Elementary results for the fundamental representation of SU(3)

arXiv:submit/1301737 [physics.gen-ph] 17 Jul 2015

Thomas L. Curtright and Cosmas K. Zachos

Department of Physics, University of Miami
Coral Gables, FL 33124-8046, USA

curtright@miami.edu zachos@anl.gov

Abstract
A general group element for the fundamental representation of SU(3) is expressed as a second order
polynomial in the hermitian generating matrix H, with coefficients consisting of elementary trigonometric
functions dependent on the sole invariant det H, in addition to the group parameter .

Consider an arbitrary 3 3 traceless hermitian matrix H. The Cayley-Hamilton theorem [1] gives


(1)
H 3 = I det H + 12 H tr H 2 ,


where in fact det H = tr H 3 /3. Note that an H 2 term
is absent in the polynomial expansion of H 3 because
of the trace condition, tr H = 0. Also note, since tr H 2 > 0 for any nonzero hermitian H, this bilinear trace
factor may be absorbed into the normalization of H, thereby setting the scale of the group parameter space.
We now write the exponential of H as a matrix polynomial. As a consequence of (1) any such exponential
can be expressed as a matrix polynomial second-order in H, with polynomial coefficients that depend on the
displacement from the group origin as a rotation angle .
Moreover, the polynomial coefficients will also depend on invariants of the matrix H. These invariants can
be expressed in terms of the eigenvalues of H, of course [2, 3, 4]. Nevertheless, while the eigenvalues of H will
be manifest in the final result given below, the deliberate diagonalization of H is not necessary. This is true for
SU (3), for a normalized H, since there is effectively only one invariant: det H. This invariant may be encoded
cyclometrically as another angle. Define







3
5
9
,
(2)
3 (det H) + O (det H)
= 31 arccos 23 3 det H 2 = 12 3 (det H) 16
whose geometric interpretation will soon be clear. Inversely
2
sin (3) = 23 +
det H = 3
3

33 + O 5

(3)

The general result for 3 3 unitary SU (3) group matrices generated by H is then
h
exp (iH) = 32 I sin ( + 2/3) sin ( 2/3)
h
+ 32 I sin () sin ( 2/3)
h
+ 32 I sin () sin ( + 2/3)

1
3

1
3

1
3

H sin ()

H sin ( + 2/3)
H sin ( 2/3)

1
2

1
2

H2

1
2

H2
H2

exp

2
3


i sin

2 cos ( + 2/3) cos ( 2/3)



i exp 2 i sin ( + 2/3)
3

2 cos () cos ( 2/3)



i exp 2 i sin ( 2/3)
3
2 cos () cos ( + 2/3)

where we have set the scale for the parameter space by choosing the normalization


tr H 2 = 2 .

(4)

(5)

This choice is consistent with the Gell-Mann -matrices. With this normalization, the Cayley-Hamilton result
(1) is just
H 3 = H + I det H .
(6)
1

Thus the group element, expressed as a matrix polynomial, depends on the sole invariant det H in addition to
the group rotation angle . Both dependencies are in terms of elementary trigonometric functions when det H
is expressed as the angle , whose geometric interpretation follows immediately from the three eigenvalues of H
exhibited in the exponentials of (4). Those eigenvalues are the projections onto three mutually perpendicular

axes of a single point on a circle formed by the intersection of the 0 = tr H eigenvalue plane with the 2 = tr H 2
eigenvalue 2-sphere. The angle parameterizes that circle. Equivalently, the eigenvalues are the projections
onto a single axis of three points equally spaced on a circle.
Two cases deserve special mention. On the one hand, the Rodrigues formula for SO (3) rotations about an
axis n
b, as generated by j = 1 spin matrices, is obtained for = 0 = det H. Thus
exp (iH)|=0 = I + iH sin + H 2 (cos 1) .

