Sunteți pe pagina 1din 5

15 May 1998

. Optics Communications 151 1998 3539


Measurement of chromatic dispersion in optical fibers
using pairs of correlated photons
J. Brendel, H. Zbinden, N. Gisin
Uniersite de Genee, Group of Applied Physics, 20 rue de lecole de medecine, CH-1211 Genea 4, Switzerland `
Received 7 December 1997; revised 20 February 1998; accepted 5 March 1998
Abstract
Photon pairs generated by spontaneous parametric downconversion are used to implement a novel technique for
measuring the chromatic dispersion of optical fibers. The time and energy correlation of the photons of a pair allow a
measurement with a high time resolution which is only limited by the detection electronics. The device gives direct access to
the differential group delays at several wavelengths around 1.3 mm. Our prototype setup is capable of measuring group delay
differences with a resolution of about 5 ps, only limited by the upper delay limit of the time interval analyzer presently 50
. ns . q1998 Elsevier Science B.V. All rights reserved.
PACS: 42.50.Dv; 42.62.Eh; 42.65.Ky; 42.81.Cn
Keywords: Nonclassical photon states; Metrological application; Frequency conversion; Chromatic dispersion
1. Introduction
Chromatic dispersion in optical fibers is usually deter-
mined by the pulse delay method or the phase shift method
w x 1 . For short fibers, where the group delay differences are
of the order of ps over a few nm, interferometric tech-
w x niques are applied due to their higher time resolution 2,3 .
The direct measurement of the group delay by the pulse
delay method or the phase shift method usually requires
several diode lasers operating at different wavelengths.
The resolution is limited by the wavelength stability, the
pulse width and the jitter of the laser pulse and the
detectors.
We present a novel technique which provides direct
access to the group delay difference at several wavelengths
comparable to the pulse delay method. In contrast to all
existing methods we use a continuous light source emitting
pairs of photons which are highly correlated in time and
energy. These photon pairs are generated by parametric
w x downconversion of laser light in a nonlinear crystal 4 .
This process can be considered as the spontaneous anni-
hilation of a pump photon with frequency v and the
p
instantaneous creation of two photons, a signal and an idler
photon with frequencies v , v , respectively, which share
s i
the pump photons energy according to
v sv qv . 1 .
p s i
Hence, only the narrow spectral width of the pump laser
limits the uncertainty of the sum energy. The time correla-
tion of the signal and idler photon has been subject to
w x several investigations 5,6 . It has been found that this
temporal correlation is only limited by the temporal spread
of the single photon wave packets, i.e. the coherence time
of the emitted light which is usually below 1 ps.
. The energy conservation rule 1 allows for a broad
emission spectrum of the downconverted light. Therefore,
in conjunction with its perfect time correlation properties
this light field is ideally suited for application in chromatic
dispersion measurements. The technique presented in this
article makes use of photon pairs as a broadband light
source, comparable to a LED, with subpicosecond
pulsewidth. Further, choosing one photon wavelength by a
filter directly determines the wavelength of the second one
. according to the energy correlations 1 . Therefore we can
0030-4018r98r$19.00 q 1998 Elsevier Science B.V. All rights reserved.
. PII S0030- 4018 98 00140- 0
( ) J. Brendel et al.rOptics Communications 151 1998 3539 36
have direct access to the group delay difference between
these two wavelengths using a single filter.
The frequency correlation also gives rise to unique
w x dispersion properties of the two-photon light field 710 .
If the central wavelength l is chosen close to the zero
dispersion wavelength, the signal and idler photons will be
almost equally delayed by the fiber even if both have a
large wavelength uncertainty. The dispersion of the signal
photons is therefore compensated by the dispersion of the
idler photons and the resulting pulse broadening will be
much smaller than expected for a classical pulse of equal
w x spectral width 7 . This cancellation is only limited by the
spectral width of the pump leading to an uncertainty in Eq.
. 1 . The effect of the time delay correlations on our
dispersion measurement technique will be examined in
Section 4.
For more than 20 years now, photon pairs generated by
parametric downconversion are the basis of innumerable
experiments revealing the quantum nature of light. But
despite their interesting properties only a few practical
applications have been suggested or implemented yet.
Examples demonstrating the potential of this two-photon
metrology are the measurement of detector quantum effi-
w x w x ciencies 1114 , quantum cryptography 15,16 and inter-
w x ferometry 17 . One major problem one had to face when
trying to realize practical applications was the lack of
compact and cheap two photon sources. This was mainly
due to the need for an argon laser as the pump. In contrast
to all previous two-photon sources our source is based on a
standard red laser diode. As a consequence we were able
to reduce the source size down to a box of 50=50=15
cm
3
and the cost of the pump laser by a factor of 10
y2
.
