Download as pdf or txt
Download as pdf or txt
You are on page 1of 3

August 1981 / Vol. 6, No.

8 / OPTICS LETTERS

401

Sagnac effect in fiber gyroscopes


H. J. Arditty and H. C. Lefovre*
Laboratoire Central de Recherche, Thomson CSF, BP 10, Domaine de Corbeville, 91401 Orsuy, France

Received April 21, 1981 We review the kinematic explanation of the Sagnac effect in fiber gyroscopes and recall that the index of the dielec-

tric medium does not have any influence. Furthermore, we present a novel electrodynamic approach that confirms
in more detail that the Sagnac phase shift of a ring-waveguide interferometer experiment with plane waves in a vacuum. is the same as for the original Sagnac

There is still a theoretical problem about the respective influences of Fizeaul and Sagnac2 effects in the calculation of the phase shift that is due to a rotation in
single-mode optical-fiber gyroscopes. As a matter of fact, a review paper 3 and more recently two letters 4' 5

beam splitter has moved through a length AIM = RQiTM

have reported conflictingresults about the involvement of the dispersion of the medium. We review the kinematic method that these authors have used and consider a simplified Sagnac ring interferometer. This allows

us to find the first-order expression of the phase shift that is due to rotation without complicated relativistic equations and to point out where there appears to be a contradiction between these different solutions. Furthermore, we propose a relativistic electrodynamic approach that shows without any ambiguity that the fiber gyroscope phase shift is independent of the medium or waveguide properties. Kinematic Method While this simplified interferometer, with a closed
circular path of counterpropagating waves in a vacuum, is at rest [Fig. 1(a)] in an inertial frame of reference, the

=nAtl in this time TM. But, in this case, the speed of light is no longer the same in both directions. Indeed, this experiment is observed in the motionless inertial frame of reference, and a Fresnel-Fizeau drag effect occurs that depends on the relative directions of light propagation and medium motion. The speeds of the corotating and counterrotating waves are, respectively, CMCR= (Cv/n) + (aRQ) and CMCTR = (cv/n) - (aRQ), 6 To first order in RQ/c, where a is the drag coefficient. the delay AtM = (LMCR/CMCR) - (LMCTRJ/CMCTR) becomes

AtM =

X4r 2 cV2

n2( - a).

(1)

The problem is to find what value has to be given to this coefficient a. The speed CMM of the light in a medium moving with a velocity v with respect to the motionless laboratory frame can be expressed by two equivalent formulas6 :
Cmm C,

n(cop)

+ 1- I v
n2

(2)

path lengths are equal and light travels at the same


speed c, in both directions around the loop: both waves return to the beam splitter in phase, after a propagation time T, = 27rR/cv, where R is the path radius. While

or CMM= CU r+ 1 nacho)
n~

oo dn)

n dw

(3)

it is rotating at a rate Q [Fig. 1(b)] and the observer is still in the motionless inertial frame, the beam splitter has moved through a length Al, = R X Q X TVin this time TV. Therefore the difference between the respective lengths LVCR and LVCTRof the corotating and counterrotating path is LVCR- LVCTR = 2A1, and because the speed of light retains the same value c,, the counterrotating signal will arrive at the beam splitter before the corotating signal. To first order in RQ/c,,
this delay is At, = (LVCR
- LVCTR)/Cv

I/'

//

R~~~~~~

(a)

AlM

47rR 2 Q/C, 2 .

((,
7

2 to a phase difference AO = w X Attv = 47rR2QcO/cv . Now we consider a similar interferometer in which the vacuum is replaced by a dielectric medium of index n.

For continuous waves of frequency co,this corresponds

While the interferometer is at rest, light travels at the


speed CM= c,/n in both directions: Both waves are still in phase after a propagation time TM = 27rR/cM = n X TV. While the interferometer is rotating [Fig. 1(c)], the
0146-9592/81/080401-03$0.50/0

7(

Fig. 1. Simplified Sagnac ring interferometer.

