You are on page 1of 4

ICTON 2012 We.A4.

Effects of Aperture Averaging and Beam Width


on Gaussian Free Space Optical Links
in the Presence of Atmospheric Turbulence and Pointing Error
It Ee Lee, Zabih Ghassemlooy, Senior Member, IEEE, Wai Pang Ng, Senior Member, IEEE
Northumbria University, Newcastle Upon Tyne, United Kingdom
e-mail: {it.ee.lee, z.ghassemlooy, wai-pang.ng}@northumbria.ac.uk
ABSTRACT
This paper investigates the performance of free-space optical (FSO) communication systems under the influence
of atmospheric loss, turbulence and pointing error. Considering the limiting Rytov-based scintillation theory
associated with the simplified field models, the Gaussian-beam wave model is introduced to characterize the
propagation properties of the optical signal through random turbulent medium. The bit-error rate (BER) and
average channel capacity are evaluated, taking into account the effects of aperture averaging and beam width. In
general, larger receiving apertures enable lower BER attainment with smaller signal-to-noise ratio (SNR), while
beam width optimization is a feasible approach in promoting capacity enhancement.
Keywords: Free-space optical (FSO) communications, Gaussian-beam wave, aperture averaging, turbulence.

1. INTRODUCTION
Free-space optical (FSO) communications is an emerging broadband wireless access candidate, which has
attracted significant attention for a variety of applications, such as horizontal-path (terrestrial) FSO systems in
complementing the existing wireless radio frequency (RF) solutions [1, 2]. The widespread deployment of such
low-cost high-bandwidth access technique is hampered by the atmospheric channel which is highly variable,
unpredictable and vulnerable to different weather conditions [2-4], and the presence of pointing error [5, 6].
The adverse effects of atmospheric loss, turbulence-induced channel fading and pointing errors on the
performance and design of horizontal FSO links have been extensively investigated. Majority of these theoretical
treatments [4-6] are based upon the classical unbounded plane- or spherical-wave approximations (the latter
often taken as a point source) [1], which are insufficient to characterize the optical wave propagation properties
through random turbulent medium [2, 3]. In addition, the influence of beam wander attributed to the presence of
large-scale turbulent eddies, which contributes to a widening of the long-term beam profile and pointing error, is
not considered in the limiting Rytov-based scintillation theory [1, 3]. Furthermore, the aperture averaging effect
of spatially partially coherent beam in mitigating the optical intensity fluctuations, has not been thoroughly
examined due to the inherent mathematical complexity of the Gaussian-beam wave model [1, 2].
In this paper, we jointly consider a spatially partially coherent laser beam and link design criteria, to examine
the FSO link impairments imposed by atmospheric attenuation, turbulence and pointing error. Taking into
account the effects of aperture averaging, beam width, link distance, and weather conditions, the BER and
average capacity are evaluated. The Gaussian-beam wave model [1, 2] characterizes the propagation properties
of the optical signal through random turbulent medium, which include the scintillation and beam wander arising
from the turbulent eddies. Our studies reveal that low BER can be achieved with smaller average electrical SNR,
by using a larger receiver aperture due to the averaging of irradiance fluctuations over the aperture, particularly
in strong turbulence scenarios. For long-distance (> 5 km) FSO links, beam width optimization is a feasible
approach in promoting capacity enhancement, since larger beam width improves the channel capacity only in the
strong turbulence regime, while imposing performance degradation for weak-to-moderate turbulence.

2. SYSTEM DESCRIPTION
2.1 System Model
We consider a single-input single-output (SISO) horizontal FSO link, employing intensity modulation with
direct detection (IM/DD).The received signal can be modelled as ‫ݕ‬௢ ൌ ݄ߛ‫ݔ‬௢ ൅ ݊௢ , whereߛ is the detector
responsivity (in A/W), ‫ݔ‬௢ ‫ א‬ሼͲǡʹܲிௌை ሽ is the optical signal intensity, ܲிௌை is the average transmitted optical
power, and ݊௢ is zero-mean additive white Gaussian noise (AWGN)with variance ߪ௡ଶ [3, 5].The channel state can
be described as݄ ൌ ݄௟ ݄௦ ݄௣ , where ݄௟ , ݄௦ and ݄௣ denote the attenuation, scintillation, and pointing errors,