2. The experimental setup
The outline of our experimental setup is shown in Fig.
1. A temperature and current stabilized laser diode
. RLT6515G operating at a wavelength of l s660.4 nm
p
serves as the pump of the two photon source. The laser
emits 15 mW optical power in a single spatial and longitu-
dinal mode. Infrared emission is filtered out by the dispers-
. . ing prism DP and the slits S . The red light is slightly
. focused into a nonlinear crystal of LiNbO LN of 10 mm
3
length, cut at fsy908, Qsy58.68. Under these angles
the phase matching conditions for degenerate parametric
downconversion of the laser light to two-photon light with
central wavelength of 1320.8 nm are fulfilled. Behind the
crystal the laser light is blocked by the color-glass edge
. filter F Schott type RG 1000 . The transmitted two-photon
. light has a spectral width FWHM of about 90 nm corre-
sponding to a coherence time of 40 fs. After having passed
the optical fiber under test, the photon pairs enter a
standard 3 dB fiber coupler. An interference filter with 10
. nm FWHM bandwidth at 1320 nm selects the spectral
range of the photons arriving at detector D1. The central
wavelength can be tuned over a 13201270 nm range by
tilting the filter angle between 08 and 308 from normal
incidence. Both detectors D1 and D2 are germanium
. avalanche photodiodes NEC NDL5131 operated in the so
w x called passively quenched Geiger-mode 18 . In this opera-
tion mode the detectors are biased about 0.5 V above the
breakdown voltage over a load resistor. In that way a
single photon will induce a huge avalanche such that
single photon counting in the 1.3 mm wavelength range is
possible. Dark count rates are about 100 kHz and quantum
efficiencies are about 15%. After amplification and pulse
shaping the output signals are fed into a time-to-
. amplitude-converter TAC . The TAC allows the determi-
nation of the difference of the photon arrival times at
detectors D1 and D2, respectively, with high resolution
within a 50 ns time interval. The time difference distribu-
tion is finally stored and analyzed by a computer.
3. Measurement procedure and results
The procedure for measuring the chromatic dispersion
is based on the time and wavelength correlation of the
photons as follows: If a photon pair has passed the test-fiber
and is split by the 3 dB coupler, the signal photon heading
towards detector D1 will only arrive if its wavelength l
s
lies within the range selected by the bandpass filter. Ac-
. cording to the energy conservation equation 1 the wave-
length of the correlated idler photon arriving at D2 is given
by
1rl s1rl y1rl 2 .
i p s
and is thus also determined by the passband of the filter.
Therefore the two detectors will respond simultaneously if
Fig. 1. Experimental setup, for details see Section 2.
( ) J. Brendel et al.rOptics Communications 151 1998 3539 37
Fig. 2. Measured time difference distributions for six different settings of the bandpass filter using 6.5 km of standard telecom fiber.
the two photons at l and l have been identically de-
s i
layed by the fiber or after a time interval given by the
difference of the group delays at these wavelengths.
An example of such a time interval distribution ob-
tained with 6.5 km of standard telecom singlemode fiber is
shown in Fig. 2. The six displayed peaks correspond to six
different settings of the bandpass filter central wavelength.
The solid lines are Gaussian fits to the data using a
nonlinear fitting procedure. The first peak from the left is
due to photons with almost the same wavelength and
serves therefore as the zero reference for the group delay.
The photon with the shorter wavelength starts the TAC.
With increasing wavelength separation it arrives later at
the detector than the higher wavelength photon causing a
shift of the coincidence peak to shorter time intervals. The
actual settings of the bandpass filter the signal wave-
. lengths , the corresponding idler wavelengths, the time
delays and widths of the coincidence peaks corresponding
to the fitting parameters are summarized in Table 1.
To derive the group delay t from our data we assume
g
that t is given by the three term Sellmeier function
g
t l sal
2
qbl
y2
qc. 3 . .
g
Then the group delay difference for the correlated photons
. . is according to Eqs. 2 and 3
2
1
2
t l yt l sa l y . .
g s g i s
/
1rl y1rl
p s
y2
1
y2
qb l y .
s
/
1rl y1rl
p s
4 .
. Deriving Eq. 3 one obtains the dispersion:
dt l .
g
y3
Ds s2aly2bl , 5 .
dl
4
' with the zero dispersion wavelength l s bra and the
0
dispersion slope
d
2
t l .
g 0
y4
S s s2aq6bl s8a. 6 .
