LASER INTERFEROMETER GRAVITATIONAL WAVE OBSERVATORY

- LIGO -
===============================
LIGO SCIENTIFIC COLLABORATION

Technical Note LIGO-T1300189–v1 2013/3/10

On the accumulated round-trip

Gouy phase shift for a general
optical cavity
Koji Arai

Distribution of this document:

LIGO Scientific Collaboration

California Institute of Technology Massachusetts Institute of Technology

LIGO Project, MS 100-36 LIGO Project, Room NW22-295
Pasadena, CA 91125 Cambridge, MA 02139
Phone (626) 395-2129 Phone (617) 253-4824
Fax (626) 304-9834 Fax (617) 253-7014
E-mail: [email protected] E-mail: [email protected]

LIGO Hanford Observatory LIGO Livingston Observatory

Route 10, Mile Marker 2 19100 LIGO Lane
Richland, WA 99352 Livingston, LA 70754
Phone (509) 372-8106 Phone (225) 686-3100
Fax (509) 372-8137 Fax (225) 686-7189
E-mail: [email protected] E-mail: [email protected]

http://www.ligo.org/
 LIGO-T1300189–v1

1 Introduction
It is well known that the stability of a two-mirror Fabry-Perot cavity can be characterized
by the product of two values called g-factors:
L
g1 = 1 − (1)
R1
L
g2 = 1 − , (2)
R2
where L is the length of the cavity, and R1 and R2 are the curvature radii of the cavity mirrors.
When the product g1 g2 fulfills the following condition, the cavity has stable eigenmodes.

0 ≤ g1 g2 ≤ 1 (3)

The product of the g-factors also provides a handy way to calculate the transverse mode spac-
ing (TMS) of the cavity. The product of the g-factors is closely related to the accumulated
round-trip Gouy phase shift (ζ) between the cavity eigenmodes1 .
√
ζ = 2 cos−1 ± g1 g2 (4)

Here the sign ± is determined by the sign of g1 , and in fact g1 and g2 have the same sign
for a stable cavity (cf. Eq. (3)). Consequently, the resulting TMS is given by the following
formula √
ζ cos−1 ± g1 g2
νTMS = νFSR = νFSR (5)
2π π
where νFSR is the free spectral range of the cavity that is given by νFSR = c/(2L).
For a general cavity case (like a ring caity or a folded cavity with intra-cavity lensing optics),
one would naturally imagine that there might be a similar simple indicator to the g-factors.
In fact, J. Rollins recently conjectured from the discussion of the folded recycling cavity of the
40m prototype that ζ can be approximately calculated with the product of the generalized
g-factors
v
u n
uY
ζ = 2 cos−1 t gi (6)
i=1

LRT
gi = 1 − (7)
2Ri
where LRT is the round-trip length of the cavity, and Ri is the curvature radius of the i-
th mirror. However, this conjecture is obviously not applicable to every cavity;
N. Smith pointed out that permutation of the mirrors in a ring cavity (like the output mode
cleaner cavity) changes the TMS while the corresponding permutation of the g-factors in the
1
The word “accumulated Gouy phase shift” means the shift of the optical phase between one of the
Gaussian modes and the mode at the next higher-order spacial mode (e.g. TEM00 and TEM01 ) caused by
Gouy phase effect. The “round-trip” one is the same quantity for a single round-trip of the optical cavity.

page 1
 LIGO-T1300189–v1

product does not change ζ. In fact, it will be shown in this document that this conjecture
is useful only when 1 − gi  1 for all i. 2
As long as the simple and exact indicator like the g-factors is not available, we need to
go back to the first principle of the beam calculation: ABCD matrix. If we calculate the
ABCD matrix of the cascaded optical system by multiplying the individual ABCD matrices,
all aspect of the system in terms of the beam parameter can be characterized. This can
be understood by looking at the formula for Gaussian beam transformation by the ABCD
matrix that is derived from Huygens propagarion integral ([1], Chapter 20):
Aqin + B
qout = , (8)
Cqin + D
where qin and qout are the q-parameters for the input and output beam respectively. A,
B, C, and D are the elements of the ABCD matrix of the optical system. This formula
means that different optical systems with the same ABCD matrix result in identical beam
transformations.
It is also known that the quantity (A + D)/2 is closely related to cavity stability. The cavity
eigenmodes are supposed to fulfill the beam-reproducing condition qout = qin . By solving this
condition, Eq.(8) with the unitarity condition (i.e. AD − BC = 1), we obtain two solutions
for the forward and backward beams:
p
A − D ± (A + D)2 − 4
q= (9)
2C
As q is a complex number3 , the stability criteria is given by
A+D
−1≤ ≤ 1. (10)
2
Equivalently, this can be expressed as
A+D+2
0≤ ≤ 1. (11)
4
A+D+2
As we will find in Section 3, for a Fabry-Perot cavity corresponds to the product
4
of the g-factors g1 g2 . Therefore Eq. (11) is equivalent to Eq. (3).
In this note, the interpretation of Eqs. (10) and (11) are extended for general cavities. It is
derived that the accumulated round-trip Gouy phase shift can be computed only
from the round-trip ABCD matrix of the cavity as:
 
