CCL:G: Combining Z-matrix with Cartesian coordinates in Gaussian

[Grigoriy Zhurko <reg_zhurko : chemcraftprog.com>]

 Sent to CCL by: Grigoriy Zhurko [reg_zhurko]~[chemcraftprog.com]
 Hello,
 I need to perform a PES scan in Gaussian, so that the distance between an atom
 and the center between two other atoms will be scanned.
 When I need to simply scan an iteratomic distance, I use OPT(MODREDUNDANT). But
 in current case I have to use a z-matrix and add a dummy atom (the center
 between two real atoms).
 It is possible to use only Z-matrix for this job, but the result becomes rather
 bad: the optimization starts with evident bugs (probably because this geometry
 has rings).
 I tried to combine the Z-matrix with the Cartesians. I created the following
 input file:
 %NPROCSHARED=3
  #P uwB97XD/6-31G(D,P) OPT(Z-Matrix,LOOSE)
  NOSYMM
  Silicon
  0 1
  X
  N 1 R12
  X 1 1.400000 2 90.000000
  N 1 R12 3 90.000000 2 180.000000
  N 1 R15 2 A215 3 -90.000000
  N 1 R16 3 A316 5 D5316
  Zn 1 R17 2 A217 5 -D5217
  6 0.000000000 2.863016000 1.101006000
  6 0.000000000 2.425840000 2.425840000
  6 0.000000000 1.101006000 2.863016000
  1 0.000000000 3.193595000 3.193595000
  6 0.000000000 2.863016000 -1.101006000
  6 0.000000000 2.425840000 -2.425840000
  6 0.000000000 1.101006000 -2.863016000
  1 0.000000000 3.193595000 -3.193595000
  6 0.000000000 -1.101006000 -2.863016000
  6 0.000000000 -2.425840000 -2.425840000
  6 0.000000000 -2.863016000 -1.101006000
  1 0.000000000 -3.193595000 -3.193595000
  6 0.000000000 -2.863016000 1.101006000
  6 0.000000000 -2.425840000 2.425840000
  6 0.000000000 -1.101006000 2.863016000
  1 0.000000000 -3.193595000 3.193595000
  6 0.000000000 -4.245855000 0.681548000
  6 0.000000000 -4.245855000 -0.681548000
  6 0.000000000 -0.681548000 -4.245855000
  6 0.000000000 0.681548000 -4.245855000
  6 0.000000000 -0.681548000 4.245855000
  6 0.000000000 0.681548000 4.245855000
  6 0.000000000 4.245855000 0.681548000
  6 0.000000000 4.245855000 -0.681548000
  1 0.000000000 -5.096481000 1.350466000
  1 0.000000000 -5.096481000 -1.350466000
  1 0.000000000 -1.350466000 -5.096481000
  1 0.000000000 1.350466000 -5.096481000
  1 0.000000000 5.096481000 -1.350466000
  1 0.000000000 5.096481000 1.350466000
  1 0.000000000 -1.350466000 5.096481000
  1 0.000000000 1.350466000 5.096481000
  Variables:
  R12 = 2.04198300
  R15 = 2.04198300
  A215 = 90.00000000
  R16 = 2.04198300
  A316 = 90.00000000
  D5316 = 180.00000000
  A217 = 90.00000000
  D5217 = 90.00000000
  constants
  R17 = 0.40000000 s 20 0.2
 This job constructs a valid geometry, the optimization starts, but it starts
 incorrectly: during the optimization, only the Z-matrix parameters are varied,
 while the coordinates of most atoms are fixed. Can I do something to make
 Gaussian optimize these atomic coordinates as well?