The calculation of Biradical with mix of HOMO and LUMO

Questions
#1

Hello everyone, I use the input file below to get C4H8 biradical. First, I get the initial guess wave function through UDFT single point energy calculation ，which needs “SCF_GUESS_MIX” and “STABILITY_ANALYSIS” to ensure the stability of the initial guess wave function. Then, I carried out structural optimization and frequency calculation, and both steps read the prior initial guess wave function by “SCF_GUESS = read”. Unfortunately, the result converges to the structure of cyclobutane rather than the C4H8 diradical. Could you help me check what’s wrong with the input file？Do you have any suggestion on how to calculate diradicals? Look forward to your reply, thank you
$molecule
0 1
C -0.92219100 1.49609700 0.46979900
C -0.80401500 0.21210200 -0.28769100
C 0.21213400 -0.80400300 0.28771600
C 1.49608600 -0.92216400 -0.46985100
H -1.74223400 2.17255800 0.27078800
H -0.09268100 1.86292200 1.06116000
H -0.54327100 0.40987500 -1.34002800
H -1.79333800 -0.25598800 -0.33446900
H -0.25592100 -1.79333500 0.33456200
H 0.40995700 -0.54319200 1.34002800
H 1.86306800 -0.09247100 -1.06086100
H 2.17240500 -1.74240000 -0.27115200
$end

$rem
BASIS = 6-311G**
GUI = 2
METHOD = M06-2X
UNRESTRICTED = true
SCF_GUESS_MIX = 2
SYM_IGNORE = true
STABILITY_ANALYSIS = true
$end

@@@

$molecule
read
$end

$rem
JOB_TYPE opt
METHOD M06-2X
BASIS 6-311G**
UNRESTRICTED = true
SCF_GUESS = read
SYM_IGNORE = true
SCF_MAX_CYCLES = 250
$end

@@@

$molecule
read
$end

$rem
JOB_TYPE freq
METHOD M06-2X
BASIS 6-311G**
UNRESTRICTED = true
SCF_GUESS = read
SCF_MAX_CYCLES = 250
$end

#2

In general this probably requires one of three approaches. In increasing order of sophistication, these are:
(1) Unrestricted DFT (as you have done) but combining maximum overlap method (MOM) with an open-shell guess ($occupied section). Alternatives to MOM are STEP and SGM algorithms (see manual). Resulting solution, if you can find it, will be heavily spin contaminated, probably <S^2> ~ 1.

(2) Open-shell singlet optimization (OPSING = TRUE), which is based on the spin purification formula E(OSS) = 2 E(BS) - E(T) where the open-shell singlet (OSS) energy is obtained from the broken-symmetry (BS) energy and the triplet (T) energy. This code will automatically perform both the broken symmetry and the triplet energies. May also need to be used with MOM.

(3) Restricted open-shell Kohn-Sham (ROKS), which can be viewed as a self-consistent (or orbital-optimized) version of the simple spin-purification formula in (2).