(7)

This is the Euler-Rodrigues result, upon identifying H = n

b J (see [5, 6]). It provides an explicit embedding
SO (3) SU (3). In fact, (7) is true if H is any one of the first seven Gell-Mann -matrices, or if H is a
normalized linear combination of 13 , or of 47 . However, for generic linear combinations of 17 , det H will
not necessarily vanish, and the general result (4) must be used.
On the other hand,

1 0 0
1
0 1 0
8 =
(8)
3
0 0 2

is the only one among Gell-Manns choices for the 3 3 representation matrices for which 6= 0, and for which
, so = /6. In addition,
two eigenvalues are degenerate. Obviously, det 8 = 32
3
28 =

1
2
I 8 .
3
3

(9)

Thus, directly from (4),

exp i/ 3
0

0

 1


2
i 3
=
exp (i8 ) = 31 2I + 38 e 3 i 3 + 31 I 38 e
0
exp i/ 3
0
,
0
0
exp 2i/ 3
(10)
as it should. Note that this particular example followed from (4) by carefully taking the limit as /6 of
the first two lines in that general expression (as necessitated by the degeneracy of the corresponding eigenvalues
of 8 ) combined with the straightforward limit of the third line. That is to say,
 
 





1
1 2 exp i cos 13 sin
1
1
1

3 (sin ) cos + 3 sin I +

lim
cos 3 sin 8 + 8
(cos 3 sin ) cos
/6
2
2


h

i exp 2 i sin  

3
1
1

ei/ 3 .

=
(11)
cos2 13 sin2 I 23 (sin ) 8 28 (cos2 3
= lim
I
+
3
sin2 )
2 3 8


/6

Finally, one readily checks that the Laplace transform of (4) gives the resolvent in the standard form as a
matrix polynomial [7, 8, 9].
1
=
I iH

et exp (iHt) dt =

1
det (I iH)

rank
XH
n=0

(iH)

Trunc

rank H1n

[det (I iH)] ,

(12)

where the truncation

(as defined in [5]) is in powers of . For the case at hand rank H = 3, again with tr H = 0


and tr H 2 = 2, so we have
1
1
=
I iH
1 + 2 + i3 det H

1 + 2 I + iH 2 H 2 ,

(13)

along with the simple Cayley transform [1] representation [2, 8, 9] of the corresponding SU (3) group elements,
1
I + iH
=
I iH
1 + 2 + i3 det H

1 + 2 i3 det H I + 2iH 22 H 2 .
2

(14)

Acknowledgements This work was supported in part by NSF Award PHY-1214521, and in part by a University of Miami Cooper Fellowship.

References
[1] The Collected Mathematical Papers of Arthur Cayley, Cambridge University Press (1889).
See Vol. I. pp 28-35 for the Cayley transform, and see Vol. II. pp 475-496 for the CayleyHamilton
theorem.
Also
see
http://en.wikipedia.org/wiki/Cayley-Hamilton theorem
and
http://en.wikipedia.org/wiki/Cayley transform.
[2] A J MacFarlane, A Sudbery, and P H Weisz, On Gell-Manns -Matrices, d- and f -Tensors, Octets, and
Parametrizations of SU(3) Commun.Math.Phys. 11 (1968) 77-90.
[3] D Kusnezov, Exact matrix expansions for group elements of SU(N) J.Math.Phys. 36 (1995) 898-906.
[4] A Laufer, The exponential map of GL(N) J.Phys.A:Math.Gen. 30 (1997) 5455.
[5] T L Curtright, D B Fairlie, and C K Zachos, A Compact Formula for Rotations as Spin Matrix Polynomials
SIGMA 10 (2014) 084. e-Print: arXiv:1402.3541 [math-ph]
[6] T L Curtright and T S Van Kortryk, On rotations as spin
J.Phys.A: Math.Theor. 48 (2015) 025202. e-Print: arXiv:1408.0767 [math-ph]

matrix

polynomials

[7] M X He and P E Ricci, On Taylors formula for the resolvent of a complex matrix
Computers and Mathematics with Applications 56 (2008) 22852288.
[8] T S Van Kortryk, Cayley transforms of su (2) representations e-Print: arXiv:1506.00500 [math-ph]
[9] T L Curtright, More on Rotations as Spin Matrix Polynomials e-Print: arXiv:1506.04648 [math-ph]