1966 OPTICS LETTERS / Vol. 21, No.

24 / December 15, 1996

Direct continuous-wave measurement of n2 in various

types of telecommunication fiber at 1.55 mm
A. Boskovic, S. V. Chernikov, and J. R. Taylor
Department of Physics, Imperial College of Science, Technology and Medicine, Prince Consort Road, London SW7 2BZ, UK

L. Gruner-Nielsen and O. A. Levring

Lycom A/S, NKT Allé 75, DK-2605, Brøndby, Denmark

Received July 5, 1996

A method for measuring the nonlinear refractive index of optical fibers with an error of less than
5% is demonstrated. The technique is based on measuring the nonlinear phase shift experienced by
a dual-frequency beat signal, permitting a simple, highly sensitive, accurate, repeatable, and easily
automated measurement procedure and sampling. Measurements of the nonlinear coefficient in standard
telecommunication, dispersion-shifted, and a number of dispersion-compensated fibers are presented.  1996
Optical Society of America

It has been realized over the past few years that
nonlinearities are playing a decisive role in currently
developing high-bit-rate transmission systems that
use solitons and nonreturn-to-zero formats and include
wavelength-division multiplexed systems. At the
same time, the idea of dispersion management in a
transmission system has been shown to be useful not
only for upgrading a currently installed system but
also for ultrahigh-bit-rate transmission. These sys-
tems may incorporate different types of fiber, including
standard telecommunication fiber (STF), dispersion-
shifted fiber (DSF), or both, in conjunction with
dispersion-compensating f ibers (DCF). To implement
or accommodate nonlinearities, one must know the
nonlinear coeff icient, n2 , for all types of fiber used,
and the effect of the constituents of the f iber on the
nonlinearity must be readily determinable. Therefore
there is increasing interest in accurate measurement
of the nonlinear refractive indices of f ibers of different
compositions.
Several methods of measuring n2 that exploit differ-
ent nonlinear effects, such as self-phase modulation
(SPM) by use of a pulsed laser,1,2 cross-phase modu-
lation,3,4 modulation instability,5 and a technique
based on four-wave mixing,6 have been proposed.
These techniques have been used mostly for STF
and DSF. DCF has been evaluated in Refs. 3 and 4;
however, the values of n2 measured in DSF and STF by
this technique are generally significantly higher than
those measured by others. These techniques require
a special signal source, have limits on the length and
dispersion of fiber that can be measured, or suffer
from a large uncertainty in the nonlinear phase shift
measurement. Here we demonstrate a technique that
has the advantages of using an ordinary pump source,
a conventional measurement apparatus, and simple
sampling with high accuracy and sensitivity.
The method is based on measurement of the nonlin-
ear phase shift induced through the SPM effect, with
a cw dual-frequency beat signal used as a pump sig-
nal.7 Neglecting dispersion, the nonlinear phase shift
experienced by the beat signal propagating along the
test f iber can be expressed as7,8
Est, Ld ­ 2a cossDvtdexpsiv0 tdexpfiwSPM cos2 sDvtdg ,
2v0 n2
wSPM ­ Leff P , (1)
c Aeff
where P is the average power of the signal, Leff is the
effective length of the fiber, n2 yAeff is the nonlinear
coefficient of the f iber, n2 is the nonlinear refractive in-
dex, and Aeff is the effective area. To determine the
nonlinear coeff icient of the fiber we measure experi-
mentally the SPM nonlinear phase shift and the corre-
sponding average power. The nonlinear phase shift is
measured in the spectral domain. As the electric f ield
Estd in Eq. (1) is a periodic function in time, its spec-
trum is descrete, consisting of harmonics of the beat
frequency as shown in Fig. 1. The nonlinear phase
shift wSPM is determined from the shape of the spec-
trum, i.e., the relative ratio of the spectral components.
In our measurements we use only the ratio of intensi-
ties of the first-order sideband to the spectral inten-
sity at the fundamental frequencies. By taking the
Fourier transform of Eq. (1), we can analytically cal-
culate this ratio in terms of Bessel functions as
I0 J0 2 swSPM y2d 1 J1 2 swSPM y2d ,
­ 2
I1 J1 swSPM y2d 1 J2 2 swSPM y2d
where I0 and I1 are the intensities of the zero- and first-
order harmonics (Fig. 1) and Jn is the Bessel function of
the nth order. The nonlinear phase shift is a function
of I1 yI0 only, which can be readily measured.
The experimental setup is similar to that de-
scribed in Ref. 7. The optical beat signal was de-
rived from two cw distributed-feedback diode lasers
operating at , 1.55 mm. The spectral separation
was adjusted through control of the laser tempera-
ture. Although the measurement result does not
depend on either the frequency separation or the
laser linewidths, the relative stability of the frequency

