Limitations due to residual interference in a fiber-based optical

frequency reference at 1.55 µm

K. Manamanni,1 T. Steshchenko,1 F. Wiotte,1 R. Le Targat,2 M. Abgrall,2 O. Lopez, 1 E.

Cantin,1 P-E. Pottie,2 A-. Amy-Klein1, V. Roncin,1 F. Du-Burck,1, *
1-Laboratoire de Physique des Lasers, CNRS, Université Sorbonne Paris Nord, 99 av. J.-B.
Clément, F-93430 Villetaneuse, France
2-LNE-SYRTE, Observatoire de Paris, Université PSL, CNRS, Sorbonne Université, 61 Avenue
de l’Observatoire, F-75014, Paris, France
* Corresponding author: [email protected]
Abstract
We present an experimental investigation of the stability limits specific to optical frequency
standards using a fiber-optic architecture and semiconductor lasers. A compact setup
composed of a semiconductor laser frequency-locked onto an acetylene transition detected
in saturated absorption has been implemented using only fiber-optic components. Fiber
optic technology allows compact and reliable solutions for various applications. However,
for high sensitivity and stability applications such as metrology, residual reflections induced
by optical index inhomogeneities in connectors and fiber-coupled components leading to
spurious interference significantly limit performance. We have examined the origin of the
interference fringes superimposed on the detected signal and the limitations they cause to
the frequency stability of the reference. The effects of temperature and beam power
fluctuations are also examined. Our results show that the frequency stability is limited in the
10-13 range by the effect of interference fringes due to use of fiber components.

