure ofcement paste, which will further the understanding of its behavior.

Aggregates

A very wide range of aggregates is used in the production of concrete.

!mpirically, several factors have been established as affectigg the workabil ity of thefresh concrete and the durabil ity of the hardened structure." These factors may besummarized as follows:

Physical characl.eristics of the aggregate particles: the particle-size distribution(gradingX the particle shape; the surface texture of the particles.

Impurities and contaminants: Clay, silt, and dust; organic matter; mica, chalk,shell; sulfates and chlorides; metall ic impurit ies, in particular lead and zinc oxides.

Mechanical and chemical stability: mechanical strength; susceptibility to reactionwith alkalis; soundness with respect to freezing and thawing.

During the early stages of hydration, the aggregate does not play an active rolein microstructural development. Thus, in the absence of significant amounts ofimpurit ies or contaminants, the physical properties of the aggregate are of mostimportance. In particular, the grading of the aggregate wilt affect the packing of theparticles in the concrete and the surface area of aggregate which must be covered bycement paste. It is difficult to characterize features such as particle shape andtexture in a quantitative manner; in particular, a surface may be described as roughon several different scales.

Examples of two commonly used aggregates are shown in Figs. 34(a) and (b).The aggregate in Fig. 34(c) is a natural gravel aggregate; the particles are generallyrounded and have a wide range of surface textures. The other agg,regate is a crushedgranite rock. In this case the particles are angular and the surface is fairly uniform.The speckled appearance indicates the polymineralic nature of this aggregate. Figure35 shows the surface of a fine aggregate sand taken by SEM; some particles are fairlysmooth, while others are pitted.

Over periods of several years, some aggregates may undergo chemical reaction.The most serious of these reactions is the aggregate alkali reaction (AAR), in whichcomponents of the aggregate react with alkalis in the cement paste to give a gel,which may imbibe water and swell, causing cracking in the concrete. The AAR ismost commonly caused by the presence of amorphous or microcrystalline silica in theaggregate, but has also been reported with carbonate aggregates.

Fig. 34. Photographs of (c) gravel and (D) crushed granite. (Courtesy of A. K.Crumbie.)

Fig. 35. Micrograph of sand particles ln the SEM (bar = 196 U.1.

Cement Paste/Aggregate Interface

The interface between cement paste and aggregate cannot be studied in isolationas it exists only as a localized region in the microstructures of concretes and mortars.This makes microscopical techniques well suited to the study o-f. this region.Examinations of fracture surfages (e.g., Ref. 53) and polished surfaces)" by SEM, andion-thinned sections by TEMr"rd have all been reported. However, because of thedifficulties in obtaining quantitative information from microscopical studies, manyworker;(e.9., Refs.55-60) have used the "model'concrete specimen, developed byFarranot to study the interface. In this specimen, cement paste is cast onto a largeflat piece of aggregate. After a suitable period of hydration, the specimen can thenbe fractured at the interface and the surfaces so produced examined by SEM. The

3# wffiy4* +tEilt

*#w=

- j*

Materials Science of concrete Materials Science of concretel4E

L.

r49

!rg. J6. Leyer of productdeposited on rggregrte surfece infour-dey-old mortrr.

Concentration and orientation of phases in the tnterta9lau:egton can also De analyzeO

by the technique developed by Grandet and Ollivier,r)-r ' which involves successive

abrasion of the cement paste side of the interface and examination of the phases

Dresent on the surface by X-ray diffraction. This type of specimen also facilitates

microstructural examination by SEM, when it is fractured along the interface. As

a model for concrete there are several disadvantages to this specimen configuration.

First, the surface of the aggregate is flat and often polished; second, the aSSregate

is not present during the mixing process, and third, there is a considerable thickness

of paste in contact with the aggregate as opposed to the narrow bands of paste

between aggregate particles in concrete.Microstructural studies indicate that a thin layer of hydration product (sl pm

thick) forms on the -slrface of the aggregate. From their studies of cement paste cast

against glass slides,o'Barnes et al. suggested that this is a duplex layer with a thin

film of calcium hydroxide adjacent to the aggregate surface surrounded by a layer

of C-S-H. However, Scrivener and Pratt! found no evidence of the calcium

hvdroxide film in their study of mortars examined as fracture surfaces (Fig. 36) and

