September 1, 1990 / Vol. 15, No.

17 / OPTICS LETTERS

947

Wavelength dependence of bending loss in monomode optical fibers: effect of the fiber buffer coating
R. Morgan, J. S. Barton, P. G. Harper, and J. D. C. Jones Department of Physics, Heriot-Watt University, Riccarton, Edinburgh EH14 4AS, UK
Received February 2, 1990; accepted June 22, 1990

The bend-induced optical loss in a monomode fiber has been studied as a continuous function of wavelength in the
range 1.2-1.6 tm. The loss was observed to be an oscillatory function of wavelength. These observations were explained by using a model based on interference between the guided core mode in the fiber and a "whispering gallery" mode ejected from the core by the bend. The wavelengths of the minima in the oscillatory function were

accurately predicted by the model when the propagation of optical power in the fiber buffer coating was taken into account.

Since the advent of optical waveguides the phenomenon of bend-induced losses has been recognized.' In many applications this phenomenon is a disadvantage

because of its adverse effect on the power budget. 2' 3

However, it has also been exploited in a range of practical devices. For example, it has been used as a transduction mechanism in fiber-optic sensors, chiefly by using multimode fibers.4 In both monomode and multimode fibers it has been used to eject guided light from the core into a receiver, thus permitting the fiber guided beam to be monitored, for example, as an aid in
splicing operations. 5' 6 The dependence of loss on

that the observed loss oscillation'5 may be understood as being due to interference between the core mode and cladding WG modes. The oscillatory power loss minima depend on X according to the interference condition
27r 1 X

Z + sp= 2m7r,

m an integer,

(1)

wavelength is especially important with developments in wavelength-division-multiplexed systems in both communications7 and sensors8 and in sensors that use multiple-wavelength techniques. 9 Considerable theoretical research has been carried out on the subject of bend losses in both multimode
and monomode fibers (see, e.g., Refs. 10-14), generally

with the assumption that the cladding of the fiber has an infinite extent. These theories have predicted a monotonic functional form for the attenuation versus the bend radius or arc length. More recent experimental and theoretical research has considered bend loss for practical monomode fibers, with finite cladding dimensions, and has shown oscillation of the attenuation with the bend parameters at a few specific
wavelengths. 15,16

where 1 is the optical path length of the WG mode, -f3Z is the relative phase at the caustic surface'5 with fi being the propagation constant of the core guided wave and Z the arc length shown in Fig. 1, and sois a phase constant. In the original theory,' 5 1was calculated as the product 2Lncl, as shown in Fig. 1(a). This approach assumes that the dominant reflection of the WG ray occurs at the cladding-buffer interface. The optical properties of the fiber used in the present experiments are summarized in Table 1. It can be seen that the refractive indices of the cladding and buffer are similar, and hence the reflectivity of the cladding-buffer
interface is weak (<1%). Most of the WG ray is thus

refracted into the buffer and experiences a strong reflection at the buffer-air interface. We accordingly calculate 1 as shown in Fig. 1(b) such that

1= 2(nc Ll + nblL2 + nb2 La), 1

(2)

In this Letter we study bend loss in monomode fibers as a function of wavelength in the range 1.2-1.6 Am. Strong oscillations of attenuation with varying wavelength are observed. We predict accurately the positions of the minima in the experimentally observed variation of bend loss with wavelength and see that the optical properties of the fiber buffer coating must be taken into account. According to the theories of Marcuse and others,'0' 4 the pure bending loss in monomode fibers is an exponentially decreasing function of bend radius and optical wavelength X. Gambling et al. 1 4 observed light emission as discrete rays from a bend and proposed that this was consistent with tunneling of "whispering gallery" (WG) rays. Harris and Castle have shown
0146-9592/90/170947-03$2.00/0

and the value of 1is thus increased. The accuracy of the position of the loss minima predicted by Eq. (1) and shown in Fig. 3 below has prompted a study of the diffractive basis of the model,

using the Fresnel-Kirchhoff theory. Considered

7 along with the research of Snyder and Love,1 the geometrical ray approach is justified at least for the laminar optical model and for sufficiently large bend radii. The experimental arrangement is shown in Fig. 2. A white-light source was used in conjunction with a monochromator to illuminate the fiber with light of wavelength tunable over the range 1.2-1.6 Am with a resolution of 5 nm and a power of approximately -60 dBm. The detector was a power meter that used lock 1990 Optical Society of America