1- Introduction
Only a few molecules present transitions in the extended fiber-optic telecommunications
range. Several frequency references realizations have been reported from 1260 to 1675 nm
using C2H2, HCN, CO, CO2, H2S, CH4, H2O [1]. The most popular molecule is acetylene
owing to the (ν1 + ν3) overtone band of acetylene 13C2H2 within the C band (1530–1560
nm). Moreover, it occupies a special place for its P(16) transition at 1542.38 nm, which has
been chosen by the CIPM as a primary wavelength standard [2].
Thus, acetylene spectroscopy in this region has been carefully studied and high stability and
accuracy references have been demonstrated, based on the detection of sub-Doppler
resonances in acetylene vapor probed in a cell [3, 4, 5] or in a cavity [6, 7, 8; 9]. In both
configurations, frequency stabilities (the relative frequency fluctuations) in the 10 -14 range
have been reached [5, 9].
In parallel to these laboratory experiments, special attention is given to the development of
compact and portable references dedicated to non-laboratory measurements. For this
purpose, 1550 nm fiber-optic components allow considerable reduction of the volume of
such acetylene-based frequency standards. In addition to the compactness, this approach
guarantees the robustness of such setups for optical alignments. However, several issues
inherent to the use of optical fibers and photonics components limit their stability
performance. Residual Fresnel reflection in connectors and diopters leading to interference
fringes, as well as the sensitivity of light polarization with the mechanical and thermal
environment are the main issues reported in the literature.
For example, Ahtee et al. [10] describe the performance of an acetylene-stabilized laser
system based on a fiber setup, in which the acetylene cell is the only part of the setup in
free space. To overcome the problem of interference, the optical path length is dithered
using a vibrating mirror in the free space part of the setup. A fractional frequency deviation
of 8.8 10- 12 -1/2 for integration time in the range 10 - 1000 s is reported.
Initiated by the development of hollow-core photonic crystal fibers (HC-PCF), more compact
acetylene cells could be realized by filling such fiber with gas and hermetically connecting it
at both ends to standard single-mode fibers [11, 12]. This makes a promising all-fiber setup
with intermediate stability performance limited principally by HC-PCF guiding issues and
air/silica reflections in the cell [13, 14].
In this work, we present the experimental setup of a fiber-based optical frequency reference
consisting of a semiconductor laser locked onto the P(16) transition of the (ν1 + ν3)
overtone band of 13C2H2 at 1542.38 nm detected in a cell. The saturated absorption signal is
detected by a modulation transfer technique in which the saturating beam is frequency
modulated and the detection is performed on the probe beam at the modulation frequency.
This results in a background-free sub-Doppler signal. A stability in the 10-13 range between 1
and 104 s is demonstrated. Various effects limiting the stability are analyzed and it is shown
that the essential limitation is due to interference whose origins are carefully identified.
2- Experimental set-up
The principle of the experiment is depicted in Fig. 1. The part of the setup limited by the
dotted line corresponds to the free space part between two collimators. The laser diode is a
Rio Planex with a linewidth of 10 kHz and an output power of 12 mW for a 120 mA current.
Its output power is split into two beams, one to saturate the absorption of the acetylene
vapor in the cell and the other propagating in the opposite direction in the cell for probing
the absorption.
The saturating beam is frequency shifted (f1 = 25 MHz) and frequency modulated at fFM =
200 kHz with a modulation index  = 2 by an acousto-optic modulator (AOM#1). The
saturating beam is amplified by an Erbium-Doped Fiber Amplifier (EDFA) and its power in
the cell is automatically controlled by a variable optical attenuator driven electronically
(VOA#1). This setup compensates for power variations induced by polarization fluctuations
upstream of a polarization beam splitter (PBS#1) which ensures a linear polarization in the
cell. We have verified that the power variations detected on Photodiode #2 were not due to
interference after PBS#1. The power of the saturating beam is 30 mW and the
corresponding saturation parameter is 0.4.
The probe beam is frequency shifted (f2 = 40 MHz) by the acousto-optic modulator AOM#2
in order to avoid the detection of interference due to the residual zero-order at AOM#1’s
output as will be discussed later. The probe power is automatically controlled in the same
way as the saturating beam by the electronically driven variable optical attenuator VOA#2.
The polarization fluctuations upstream of PBS#2 are then compensated. An optical
circulator is used to extract the saturating beam at the cell’s output. The probe power in the
cell is 2 mW. The probe power modulated by the absorption in the cell at f FM = 200 kHz is
transmitted into PBS#1 and measured by Photodiode#1. The probe and pump polarizations
are maintained orthogonal thanks to the polarization controller PC#2 upstream of PBS#2.
The free-space part of the setup is composed of two adjustable collimators (Schäfter +
Kirchhoff), the cell filled with acetylene (13C2H2) with a pressure of 2.5 Pa, and two mirrors
for beam alignment (not shown in Fig. 1). The radius of both beams is about 0.4 mm for a
fiber-to-fiber optical power coupling rate of 70 %. The cell is made and filled by the Institute
of Scientific Instrument (Brno, Czech Republic). Its diameter is 3 cm and its length is 30 cm.
An anti-reflection coating is deposited on each optical window of the cell.
The setup is into a box made of high-density polyurethane plates ensuring basic thermal and
acoustic insulation, named hereafter “thermal box”. The temperature in the lab is controlled
at 19 ± 1° C by an air conditioning system.
The error signal is detected at fFM = 200 kHz by a lock-in amplifier. The detected signal is
applied to a proportional-integral controller (PI controller) in order to generate the laser
frequency correction signal. Two integrators are cascaded. The first one generates a fast
correction signal acting on the laser current in a correction band extending up to 2 kHz. The
second integrator corrects for slow fluctuations by acting on the set-point of the laser
temperature control.
AOM #1 EDFA
Pump VOA # 1
Laser Diode
(RIO) f1 + FM (fFM)
Probe
PI Controller

Photodiode
AOM #2 PC#1
#3
f2 PC#2
VOA # 2 PBS #2
13C Photodiode
2H2 PBS #1
I,T C

#1

Photodiode
fFM PI Controller
#2
2xPI
Lock in
Controller

Fig. 1. Laser stabilization setup (AOM: Acousto-optic modulator; PBS: Polarization beam splitter; VOA: Variable
optical attenuators; PI controller: Proportional-integral controller; PC: Polarization controller).