as polished sections with bse (Fig. 37). Even with the higher resolution possible with

transmission electron microscopy, there appears to be only qSlr.nple layer of product

in contact with the aggregate in the work of Javels et al.""o (See Fig. 3E.) The

deposition of this product layer does not entail any chemical reaction with the

aggregate, which merely acts as a nucleation site. However, at later ages there may

bi1 chemical reqgt_i^on between this layer and some carbonate aggregates, producing

a stronger bond.r6'orBeyond this layer of product there is a region of high porosity which is known

as the naureole de transition.' This arises because the large anhydrous cement grains

cannot pack ctose to the aggregate particles. In composite specimens this region is

charactirized by the prefeired orientation of calcium hydroxide with--th-e^ c axisparallel to the aggregate surface, and by inc.eased concent;ations of AFt55-s8 which

extend some 50 pm into the cement paste. On fracture surfaces of concretes and

mortars, this region can be seen through openings in the surface layer where it is

incomplete or has remained attached to the aggregate particle (Fig. 39). In this region

there are many "rosettes" of C-S-H from the hydration of small grains and a large

number of long AFt rods. In bse images of polished sections, the distribution of

cement grains Jan be observed directly or deduced from relics in mature specimens.

Figure 4-0 shows the interfacial region in a one-day-old mortar. There are few large

griins close to the interface and, in this case, therl appears to be a concentration of

imall grains in the crevices of the aggregate particles, which probably accumulated

as the-mortar was mixed. The numiioul small grains close to the interface lead to

many hollow hydration shells being observed in this interfacial region which can still

be sien in mature specimens (fig.4f ). It is perhaps no coincidence that Hadley first

observed hol-low hydration stretti (Hiatey grains) while studying interfacial regions

il-;;n;;"t..5{ ns-oiscussea eailiet, t'treie tenas to be a higher proportion of

interstitial phases exposed on the surfaces of smaller grains, which- would account for

trte rrigii c,in..nir"tion oi nrt observed in this teglon. In. bse image::.crystary- 9fcalciuit hydroxide can be seen in the interfacial region (e.g., Fig. 40), but tt ts

difficult to identify any preferred orientation.

Fig. 37. Microstructure et edgeof send grein In one-day-oldmortr. There is e layer of C-S-H product elong the surface ofaggregete.

Fig. 38. Contact zone betweeucalcite and cement paste, ion-beem-thinned section of r one-year-old sample. (Courtesy J-POl l iv ier . )

Fig. 39. Area of frecture surfecein e one-day-old mortar, edJacenlto en aggregete perticle. Beyondthe surface leyer (some of whlchedhered to the eggregate durlngfncture) can be seen 'rosettes' ofC-S-H from the hydration ofsmall cement grains and long rodsof AFt.

Itl

llI

tll

9fin

r50 Materials Science of Concrete Materials Science of Concrete l5l

Fig. 40. (Left) Backscattered electron image of interfacial zone of a one-day-old mortar. There are few large cement grains close to the aggregate surface.There are many hollow shells of hydration product in the interfacial region. Inthe crevices of the aggregate surface, small grains of cement have accumulatedduring mixing, leading to more product in these regions. Several massive regionsof calcium hydroxide (l ight gray) are also apprrent with e range of orientations.

Fig.4l . (Right) High-magnif icat ion bse image of the interfacial region in a l0-week-old composi te specimen (Ref. 65). Even in th is mature paste, hol low shel lsof hydration product from the reaction of small grains can sti l l be seen.

As the microstructural constituents can be distinguished in bse images ofpolished sections of concretes and mortars, this technique can be used to characterizethe microstructure of the interfacial region quantitatively in real concretes andmortars. With an automatic image analyzer, the relative amounts of the constituentscan be 4gasured in bands spreading out from the aggregate surface, as i l lustrated inFig. 42.* However, with random sections of concrete, the angle at which theaggregate surface intersects the plane of the section is unknown, and the normaldistance from the aggregate surface to a point in the sectio_q can only be estimated.