948

OPTICS LETTERS / Vol. 15, No. 17 / September 1, 1990

R\ / / , //

\\
l

\
Wnd | l l l
V

To establish the role of optical power propagating in the buffer coating, the experiments were repeated with a suitable index-matching gel applied to the fiber bend region. The optical power propagating in the buffer coating was thus effectively removed by reduc-

ing reflectionsthebuffer at interface.

Ancl

BENDING RIG

(a)

(b)

Fig. '2. Experimental arrangement. IEEE, parallel interface.

15 -

Fig. 1. Geometry of the WG mode interference model for (a) unbuffered and (b) two-component buffered fiber. R, bend radius; d, separation of the fiber axis and radiation caustic' 5 ; nco, ncl, nbl, nb2,refractive indices of the fiber core, cladding, and inner and outer buffers, respectively; Z, 'y,core arc length and angle, respectively; L, L1, L2 , L3, WG ray lengths; y,, Y2,Y3, outside radii of cladding and inner and outer buffers, respectively.

(a)

10

51

Table 1. Fiber Parameters

I

Cutoff wavelength, Xo 1.232 ,m Mode field radius 4.89 Am Cladding (pure silica) refractive index, n,1 1.445 (at 1.55 Am) Profile height parameter, A 2.89 X 10-3

.5 1.2 1.3 1.4 WAVELENGTH Vpm 1.5 1.6

Numerical aperture
Cladding diameter

0.11
125 ,m
15

Buffer coating Inner layer Outside diameter

Refractive index, nbl

(b)

205,m
1.54 d: 0.03
10

Outer layer Outside diameter Refractive index, nb2

250 um 1.53 d 0.03

m 0

!Z

in detection with a sensitivity of -90 dBm and a resolution of 0.01 dB. The bending rig would accept mandrels covering a range of radii, with provision for mounting the fiber so that the bend arc length (expressed as an angle) could be adjusted, under controlled tension sufficient to ensure that the curvature of the mandrel was followed. For each bend loss measurement, a zero-bend reference power level (i.e.,with the fiber straight) was measured as a function of wavelength. The bend arc length was then increased to the desired value, and the power level was again measured as a function of wavelength, and the net loss due to the bend was obtained by subtraction of the reference power.

mb~~~~rT1'zY'''! '~~~~~~~ Cf
,s 1.2 1.3 1.4 WAVELENGTH pm 1.5 1.6

Fig. 3. Bend loss as a function of wavelength for the fiber of Table 1. (a) Bend diameter 13 mm, arc length 1200. (b) Bend diameter 10 mm, arc length 600. The curves are spline fits to the experimental data, shown as squares. The dashed arrows and solid arrows are the theoretically predicted positions of loss minima using the models of Figs. 1(a) and 1(b),

respectively.

September 1, 1990 / Vol. 15, No. 17 / OPTICS LETTERS

10

949

zE
0

1.2

1.3

14 WAVELENGTHPm

15

16

Fig. 4.

Bend loss as a function of wavelength for the fiber of

Table 1; bend diameter 10mm, bend length 1800. Spline fit to the experimental data with (dashed curve) and without (solid curve) index-matching gel applied to the exterior of the fiber. Measurements were made as a function of wave-