−1 A+D
ζ = sgnB · cos , (12)
2

where the value range of cos−1 x is defined to be

0 ≤ cos−1 x < 2π (−1 < x ≤ 1). (13)

2
Therefore, the product of the generalized g-factors is always positive. There is no sign ambiguity.
3
cf. q = z + izR , where zR is the Rayleigh range.

page 2
 LIGO-T1300189–v1

Eq. (12) can equivalently be expressed as

r !
−1 A+D+2
ζ = 2 cos sgnB · . (14)
4

These expressions requires the calculation of the ABCD matrix for the round-trip path, and
therefore more complicated to calculate than the product of the g-factors. However, it still
provides concise evaluation of the stability on the general cavities.

2 Derivation
The derivation of Eqs. (12) and (14) is based on the following two points:

1. The accumulated Gouy phase shift between an arbitrary two points of a Gaussian beam
can be computed only from the ABCD matrix between the points and the parameter
of the input beam. Since the detail of this point is proven elsewhere [2], we will use
this result without re-derivation.

2. The beam parameters of a cavity eigenmode id necessarily reproduced after a round

trip in the cavity. Therefore, the cavity eigenmode must fulfill the beam-reproducing
condition qout = qin . This means that the input and output beams have the same waist
radii, same waist positions, and thus the same wavefront radii of curvature.

Erden et al. [2] described how the beam parameters (i.e. the beam radius, wave front radius
of curvature, and Gouy phase) are described by the elements of the ABCD matrix and
the parameters of the input beam Eqs. (13)∼(15) in [2]. Here are the excerpted equations
rewritten: 2
B 2 λ2

2 2 B
ωout = ωin A + + 2 2 (15)
rin π ωin
! !
D B BDλ2
C+ A+ + 2 4
1 rin rin π ωin
= ! (16)
rout B B 2 λ2
A+ + 2 4
rin π ωin

B
tan ζ = ! (17)
2
B πωin
A+
rin λ

Note that the ABCD matrix in [2] is defined differently from the equations above. The
equations here have been modified so that the convention of the elements A, B, C, D agrees
with the usual conventions in the optics book like [1]. (i.e. BSiegman = BErden /λ, CSiegman =
CErden λ )

page 3
 LIGO-T1300189–v1

Indeed, the consequence of their paper, shown as Eq. (17) here, is a useful expression. This
expression enables us to calculate the accumulated Gouy phase shift, ζ, from rin , ωin and
the elements of the ABCD matrix, no matter what optical elements are in the path. This
may be particularly useful for the calculation of ζ for the Gouy phase telescope of the wave
front sensing (WFS) systems.
Erden et al. conversely derived the elements of the ABCD matrix from the beam parameters.
ωout B
A= cos ζ − (18)
ωin rin
πωin ωout
B= sin ζ (19)
λ !
1 B Bλ2
A+ + 2 4
A rin rin π ωin
C= − !2 (20)
rout B B 2 λ2
A+ + 2 4
rin π ωin
1 + BC
D= (21)
A
The last equation comes from the unitarity of the ABCD matrix.
A+D
After simplification, the quantity can be expressed as follows
2
   
A+D 1 ωin ωout 1 1 1 πωin ωout
= + cos ζ + − sin ζ (22)
2 2 ωout ωin 2 rout rin λ

For the round-trip Gouy phase shift of a cavity, the formula is significantly simplified because
of the beam-reproducing condition. We therefore find
A+D
= cos ζ (23)
2
and
A+D+2 cos ζ + 1 ζ
= = cos2 (24)
4 2 2
Because of the multi-value nature of the inverse cosine function there is a sign ambiguity:

cos(|ζ|) = cos(−|ζ|) (25)

The key to solve this ambiguity is the element B. As the sign of B is determined by sin ζ,
sgnB indicates the sign of ζ. Therefore we obtain the following expressions
 
