The delimiters recognized by LaTeX include
\left \{ ... \right \[ produces a legal pair.
In cases where only one delimiter is desired, it
is possible to make the matching delimiter "invisible" by typing a
period (.) after the command, i.e.,
\left. or \right..
In an eqnarray environment the matching \left and \right cannot be split between lines and it may be necessary to use an "invisible" \right. and \left. to terminate and begin parts on different lines. In this case a second problem may arise, since the size of the delimiters will be chosen only for the local part, so that the size of the visible "left" and "right" delimiters might not match. The solution is to trick LaTeX into thinking that both parts have the same vertical height. This can be done by placing a strut, that is a zero-width \rule. It can also be accomplished with the \vphantom command, which I have not found documented, but which appears to work.
\vphantom{construct}
creates a zero-width object with the height of construct. The
argument can contain things such as \frac
or the variable size math symbols and should
be chosen according to what is in the section with the delimiter you want to match.
\[ \left\{
\begin{array}{ccc}
a & b & c \\
d & e & f \end{array}
\right\} \]
This should be displayed something like (insofar as it can be rendered in
"ascii art"):
( a b c )
- -
( d e f )
Note that the \[ ... \] set this off as
Display Math, and that the Array Environment
is used to generate the three centered columns inside the braces.
\[ f(x) =
\left\{ \begin{array}{l}
0, x < 0 \\ 1, x = 0 \\
2, x > 0 \end{array} \right. \]
will be displayed as
( 0, x < 0
f(x) = - 1, x = 0
( 2, x > 0
Note that the "invisible" \right delimiter is specified using
a "period".