0 0
2
dl
. Fitting the function 4 to the experimental data we
obtain the parameters as0.0816 psrnm
2
and bs2.498
Table 1
Experimental data corresponding to the fitting curves shown in Fig. 2
. . . . . Signal wavelength nm Idler wavelength nm Delay ps Width FWHM ps
1320 1321.6 0 227
1314.5 1327.2 13.99 229
1307.2 1334.7 33.08 230
1297.6 1344.8 51.35 230
1286.2 1357.3 68.42 227
1271.9 1373.6 61.87 219
( ) J. Brendel et al.rOptics Communications 151 1998 3539 38
Fig. 3. Differential group delay and dispersion of the standard
telecom fiber.
=10
11
ps nm
2
. They imply a zero dispersion wavelength
of l s1322.7 nm and a dispersion slope of S s0.099
0 0

2
. psr nm km . With these parameters our measured data
yield the group delay and dispersion curves as shown in
Fig. 3. They are in good agreement with the results
obtained with a commercial device based on the baseband
phase comparison method using four laser diodes of differ-
. ent wavelengths Anritsu ME9301A .
Next, we repeated the measurement described above
with an optical fiber whose zero dispersion wavelength is
shifted to the 1550 nm band. This fiber allows us to
examine the case where the central wavelength l of the
photon pair light is far from the zero dispersion wave-
length l of the fiber under test. The length of the fiber is
0
2.237 km. The results are summarized in Table 2. Due to
the huge dispersion of this fiber the zero dispersion refer-
. ence has been measured with a short 0.5 m piece of
standard optical fiber as indicated in the first line of Table
2. The width of the time difference distributions are now
significantly larger compared to the standard fiber. This
effect is mainly due to the two-photon dispersion cancella-
tion properties which will be examined in more detail in
the following section. Fitting the data with the same
.. Sellmeier function as above Eq. 3 we find: as0.0159
psrnm
2
and bs9.172=10
10
ps nm
2
. Since our tech-
Fig. 4. Differential group delay and dispersion of the dispersion
shifted fiber.
nique does not provide data at wavelengths close to 1550
nm, a reasonable fit can only by obtained when fixing the
zero dispersion wavelength to 1550 nm before fitting the
data. The resulting group delay and dispersion are plotted
in Fig. 4.
The main source of error in our method is the accuracy
of the wavelength set by tilting the filter angle. From this
we estimate an error of "0.5 nm when determining the
zero dispersion wavelength which exceeds all statistical
errors. This error can be almost completely eliminated
using commercial devices with up to 0.05 nm wavelength
resolution. Otherwise, the achieved accuracy is only lim-
ited by the electronics and the measuring time. The accu-
racy can also be further improved by increasing the num-
ber of wavelengths used in the measurement.
4. Two-photon dispersion cancellation
To examine the two-photon dispersion cancellation
properties mentioned in the introduction, we first refer to
the width of the detected coincidence peaks as indicated in
Table 1. The width of the time interval peak found for
signal and idler photons of equal wavelengths indicates the
'
time jitter of the single detectors Dt s227 2 f160 ps.
D
Table 2
Measured data for 2.237 km of dispersion shifted fiber
. . . . . Signal wavelength nm Idler wavelength nm Delay ps Width FWHM ps
. . 1320 1321.6 0 ref. fiber 264 ref. fiber
1314.5 1327.2 416.24 756
1307.2 1334.7 965.35 735
1297.6 1344.8 1740.25 718
1286.2 1357.3 2808.04 952
1271.9 1373.6 3756.69 869
( ) J. Brendel et al.rOptics Communications 151 1998 3539 39
The same result is obtained when replacing the test fiber
by a short fiber. Notice, that this width remains unchanged
when l and l are symmetrically apart from the zero
s i
dispersion wavelength l , like e.g. l s1286.2 nm and
0 s
l s1357.3 nm.
i
This result should be compared to equivalent classical
light pulses with negligible temporal width and Dls10
nm bandwidth. The width of the classical pulses leaving
w x the test fiber can be approximated by DtsDLDl ps ,
. where D is the dispersion of the fiber in psr nm km and
L the fiber length in km. With our measured dispersion
. . data D 1286.2 nm s 3.8 psr nm km and D 1357.3
s i
. .. . nm s y3.34 psr nm km we obtain Dt 1286.2 nm
s
. s 247 ps and Dt 1257.3 nm s y217 ps. From this we
i
deduce an expected total width of the classical two detec-
tor correlation signal Dt by
classic
2 2 2
( Dt s Dt qDt q2Dt s400 ps. 7 .
classic s i D
In contrast, due to the energy correlation of the photon
pairs within the two 10 nm bandwidth intervals, the result-
ing two-photon dispersion has to be approximated by the
w x sum of the dispersion at the two pass bands 7 . Under this
assumption the total width of the two photon time differ-
ence distribution is
2
2
( Dt s Dt qDt q2Dt s228 ps. 8 . .
two- photon s i D
As a consequence of the two-photon dispersion cancella-
tion the pulsewidth remains almost unchanged, in accor-
dance with our experimental data.