(a) Inter-

ferometer at rest, (b) vacuum interferometer in rotation, (c) dielectric medium interferometer in rotation.
1981, Optical Society of America

402

OPTICS LETTERS / Vol. 6, No. 8 / August 1981

The first formula [Eq. (2)] depends on the light frequency up in the proper moving frame of the medium, and the second formula [Eq. (3)] depends on the light
frequency wo in the laboratory frame. There is a Dop-

that is due to pler frequency shift between wp and cwo the frame translation (cop - w 0 )/wo = nv/c,. The
equivalence of Eqs. (2) and (3) can be established by expanding n(wp)- to first order in (cop- wo) and using

the expression of this Doppler shift, yielding n(cp)-c = n(o 0 )-' - (co/n)(dn/dX)v. In the original Fizeau experiment, Eq. (3) has to be used because the light
frequency co 0 in the laboratory frame rather than wp can be easily measured. On the other hand, in a Sagnac

ence. In this equation, the effect of the rotation is to add a term that can be considered a perturbation. The solutions correctly identify the modulus and the phase of the two counterpropagating guided modes. The effect of the perturbation on the phase explains the Sagnac phase shift. In Refs. 8 and 9 we detail the method used to obtain this propagation equation in a rotating frame. It involves four-dimensional tensorial calculations and the determination of the constitutive tensor8 - 11 that connects the contravariant components of the derived field

tensor (H, D) with the covariant components of the


macroscopic field tensor (B, E) when there is a comov-

interferometer with a comoving medium, it is simpler to use Eq. (2). The light frequencies of the counterpropagating waves are identical in the frame of the 2 )RQ and medium. So CMCR = cv/n(wp) + (1 - n2 CMCTR= cv/n(cO) - (- n- )RQl. Therefore the value of a is (1 - 1/n2 ), and Eq. (1) becomes AtM = 47rR2 Q/
C, 2 .

This delay Atm is independent of the index of the medium and is equal to the delay At,, in the case of a

ing dielectric medium. After we remove the harmonic dependence exp(iCctR) of the field, where X is the wave frequency and tR is the temporal coordinate, the spatial differential equation of the field amplitude expressed in usual cylindrical coordinates (rR, OR, ZR), to first order in rR Qe,,, is, in a frame FR rotating with a rate Q about the axis zzz'l of an inertial frame F1 ,
a2 E

vacuum. The Fresnel-Fizeau drag serves to compensate to first order in RQc,, the effect of the increase in path difference LMCR - LMCTR = n(LVCR - LVCTR)In the laboratory frame, the wave reflected on the beam splitter of the Sagnac interferometer suffers a Doppler frequency shift Acweven with a corotating
source. Using Eq. (3) would necessitate taking different index values n(wO) and n (co + Ac) for the transmitted

,OrR

2iw 8 2E n 2 cc 2 1 a2 E 2+ 2+ c2 E+ 2 ++ - 2 2 2
rR OC0R OaZR

C,,

c,,

X=E xQ x-= =

(4)

and the reflected waves, respectively. That would forbid deducing the result from Eq. (1), which assumes that the index n has the same value in the expressions
of CMCR and CMCTR. The error of Vali et al. in Ref. 4 in Eq. (1) and was to use a = (1 - 1/n2 - coo/ndn/dcw)

with the assumption that the radius of curvature R of the waveguide is much larger than its lateral dimensions. This equation is similar to the one proposed by
Post, 3 who gave a perturbation term (2ijrCO Q/C, 2 X

OE/MOR). This dependence on the relative magnetic


permeability A, comes from an incorrect expression of his constitutive tensor. This was corrected by Anderson et al.10 and also by Shiozawa.11 We can notice that

thus to omit the Doppler effect of the reflected wave.


However, in an earlier review paper of the Sagnac effect,

Post3 pointed out this problem, which has persisted since the experiments of Haress in 1912.7 Post postulated3 that the drag coefficient a does not necessarily have to be identical with the usual Fresnel-Fizeau 2 - cw/ndn/dcw), and he used the coefficient (1 - 1/n correct coefficient (1 - 1/n 2) for the Sagnac experiment to match the results of the kinematic and the electromagnetic theories. We have seen how to explain this result simply by comparing Eqs. (2) and (3). Leeb et al. 5 used Post's result, but their Letter focuses on the right interpretation of another usual method of testing
a fiber-rotation sensor in which only the fiber spool

for optical frequencies in a dielectric medium, Ar is, in any case, equal to unity with good precision, and therefore the experimental verification of either of the two results is impossible in such an experiment. Now the boundary conditions imposed by a weakly
guiding bent single-mode fiber with an axis of curvature