In order to avoid this problem, Scrivener and Gartneror prepardd specimens inwhich single pieces of aggregates were cast into blocks of cement paste. Sectionswere then cut from these specimens, normal to the aggregate surfaces. These sectionswere analyzed by the technique described above, averaging the results for each bandover l0 fields covering about 200 pm length of the aggregate surface. With thisnumber of fields there was considerable scatter in the results. but there was a cleardecrease in the amount of anhydrous material and increase in the porosity in aninterfacial zone some 30 to 50 pm deep (Fig. a3). A slight increase in the amount ofcalcium hydroxide close to the interface was also observed but this was not signifi-cant with regard to the scatter in the results. In terms of simulating concrete, thesespecimens have an advantage over composite specimens which use a flat piece ofaggregate, by having an unprepared aggregate surface. However, the aggregate isstill not present during mixing. The accumulation of small particles at the aggregatesurface seen in Fig. 40 indicates that the movement of aggregate particles through thepaste during mixing has an effect on the microstructure of the interfacial zone.

Fig. 42. Measurement of microstructural gradients in the interfacial region. (c)

Bse image with aggregate part icle at the top of the micrograph; (6) porosity detectedin cement paste; (c) black l ines indicate the bands in which the microstructural

consti tuents are measured.

Ftg. 43. Microstructural gradient in the interfacial region of 10-week-oldcomposite specimens examined by Scrivener and Gartner (Ref. 65). The resultsplotted ere the average measurements from 10 fields in each of three differentaggregates. (a) Anhydrous material, (D) porosity.

Materials Science of Concrete Materials Science of Concrete

ogtrocootroooo

l l ! l l l

r r r r r ar t t

r52

-3--

153

Concrete (ogg/cem 4:1)zgdoys w/c = O,1

ANHYDROJS Do"--o

o

pm

ANHYDROUS

Otr_toooooooootroootroo

PORO SI TY

aatt t r l r l l t l l l t r r r r

b) c)

Fig.44. Microstructural gradients in the interfacial region of 28-day-oldconcretes, average of 50 f ields. (a) Anhydrous material, (D) porosity, (c)

standard errors.

Despi te the problem of the unknown angle of intersect ion of the aggregate wi ththe plane of the section, it might be possible to apply this method to concrete if ,4large number of areas are analyzed. To investigate this possibil i ty, Scrivener et al.mlooked at the variabil ity of the results when a large number of f ields, selected in anuniform randomn manner, are analyzed. Some of the results from this work areshown in Fig.44. In these graphs, the distances from the aggregate surface measuredin the plane of the section have been estimated from the average angle of intersectionof the aggregate surfaces with a random section. The proportions of the constituentschange smoothly and the standard errors for the measurements are all around l0% ofthe measured values. From this preliminary study it appears that this method couldbe developed as a means of characterizing the interfacial zone in real concretes andused to investigate the effect that factors such as mix design or direction of castinghave on this region.

All the quantitative studies, both on composite specimens and in concretes,indicate that the width of the interfacial zone is between 30 and 50 pm. From Fig.4 it can be seen that only about l5 to 20% of anhydrous grains are larger than 50 pm.This seems to indicate that the essential feature determining the interfacial zone isthe packing of the anhydrous cement grains against the "walln of aggregate. However,

the type ot'specimen configuration, and especially the presence of the aggregate

during mixing, may affect the gradients of microstructure in this zone.

Effect of Silico Fume on the Interface

There is considerable interest in the incorporation of sil ica fume in mixes to

oroduce high-strength concrete. This replacement material appears to have a

brofound effect on the microstructure of the interfacial region, consistent with the

microstructure observed in sil ica fume/cement pastes. Figure 45 shows the

microstructure of the interfacial zone in a concrete containing sil ica fume; the

eenerally dense appearance of the microstructure is apparent. However fairly large

6ores are sti l l present in the form of hollow hydrate shells. In contrast to theiquivalent paste, the sil ica fume was well dispersed in the concrete, due to crushingor shearing -b-etween

the aggregate parlicles, and no clumping of the sil ica fume was

observed."'o' Scrivener and Bentur*t analyzed microstructural gradients of the

interfacial zones in high-strength concretes with and without silica fume. Thecradient in the amount of anhydrous material was in both cases similar to that5bserved in the other studies described (Figs. 43(a) and 44(a)). This is consistentwith the view that the particle-size distribution of the cement determines the widthof the interfacial zone. The porosity gradient was much shallower in the concretecontaining sil ica fume, as indicated by the appearance of the micrograph (Fig. 45);this suggests that the small particles of sil ica fume accumulate in the interfacialregion during mixing.

