Home
Search
Collections
Journals
About
Contact us
My IOPscience
Quantum cryptography over 23 km in installed under-lake telecom fibre
This article has been downloaded from IOPscience. Please scroll down to see the full text article.
1996 Europhys. Lett. 33 335
(http://iopscience.iop.org/0295-5075/33/5/335)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 129.194.8.73
The article was downloaded on 13/07/2011 at 08:19
Please note that terms and conditions apply.
EUROPHYSICS LETTERS
10 February 1996
Europhys. Lett., 33 (5), pp. 335-339 (1996)
Quantum cryptography over 23 km in installed under-lake
telecom fibre
A. MULLER,H. ZBINDEN and N. GISIN
Group of Applied Physics, University of Geneva - l211 Geneva 4, Switzerland
(received 13 September 1995; accepted in final form 20 December 1995)
PACS.03.65Bz
Foundations,theory of measurement,miscellaneoustheories(including
Aharonov-Bohm effect, Bell inequalities, Berry’sphase).
PACS. 42.50-p - Quantum optics.
PACS. 42.81Gs - Birefringence, polarization.
-
Abstract. - A quantum cryptographic channelusing a 23 km long installed standard telecom
optical cable is reported. The key is encoded in the polarization of very weak laser pulses of
average photon number 0.12. The measured error rate islower than 3.4%. The ability of the
system to establish a quantum key at standard telecom wavelength (1300 nm) using installed
cables is thus assessed.
Quantum mechanics is often presented as a limitation to achieve certain tasks, for instance
the simultaneousdetermination of position and momentum of atomicparticles. However,
despite of or thanks to the limitations due to
Heisenberg’s uncertainty principle, quantum
mechanics has made possible many useful applications, like the laser and modern electronics.
More recently, the idea of using Heisenberg’s uncertainty relations for secure communications
has been investigated, both theoretically [:l.]-[3]and experimentally [4]-[8].The idea is that
whenever a third party taps into a quantum communication channel, his action necessarily
perturbs the signal, as a quantum measurement necessarily perturbs the system, provided the
bits of the key are represented by incompatible quantum states. Thisunavoidable perturbation
can then be detected by controlling the correlation between the transmitted and the received
signals. Last but not least,
the quantum channel is not used to transmit information, but
only t o establish a coding key. Hence, in case the key is found corrupted, no information
is lost. After the secrecy of the key has been checked, thanks to the very basic principles
of quantum mechanics, it can safely be used to encode the relevant message using standard
cryptography based on the well-established information theory of Shannon [g]. This new field
of research is called quantum cryptography. Its future depends on the
possibility to implement
such systems on standard telecom fibres over a reasonable range of some tens of kilometres.
It depends also on the evolution of current public key crypto-systems whose safety is still not
proven mathematically [lo].
In practice, the quantum communication channel consists of laser pulses attenuated in a
way such that the average photon number is well below unity. Therefore the probability of
a pulse with 2 or more photons is negligible. The key can be encoded in the polarization
@ Les Editions de Physique
336
EUROPHYSICS LETTERS
of these pseudo single-photon pulses [5], [7] or in their phase [6], [ll],using interferometers
at both ends of the channel . In this letter, the possibility of using the polarizationcoding
with 1300 nm low-intensity light pulses is investigated. This choice is motivated by conceptual
simplicity and by the fact that control of the polarization is also needed for high-visibility
interferometry. The standard installed telecom fibre used, joins Geneva to Nyon, most of its
23 km length is buried under the lake of Geneva in Switzerland.
In the polarization scheme, the transmitter, traditionally called Alice, consists of a source
of attenuated pulses with a random polarization oriented along
one of the four directions,
0", 45",go", 135". When creating a key, the receiver, called Bob, chooses randomly between
two orientations of his polarization analyser (0" ,45"). Both armsof the analyser are connected
to a photon counter. In a first stage, the outgoing electrical pulses are combined, timed and
sent to a computer. In a second stage, Alice and Bob publicly compare the chosen bases of
emission and detection, but without
disclosing the polarization states transmittedor measured.
Thus they can eliminate the incompatible measurements, keeping only the compatible ones
which should be ideally 100% correlated. This correlation is then checked on a sample basis.
An eavesdropper that would measure the polarization of all the photons would introduce a
25% reduction in the correlation which could easily be detected. More sophisticated analyses
of eavesdropping strategies can be found in references [12], [13].
A fibre optic implementation of the polarization transport faces two main difficulties. The
first oneis that a long enough fibre depolarizes the guidedlight. This is due to thepresence in a
real fibre of two group velocities which depend on the stateof polarization. This phenomenon,
called Polarization Mode Dispersion (PMD), tends to depolarize the light. However, if the
polarization mode delay is smaller than the coherence time of the laser source, the light at the
output of the fibre is still polarized. [14]. In practice, semiconductor lasers have a coherence
time of the order of 10 PS. Some of the installed fibres have a polarization mode dispersion of
the same order of magnitude and can thus not be used for quantum cryptographic purposes.
