By Wendell Dubberke
Since the Duggan test method page was completed prior to this page, a considerable amount of information regarding ettringite in concrete has already been written there. You may want to visit that page before continuing here.
The Duggan test method can be used to predict the potential for the generation of excessive ettringite in a portland cement concrete (PCC) mix design.
Delayed ettringite formation (DEF) is sometimes called an internal sulfate attack. An external source of sulfur is not required for this type of early PCC deterioration to occur.
The type of ettringite formed can appear to be crystalline, but when analyzed by qualitative x-ray diffraction (QXRD), it is amorphous. In some cases, crystalline ettringite can also form. Extreme care must be exercised when quantifying and qualifying ettringite in concrete when using the x-ray diffractometer. Some purists claim that if it can not be identified by QXRD, then it is not ettringite, which is technically true. But what if this ettringite is a gel (or proto-crystalline), similar to the relationship of silica gel to quartz?
Element mapping with the scanning electron microscope (SEM) shows this material (ettringite gel?) to be mainly a calcium sulfoaluminate. True crystalline ettringite has approximately 32 water molecules tied to each calcium sulfoaluminate molecule. What the water arrangement in an ettringite gel would be is unknown, but researchers in California have worked on that problem. The SEM can not be used to quantify water in the ettringite gel because hydrogen is too light to detect. Thermogravimetric analysis (TGA) should be able to solve this problem.
This fiberous, opaque, amorphous ettringite can be easily seen in micro-cracks and in air voids of modern PCC. It is especially prominent in some PCC pavements suffering from very early ( 4 years in some cases) deterioration.
At Iowa State University (ISU) in Ames, Iowa, the x-ray diffractometer was used in an attempt to identify this material (ettringite gel) that was accumulating in the air voids of early deteriorated portland cement concrete pavements (PCCP). The ettringite gel material (fibous and like a thick paste) was removed from the air voids and put on a quartz sample holder. A one hour run, concentrating on the zone where the ettingite peaks should occur, showed only a halo associated with amorphous material.
Later, a PCCP researcher from Minnesota used the SEM and diffractometer to analyze 20 year old, severly deteriorated PCC. From the fractured surface of the concrete, hard balls of ettringite were removed. These hard balls of ettringite were from air voids over 100 microns in diameter. Some of the ettringite balls were clear and some were opaque. The clear ettringite balls and the opaque balls were pulverized and analyzed by XRD. Both tests showed the material to be amorphous. The ettringite balls, after being pried out of their air void sockets, had the appearance of the surface of a human brain.
I is my opinion that the ettringite fibers, found in air voids, are growing at the base rather than at the top and that individual fibers come together to form bundles. When looking at air voids completely full of ettringite, it appears that a squashing from the outside edge has taken place. Also, when examining (with the SEM) laboratory concrete, containing 15% fly ash, a few weeks after fabrication, fly ash spheres, lifted off the surface of the air voids, could be seen sitting on top of ettringite bundles. Initially, it was thought that ettringite might be extruding from pores in the air void surface. Another explanation would be that the ettringite is forming at the surface and pushing the ettringite fiber up. This explanation is not as absurd as it first sounds. Fresh water ice domes occur in salty lakes in the upper Andes. Fresh water seeps upwards through the base material of the shallow, salt water lakes, where it then combines with bulk ice, forcing the ice islands to rise above the surface of the cold (below freezing temperature) salt water.
PCC researchers in Demark have analyzed this PCCP ettringite material for a longer period of time on XRD equipment and have said that they were able to identify the ettringite peaks. It was not disclosed if these were small peaks on top of an amorphous halo.
Researchers from England, Schlumberger Research have published reports that indicate phosphonate concrete retarders can "poison" ettringite gel and hinder its developement into crystalline ettringite. Are other dispersent/surfactant concrete additives able to do likewise?
After observing this amorphous ettringite over time, identification can usually be made from its physical appearance. On occasion, a thin coating of silica gel can cover ettringite and cause some confusion when doing element mapping with the SEM. An element map of the cross section will usually solve the problem as well as determine the sequence of events.
Some PCCP petrographers, using optical equipment, are apparently confusing ettringite gel for silica gel. The SEM can readily separate ettringite gel from silica gel, contrary to the opinion of some petrographers. In fact, the SEM will quickly and easily identify the alkalis in the silica gel. In addition to sodium and potassium, calcium was seen quite often as the major metal in silica gel.
Many samples of deteriorated PCCP, from other states, were analyzed with the equipment at ISU. Looking at modern, early deteriorated PCCP from southern states was particularly helpful in determining the role of freeze/thaw in Iowa's early deteriorating PCCP.
