Computational costs with the use of the initial guesses formed by the divide-and-conquer approach are examined. For conventional density functional theory calculations, their computational costs rapidly increase as the system sizes increase, therefore, an efficient methodologies to construct a fine initial molecular orbital guess or a fine density matrix (DM) is quite variable in order to enhance the fast convergence. We examined the DM construction approaches by changing the subsystem buffer sizes and number of divisions, and their computational costs were numerically investigated in order to clarify their characteristic features. (C) 2015 Elsevier B.V. All rights reserved.