Below is an old but working example. See the comment about the units (Bohr's units in SS card v. Angstrom in Overlap card) - it used to be a bug in FEFF6 - and it may still be unless Bruce/Matt changed the code in the recent distribution. If rescaling the distances in your SS card does not help, swap the SS and the OVERLAP cards (in my example SS preceeds OVERLAP - may be it matters).

TITLE   Cu metal in aqueous solution
TITLE   Structure is from Beagley et al, J. Phys.: Cond. Matt. 1, 2395 (1989)
TITLE   Distance units are in Angstroms in OVERLAP cards
TITLE   Distance units are in Angstroms/0.529177249 in SS cards

HOLE 1   1.0     Cu K edge ( 8.979 keV), s0^2=1.0

*         mphase,mpath,mfeff,mchi
CONTROL   1      1     1     1
PRINT     1      0     0     0

POTENTIALS
*   ipot   z label
       0   29   Cu
       1   8    O
       2   8    O
       3   1    H
       4   8    O

SS 1 1 4   3.760555
SS 2 2 2   4.327473
SS 3 4 11 7.464418

* note that the distances above correspond to the three shells:
*1.99, 2.29 and 3.95 A. However, because of the bug in FEFF6,
*these distances should are used in FEFF6 in the units of Bohr's radius.
*Therefore, for the program to work properly, you should divide the true
*shell distance by a0 = 0.529177249 to enter in SS cards, as in the example above
*The overlap distances below stay in Angstroms, as they should.

OVERLAP 0
* Cu - O_eq
1 4 1.99
* Cu - O_ax
2 2 2.29
OVERLAP 1
* O_eq - H
3 2 0.957
* O_eq - Cu
0 1 1.99
OVERLAP 2
* O_ax - H
3 2 0.957
* O_ax - Cu
0 1 2.29
OVERLAP 3
* H - O_eq
1 1 0.957
* H - H
3 1 1.513
* H - O_2sh
4 1 0.957
OVERLAP 4
* O_2sh - H
3 2 0.957

I have read some references of Kelly Shelly, and I have got some idea of how to
construct a structure without difinite crystalline structure. But when I try to
apply to my own question, meet some difficulties.

In my case, four nitrogen atoms is surrounding a iron atom. So I use OVERLAP and SS card as
to define such system as described in the FEFF manual:
--------------------------------------------------
POTENTIALS
* ipot Z element
0 26 Fe
1 7 N
* Determine approximate overlap for central and 1st nn in Cu
OVERLAP 0 *determine overlap for central atom of Fe
*iphovr novr rovr * ipot, number in shell, distance
1 4 2.15
OVERLAP 1 *determine approximate overlap for 1st shell N
*iphovr novr rovr * ipot, number in shell, distance
0 1 2.15
*SS index ipot deg rss
SS 1 1 4 2.15
--------------------------------------------------
but feff8 stopped and said:
Absorbing atom coords not specified.
Cannot find multiple scattering paths.
since the OVERLAP does't need ATOMES card, where can I define the coordination of absorbing atom?
Is there any other CARD needed for such situation?

=========================================
Hi Jun,
That keyword is explained here:
http://feff.phys.washington.edu/feff/wiki/index.php?title=OVERLAP
I think the problem is that you have not specified the SS keyword:
http://feff.phys.washington.edu/feff/wiki/index.php?title=SS

I use the overlap/ss thing in feff very infrequently. My preference
is to start from a known structure that is somehow similar to the
material I am investigating. For example, if I were to look at
amorphous germanium, I would use crystalline germanium as my starting
point. I would then construct the model in such a way that I
parameterize the differences between the starting structure and what I
expect to find in the data.

I'd recommend reading papers by Scott Calvin or Anatoly Frenkel on
nanoparticles or by Shelly Kelly on uranium in biogeochemical systems
for excellent examples of how that sort of analysis is done.

Hope that helps,
B