respectively [5]. The received electrical SNR is defined asܴܵܰሺ݄ሻ ൌ ʹܲிௌை ߛ ଶ ݄ଶ Τߪ௡ଶ [5].
2.2 Beam Model
We consider a lowest-order transverse electromagnetic (TEM00) Gaussian-beam wave [1] with effective beam
radius (spot size) ‫ݓ‬଴ . After propagating through the channel with link distance‫ܮ‬, the receiving beam size is
ଵΤଶ
‫ݓ‬௅ ൌ ‫ݓ‬଴ ൫Ĭଶ௡ ൅ ߦȁଶ௡ ൯ , where Ĭ௡ ൌ ሺ‫ܨ‬଴ െ ‫ܮ‬ሻΤ‫ܨ‬଴ and ȁ௡ ൌ ߣ‫ܮ‬Τߨ‫ݓ‬଴ଶ are the normalized components,

978-1-4673-2229-‹,((( 1
ICTON 2012 We.A4.2

ߞ ൌ ߞ௦ ൅ ʹ‫ݓ‬଴ଶ Ȁߩ଴ଶ is the global coherence parameter; ‫ܨ‬଴ is the phase front radius of curvature at the transmitter,
ߞ௦ is the coherence of the beam, ߩ଴ ൌ ሾͲǤͷͷ‫ܥ‬௡ଶ ሺʹߨΤߣሻଶ ‫ܮ‬ሿିଷΤହ is the coherence length of a spherical wave.
The point-receiver scintillation index is given by[1, 3]

Τ ߩ ͷ ͳ ൅ ʹȣ௅ ͳͳ ହΤ଺
ߪூଶ ሺͲሻ ؆ ͶǤͶʹߪோଶ Ȧହ௅ ଺ ଶ ൅ ͵Ǥͺ͸ߪோଶ ൜ͲǤͶͲሾሺͳ ൅ ʹȣ௅ ሻଶ ൅ ͶȦଶ௅ ሿହΤଵଶ FRV ൤ WDQିଵ ൬ ൰൨ െ Ȧ ൠ (1)
‫ݓ‬௅ ͸ ʹȦ௅ ͳ͸ ௅
where ߪோଶ ൌ ͳǤʹ͵‫ܥ‬௡ଶ ሺʹߨΤߣሻ଻Τ଺ ‫ܮ‬ଵଵΤ଺ is the Rytov variance for a plane wave, ȣ௅ ൌ ͳ ൅ ሺ‫ܮ‬Τ‫ܨ‬௅ ሻ and
Ȧ௅ ൌ ߣ‫ܮ‬Τߨ‫ݓ‬௅ଶ are the receiver beam parameters, ߩ is the separation distance between two points, and ‫ܥ‬௡ଶ is the
refractive-index structure parameter. The phase front radius of curvature at the receiver is defined by
‫ܨ‬௅ ൌ ‫ܮ‬൫Ĭଶ௡ ൅ ߞȦଶ௡ ൯ൗ൫߶Ȧ௡ െ ߞȦଶ௡ െ Ĭଶ௡ ൯, with ߶ ‫ ؠ‬ሺȣ௡ ΤȦ௡ ሻ െ ሺȦ௡ ‫ݓ‬଴ଶ Ȁߩ଴ଶ ሻ.
The aperture-averaging factor is given by[1]

ߪூଶ ሺ‫ܦ‬ሻ ͳ͸ െ‫ܦ‬ଶ ‫ ݔ‬ଶ ߩ଴ଶ ߩ଴ଶ ߶ ଶ
‫ ீܣ‬ൌ ଶ ൌ න ‫ ݔ‬ቂFRVିଵ ሺ‫ݔ‬ሻ െ ‫ݔ‬ඥͳ െ ‫ ݔ‬ଶ ቃ H[S ቈ ቆʹ ൅ െ ቇ቉ ݀‫ݔ‬ (2)
ߪூ ሺͲሻ ߨ ߩ଴ଶ ‫ݓ‬଴ଶ Ȧଶ௡ ‫ݓ‬௅ଶ