length. Nevertheless, the measurement as a function of wavelength is likely to be more accurate because it does not require mechanical adjustment of the fiber during the experiment. In particular, it is difficult to envisage a practical technique for the continuous variation of bend radius. It is clear that our observed bend loss is a strong function of wavelength. Furthermore, the fiber and bend geometries used are typical of practical arrangements for local light detectors. This suggests that optimum design of such devices must take into account the optical properties of the buffer coatings of the fibers to be used. The strong wavelength dependence of loss is also a complicating factor in the design of intensity-modulated fiber-optic sensors that use a second wavelength to provide a reference intensity, because such devices assume that the bend loss is approximately equal for both wavelengths.9 The authors thank David Ferguson and C. P. Botham of British Telecom Research Laboratories (BTRL) for many helpful discussions. This research was partially supported by BTRL and the Science and Engineering Research Council (UK), whom R. Morgan thanks for the provision of a research studentship. References
1. L. Lewin, D. C. Chang, and E. F. Kuester, Electromag-

length from 1.2 to 1.6 ,m in 5-nm steps, for bend radii from 2.5 to 7.0 mm in~0.5-mm steps, and for bend
angles 600 to 1800. Two sets of experimental data are shown in Fig. 3. The general form is an increase of loss

with wavelength onto which is superimposed an oscillatory component whose spacing increases with wavelength. As expected, the loss is higher at smaller bend radii. The model described above was used to calculate the wavelengths corresponding to minima in the loss function, i.e., phase matching of the core and cladding modes, for each combination of bend radius and angle. Satisfactory fits were obtained by using the fiber data
of Table 1, assuming values of 1.51 and 1.515 for the

netic Waves and Curved Structures (Peregrinus, London, 1977).

2. D. Gloge, IEEE Trans. Microwave Theory Tech. MTT23, 106 (1975). 3. D. Marcuse, Bell Syst. Tech. J. 55, 1445 (1976).

4. J. N. Fields, C. Asawa, 0. G. Ramer, and M. K. Barnowski, J. Acoust. Soc. Am. 67, 816 (1980). 5. C. M. de Blok and P. Mathiesse, Electron. Lett. 20, 109 (1984). 6. P. V. Andrews and D. Owens, Proc. Soc. Photo-Opt. Instrum. Eng. 522, 18 (1985).

inner and outer buffer refractive indices, respectively. For comparison, the positions of loss minima calculated by omitting the effect of the buffer coating are
also shown in Fig. 3 (assuming that the WG mode is

7. H. Mahlein, Fiber Integrated Opt. 4,4 (1983).

8. A. Dandridge and A. D. Kersey, Proc. Soc. Photo-Opt. Instrum. Eng. 798, 158 (1987).

reflected at the cladding-buffer interface) and clearly do not reproduce the spacing of the observed turning points. As direct confirmation of the importance of the buffer-air interface in controlling bending loss, experiments were also carried out with refractive-index-matching gel applied to the buffer coating. The
results are shown in Fig. 4, and it may be seen that the

9. B. E. Jones and R. E. Spooncer, Proc. Soc. Photo-Opt.

Instrum. Eng. 514, 223 (1984). 10. D. Marcuse, J. Opt. Soc. Am. 66, 216 (1976). 11. D. Gloge, Appl. Opt. 11, 2506 (1972). 12. W. A. Gambling, H. Matsumura, and C. M. Ragdale, Opt. Quantum Electron. 11, 43 (1979).

13. W. A. Gambling, H. Matsumura, C. M. Ragdale, and R.

A. Sammut, IEE J. Micro Opt. Acoust. 2, 134 (1978). 14. W. A. Gambling, D. N. Payne, and H. Matsumura, Electron. Lett. 12, 567 (1976).

oscillations in the loss function are strongly damped, thus confirming that the WG mode had been strongly attenuated. These results show similarity with those recently reported by Valiente and Vassalo,'8 who studied bend loss in a fiber with a lossy buffer coating.

15. A. J. Harris and P. F. Castle, IEEE J. Lightwave Technol. LT-4, 34 (1986). 16. S. K. Yao, C. K. Asawa, and G. F. Lipscomb, Appl. Opt. 21, 3059 (1982). 17. A. W. Snyder and J. D. Love, IEEE Trans. Microwave

Our experimental research was principally directed at determining the bend loss as a function of wavelength. However, the theoretical model is equally applicable to bend loss as a function of bend radius and

Theory Tech. MTT-23, 134 (1975).

18. I. Valiente and C. Vassalo, Electron. (1989). Lett. 25, 1544