The spectral analysis of the demodulated signal at 200 kHz, used for frequency stabilization
of the laser on the acetylene transition at 1542.38 nm is shown in Fig. 2. The green curve is
the detection noise that fixes the sensitivity of the measurement. The black curve is
obtained by reducing the servo bandwidth (the frequency range in which the correction is
active) down to 200 Hz. The trace thus corresponds to the open-loop frequency noise for
Fourier frequencies higher than 200 Hz from which it reaches the 1/f2 trend. The red curve is
the noise in closed loop. We also plot in Fig. 2 the probe beam intensity noise measured by
shifting the laser frequency out of the transition. This intensity noise is converted into
frequency noise (blue curve) from the slope of the detected signal (inset of Fig. 2).
90

80

Frequency Noise PSD (dBHz2/Hz) 70

60
Intensity noise
50

40
6

Voltage (V)
30
4
20 2
Detection noise
10 0
0 -2

-10 -4
Frequency
-20 -6
1 2 3 4 5
10 10 10 10 10
Frequency (Hz)

Fig. 2. Power spectral density of the error signal for frequency stabilization of the laser onto the acetylene
transition at 1542.38 nm. The black curve is the open-loop noise (for frequency > 200 Hz – see text), the red
curve is the closed loop noise, the blue curve is the intensity noise contribution limiting the noise rejection, the
green curve is the detection noise giving the sensitivity of noise measurements. Inset: trace of the error signal
corresponding to the detected molecular profile.

The noise reduction above 50 kHz observed in all these curves is due to the mixer
output filter. The measured closed-loop noise reaches the open-loop noise curve at the
frequency 2 kHz, which therefore corresponds to the servo bandwidth. The probe beam
intensity noise is white noise at the 50 dBHz2/Hz level. Throughout the correction band,
this noise level is higher than the noise measured in closed loop. Consequently, the actual
frequency noise of the corrected laser copies this intensity noise, which limits the
possibilities of correction.
The error signal is shown in the inset of Fig. 2. Its peak-to-peak width is 630 kHz and its
amplitude is 12 V leading to a frequency-to-voltage conversion factor of 2 10-5 V/Hz. The
sensitivity of the laser frequency to the input voltage is 37.5 MHz/V and the cut-off
frequency of the PI corrector is 1.6 Hz. This results in a servo bandwidth of 1.2 kHz in good
agreement with the experimental value obtained in Fig. 2.
3- Stability results
The stability of our frequency standard at 1542.38 nm is evaluated from a comparison with
a metrological reference provided by LNE-SYRTE (Observatoire de Paris) in the framework of
the national metrological network REFIMEVE+ [15]. A continuous wave Erbium-doped fiber-
laser emitting at ref = 1542.14 nm (channel 44 of WDM ITU-T) is stabilized onto an ultra-
stable Fabry-Perot cavity [16]. For long-term stability, it is referenced onto a hydrogen-
maser, ensuring a stability better than 10-15 until 105 s of integration time. The laser is
transferred to our lab through a 43 km long single mode fiber. The propagation noise along
the fiber, mainly thermal and acoustic noise, is actively compensated [17]. In our lab, the
signal is regenerated by phase locking a local laser to the incoming reference signal and is
then distributed to our experiment using a multibranch laser station [18]. As a result,
the transfer does not degrade the stability and accuracy of the transmitted optical
reference [17,18].
The frequency shift between the metrological reference and the C2H2 stabilized laser is
about 30 GHz. For stability measurements, both sources are modulated in the same Mach-
Zehnder modulator at fs = 15 GHz, generating two sets of sidebands. By tuning the
modulation frequency fs, we obtain a beat-note at 60 MHz that is filtered and sent to a
frequency counter. In order to quantify the frequency stability of the reference, Allan
deviations in relative value σy (fractional Allan deviation) of the beat-note are drawn in Fig.
3.
The three curves in Fig. 3 reflect the impact of temperature and power fluctuations on the
frequency stability of the setup described in Fig. 1. For integration times less than 10 s,
all three curves exhibit the same fractional stability of 8 10-13 -1/2, consistent with
the limita on of fre uency correc on due to the laser white amplitude noise iden ed in ig. .
ndeed, this noise of 50 dBHz2/Hz leads to an Allan deviation of the form 10-12  -1/2 [19], in
good agreement with the experimental stability obtained in Fig. 3.
For integration times higher than 10 s, the evolutions of the stability curves are different,
but some bumps are observed on the three curves. We will show in the next section that
this behavior is due to interference fringes coming from parasitic reflections on fiber
components.