But the present day trend is toward a specification of 0.5 PS/&
(the strange unit &
reflects the fact that polarizationmodedispersion
is a diffusive process, like the random
walk). Hence, modernoptical fibres donotsignificantlydepolarizelight
over a length of
several tens of kilometres. The second difficulty is that evenfor coherentenoughsources
the polarization at the output of the fibre is time dependent. This is caused both by slight
fluctuations of the birefringence (phase velocity difference) along the fibre and by variations
of the geometry of the optical path, and hence the parallel transport of the polarization state.
The first effect is due toslight changes of stress on thefibres in real opticalcables. The second
one is due to vibrations that affect the Berry topological phase [15]. However, experiences
show that the rateof these changes is extremely slow compared to the usual timesinvolved in
optical communications: tens of minutes or hours, depending on the thermal and mechanical
stability of the environment[16]. Consequently, polarization tracking can compensate
for these
polarization fluctuations.
Depolarization and polarization fluctuations are the main causes
of optical noise. The noise
of the detector should be added to this
optical noise in order to get the overall bit error
rate (BER), defined as the ratio of the number of wrong detections to the total number of
detections.
BER =
wrong detections
total number of detections
= BERoptical
+ BERdetector .
(1)
The BERdetector can be evaluatedas follows:
BERdetector =
dark count .61dark count .SI- photonsinjected
. loss . 77 '
+
(2)
.
A. MULLER
et al.:
QUANTUM CRYPTOGRAPHY OVER
23 km
ETC.
337
emitter
I
I
Fig. 1. - Experimental set-up. The 1300 nm laser was pulsed by an electrical pulser (Avtech). Half
of the outgoinglightwas then splitted to have a timingreference. The other halfwascircularly
polarized by a X/4 wave plate and polarized by a rotating polarizer (OZ Optics). The loops indicate
a polarizationcontroller. The polarizationcontrolwasperformedby
three fibreopticloops (X/4,
X/2, X/4) and a set of four pigtailedliquid-crystalcells (A01 PCLC-2). The ”Att.” was a 65 dB
attenuator (Wandel & Goltermann). The fibre link in the lab was simulated by a coil of 26 km and
in the field was an underwater cable of 22.7 km. The PBS was a pigtailed Polarization Beam Splitter
(OZ Optics). The Detectors (Dl,D2) were singlephoton countersbasedongermaniumavalanche
photo diodes (NEC-NDL 5102P) cooled with liquid nitrogen, working in Geiger mode
and passively
quenched. The signal coming from the detectors was amplified, discriminated and combined by an
”or” gate. The TPHC, a Time-to-Pulse-Height Converter (EGG-Ortec), recorded the time of arrival
of the photons using the signal received by a standard PIN detector as time reference. By using a
delay line, it was possible to measure the polarization state of the incoming photon. The MCBA is a
Multi Channel Buffer Analyser (EGG-Ortec) which analysesthe amplitudes produced by the TPHC.
where ST is the detection time window, dark count is the number of counts per unit time not
due to incoming photons, photonSinjectedis the number of photons per pulseat the transmitter
output, loss is the sumof the losses in the transmission cable and in the receiver’s polarization
splitter, and 71 is the overall quantum efficiency of the detector.
In the experiment reported here, a 1300 nm pulsed laser with 1 ns pulse width and 1.1 MHz
pulse rate has been used, see fig. 1. This laser was also used in continuous mode (CW) to align
the system. A manual rotating polarizer placed after a X/4 wave plate allowed for polarization
modulation. A polarization controller compensated the polarization modification due to the
fibre link. This is necessary in order to align the emission and measurement axis. The pulses
were attenuated to obtain 0.12 photons per pulse sent into the fibre link.
At the receiver
a pigtailed polarization splitter
was used as the analyser. The photons were detected by a
liquid-nitrogen-cooled Ge avalanche photodiode
working in Geiger mode [l71 and passively
quenched. The characteristics of this photon counter are as follows: the diode was biased with
24.1V (breakdown N 24 V), the dark-count rate was 700 per second and quantum efficiency
(v) 0.2% ( l ) . Assuming a 10 dB loss in the channel, this corresponds to BERdetectorM 3% (see
relation (2)). Counting electronics and a personal computer collected the data.
(l) The quantum efficiency could have been improved
up to10% by a larger bias,but at theexpense
of the dark-count rate. The above parameters were found optimum for low BER.
338
EUROPHYSICS LETTERS
600
a)
500 -
g400-
/jll
2
F
300
S
g200
~
100-
L
N
00
f - 0 4 - k - 4
6
I i1
8
l0
10
0
2
4
6
8
1
0
0
2
4
6
8
1
0
Fig. 2. - These results present the time histogram of the detected photons in three different configurations. The horizontal scale is the channels of the MCBA converted into nanoseconds and grouped
into bins of 0.13 nanoseconds. The vertical scale is the corresponding number of detected photons.