where ‫ ݔ‬ൌ ߩΤ‫ ܦ‬, with ‫ ܦ‬being the receiver aperture diameter. The aperture-averaged scintillation index ߪூଶ ሺ‫ܦ‬ሻ
can be obtained by multiplying the ߪூଶ ሺͲሻ in (1) with the ‫ ீܣ‬in (2)[1].
2.3 Combined Channel Fading Model
The channel state distributions of ݄ ൌ ݄௟ ݄௦ ݄௣ for both weak and strong turbulenceregimes are given by (3) and
(4), respectively, as defined and elaborated in [5].
ஶ ଵ ଶ
ߦଶ ͳ ቂ݈݊ሺ݄௔ ሻ ൅ ߪூଶ ሺ‫ܦ‬ሻቃ
క మ ିଵ ଶ
݂௛ ሺ݄ሻ ൌ ݄ න H[S ቐെ ቑ ݄݀௔ (3)
ሺ‫ܣ‬଴ ݄௟ ሻకమ క మ ାଵ
݄௔ ߪூ ሺ‫ܦ‬ሻξʹߨ ʹߪூଶ ሺ‫ܦ‬ሻ
௛ Τ஺ బ ௛ ೗

ʹߦ ଶ ሺߙߚሻሺఈାఉሻΤଶ కమ ିଵ ሺఈାఉሻΤଶିଵିక మ
݂௛ ሺ݄ሻ ൌ ݄ න ݄௔ ߈ఈିఉ ൫ʹඥߙߚ݄௔ ൯ ݄݀௔ (4)
ሺ‫ܣ‬଴ ݄௟ ሻకమ īሺߙሻīሺߚሻ
௛ Τ஺ బ ௛ ೗

3. PERFORMANCE ANALYSIS
3.1 Bit-Error Rate
$VVXPLQJ ³´ DQG ³´ ELWV DUH WUDQVPLWWHG ZLWK HTXDO SUREDELOLW\ WKH DYHUDJH %(5 FDQ EH GHWHUPLQHG E\
averaging their conditional bit error probabilities over the probability density function (pdf) of ݄, given by [4]:

ܲ௘ ൌ න ݂௛ ሺ݄ሻܳ ቀඥܴܵܰሺ݄ሻቁ ݄݀ (5)




where ܳሺ‫ݕ‬ሻ ൌ ൫ͳΤξʹߨ൯ ‫׬‬௬ H[Sሺെ‫ ݐ‬ଶ Τʹሻ݀‫ ݐ‬is the Gaussian-Q function.

3.2 Average Channel Capacity


The average channel capacity ‫ۄܥۃ‬is defined as the maximum mutual information between the input to the
channel ܺ, and output from the channel ܻ, over all input distributions [3]. The mutual information becomes


݂ሺ‫ݕ‬ȁ‫ݔ‬ሻ
‫ܫ‬ሺܻǢ ܺሻ ൌ ෍ ܲ௑ ሺ‫ݔ‬ሻ න ݂ሺ‫ݕ‬ȁ‫ݔ‬ሻORJଶ ቈ ቉ ݄݀݀‫ݕ‬ (6)
ିஶ σ௠ୀ଴ǡଵ ݂ሺ‫ݕ‬ȁ‫ ݔ‬ൌ ݉ሻܲ௑ ሺ݉ሻ
௫ୀ଴
where ܲ௑ ሺ‫ݔ‬ሻ is tKHSUREDELOLW\RIWKHELWEHLQJ³´RU³´, with ܲ௑ ሺͲሻ ൌ ܲ௑ ሺͳሻ ൌ 0.5, and ݂ሺ‫ݕ‬ȁ‫ݔ‬ሻ is given by
ͳ ‫ݕ‬ଶ
‫ۓ‬ H[S ቈെ ଶ ቉ ‫ ݔ‬ൌ Ͳ
ۖ ඥʹߨߪ௡ଶ ʹߪ௡
݂ሺ‫ݕ‬ȁ‫ݔ‬ሻ ൌ ஶ (7)
‫ͳ ۔‬ ሺ‫ ݕ‬െ ʹܲிௌை ߛ݄ሻଶ
ۖ න H[S ቈെ ቉ ݂௛ ሺ݄ሻ݄݀ ‫ ݔ‬ൌ ͳ
‫ە‬ඥʹߨߪ௡ଶ ଴ ʹߪ௡ଶ