PCCP from I-20 east and west of Monroe, La was analyzed with ISU's equipment. I-20 began to deteriorate in less than 5 years. Both sections east and west of Monroe deteriorated equally, but only one section contained fly ash. The coarse-aggregate in the concrete contained some argillaceous (shaley) particles. These particles would not perform well in Iowa PCCP (freeze/thaw cycles & deicing salts) but apparently can perform well in Louisiana PCCP as service records attest. Images and element maps from the SEM showed a considerable amount of ettringite in the PCCP void system. Other petrographers, who analyzed this PCCP, said there was little or no ettringite observable even though a single SEM element map, published in their report, showed air voids full of ettringite. A water reducer/retarder was used on the projects.
PCCP from Wisconsin was also analyzed. The type of deterioration was quite different from the other early deteriorated PCCP. The early deterioration in the Wisconsin pavement took the form of a 5 inch deep "V" trough at the sawed joint. A mostly igneous gravel (good service record) was used for coarse-aggregate in the PCCP mix. The cement was relatively high in the amount of sulfur and potassium. SEM analysis showed the air voids to be full of ettingite. Wisconsin DOT engineers said other roads, built at the same time with the same design, but with the use of a carbonate coarse-aggregate, were performing adequately. While it is possible to have early PCCP deterioration problems caused by delayed ettringite formation (DEF), without the application of external heat, the application of heat can take a marginal situation over the edge particularly if the concrete matrix looses its integrity. The application of heat, in the Duggan test method pretreatment process, probably produces micro-cracks, in the matrix, that allow water to get to unreacted ettringite. Could the sawing of PCCP, made with igneous coarse-aggregates, and a mix susceptible to the generation of excessive ettringite relate to the cause of this very selective early PCCP deterioration? Anyone who has ever sawed concrete, made with igneous aggregates, has seen red hot particles. Materials in the sand fraction can also glow red-hot when sawed.
Some concrete investigators say that a general expansion of the matrix occurs when PCCP fails due to DEF. In-house studies at the Iowa DOT would support that conclusion. However, when observing aggregate particles in a polished PCC surface, ettringite does not have to completely surround the particles. Quite often the aggregate particles will remain in contact with part of their sockets. The amount of ettringite, if any, will depend on where the slice bisects the aggregate/paste.
PCC box beams from Texas were also studied using the equipment at ISU. These beams were made with type III cements from two sources. The coarse-aggregate was crushed limestone containing a minor amount of porous chert. The coarse aggregate had been used previously and had an acceptable service record. SEM images showed expansion taking place in a preferential (transverse) direction. Micro-cracks. filled with ettringite, were parallel to the edges of the beam. Porous chert becomes expansive when used in PCC, however, good PCC can tolerate some porous chert with no ill effects. The interior tensile strength of the matrix will exceed the expansive energy of the reacive chert, particuarly if the pore system in the matrix can handle the silica gel. The situation near the surface is different and popouts can occur. When general expansion of the matrix (due to DEF) occured in these box beams, some of the micro-cracks cut across some reactive chert particles. In these cases, the matrix and aggregate particle were both expansive. SEM element mapping of these areas showed silica gel in the reactive chert crack (just to its border) and ettringite in the matrix crack (just to its border with the chert particle. Neither the silica gel or the ettringite intruded into each others space, indicating nearly equal expansion energy.
The type III cement used in the fabrication of these Texas beams, like most other type III cements, contains more sulfur than type I cements. Type III cements are made to a finer grind to attain high early strengths when used in PCC. Most likely, fine-grind cements would have a greater problem with flowability (out of the silos) than coarser grinds. A potential problem (DEF in concrete) could occur if the extra sulfur in the type III cement is tied to potassium as arcanite. The amount of arcanite (K2SO4) in cement relates directly to the amount of syngenite generated in cement silos. How much (and how aggressive) grinding aid is needed to control flowability under these conditions? The use of a type III cement, containing a significant amount of grinding aid, in a mix containing a super-plasticizer could lead to DEF related problems. An Iowa PCC pavement, made with type III cement, failed prematurily.
The states of Ohio and New York have used type K (shrinkage compensating or expansive) cement in pavements and structures. Ohio's experience has been positive. New York has experienced some minor failures with structures. Filling the concrete pore system with ettringite while maintaining matrix integrity could be beneficial for initial strength and stopping chloride penetration. How do we guarantee long-term matrix integrity? Can we guarantee that all of the ettringite will form during the plastic stage of hydration or that water will never reach any unreacted ettringite in the future? Would remedial techniques, such as furnishing abundant reservoirs (air voids) for DEF, solve the problem? Ohio PCC, made with type K cement, was analyzed with the SEM. Image analysis showed the sample to contain nearly 10% air voids. Type K cement works for Ohio PCC.
ASTM specifications allow for the additional sulfur in type III cement because of the additional surface area of the C3A particles due to fine grinding. Gypsum, hemi-hydrate (bassanite) and/or anhydrite (calcium sulfates, some hydrous) are introduced into the operation during clinker grinding. The potassium sulfate (when present) is usually derived from the cement clinker.