The graphic format is a character string where each character is represented
by 2 bytes where all characters are part of a specific
double-byte characer set. 
For information on the UCS-2 (Unicode) format which also uses double-byte
characters, see UCS-2 Format. 
Fields defined as graphic data do not contain shift-out (SO) or shift-in (SI) characters. The difference between single byte character and double byte graphic data is shown in the following figure:

The length of a graphic field, in bytes, is two times the number of graphic characters in the field.
The fixed-length graphic format is a character string with a set length where each character is represented by 2 bytes.
For information on the variable-length graphic format, see Variable-Length Character, Graphic and UCS-2 Formats.
You define a graphic field by specifying the
GRAPH or VARGRAPH keyword in a free-form definition or by specifying G in the Data-Type entry
of a fixed-form specification. You can also define one using the LIKE keyword
on the definition specification where the parameter is a graphic field. 
You can specify the default CCSID for graphic fields
using the CCSID(*GRAPH) keyword
on a Control statement or a /SET directive.
You can also specify the CCSID explicitly using the
Definition statement CCSID keyword. 
The default initialization value for graphic data is X'4040'. The value of *HIVAL is X'FFFF', and the value of *LOVAL is X'0000'.