0146-9592/96/241966-03$10.00/0  1996 Optical Society of America

 December 15, 1996 / Vol. 21, No. 24 / OPTICS LETTERS 1967

used fiber with a length of the order of 500 m and a

wavelength separation of ,0.2 nm. Despite the fact
that the nonlinear coefficient of the STF is considerably
lower than that of a DSF, a nonlinear phase shift
of 0.17 could be achieved with an input power of
130 mW. This phase shift was sufficient to permit
high accuracy of the measurement.
DCF’s are designed to have exceedingly high normal
dispersion at 1550 nm, at present up to 300 psy(nm
km). One way to overcome any problems with
dispersion in the measurements is to reduce the
beat-frequency separation, which shows a square-law
dependence as shown in inequality (2). However, for
Fig. 1. Typical SPM spectrum generated by the propa- convenience we preferred to use a conventional diffrac-
gation of a beat signal in a f iber. tion grating spectrum analyzer with , 0.1-nm reso-
lution, which required no critical alignment and had
separation, which was better than the resolution of our a high accuracy of spectral power measurement and
spectrum analyzer (0.1 nm), permitted high-accuracy a high dynamic range. Although the closer the two
measurement. The beat signal was then amplif ied frequencies the lower the dynamic range for measuring
in a preamplifier and transmitted through a 2-nm the f irst-order sideband, the compromise of having
bandpass f ilter to suppress the amplified stimulated ,200-m-long fiber under measurement and a wave-
emission and a polarizer to a following high-power length separation of 0.16 –0.18 nm was found. This
erbium amplifier. The polarizer was used to ensure permitted a maximum nonlinear phase shift of ,0.5,
identical polarization and equal intensities of the which could be readily measured by this technique,
two frequencies, which were adjusted by polarization which can detect wSPM as small as 0.1. The accuracy
controllers. The amplified stimulated emission level of the measurement is depicted in Fig. 2(b). To ensure
after the power amplifier was typically less than that dispersion effects did not inf luence the result, we
0.2% of the total output power. The power amplifier measured the slope for different wavelength separa-
(IRE-POLUS Group) was able to deliver output powers
as high as 28 dBm (,600 mW). The amplif ied beat
signal was then directed into a test fiber via a variable
attenuator, which was used to control the input power.
Both the launched power and the spectrum were moni-
tored with an accurately calibrated powermeter (An-
ritsu, M99020B) and a spectrum analyzer (Anritsu,
MS9701B optical unit and MS9030A display unit).
A typical curve measured for a DSF of 5.15-km
length is shown in Fig. 2(a). The plot could be well
interpolated by a linear fit, and the slope is determined
with standard deviation of typically less than 0.5%.
From the measured slope, an average value of the non-
linear coefficient of the fiber is calculated from Eq. (1).
Measurements of the same fiber were repeated a num-
ber of times to ensure high repeatability of the results
[Fig. 2(b)].
To neglect dispersion in different types of fiber that
have different chromatic dispersions at 1550 nm in the
measurement algorithm above one must satisfy the
following criterion:

DwSPM k2 Dv0 2 L ,, 1 , (2)

where k2 ­ ≠2 ky≠v 2 is the f iber’s chromatic dispersion.

This condition is easily achieved in DSF’s up to a few
kilometers long with an input beat-frequency signal
separation of 0.3– 0.5 nm. In this case an average
power of as much as 100 mW coupled into the f iber
is suff icient to yield a nonlinear phase shift of 1 –2
[Fig. 2(a)]. The high linearity of the nonlinear phase
shift versus power dependence indicates that we are Fig. 2. (a) Plot showing a typical measurement of the
well below the limit set by dispersion [Eq. (2)]. nonlinear phase shift against the power together with a
STF’s have an anomalous dispersion of ,17 psy(nm linear f it and (b) histogram of several measurements of the
km) at 1550 nm. To meet the condition of Eq. (2) we nonlinear coeff icient for a particular DCF.
1968 OPTICS LETTERS / Vol. 21, No. 24 / December 15, 1996

Table 1. Summary of All Results Obtained and Some Relevant Parameters for the Fibers Investigated

Mol % Ge Core D at 1.55 mm Core Æ Mode n2 yAeff 3 1029 Aeff n2 3 10216

Type of Fiber # (F Cladding ) [ psy(nm km)] s mmd Field Æ s mmd s W21 d s mm2 d scm2 yWd