4. NUMERICAL RESULTS AND DISCUSSION


Figure 1(a) illustrates the average BER in terms of the ܲ)62 under varying level of fog conditions at ‫ ܮ‬ൌ NP,
whereby a range of ܸ and ‫ ܦ‬is examined. It is noted that lower values of BER can be achieved with a smaller
ܲ)62 for a larger‫ܦ‬, in which ܲ)62 reduction of൐ G%is possible for ‫ ܦ‬ൌ PPand a BER of 10-9,
compared with a point aperture (‫ ܦ‬ൌ PP) [1] under the light fog condition. The system requires higher
ܲ)62 under moderate fog condition to compensate for the increase in the atmospheric attenuation
(ߪ ൌ G%NP), in which greater values of ܲ)62 of ~13-orderis observed for a BER of10-9, albeit the

2
ICTON 2012 We.A4.2

introduction of ‫ ܦ‬ൌ PP. Under the extreme low-visibility case with ܸ ൌ P and ߪ ൌ G%NP, the
FSO link experiences system annihilation, which is non-recoverable albeit optimization, since the absorption and
scattering effects attenuate and destroy the optical beams. Next, Figure 1(b) depicts the variation of the average
BER against the ܲ)62 for the strong turbulence regime, in which the aperture averaging effect is evaluated at
‫ ܮ‬ൌ NP, while the impact of scintillations observed for ‫ ܮ‬ൌ ሼሽNP. In general, the FSO link has
a lower ܲ)62 requirement relative to the low-visibility cases, mainly due to the significantly lower attenuation
ߪ ൌ G%NP. The introduction of a larger ‫ ܦ‬enhances the link performance, in which ܲ)62 reduction
൐ G%is observed for a BER of 10-9 with ‫ ܦ‬ൌ PP, as compared with a point aperture.
At ‫ ܮ‬ൌ ሼሽNP, the increase in scintillation level (signified by larger ߪோଶ ) results in skewing of the BER
curve, exhibiting much smaller step size in the BER reduction with respect to the same increment in ܲ)62 .
Figure 2(a) presents the average BER as a function of the SNR, under the light fog condition at ‫ ܮ‬ൌ NP,
taking into account the effect of ‫ ܦ‬and comparing it with the AWGN case. A two-fold reduction in BER is
observed at 615 ൌ G% with ‫ ܦ‬ൌ PP relative to a point aperture, thus bringing the BER closer to the
non-turbulent case, albeit deviating from the reference with a 2-dB gap in SNR for achieving a BER of 10-9.
Performance enhancement due to aperture averaging is prevalent in the strong turbulence regime (Fig. 2(b)),
whereby the scintillation reduction potentially decreases the BER by more than four-order of magnitude
for615 ൒ G%with ‫ ܦ‬ൌ PP. These observations can be justified by the shifting of the relative
frequency content of the irradiance power spectrum toward lower frequencies due to aperture averaging [1];
in essence, averaging out the fastest fluctuations, thus resulting in scintillation reduction and lower BER
attainment.
In Figure 3, the variation of the average channel capacity is evaluated with respect to the SNR, under (a) light
fog condition at ‫ ܮ‬ൌ NP, and (b) clear weather scenarios at ‫ ܮ‬ൌ ሼሽNP. In Fig. 3(a), we observe
a faster increase in the values of ‫ ۄܥۃ‬with respect to the SNR, in which the maximum capacity is approached
at615 ൒ G%. Near-identical behaviour is noted for all‫ܦ‬, albeit observing an increase in ‫ܣ‬଴ and reduction in
ߪூଶ ሺ‫ܦ‬ሻ. The channel capacity strongly depends on the turbulence (Fig. 3(b)), which imposes a higher scintillation
level and pointing error loss, signified by larger ߪூଶ ሺ‫ܦ‬ሻ and ߦ values, respectively [5]. For‫ ܮ‬ൌ NP, the
channel capacity increases at a slower rate for‫ ܦ‬ൌ PP, and asymptotically approaches a limit of
0.98 bits/channel use albeit continual increase in the SNR. Nevertheless, capacity enhancement can be achieved
with a larger‫ܦ‬, in which the optimal value is approached at 615 ൌ G%with ‫ ܦ‬ൌ PP. Furthermore, we
observe a larger penalty on the channel capacity at‫ ܮ‬ൌ NP, which is particularly pronounced at higher
electrical SNR. The channel capacity increases at a much slower rate for all cases of‫ܦ‬, and approaches their
respective asymptotic limit well below the optimal capacity. However, improvement in the channel capacity of
൐  can be made possible for 615 ൒ G%with ‫ ܦ‬ൌ PP, as compared to a point aperture.
0 0
10 10 L = 1.0 km (1.00); D = 40 mm
L = 1.0 km (1.00); D = 80 mm
-1 V = 642 m (23.31 dB/km); D = 40 mm -1 L = 1.0 km (1.00); D = 200 mm
10 10
V = 642 m (23.31 dB/km); D = 80 mm L = 1.0 km (1.00); D = 400 mm
V = 642 m (23.31 dB/km); D = 200 mm L = 4.5 km (15.69); D = 200 mm
-2 -2
10 V = 642 m (23.31 dB/km); D = 400 mm 10 L = 7.5 km (40.02); D = 200 mm
V = 480 m (36.12 dB/km); D = 200 mm
-3 -3
10 V = 150 m (115.58 dB/km); D = 200 mm 10
Average BER, Pe
Average BER, Pe

