Here are a couple of functions that are similar to the bisect ones.  decimate finds a better minimum point by checking points which neighbor our original guess.  f is the function that you want to minimize, center_x etc is the center point, ie our first guess, radius is the farthest distance from the center that you want to check, and new_x etc stores the x, y, and z values of the new minimum point.
> decimate:=
> proc(f,center_x,center_y,center_z,radius,new_x,new_y,new_z)
>   local i,j,k,min_point_x, min_point_y, min_point_z, min_f, temp_f;
>   min_point_x:=center_x;
>   min_point_y:=center_y;
>   min_point_z:=center_z;
>   min_f:=f(center_x,center_y,center_z);
>   for i from -5 to 5 do 
>      for j from -5 to 5 do
>         for k from -5 to 5 do
>            temp_f:= f(center_x + i/10*radius, center_y + j/10*radius,                                                                                                     center_z + k/10*radius);
>            if(temp_f < min_f) then
>               min_point_x:=center_x + i/10*radius;
>               min_point_y:=center_y + j/10*radius;
>               min_point_z:=center_z + k/10*radius;
>               min_f:=temp_f; 
>               
>               
>             end if;
>         od;
>     od;
>   od;
>   new_x:=min_point_x;
>   new_y:=min_point_y;
>   new_z:=min_point_z;
> 
> end:
decimate_solve iterates the process and finds a new x,y,x.   f is the function that you want to minimize, center_x etc is the center point, ie our first guess, init_rad is the farthest distance from the center that you want to check, final_rad is the farthest that you want to be from the actual minimum, and new_x etc stores the x, y, and z values of the new minimum point.
> decimate_solve:=
> proc(f,center_x,center_y,center_z,init_rad,final_rad,
>       new_x,new_y,new_z)
>  local nx,ny,nz,temp_x,temp_y,temp_z,temp_rad;
>  temp_x:=center_x;
>  temp_y:=center_y;
>  temp_z:=center_z;
>  temp_rad:=init_rad;
> 
>  while (temp_rad >= final_rad) do
>   unassign('nx','ny','nz');
>   decimate(f,temp_x,temp_y,temp_z,temp_rad,nx,ny,nz);
>   temp_x:=nx;  
>   temp_y:=ny;
>   temp_z:=nz;
>   temp_rad:=temp_rad/10;   
> end do;
> 
>  new_x:=temp_x;new_y:=temp_y;new_z:=temp_z;
> end;

  decimate_solve := proc(f, center_x, center_y, center_z, init_rad,

final_rad, new_x, new_y, new_z)
local nx, ny, nz, temp_x, temp_y, temp_z, temp_rad;
    temp_x := center_x;
    temp_y := center_y;
    temp_z := center_z;
    temp_rad := init_rad;
    while final_rad <= temp_rad do
        unassign('nx', 'ny', 'nz');
        decimate(f, temp_x, temp_y, temp_z, temp_rad, nx, ny,
            nz);
        temp_x := nx;
        temp_y := ny;
        temp_z := nz;
        temp_rad := 1/10*temp_rad
    end do;
    new_x := temp_x;
    new_y := temp_y;
    new_z := temp_z
end proc

Here's an example.  We'll find the minimum of g:
> unassign('nx1','ny1','nz1');
> g:=proc(x,y,z) x^2 +(y-2.2)^2 +z^2 end;

         g := proc(x, y, z) x^2 + (y - 2.2)^2 + z^2 end proc

> decimate(g,1,1,-1,5,nx1,ny1,nz1):
> nx1;
> ny1;
> nz1;
This is an okay approximation, but we really want y to be 2.2.  decimate_solve does a better job:
> unassign('nx1','ny1','nz1');
> decimate_solve(g,1,1,-1,5,.1,nx1,ny1,nz1):
> nx1;
> ny1;
> nz1;

                                 .04


                               x_is, 0


                               y_is, 2


                               z_is, 0


                                 1/2


                                  0.


                               x_is, 0


                              y_is, 11/5


                               z_is, 0


                                 1/20


                                  0


                                 11/5


                                  0

That works!  Yay!  I think you'll want to use this on the function for energy.  For you, the x, y, and z will be the values for T[[0,1],8], T[[1,2],8],T[[0,2],8].   You might want the initial radius to be 1, and the final to be .00001, or something similar.  Remember to use unassign! :)   Melanie
> 
>