Single Mode 1 – 1 17.0 8.3 10.5 0.27 80 2.2

DSF 1 5.5 2 0.2 4.44 8.62 0.424 55.1 2.33
2 – 1 1.0 – 9.00 0.385 60.0 2.31
3 6.0 2 1.5 4.35 7.80 0.495 46.2 2.29
4 4.5 1 0.8 4.41 8.52 0.421 53.8 2.30
5 – – – 7.83 0.521 45.5 2.37
DCF 1 21 2 72 2.07 4.92 1.47 18.2 2.67
2 18 2 62 2.15 5.62 1.02 24.0 2.45
3 16 2 54 2.32 5.49 1.12 22.7 2.54
4 13 2 35 2.86 5.74 1.04 24.5 2.55
DCF 5 16 (3.6) 2 130 2.52 5.21 1.11 22.2 2.46
6 16 (3.6) 2 108 2.51 5.01 1.30 19.4 2.52
7 16 (3.6) 2 90 2.50 4.71 1.55 16.9 2.63
8 16 (3.6) 2 58 2.66 4.55 1.78 15.4 2.75

tions in the range of 0.15 –0.25 nm for the f iber with a
dispersion of 108 psy(nm km). No dependence of the
slope on the frequency separation within this range
was found. Also, measurements with a 50-m piece of
this fiber were made, and the same value was obtained
for the nonlinear coefficient.
Another problem that has to be addressed for a
short piece of fiber is the dependence of the nonlinear
refractive-index on the polarization state. In non-
polarization-maintaining f ibers the evolution of the
polarization state leads to an average number that
is 8y9 of that for the linear polarization state.7,9 To
confirm that the input polarization state does not
inf luence the results in short pieces of fiber we took
measurements for various input polarization states
in a 50-m-long DCF. No significant dependence of
n2 on the input polarization was observed within the
measurement error of ,5%.
The results of all measurements of various types
of f iber are summarized in Table 1. For STF the
nonlinear coeff icient for random polarization of 0.27 3
1029 W21 was measured and corresponded to n2 ­
2.2 3 10216 cm2 yW, assuming that the effective area
was 80 mm2 .
In DSF the nonlinear coeff icient is ,0.4 3 1029 W21 ,
almost two times greater than that for STF, mainly
because of a smaller effective area. It was also found
that the nonlinear coeff icient in DSF could change from
fiber to f iber by as much as 20%. The effective area
for DSF was calculated from the measured mode-f ield
diameter (Peterman II).10
For DCF’s the value of the nonlinear coefficient
is significantly greater. We measured a number of
different DCF’s to determine the effects of various
dopants, their concentrations, and f iber geometry on
the value of the nonlinear coeff icient. DCF’s normally
have high Ge concentrations in the core. Ge doping in-
creases the value of n2 in the core, but the value of n2 in
a single-mode fiber is an average of that in the core and
in the cladding, balanced by the conf inement factor. F
doping of the cladding was found to decrease n2 . The
effective area was calculated from the near-f ield mode-
field distribution obtained through the Hankel trans-
form of the measured far-f ield distribution. The value
was confirmed by numerical simulation, which deter-
mined the relationship between the mode-field diame-
ter and the effective area.11 From the data obtained,
the contribution of Ge and F doping has been evalu-
ated as11
n2 ­ n2 sSiO2 d 1 KsGeO2 dXGeO2 1 KsF dXF , (3)
where n2 sSiO2 d ­ 2.16 3 10216 cm2 yW and KsGeO2 d ­
0.033 3 10216 cm2 ys%Wd, KsF d ­ 20.017 3 10216 cm2 y
s%Wd, and XGeO2 and XF are the mole concentrations
of GeO2 and F , respectively, in percent.
References
1. K. S. Kim, R. H. Stolen, W. A. Reed, and K. W. Quoi,
Opt. Lett. 19, 257 (1994).
2. Y. Namihira, A. Miyata, and N. Tanahashi, Electron.
Lett. 30, 1171 (1994).
3. T. Kato, Y. Suetsugu, M. Takagi, E. Sasaoka, and M.
Nishimura, Opt. Lett. 20, 988 (1995).
4. T. Kato, Y. Suetsugu, and M. Nishimura, Opt. Lett. 20,
2279 (1995).
5. M. Artiglia, E. Ciaramella, and B. Sordo, Electron.
Lett. 31, 1012 (1995).
6. L. Prigent and J.-P. Hamaide, IEEE Photon Technol.
Lett. 5, 1092 (1993).
7. S. V. Chernikov and J. R. Taylor, Opt. Lett. 21, 1559
(1996).
8. R. H. Stolen, in Optical Fiber Communications, S. E.
Miller and A. G. Chynoweth, eds. (Academic, New York,
1979), pp. 125–150.
9. S. G. Evangelides, L. F. Mollenauer, G. P. Gordon,
and N. S. Bergano, IEEE J. Lightwave Technol. 10, 28
(1992).
10. Y. Namihira, Electron. Lett. 30, 262 (1994).
11. L. Grüner-Nielsen, A. Boskovic, S. V. Chernikov, and
J. R. Taylor, presented at the European Conference
on Optical Communications (ECOC’96), Oslo, Norway,
September 15 – 19, 1996.