−1 A+D
ζ = sgnB · cos , (26)
2

where the value range of cos−1 ζ is

0 ≤ cos−1 ζ < 2π (−1 < ζ ≤ 1). (27)

page 4
 LIGO-T1300189–v1

ζ
If we use the inverse function of Eq. (24) for the value range above, we find cos is always
2
positive and the ambiguity is hidden in the sign of the square-root term. Therefore the sign
needs to be inside of the inverse cosine:
r !
A + D + 2
ζ = 2 cos−1 sgnB · . (28)
4

3 Example 1: Fabry-Perot cavity

A two-mirror Fabry-Perot caity is chosen for the first example in order to confirm the well-
known stability criteria using mirror g-factors.
The ABCD matrix for a space with the length of L:
 
1 L
S(L) = (29)
0 1

The ABCD matrix for a mirror with the curvature radius of R:

 
1 0
F(R) = (30)
−2/R 1

Let’s assume we have a linear cavity with the length of L and the curvature radii of the
input and end mirrors as R1 and R2 . The round-trip ABCD matrix is calculated as
 
A B
= F(R1 ) S(L) F(R2 ) S(L) (31)
C D
2L2
 
2L
1− 2L −
=
 R 2 R2 
2 . (32)
 
2 2 4L 2L 4L 4L
− − + 1− − +
R2 R1 R1 R2 R2 R1 R1 R2

From this, we can calculate the cosine of the round-trip Gouy phase shift as
A+D
cos ζ = (33)
2
L L L L
=1−2 −2 +2 (34)
R1 R R1 R2
  2 
L L
=2 1− 1− −1 (35)
R1 R2
≡ 2g1 g2 − 1 (36)
or the quantity equivalent to the product of the g-factors
ζ A+D+2
cos2 = (37)
2  4  
L L
= 1− 1− (38)
R1 R2
≡ g1 g2 (39)

page 5
 LIGO-T1300189–v1

The sign of the g-factor can be extracted from the element B:

" !#
L
sgnB = sgn 2L 1 − (40)
R2
= sgn g2 (41)
Therefore we obtain
√
ζ = 2 cos−1 (sgn g2 · g1 g2 ). (42)
This is equivalent to the famous expression Eq. (4).

4 Example 2: aLIGO OMC

The second example is the aLIGO output mode cleaner (OMC). The cavity of the OMC is
a bow-tie style ring cavity formed by four mirrors: two flat input/output mirrors (FMs) and
two curved mirrors (CMs) with the curvature radius of R = 2.575m (as the average number
of measured values). The cavity mirrors are arranged in the order of FM1-FM2-CM1-CM2-
FM1. The distances FM1-FM2 and CM1-CM2 are both L1 = 0.2816 m, while the distances
FM2-CM1, and CM2-FM1 are both L2 = 0.2844 m. This means the round-trip length is
LRT = 1.132 m, and the FSR is νFSR = 264.8 MHz. The angle of incidence on each mirror is
θ = 4.042 deg (= 0.0706 rad).
The round-trip ABCD matrix C is defined as
C± = M± S(L2 )M± F(R± ) S(L1 )M± F(R± ) S(L2 )M± S(L1 ). (43)
Because of the non-zero incident angle, the system naturally involves astigmatism. The signs
+ and − are applied to the vertical and horizontal mode, respectively. The reflection matrix
M± is defined as  
±1 0
M± = , (44)
0 ±1
which correspond to the phase flip of the horizontal modes when the beam is reflected with
non-zero incident angle in the horizontal plane. This matrix becomes important for an odd-
number-mirror ring cavity, while we can ignore them for our four mirror cavity. The curved
mirror therefore behaves as an astigmatic mirror with an effective radius of curvature of
R± = R · (cos θ)±1 . (45)

The results of the mode calculations are shown below.

Horizontal mode
 
0.3892 0.7242
C− = (46)
−1.381 0.0004690
Thus we obtain
A+D
ζ = sgnB . cos−1 = 1.375 rad ⇒ 39.38 deg (47)
2
page 6
 LIGO-T1300189–v1

ζ
νTMS,H = νFSR = 0.2188 νFSR = 57.94MHz (48)
2π

Vertical mode
 
0.3864 0.7223
C+ = (49)
−1.387 −0.0004043
Thus we obtain
A+D
ζ = sgnB · cos−1 = 1.378 rad ⇒ 39.49 deg (50)
2
ζ
νTMS,V = νFSR = 0.2194 νFSR = 58.10MHz (51)
2π

5 Example 3: Rollins conjecture

J. Rollins conjectured that the accumulated round-trip Gouy phase shift ζ of a multi-mirror
cavity can be approximated by the product of the generalized g-factors gi , as follows:
v
u n
uY
ζ = 2 cos−1 t gi (52)
i=1

