Review of the splitting-test standards from a fracture mechanics point of view C. Rocco a , G.V. Guinea b, *, J. Planas b , M. Elices b a Facultad de Ingenierõ Âa, Universidad Nacional de la Plata, La Plata, Argentina b Departamento de Ciencia de Materiales, Universidad Polite Âcnica de Madrid, Madrid, Spain Received 3 March 2000; accepted 5 September 2000 Abstract This article analyzes by means of fracture mechanics the current splitting-test standards for concrete. The cohesive crack model, which has shown its utility in modeling the fracture of concrete and other cementitious materials, has been used to assess the effect of the specimen size, the specimen shape and the width of the load-bearing strips on the conventional splitting tensile strength, f st . The results show that, within the ranges recommended in the standards, the values of the splitting tensile strength can differ by up to 40%, and, consequently, f st can hardly be assumed to be a material property. Empirical formulae for concretes with different compressive strength (10±80 MPa) and maximum aggregate size (8 ± 32 mm) have been used to show that f st is nearly specimen independent only for certain compositions, such as high- strength concretes, or when the aggregate size is under 16 mm. New closed-form expressions for f st are given in this paper to incorporate the effect of material properties, and some recommendations are drawn to minimize the influence of the width of the load-bearing strips. D 2001 Elsevier Science Ltd. All rights reserved. Keywords: Tensile properties; Fracture; Concrete; Modeling 1. Introduction The splitting tensile test is used worldwide to measure the tensile strength of concrete. This test was first proposed by Lobo Carneiro and Barcellos during the Fifth Conference of the Brazilian Association for Standardization in 1943 [1], hence, its other denomination Ð the Brazilian test; now, it is a standardized test method included in the major interna- tional concrete standards such as ASTM C-496, ISO 4105, BS 1881-117 and others [2±4]. The main advantage of the splitting test is that only external compressive loads are required. A cylindrical or prismatic specimen is compressed along two diametrically opposed generators so that a nearly uniform tensile stress is induced in the loading plane. To prevent local failure in compression at the loading generators, two thin strips, usually made of plywood, are placed between the loading platens and the specimen to distribute the load. The induced tensile stress state causes the specimen to fail by splitting. The maximum value of the tensile stress, computed at failure from the theory of elasticity, is the splitting tensile strength, f st , ordinarily assumed in the standards to be a material property. It is important to note that specimens of different geometry, size and width of the load-bearing strips are prescribed in the various standards, whereas the splitting tensile strength is computed with a single formula that does not take account of these variables. The effect of the geometry and the width of the load-bearing strips have been analyzed theoretically since the early day of splitting test application [5±8], usually within the framework of the classical theory of elasticity. These works have shown that when the specimen types and the load-bearing strips recom- mended in the standards are used, the splitting strength determined via the standard equation only varies by 4%. A more refined analysis of the splitting test by the authors based on the cohesive crack model [9] has demon- strated that the results from the classical theory of elasticity are not applicable to cementitious materials such as mortar or concrete. From this analysis, it has been shown that the standardized f st can vary significantly with the specimen * Corresponding author. Departamento de Ciencia de Materiales, ETSI Caminos, Ciudad Universitaria s/n, 28040 Madrid, Spain. Tel.: +34-9-1- 336-6754; fax: +34-9-1-336-6680. E-mail address: gguinea@mater.upm.es (G.V. Guinea). Cement and Concrete Research 31 (2001) 73 ± 82 0008-8846/01/$ ± see front matter D 2001 Elsevier Science Ltd. All rights reserved. PII:S0008-8846(00)00425-7