z'IzI give an approximate solution for a system at rest


in F1 (Refs. 12-14): Ej(r 1 , 01, zi) = A(r 1 , zi) exp(-if3ROI), wherethe amplitude A(r1 , zj) = a expj[-zj 2 2 2 - (ri - R - AR)]/2wo j and (r1 , 01, zj) are the cylindrical spatial coordinates of FI, a is the maximum amplitude of the field, j3is the phase-propagation constant of the mode of the straight fiber, wo is the average radius

rotates while the beam splitter stays at rest. However, they pointed out not only that both configurations are equivalent but also the important fact that, because locally the medium sees the same frequency for both
counterpropagating waves, the presence of a dispersion

of this pseudo-Gaussian mode,'2 -' 4 and AR is the centrifugal shift in modulus produced by curvature' 3 and is equal to wo4 n2k 2/Rc, 2 . Using a method of perturbation, 8'9 we can deduce the general solution of the differential Eq. (4) from the solution El (r 1 , CI, zj) obtained in the particular case in which Q = 0. We look for solutionsof the form ER (rR, OR, ZR) = El(rR, OR, ZR) [1 + AA(rR)Iexp(iAKOR), where AA is the amplitude

term in the Sagnac phase shift is not conceivable. Electrodynamic Method The above kinematic treatment allows one to determine that there is no effect of a medium on the Sagnac

phase shift in the approximation of a circular light path


of a specific.radius. But for a real fiber gyroscope, a more rigorous approach is to consider a toroidal wave-

perturbation and AK is the propagation constant perturbation related to the phase perturbation. Inserting this solution in Eq. (4) gives
2OR |c >Q C,,
2

guide in rotation. Relativistic electrodynamics gives the propagation equation in a rotating frame of refer-

+ (dA/drR)(dAA/drR) rR ) A(rR, ZR)


2

+ d2 A

0
(5)

drR2

August 1981 / Vol. 6, No. 8 / OPTICS LETTERS

403

In the general case of a dielectric and also of a metallic


waveguide, the perturbation AK is given by cQcC, 2 AK/Rm 2 = 0, where RM is a mean radius. In the simple case of the pseudo-Gaussian mode of a single-mode fiber, the value of Rm is (R + AR), and therefore AK = wQ(R + AR) 2 /C,2. The amplitude perturbation is found to be AA(rR) = s X rR, with s = 4wo 2cIQ/c,, 2 .

The amplitude perturbation produces a lateral shift of the mode as in fiber bending, but it depends on the respective directions of rotation and wave propagation through the sign of the product 3Q. A corotating propagation gives a centrifugal shift, and a counterrotating propagation gives a centripetal shift. This effect is usually completely negligible: With, for example, R = 10 cm, wo = 3 um, n = 1.45, c = 2 X 1015 Hz, and Q = 1 rad X s- 1 , the lateral shift that is due to the Sagnac
effect is only AR' = s X w0 2
t6 X 10-11 ,m.

independent of the medium or waveguide properties. Assuming that the value of the Sagnac phase shift can be measured with good absolute precision, the stability of the scale factor of a fiber rotation sensor will depend only on the stability of the mean radius of curvature (R + AR) and the light frequency c. It will also require the stability of the frequencybandwidth to avoid change in the contrast of the interferences when a broadspectral-range source is used. We would like to thank H. J. Shaw for helpful discussions and R. A. Bergh and M. Digonnet for useful suggestions. * Now at Edward L. Ginzton Laboratory, Stanford University, Stanford, California 94305. References
1. H. Fizeau, "Sur les hypotheses relatives Al'6ther lumineux, et sur une experience qui parait d6montrer que le mouvement des corps change la vitesse avec laquelle la
lumibre se propage dans leur int6rieur," C. R. Acad. Sci. 33, 349-354 (1851). 2. G. Sagnac, "L'6ther lumineux demontr6 par l'effet du vent

The shift

that is due to bending is AR - + 0.2 Aum. The problem is not the same for the perturbation of the phase. The relative change in the phase AO/q5 that is due to rotation is also small, but since the phase is cyclical, its absolute variation AO is measured with respect to 27rin the interferences. The value AkI/27r accumulates with the length of the propagation path.
After N turns of counterpropagation with the respective propagation constants [+I3 1 - (AK/R)] and [-1I1 (AK/R)] and the same frequency cc,the phase difference

relatif d'6ther dans un interferomAtre en rotation uniforme," C. R. Acad. Sci. 95, 708-710 (1913).