LRT
gi ≡ 1 − (53)
2Ri
where LRT is the round-trip length of the cavity, and Ri is the curvature radius of the i-th
mirror.
A+D+2
The first guess is that if all of δgi = 1 − gi are small enough, may give us a
4
good approximation of the the generalized g-factor product. Let’s prove it in the following
subsections.
In the following derivation it is convenient to define the ABCD matrix of a curved mirror
and a space as
   
1 Li 1 Li
Pi = F(Ri ) S(Li ) =  2 2Li  =  δfi δfi  (54)
− 1− − 1− Li
Ri Ri LRT LRT
where
2LRT
δfi ≡ = 4δgi (55)
Ri

Proposition:
The ABCD matrix of the optical system Pn Pn−1 . . . P1 is represented in the following form
for n ≥ 2, up to the first order of δfi .
 
An Bn
Pn Pn−1 . . . P1 = , (56)
Cn Dn
page 7
 LIGO-T1300189–v1

where
n i−1
!
X X δfj
An = 1 − Li (57)
i=2 j=1
LRT
n n−1
" i
! n
!#
X X δfi X X
Bn = Li − Lj Lk (58)
i=1 i=1
LRT j=1 k=i+1
n
X δfi
Cn = − (59)
i=1
LRT
n i
!
X δfi X
Dn = 1 − Lj (60)
i=1
LRT j=1

Proof:
For n = 2, we directly calculate P2 P1 :
  
1 L2 1 L1
P2 P1 =  δf2 δf2 L2   δf1 δf1 L1  (61)
− 1− − 1−
LRT LRT LRT LRT
 
δf1 δf1
 1 − L2 L L1 + L2 −
L
L1 L2 
=  RT RT  + O2 (δfi ) (62)
δf1 + δf2 δf1 δf2 
− 1− L1 − (L1 + L2 )
LRT LRT LRT

Assuming the proposition is fulfilled for An , Bn , Cn and Dn , we calculate Pn+1 Pn . . . P1 as

 
1 Ln+1  
A n Bn
Pn+1 Pn . . . P1 =  δfn+1 δfn+1 Ln+1  + O2 (δfi ) (63)
− 1− C n D n
LRT LRT
 " !# 
Pn δfi Pn Pn δfj
 An − Ln+1 i=1 Bn + Ln+1 1 − i=1 Li j=i
LRT LRT

=  + O2 (δfi )
 
 δfn+1 δfn+1 Pn 
Cn − Dn − i=1 L i
LRT LRT
(64)
 
An+1 Bn+1
= + O2 (δfi ) (65)
Cn+1 Dn+1

Q.E.D.

page 8
 LIGO-T1300189–v1

Consequence
From Eqs. (57) and (60), we obtain
A+D+2 ζ
(= cos2 ) (66)
4 2 " !#
n n
X δfi X
=1− Li + O2 (δfi ) (67)
i=1
4LRT j=1
n
X δfi
=1− + O2 (δfi ) (68)
i=1
4
n
X LRT
=1− + O2 (δfi ) (69)
i=1
2Ri

This quantity can be compared with the product of the generalized g-factors.
n n  
Y Y LRT
gi = 1− (70)
i=1 i=1
2Ri
n  
Y δfi
= 1− (71)
i=1
4
n
X δfi
=1− + O2 (δfi ) (72)
i=1
4
n
X LRT
=1− + O2 (δfi ) (73)
i=1
2Ri

This agrees with Eq. (69).

This means that Rollins conjecture is true if all of the curvature radii of the mirrors
are long enough relative to the round-trip length of the cavity (i.e. δgi  1 for all
i). Note that this result means that permutation of the mirrors does not change the resulting
round-trip Gouy phase significantly up to the first order in this small δgi regime. Also note
that all of the g-factors are kept in the positive number. So there is no ambiguity of the
Gouy phase shift.

6 Acknowledgement
The author appreciates Nicolas Smith and Jameson Rollins for their useful discussions, care-
ful confirmation of the derivations, and profound meditation about the physics behind the
calculation.

References
[1] A. E. Siegman, Lasers (University Science Books, 1986).
page 9
 LIGO-T1300189–v1

[2] M. Fatih Erden and Haldun M. Ozaktas, Accumulated Gouy phase shift in Gaussian
beam propagation through first-order optical systems, J. Opt. Soc. Am. A, 14 (1997)
2190-2194.

[3] K. Arai, S. Barnum, P. Fritschel, J. Lewis, S. Waldman, Output Mode Cleaner Design,
LIGO Document T1000276-v5, (2013).

page 10