Before starting the td calculation, the initial states (that are read from the restart/gs directory, which should have been generated in a previous ground-state calculation) can be "transformed" among themselves. The block TransformStates gives the transformation matrix to be used. The number of rows and columns of the matrix should equal the number of the states present in the time-dependent calculation (the independent spin and k-point subspaces are all transformed equally); the number of columns should be equal to the number of states present in the restart/gs directory. This number may be different: for example, one could have run previously in unocc mode in order to obtain unoccupied Kohn-Sham states, and therefore restart/gs will contain more states. These states can be used in the transformation.
Note that the code will not check the orthonormality of the new states!
Each line provides the coefficients of the new states, in terms of
the old ones. The coefficients are complex, but the imaginary part will be
ignored for real wavefunctions.
Note: This variable cannot be used when parallel in states.