Overall Microstructure of Concrete

The microstructure of concrete is that of a composite material, composed ofaggregate, cement paste, and the interface between them. The complexity behind thisstatement can be appreciated by observing the sequence of micrographs shown in Fig.46, each of which was taken of the same area, at about four times the magnification

Fig. lS. Interfacial region of sil ica fume concrete. The microstructure is denseand there lre no clumps of sil ica fume as seen in the cement/sil ica fume paste(Fig. 30). The hydrati,on of some small cement grains has lefi some enclosid pores.

o

s-

E

e&

oc4

z

Materials Science of Concrete Materials Science of Concrete

tOdistf f ie l rm intsloce

r l l r l r laatr l l

154

>-.

r55

rillthe picture there is very little anhydrous material present. At x 400 themicrosttucture of the paste can be seen in more detail and the distribution of thelarger pores can be seen. The next micrograph (x 1600) shows the microstructurebetween two hydrating cement grains and the final micrograph (x 6000) gives animpression of the morphology of the hydration products.

The characterization of the concrete microstructure entails the characterizationof the cement paste and the paste/aggregate interface, which have already beendiscussed. In addition, the distribution of the aggregate particles and the incidenceof features such as air voids and bleeding must be considered. In the case of realconcrete structures, the microstructure may vary throughout the cast section. Inparticular it is essential to take account of the microstructure of the material near thesurface, whic! p-lays a crucial role in protecting the reinforcing steel.

Kreijger""* discussed the phenomenon of the concrete skin, which occurswhenever concrete is poured against a barrier or is leveled flat. This skin consists ofan outer layer of cement paste some 0.1 to 0.3 mm thick, beneath which is a 3 to 5mm thick layer of mortar. The water/cement ratio also varies in these surface layers.In addition, the gradients will also occur in the mois.t^u;p stare of this layer as waterevaporates from the surface, as discussed by Parrott."'''u These effects will producemicrostructural gradients extending some 50 mm into the concrete.

The fact that these phenomena produce variations of microstructure in spaceindicates that quantitative microscopy (both optical and with backscattered electrons)can play an important role in characterizing and understanding the surface region ofconcrete.

The microstructure of concrete continues to change throughout its l i fe, due notonly to continuing hydration but also to the effects of the environment. Inparticular, the ingress of other compounds such as carbon dioxide may have profoundeffects on the microstructure. The diffusion of such compounds wil l also producegradients in the microstruc-ture. I^n environments such as seawater, several differentionic species (e.g., Cl-, SO?-, Mg2+) diffuse at different rates and the gradients theyproduce must be considered individually.

Despite these many complexities, "good" microstructural characterization ofconcrete is not an unrealistic goal. In qualitative terms, the way in whichmicrostructure develops during the hydration of cement and concrete is fairly wellunderstood. What are needed now are more quantitative descriptions of themicrostructure, which can be related to its behavior. They wil l not necessarily entailcomplete quantification of every detail, but only by understanding the overall contextcan useful methods of characterization be established.

Acknowledgments

The author would like to thank Dr. K. D. Baldie, Ms. M. P. Cook, Ms. A. K.Crumbie, and Prof. P. L. Pratt for helpful discussions and the Warren Research Fundof the Royal Society for financial support.

References

lE. Hornbogen, 'On the Microstructure of Alloys,' Acta Metall., 32, 615-27(t984).

rlFig. 46. Microstructure of concrete (180 days): (c) nomlnal magnlfication (NM) x?5. A large particle of granite eggregate and seieral sma[ part'icles of send cenbe seen In e matrix of cement pasie. lr; NM x 100; the dis;ibutioo or

"ohio-i*,cemen-tin the paste cen !e seen-more cleerly. There is very littte anhydrouimaterial in the narrow-ribbon of paste betwlen two sand giains to the right ofthe microgrrph. (c) NM x 400; distribution of the catciui hydroxide an-d largepores in the cement paste can now be seen. (d) NM x 1600 microstructurebetween two partielly reected cement grains. (e) NM x 6000 detairedmlcrostructure of hydration products.

of the precedlng one (as indicated -by ttre white boxes). At the lowest magnification

(nominally x 25) a_large particle of lranite aggregate and several sand gra'ins can beseen in a matrix of cement paste. At a nom-i-nal-magnification oC

" 16o tne bright

anhydrous cement cores can be seen more clearly. T-heir distribution is uneuen; io.example, in the narrow ribbon of paste between the two sand grains on the right of

Materials Science of concrete Materials Science of concreter56

.E-.

t57

-tr. ts. w. 'r'ayror, chemistry of cement Hydration; pp. g5-l l0 in proceedings ofthe 8th International congress on the chemistry of cement, vol. I. Finep, Biazil,r986.