In a ) the photons are vertically polarized, in b) they are horizontally polarized and in c ) they are
diagonally polarized. One can observe that the number of wrong counts is very small in a ) and b), in
contrast to c ) , where the incoming photons are found equally distributed.
The device was first tested in the lab with a 26 km long fibre coil. In the field test the
quantum channel was a fibre installed in a cable joining Genevato Nyon. It contains 40 fibres,
many of them are presently in use at a bit rate of 2.5 Gbit/s. The transmitter was at Nyon
and the receiver at Geneva. Most of the cable is underwater, its attenuation is 8.2 dB and its
length 22.7 km. The cable arrives at both ends in the basement of a building, nothing more
than the usual packaging of the cable was used to stabilize it.
A polarization separation of 23 dB has been obtained using an intense CW light source,
both in the lab with the
26 km long fibre and in the field with the 22.7 km cable. This
corresponds to BEhptical = 0.5%.
In fig. 2, the results obtained using low-intensity pulses (containing 0.12 photons) polarized
vertically, horizontally and at 45 degrees with respect to the detection polarization splitter
axis are presented. The first two cases simulate a compatible preparation and measurement
basis, whereas the pulses emitted at 45" simulate the incompatible measurements. Table I
summarizes the measured bit error rates. These measurements show a very low BER (3.4%).
This is in agreement with the BER expected according to relation (1). More important, this
BER is low enough that standard algorithmof errors correction and privacy amplification can
be applied and guarantee the privacy of the key. The detector count rate, which is related to
the key creation rate, was of the order of 30 counts per second. This is low, but it should be
noted that using active quenching and a better electronic it could be increased of at least 2
orders of magnitude.
TABLE
I.
Fibre type
error
Bit
0"
in 26 km
under
22.7 km
the lab
the lake
rate
90"
0.041
0.043
f0.003
f0.003
0.0330.034
f0.003
f0.003
A. MULLER
et al.:
QUANTUM CRYPTOGRAPHY OVER
23 km ETC.
339
The stability of the polarization alignment was found to be excellent for most of the time.
Indeed, all the results presented in fig. 2 and table I were done over about one hour without
realigning the system. Thanks to this stability,
we could test the polarization modulation even
with our manual (hence slow) rotating polarizer. But it should be reported that at certain
times the signal showed a high polarization instability, typically 2~ in 3 seconds. This has
been possibly causedby a road construction in the neighbourhood.A fast enough polarization
controller could probably overcome even such large fluctuations.
In conclusion, the feasibility of a quantum cryptographic
system using polarization coding in
the second telecomwindow around 1300 nm on installed optical cables has been demonstrated.
The measured bit error rate (3.4%, before any error correction)is low enough to guarantee the
privacy of the quantum communication channel. Consequently, the issue is no longer whether
the quantum technology works, but whether it can be made robust and practical enough for
use over an optical network.
***
Many thanks are due to the Swiss Telecom PTT and to the people who have made the
access to the cable possible and even very pleasant.
REFERENCES
[l] WIESNER
S., Sigact News, 15 (1983)78.
[2] BENNETTC.H., BESSETTEF. andBRASSARD G., Proceedings of the IEEEInternational
Conference on Computer Systems and Signal Processing, Bangalore (1984), p. 175.
[3] EKERTA., Nature, 358 (1992) 14.
[4] BENNETTC. H., BRASSARD
G. and EKERTA., Scientific American, October issue (1992) 26.
[5] MULLERA . , BREGUET
J. and GISIN N., Europhys. Lett., 23 (1993) 383.
[6] TOWNSEND
P. D., RARITYJ. G. and TAPSTER
P. R., Electron. Lett., 29 (1993) 1291.
[7] FRANSON
J. D. and JACOBS
B. C., Electron. Lett., 31 (1995)232.
[8] BENNETTC. H., BESSETTEF., BRASSARD
G., SAVAILL. and SMOLINJ., J. Cryptology, 5
(1992) 3.
[g] SHANNON C. E.,
Bell Syst. Tech. J., 28 (1949) P 656.
[lo] GROLLMAN
J. and SELMAN
A. L., SIAM J. Comput., 17 (1988) 309.
[l11 TOWNSEND
P. D., Electron. Lett., 30 (1993) 809.
[l21 EKERTA. K., HUTTNER
B., PALMA
M. G. and PERES
A., Phys. Rev. A , 50 (1994) 1047.
[l31 HUTTNER
B. and EKERTA. K., J. Mod. Opt., 41 (1995)2455.
[l41 GISINN. and PELLAUX
J. P., Opt. Commun., 89 (1992) 316.
[l51 TOMITA
A . and CHIAOR. Y . , Phys. Rev. Lett., 57 (1986) 937.
[l61 DE ANGELISC., GALTAROSSA
A., GIANELLO
G., MATERAF. and SCHIANO
M., J. Lightwave
Technol., 10 (1992)552.
[l71 OWENSP. C., RARITYJ. G., TAPSTER
P. R., KNIGHTD. and TOWNSEND
P. D., Appl. Opt.,
33 (1994) 6895.