Numerical values of the Chandrasekhar function are needed with high accuracy in evaluations of theoretical models describing electron transport in condensed matter. An algorithm for such calculations should be possibly fast and also accurate, e.g. an accuracy of 10 decimal digits is needed for some applications. Two of the integral representations of the Chandrasekhar function are prospective for constructing such an algorithm, but suitable transformations are needed to obtain a rapidly converging quadrature. A mixed algorithm is proposed in which the Chandrasekhar function is calculated from two algorithms, depending on the value of one of the arguments.Program title: CHANDRASCatalogue identifier: AEMC_v1_0Program summary URL: http://cpc.cs.qub.ac.uk/summaries/AEMC_v1_0.htmlProgram obtainable from: CPC Program Library, Queenʼs University, Belfast, N. IrelandLicensing provisions: Standard CPC licence, http://cpc.cs.qub.ac.uk/licence/licence.htmlNo. of lines in distributed program, including test data, etc.: 567No. of bytes in distributed program, including test data, etc.: 4444Distribution format: tar.gzProgramming language: Fortran 90Computer: Any computer with a FORTRAN 90 compilerOperating system: Linux, Windows 7, Windows XPRAM: 0.6 MbClassification: 2.4, 7.2Nature of problem: An attempt has been made to develop a subroutine that calculates the Chandrasekhar function with high accuracy, of at least 10 decimal places. Simultaneously, this subroutine should be very fast. Both requirements stem from the theory of electron transport in condensed matter.Solution method: Two algorithms were developed, each based on a different integral representation of the Chandrasekhar function. The final algorithm is edited by mixing these two algorithms and by selecting ranges of the argument ω in which performance is the fastest.Restrictions: Two input parameters for the Chandrasekhar function, x and ω (notation used in the code), are restricted to the range: 0⩽x⩽1 and 0⩽ω⩽1, which is sufficient in numerous applications.Unusual features: The program uses the Romberg quadrature for integration. This quadrature is applicable to integrands that satisfy several requirements (the integrand does not vary rapidly and does not change sign in the integration interval; furthermore, the integrand is finite at the endpoints). Consequently, the analyzed integrands were transformed so that these requirements were satisfied. In effect, one can conveniently control the accuracy of integration. Although the desired fractional accuracy was set at 10−10, the obtained accuracy of the Chandrasekhar function was much higher, typically 13 decimal places.Running time: Between 0.7 and 5 milliseconds for one pair of arguments of the Chandrasekhar function.