- .tH f. W, TaVlor, "Proposed Structure for C-S-H Gel, ' ,/. Am. Ceram. Soc.,69161 494-67 (t986).

_ ^ ̂ 's.. Diqmond, "Cement Paste Microstructure-An overview at Several Levels"; pp.

2-30 in Hydraulic Cement Pastes: Their Structure and properties. cemenr -aConc.rete Association, Slough, U.K. 1976.

'H. M. Jennings, 'The Deveroping Microstructure in portland cement'; pp.349-96 in Advances in cement Technology. Edited by S. N. Ghosh. pergamon, NewYorb.19E3.

oK. L. scrivener, "The Development of Microstructure during the Hydration of

Portl4nd Cement; Ph.D. Thesis, University of London, 19g4.'D. H. Campbell, Microscopical Examination and Interpretation of Portland

Cempnt and Clinker. Portland Cement Association, Skokie, IL, 19E6._ 'G. R. Long, 'The Use and Application of Optical Microscopy in portlandCement Clinker Researchn; pp. 39-50 in Characterization and Performance predictionof cement and concrete. Edited by J. F. young. Engineering Foundation, NewYork. 1982.

vE. Fundal, 'Optical Microscopy of Cement Clinkers and Raw Mixes.' FLSRevipw No. 25. F. L. Smidth, Copenhagen.

'"N. Thaulow, A. Damgaard Jensen, S. Chatterj i, p. Christensen, and H.Gudmundsson, "Estimation of the Compressive Strength of concrete Samples byMe11_s_of Fluorescence Microscopy," Nordisk Betong, i-n, Sl-SZ (19S2).

"H. N. walker and B. F. Marshall, "Methods and Equipment used in preparingand Examining Fluorescent Ultrathin Sections of Portland Cement Concrete." Cem.Con,cg. Aggs.. CCAGDP, 1, 3-9 ( 1979).

^'D. D. Double, "Some Studies of the Hydration of portland cement Using HighVoltqge (! MeV) Electron Microscopy,' Mater. Sci. Eng.,12,29-34 (1973).

' 'D. D. Double, A. Hellawell and S. J. perry, '1he Hydration of portlandCeqglt , ] Proc. R. Soc. (London), Sect. A. ,359,435-51 (197S).

'-H. M. Jennings and P. L. Pratt, "on the Reactions Leading to calcium Sil icate

Ijld.r,atgr calcium Hydroxide and Ettringite during the Hydration of Cemenr"; pp.II. l4l-II. l46 in Proceedings of the 7th International Congress on the ChemistiyofCemgnt, Yol. II. Editions Septima, Paris, 19E0.

'-K. L. Scrivener and P. L. Pratt, 'characterisation of portland cement Hydrationby Electron optical rechniques"; pp.35l-6E in Electron Microscopy of Materials.Edited by w. Krakow, D. A. Smith, and L. w. Hobbs. Materials Risearch societySym,posium Proceedings, Yol. 31, 1983.

'"T. N. Tiegs., 'Investigation of Ion-Thinned rricalcium Silicate pastes byTransmission Electron Microscopy'; M. sc. Thesis, university of lllinois, Urbana-Cha.qrpaign, 1975.

t'R. Javels, J. C. Maso, and J. P. Oll ivier, 'Realisation de lames ultra-mince demortier pour observation directe au microscope electronique par transmission," Cera.Cong. Res., 4, 167-76 (1974\.

'oR. Javels, J. C. Maso, J. P. Ollivier, and B. Thenoz. "Observation directe aumicroscope electronique par transmission de la liason pate de ciment-granulat dansdes mortier de calcite et de quartz,' Cem. Concr. Res., 5, 285-94 (197t).