-4 -4
10 10

-5 -5
10 10

-6 -6
10 10

-7 -7
10 10

-8 -8
10 10

-9 -9
10 10
-5 0 5 10 15 20 25 30 35 -20 -15 -10 -5 0 5 10 15 20
(a) (b) Transmit Power, PFSO (in dBm) Transmit Power, PFSO (in dBm)

Figure 1. Average BER in terms of the 3)62 , (a) under different fog conditions at / NP; and (b) in the
strong turbulence regime at / ሼሽNP. In the figure legend, the numerical values (in parenthesis)
refer to (a) the attenuation coefficient ı; and (b) the Rytov variance ı5 .
-1 -1
10 AWGN channel 10 AWGN channel
D = 40 mm D = 40 mm
-2 D = 80 mm -2 D = 80 mm
10 10
D = 200 mm D = 200 mm
D = 400 mm D = 400 mm
-3 -3
10 10
Average BER, Pe

Average BER, Pe

-4 -4
10 10

-5 -5
10 No turbulence 10
No turbulence
-6 -6
10 10

-7 -7
10 10

-8 -8
10 10

-9 -9
10 10
0 2 4 6 8 10 12 14 16 18 0 5 10 15 20 25 30 35
(a) (b) Average Electrical SNR (in dB) Average Electrical SNR (in dB)

Figure 2. Average BER as a function of the SNR, under (a) light fog and (b) clear weather conditions, at
/ NP, taking into account the effect of ' and comparing with the AWGN case.

3
ICTON 2012 We.A4.2

1.0 1.0
D = 40 mm (1.2638 x 10 -4, 3.0719 x 10 -1)
D = 80 mm (5.0541 x 10 -4, 1.6806 x 10 -1)
0.9 0.9 D = 200 mm (3.1544 x 10 -3, 7.0706 x 10 -2)

Average Channel Capacity, <C> (in bits/channel use)

Average Channel Capacity, <C> (in bits/channel use)


D = 400 mm (1.2555 x 10 -2, 3.5934 x 10 -2)
0.8 0.8 D = 40 mm (5.4376 x 10 -5, 5.4946)
D = 80 mm (2.1749 x 10 -4, 3.0060)
D = 200 mm (1.3585 x 10 -3, 1.2647)
0.7 0.7
D = 400 mm (5.4223 x 10 -3, 6.4274 x 10 -1)

0.6 0.6
VR2 = 1.00
0.5 0.5 (L = 1.0 km)

0.4 0.4

0.3 0.3

0.2 D = 40 mm (1.2639 x 10-4, 3.3012 x 10-2) 0.2


-4 -2
D = 80 mm (5.0547 x 10 , 1.8060 x 10 )
0.1 D = 200 mm (3.1548 x 10-3, 7.5984 x 10-3) 0.1 VR2 = 40.02
(L = 7.5 km)
D = 400 mm (1.2557 x 10-2, 3.8616 x 10-3)
0.0 0.0
-6 -4 -2 0 2 4 6 8 10 12 14 16 18 20 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20
(a) (b) Average Electrical SNR (in dB) Average Electrical SNR (in dB)

Figure 3. Average channel capacity with respect to the SNR, under (a) light fog condition at / NP; and
(b) clear weather scenarios at / ሼሽNP. Numerical values in the figure legend refers to ൫‫ܣ‬଴ ǡ ߪூଶ ሺ‫ܦ‬ሻ൯.
1.00 1.0
w0 = 50 mm w0 = 100 mm
w0 = 100 mm w0 = 200 mm
0.95
w0 = 200 mm w0 = 400 mm