3. E. J. Post, "Sagnac effect," Rev. Mod. Phys. 39,475-494


(1967). 4. V. Vali, R. W. Shorthill, and M. F. Berg, "Fresnel-Fizeau

is Aq5= 47rNAK = 4NwQa/C, , which is precisely the result of the kinematics method with a = 7r(R + AR) 2 for the area of the surface enclosed by the mean light path corresponding to the maximum of the modulus of
the guided mode. Only the light frequency c and speed in a vacuum c, are involved in this result. The medium

effect in a rotating optical fiber ring interferometer,"


Appl. Opt. 16, 2605-2607 (1977). 5. W. R. Leeb, G. Schiffner, and E. Scheiterer, "Optical fiber

gyroscopes: Sagnac or Fizeau effect?" Appl. Opt. 18,


1293-1295 (1979). 6. See, for example, M. A. Tonnelat, Principles de la thgorie

or waveguide dispersion and the phase or group velocities do not have any influence to first order in rR Q/c,,, regardless of the type of waveguide, even for metallic
guides in which the dispersion can be important and the

electromagnetique et de la relativite (Masson, Paris, 1959) [Principles of Electromagnetic Theory and Relativity (Gordon and Breach, London, 1966)].
7. F. Haress, "Die Geschw. d. Lichtes in bewegten Korpern," Thesis Dissertation (University of Jena, Erfurt, Germany, 1912). 8. H. C. Lefevre, "Gyromatre interferom6trique A fibre op-

phase velocity higher than c,.

6 pulses15"1 instead of continuous waves does not modify

The use of light

this result. Pulses can be considered a linear superposition of continuous waves of different frequencies.
Each wave suffers a different Sagnac phase shift, which results in a decrease of contrast as in any interferometer

tique," Doctoral Thesis (University of Paris-Orsay,


1979). 9. H. J. Arditty and H. C. LefAvre, "Electromagn6tisme des

using a large-bandwidth light source.


We have treated only the case of a toroidal fiber bent

milieux dielectriques lin6aires en rotation et application Ala propagation d'ondes guidees," to be published. 10. J. L. Anderson and J. W. Ryon, "Electromagnetic radiation in accelerated systems," Phys. Rev. 181, 1765-1775 (1969). 11. T. Shiozawa, "Phenomenological and electron-theoretical

in a plane orthogonal to the axis of rotation, but this


approach gives, through a more complicated calculation, the same result for a closed path of arbitrary shape using

the area of an equivalent apparent surface. Conclusions Whereas the kinematic method simply explains the Sagnac phase shifts, the electrodynamic approach gives the complete expressions for the counterpropagating single transverse modes of a dielectric-waveguide ring interferometer. It shows, without ambiguity, that the equivalent area used in the formula for the phase shift of the fiber gyroscope is the mean area of the guided closed path. Furthermore, using the rest frame of the interferometer avoids the need to take into account Fizeau and Doppler effects. We find directly from the propagation equation that the phase perturbation is

study of the electrodynamics of rotating systems," Proc. IEEE 61,1694-1702 (1973).


12. K. Petermann, "Fundamental mode microbending loss

in graded-index and W fibres," Opt. Quantum Electron.


9,167-175 (1977). 13. W. A. Gambling, H. Matsumura, and C. M. Ragdale, "Field deformation in a curved single mode fibre," Electron. Lett. 14, 130-132 (1978). 14. D. Marcuse, "Loss-analysis of single-mode fiber splices," Bell Syst. Tech. J. 56, 703-718 (1977).

15. H. J. Arditty, H. J. Shaw,M. Chodorow,and R. Kompfner, "Re-entrant fiber optic approach to rotation sensing," Proc. Soc. Photo-Opt. Instrum. Eng. 157, 138-148
(1978). 16. C. C. Cutler, S. A. Newton, and H. J. Shaw, "Limitation

of rotation sensing by scattering," Opt. Lett. 5, 488-490


(1980).

You might also like