19n. l. Oatgteish, p. L. pratt, and R. I. Moss, "Preparation Techniques and the

Microscopical Examination of Portland Cement Paste and C35," Cem' Concr. Res., 10,

665:75 (l9Eo).---nB.i. O"igt.irtt and K. Ibe., 'Thin Foil Studies of Hydrated Portland Cement,'

Cet4;Concr. Res' , I1,729-39 (198 l ) .-- 21K. L. Scrivener and P. L. Pratt, 'Backscattered Electron Images of Polished

Cement Sections in the Scanning Electron Microscope'; pp.-145-55 in Proceedings of

inr'?i-fr International Confere-nce on the Microscopy of Cements, International

CeUpnt Microscopy Association, 1984.--'-iLk. L. Scrivenir, H. H. Patel, P. L. Pratt, and L. J. Parrott, 'Analysis of Phases

in Cement Paste Using Backscattered Electron Images"; e9' 61.-79 in Microstructural

ilr"top'n"nt during Hydration of cement.. Edited by Leslie J. Str_uble and P. w.

;;;. Materials deseirch Society Symposium Proceedings' Vol. E5' l9E7'--- 'z3R.

H. Bogue, Ind. Eng. Chem. Analyt. Edn., I l4j 192 (1929).uL. Struble, "A Review of Clinker Analysis by QXRD'; pp. 3l-37 in

Characterisation and Performance of Cement and Concrete. Engineering Foundation,

NeW-York, 1983."- 2sR. L. Berger. G. J. C. Frohnsdorff, P. H. Harris, and P' D' Johnson,

"Application of X-Ray Diffraction to Routine Mineralogy Analysis of Portland

ie'nienf; pp.234-53 in Structure of Portland Cement Paste and Concrete. Highways

Resg?rch Board SR90, 1966'----Xlw. A. Gutteridge, "Quantitative X-ray Powder Diffraction in the Study of Some

Cementive Materials"; pp. l l-23 in Chemistry and Chemically Related Properties of

Cirn"nt. Edited by F. P. Glasser. Proceedings of the Brit ish Ceramic Society' Vol.

35, J-9E4.--'27K. L. Scrivener. "The Microstructure of Anhydrous Cement and its Effect on

Hydration'; pp.39-46 in Microstructural Development during Hydration of Cement.

f-dituA UV I-iitle J. Struble and P. W. Brown. Materials Research Society Symposium

Proqgedings, Vol . 85, 1987.AW. GirttarO and E. Nagele, "The Hydration of Cement Studied by Secondary Ion

Mass-spectrometry (SIMS'1,n Cem. Concr. Res., 13, E46-56 (19E3).2eC. W. Groves- "Poriiand Cement Clinker Viewed by Transmission Electron

Microscopv.' Cem. Concr. Res., 12,619-25 (1982).30K. i. Scrivener and P. L.'Pratt, "Microstructural Studies of the Hydration of

c3A and caAF Independently and In cement Paste'; pp.207-20 in The chemistry

aiA Cnemiially Relaied Properties of Cement. Edited by F. P. Glasser. Proceedings

of thp gritistr Ceramic Society, Yol. 35, 19E4.3lD. Menetrier, L Jawed,-f. S. Sun, and J. Skalny, "ESCA and SEM Studies on

Earlv C"S Hvdration." Cem. Conu. Res., 9, 473-82 (1979).gg.-J. patgteistr,-p. L. pratt, and E. Toulson,'Fractographic Studies of Early

Hydration products in Cement Paste,' J. Mater. Sci., 17,2199-2207 (1982).'3b:t.

D"it i irn, A. Ghose, H. lvf. Jennings, and P. L. Pratt, 'The Correlation of

MicrostructurJs wiin Setting and Hardening in Cement Pqs-tg"; pp. 137-43 in

International Conference on Concrete at Early Ages, Yol. l, 1982.sn. N.-ff.Ot.V,

;fttu Nature of the Pasti-Aggregate Interface'; Ph.D. Thesis,

SchqOl of Civil Engineering, Purdue University, West Lafayette,-IN, 1972' -SP. L. Pratt a;d A. G-hose, 'Electron Microscope Studies of Portland Cement

Microstructures during Setting and Hardening,' Phil. Trans. R. Soc. ( London), Sect.

, { ,310,93-103 (1983).

Materials Science of Concrete Materials Science of Concretel5E

Lr-

159

lrIIil

--H. F. W. Taylor, 'The Reactions of Cement Compounds with Water"; inProceedings of the lOth International Symposium on the Reactivity of Solids, Dijon.Else;yier, Amsterdam, I 984.