Average Channel Capacity, <C> (in bits/channel use)


Average Channel Capacity, <C> (in bits/channel use)

0.9
0.90

0.85 0.8

0.80
0.7
0.75

0.70 0.6

0.65

0.5
0.60

0.55
0.4

0.50 0 5 10 15 20 25 30 35 40
0 5 10 15 20 25 30
(a) (b) Rytov Variance, VR2 Rytov Variance, VR2

Figure 4. Effects of beam width on the average channel capacity under the influence of turbulence strength as
signified by ߪோଶ , for clear weather condition at (a) ‫ ܮ‬ൌ NP and (b) ‫ ܮ‬ൌ NP.
In Figure 4, the effects of beam width on the average channel capacity under the influence of turbulence are
examined under clear weather conditions at ‫ ܮ‬ൌ ሼሽNP. In the weak turbulence regime (ߪோଶ ൑ ͳǤͲ), near-
optimal capacity ൐ bits/channel use is approached with smaller ‫ݓ‬଴ , whereas performance degradation is
observed for higher values of ‫ݓ‬଴ . Parameter ‫ݓ‬଴ should be increased accordingly under severe turbulence
conditions (ߪோଶ ൐ ), to compensate the scintillation and pointing error effects, thus improving ‫ۄܥۃ‬. Hence,
beam width optimization is a feasible approach in promoting capacity enhancement for long-distance FSO links,
since larger ‫ݓ‬଴ improves the channel capacity only in the strong turbulence regime, while imposing slight
performance degradation under weak-to-moderate turbulence conditions, and vice versa.

5. CONCLUSIONS
We have jointly considered a spatially partially coherent laser beam and various link design criteria, to examine
the FSO link impairments imposed by atmospheric attenuation, turbulence and pointing error. The BER and
average channel capacity are evaluated, taking into account the effects of aperture averaging and beam width.
The Gaussian-beam wave model is used to characterize the propagation properties of the optical signal through
random turbulent medium. Our studies revealed that low BER values can be achieved for smaller SNR, by using
larger receiver aperture due to the averaging of irradiance fluctuations. Beam width optimization is a feasible
approach in promoting capacity enhancement, since larger beam size improves the channel capacity only in the
strong turbulence regime, while imposing slight performance degradation for weak-to-moderate turbulence.

REFERENCES
[1] L. C. Andrews and R. L. Philips, Laser beam propagation through random media. Washington, DC: SPIE,
2005.
[2] -&DQJDQG;/LX³$YHUDJHFDSDFLW\RIIUHH-space optical systems for a partially coherent beam
propagating through non-.ROPRJRURYWXUEXOHQFH´Opt. Lett., vol. 36, no. 17, pp. 3335-3337, Sep. 2011.
[3] R. Yongxiong, D. Anhong, L. Bin and G. Hong, ³&DSDFLWLHVIRUORQJ-distance free-space optical links
XQGHUEHDPZDQGHUHIIHFWV´IEEE Photon. Technol. Lett., vol. 22, no. 14, pp. 1069-1071, Jul. 2010.
[4] :23RSRRODDQG=*KDVVHPORR\³%36.VXEFDUULHULQWHQVLW\PRGXODWHGIUHH-space optical
communicatLRQVLQDWPRVSKHULFWXUEXOHQFH´J. Lightw. Technol., vol. 27, no. 8, pp. 967-973, Apr. 2009.
[5] $$)DULGDQG6+UDQLORYLF³2XWDJHFDSDFLW\RSWLPL]DWLRQIRUIUHH-space optical links with pointing
HUURUV´J. Lightw. Technol., vol. 25, no. 7, pp. 1702-1710, Jul. 2007.
[6] &/LX<<DR<;6XQ--;LDRDQG;+=KDR³$YHUDJHFDSDFLW\RSWLPL]DWLRQLQIUHH-space optical
FRPPXQLFDWLRQV\VWHPRYHUDWPRVSKHULFWXUEXOHQFHFKDQQHOVZLWKSRLQWLQJHUURUV´Opt. Lett., vol. 35,
no. 19, pp. 3171-3173, Oct. 2010.

You might also like