The dispersion can only be cancelled as long as the
dispersion function has opposite signs at the two pass-
bands, i.e. as long as the central wavelength of the photon
. pairs is close to the zero dispersion wavelength lfl .
0
To examine the case where the dispersion has equal signs
we refer to the results obtained with the dispersion shifted
fiber as summarized in Table 2. To compare the pulse
broadening of the two photon light with classical light we
. take as an example the values D 1297.6 nm s y19.1
s
. . . psr nm km , and D 1344.8 nm s y14.6 psr nm km
i
. from our measured dispersion data. With Eq. 7 we
calculate the expected total time jitter for classical light
. Dt s583.5 ps. In contrast, Eq. 8 which takes into
classic
account the two-photon energy correlations yields
Dt s787 ps which is close to our experimental
two- photon
value of 718 ps. Therefore our results show that the
two-photon quantum correlations can give rise to a disper-
sion enhancement effect compared to a classical light field
of equal spectral distribution.
5. Conclusions
We have introduced a novel dispersion measurement
technique which takes advantage of the temporal and
spectral correlations of photon pairs in a non-classical light
field. It demonstrates that two-photon metrological applica-
tions based on compact two-photon sources are straightfor-
ward. Our system is capable of measuring the group delay
in the 1310 nm region where the results agree well with a
commercial device. The instrumentation is rather simple
and cheap, but photon counting at 1310 nm is required.
Today Ge-detectors have to be kept at liquid nitrogen
temperature which is still a drawback for practical applica-
tions. Current experiments show that photon counting at
about y508C may be possible using InGaAs detectors
w x 19 . Also, the measuring time to achieve high resolutions
comparable to commercial devices is still long about 50
. min with our present laboratory setup . An advanced de-
vice is currently being developed which will reduce the
measuring time down to a few minutes and will also
extend the wavelength region towards 1550 nm.
Acknowledgements
This work is supported by the Swiss Priority Program
. Optics PPO Quantum Optics and the Training and
. Mobility of Researches TMR Programme The Physics
of Quantum Information, sponsored by the Swiss OFES.
References
w x . 1 L.G. Cohen, J. Lightwave Techol. LT-3 1985 958.
w x 2 L.G. Cohen, J. Stone, Electron. Lett. vol. 18, p. 564.
w x 3 L. Thevenaz, J. Pellaux, J.V. der Weid, J. Lightwave Techol.
. 6 1988 1.
w x . 4 C.L. Tang, in: H. Rabin, C.L. Tang Eds. , Quantum Elec-
tronics, Academic, New York, 1975.
w x . 5 D.C. Burnham, D.L. Weinberg, Phys. Rev. Lett. 25 1970
84.
w x . 6 C.K. Hong, Z.Y. Ou, L. Mandel, Phys. Rev. Lett. 59 1987
2044.
w x . 7 J.D. Franson, Phys. Rev. A 45 1992 3126.
w x 8 A.M. Steinberg, P.G. Kwiat, R.Y. Chiao, Phys. Rev. A 45
. 1992 6659.
w x 9 A.M. Steinberg, P.G. Kwiat, R.Y. Chiao, Phys. Rev. Lett. 68
. 1992 2421.
w x 10 T.S. Larchuk, M.C. Teich, B.E.A. Saleh, Phys. Rev. A 52
. 1995 4145.
w x . 11 D.N. Klyshko, Sov. J. Quantum Electron. 10 1980 1112.
w x 12 J.G. Rarity, K.D. Ridley, P.R. Tapster, Appl. Optics 26
. 1987 4616.
w x . 13 A.V. Sergienko, A.N. Penin, Appl. Optics 30 1991 3582.
w x . 14 P.G. Kwiat et al., Appl. Optics 33 1994 1844.
w x . 15 A.K. Ekert, Phys. Rev. Lett. 67 1991 661.
w x 16 A.K. Ekert, J.G. Rarity, P.R. Tapster, G.M. Palma, Phys.
. Rev. Lett. 69 1992 1293.
w x 17 E. Mohler, J. Brendel, R. Lange, W. Martienssen, Europhys.
. Lett. 8 1989 511.
w x . 18 P.C.M. Owens et al., Appl. Optics 33 1994 6902.
w x 19 G. Ribordy, J.D. Gautier, H. Zbinden, N. Gisin, Appl. Op-
tics, submitted, 1997.

S-ar putea să vă placă și