''K. L. Scrivener, "A Study of the Microstructure of Two Old Cement Pastes'; pp.389-93 in Proceedings of the Eth International Congress on the Chemistry of Cement,Vol.JII. Finep, Rio de Janeiro, Brazil, 1986.

*H. Funk, "Two Different ways of Hydration in the Reaction of p-Ca2SiO4 withWater at 25 to 120"C'; pp.29l-95 in Proceedings of the 4th International Symposiumon the Chemistry of Cement, Yol. I. Washington, National Bureau of StandardsMoqpgraph 43, 1962.

"J. D. C. McConnell, "The Hydration of Larnite and Bredigite and the Propertiesof th^e Resulting Gelatinous Mineral Plomberite," Mineral. Mag.,30,672-80 (1955).

*H. F. W. Taylor and D. E. Newbury, nAn Electron Microprobe Study of aMatnre Cement Paste,'Cem.Concr. Res., 14, 565-73 (19E4).

:lK. L.Scrivener and H. U. Jensen; unpublished work.a2A. Ghose and P. L. Pratt, 'studies of the Hydration Reactions and

Microstructure of Cement Fly Ash Pastes'; pp. 82-91 in Proceedings of a Symposiumon the Effects of Fly Ash in Cement and Concrete, Materials Research Society,Pi t tqpurgh, l9El .

"Y. Halse, D. Goult, and P. L. Pratt, 'Calorimetry and Microscopy of Fly Ashand Sil ica Fume Cement Blends"; pp. 403- l7 in The Chemistry of Chemically RelatedProperties of Cement. Edited by F. P. Glasser. Brit ish Ceramic Society, Stoke-onTrent- 19E4.

4Y. Halse, P. L. Pratt, J. A. Daziel, and W. A. Gutteridge, "Development ofMicrostructure and Other Properties in Fly Ash/OPC Systems," Cem. Concr. Res., 14,49r-98 f l984).

asH. Tanai.a, Y. Totani, and Y. Saito, 'structure of Hydrated Glassy Blast FurnaceSlag in Concrete"; pp.963-77 in Proceedings of the lst Conference on Fly Ash, Sil icaFur49, Slag, and Natural Pozzolans in Concrete, Yol. 2. Montebello, 1983.

*A. M. Harrison, N. B. Winter, and H. F. W. Taylor, "An Examination of SomePure and Composite Portland Cement Pastes using Scanning Electron Microscopywith X-ray Analytical Capabil ity'; pp. 170-75 in Proceedings of the 8th InternationalCongress on the Chemistry of Cement, Vol. IV. Finep, Rio de Janeiro, Brazil, 1986.

"'E. Dinsoyr, T. Mosberg, and J. Young, nlnfluence of Aggregates on the Strengthand Elastic Modulus of High Strength Mortars Containing Microsil ica'; pp.2ll-18in Very High Strength Based Cement-Based Materials. Edited by J. Francis Young.Materials Research Society, Pittsburgh, 1985.

*M. Regourd,'Microstructure of High Strength Cement Based Systems"; pp. 3-17in Yery High Strength Cement-Based Materials. Edited by J. Francis Young.Mate^rials Research Society, Pittsburgh, 1985.

aeK. L. Scrivener and A. Bentur, ;Quantitative Characterization of the TransitionZong in High Strength Concretes," Adv. Cem. Res., I t4l230-37 (1988).

"8. B. Mandlebrot, The Fractal Geometry of Nature. lV. H. Freeman & Co., NewYorf-, 19t3.

ttD. N. Winslow, "The Fractal Nature of the Surface of Cement Paste,' Cem.Cong4.Res., 15, El7-24 (1985).

"L. Collis and R. A. Fox (eds.), "Aggregates: Sand, Gravel & Crushed RockAggregates for Construction Purposes'; Report of a Working Party of the EngineeringGroup of the Geological Society of London. The Geological Society, London, 19E5.

t'8. D. Barnes, S- Diamoncl' ano w' L

Interfacial zone Around A;;;;;;;;t in portland cement Mortar,n J. Am.ceram' soc.,

ut Wt;l!tttti'*r and p. L. pratt, "A preliminary Study of the Microstructure of

theCement/AggregateBondinMortarsn;pp.466-7| inProceedingsofthe8thInternational congress or iir" iii".iiiry of cement, vol. III. Finep, Rio de Janeiro,

3t^3\li.3rt"u;0", and J. p. ollivier, ',Nouvelle methode d'etude des interface ciments-

eranulats"; pp. VtI.E5-'8b-in pioceeaings ,of the 7th International Congress cjn the

b",*,{ii*,'';'1.":i.Xll;,11,'"'"01':t1,ii3'lil"ii*;,'!f!r ,,a iate de cimentoortlandau voirin"i" des granulats"; Thesis, L'Universite Paul Sabatier de Toulouse'

FraDge, l98l '),J. Grandet and J. P. Oll ivier, "orientation des hydrates au contact des granulats";

oo. VII.63-6E in proceldings of itr.:ttt.tnte*ational Congress on the Chemistrv of

6""nilt l , v"t. tt. rait ions Septima, Paris' 1980'top. J. M. fUonterro anO i. rc. Mehta, 'Ettringite Formation on the Aggregate-

c"r$i }l'iii"ih[l T,'f.' I f,ll;,' i;i T- g: $ilil;.,'rhe A g g re gate - Mo rtarInterface," Cem. Concr. Res', 15, 9-53-58 (1985)'

nL. struble "no

s. Mindi.i, ;'ruro.pnorogy of the cement-Aggregate Bond," Ittl.

t . Ceir . Comp. Lightw. Concr ' ,5, 79-86 (1983) 'orJ. Farran, "Contri luiio.i

-nline.atoiique' a I 'etude de I 'adherence entre les

coostituants hydrates des ciment et les materiaux enrobes," Rev' Mater' Construct"

No. 490-491 -492 (t956).^'" utB: D:'Ba;;;, s. ' i iamond, and w. L. Dolch, "The Contact zone Between

portland cement purt" *Jciurr'.Aggr"g"te' Surfaces," Cem. concr. Res., 8, 233-43

(t9?69| J. M. Monterio and p. K. Mehta, "Interaction Between Carbonate Rock and

Cement Paste,n Cem' Concr. Res', 16, Df ,-34, (1986)'

sK. L. Scrivener and p. L. pratt, "rne citaracierisation and Quantif ication of

cement and concrete Microstructures"; pp.6l-68 in Pore Structure and Mater ia ls

it"i leii i . i , voL. t. Ctt"ptan &-Hall, Lo.ndon' 1987'orK. L. Scrivener and E. M. Gartner, "Micrbstructural Gradients in Cement Paste

Around Aggregate i.'i,-"rti"f.r"; pp. ti-tO in Bonding In Cementious Composites'

Edited bv S. Mindesr ;fit:'i:3f;il. r"rateriars Reseaich Societv, Pittsbursh'""'q.."L."ilj;;;, A [. Cr.rt-i.,

"nO p]'L.E"it, "n Study of the Interfacial

Region Between Cement Paste and Aggreqate in ioncretes"; pp' 87-E8 in Bonding in

cementious composites. Edited bv s. Mindesr "na

s. P. sh;ir. Materials Research

Siiie,rSytp*ium Proceedings l-l 4 Pittsburch' I 9E8'o,p. c. Kreijger, "fi;:skfi;"f cbncrete ilrl"irrt N""ds,^ Mag. concr. Res.,39'

lr40Ir22-23 (1987)..P.c.Krei jger, 'TheSkinofConcrete:Composi t ionandPropert ies, ,Mater.

Conslruct.,17 il001 275-E3 (1984)'oyl,. J. parrott, "rrr"u*rJ-"* and Modelling of Moisture, Microstructure and

Propert iesinDryingConcre-te";pp'135-42- inPoreStructureandMater ia lsi;;il;ii;;, %r. r. ciapman & Hall, Lo-nd-on' l9E7'

,,L. J. parrott, D. c. Killoh, and R. c. p.ih,il.ment Hy^dration under Partially

Saturated Curing Conditions'; pp' a6--5o.ii';;qt;;G "l

tjt" Eth International

Congress of the Chemistry on Cement, Vof . fif.'iitipl-iiio Oe Janeiro' Brazil' l9E6'

L

l

l 'l

l

ill

Materials Science of ConcreteMaterials Science of Concrete160

.L-

l6 l