1  1/2
Overlapping Allan Deviation sy

-12
10

-13
Air cond. ON & Beam power stab. ON
10
Air cond. OFF & Beam power stab. ON  -1/2
Air cond. ON & Beam power stab. OFF
0 1 2 3 4
10 10 10 10 10

 (s)

Fig. 3. Stability characterization of the fiber-based frequency standard at 1.55 µm with fractional Allan
deviation (beat-note with the optical reference transferred through the REFIMEVE+ network). Three
measurements are achieved, two concerning the thermal sensitivity thanks to the air conditioning system in
the lab (blue and red curves) and one concerning the power sensitivity thanks to power beam stabilization
system using VOAs (black curve).
The blue solid line curve is obtained for a temperature in the lab controlled by the air
conditioning system and when the power of both beams are automatically controlled by
VOA#1 and VOA# in such a way that it is stable at the two PBS’s outputs. or integration
times greater than 10 s, a characteristic shape resulting from periodic variations of the laser
frequency with a period slightly greater than 2000 s is clearly visible. Note that the level of
stability always remains in the 10-13 range and that the bumps due to periodic
frequency fluctuations are superimposed on a general decreasing trend that reaches a
stability level in the 10-14 range at 104 s.
The origin of these periodic stability variations, analyzed in Section 4, is attributed to
interference fringes due to parasitic reflections on the fibered components of the setup.
The red curve in long dashes is obtained when the air conditioning system is turned off,
which generates temperature fluctuations during the long-term stability measurements. We
will show in Section 5 hereafter that the temperature sensitivity of our setup is clearly
caused by the thermal sensitivity of AOM#1.
The black curve in short dashes shows the sensitivity of the setup to beam power stability. It
is obtained when the air conditioning system is on but without power stabilization at the
output of PBS’s. The sensitivity to beam power is analyzed in Section 6.
4- Interference analysis
The most significant limitation observed in fiber-based frequency references comes from
beam reflections in the setup, at the interfaces of the fiber components and on the
collimators [10]. These various reflections result in interference fringes added to the output
signal of the photodiode monitoring the probe beam. However, only those modulated at
fFM = 200 kHz are detected by the lock-in amplifier and superimposed to the error signal.
These interference fringes lead to fluctuations in the zero-crossing of the detected signal
and thus, to an instability of the optical frequency of the standard.
When the acousto-optic modulator AOM#2 shifting the probe beam frequency by f2 = 40
MHz is removed, significant interference fringes representing 8% of the amplitude of the
error signal appear (Fig. 4(a)). They are necessarily due to the intense saturating beam, the
only one to be modulated in the modulation transfer method. However, we observed that
these interference fringes exist only when the probe beam is also received by the
photodiode. Consequently, these interference fringes do not result from the direct
detection of the residual reflections of the saturating beam, which would then exist
independently of the probe beam. They result from the homodyne mixing on the
photodiode of the probe beam and the residual zero-order output beam of AOM#1, whose
intensity is measured at 50 dB below that of the first-order output beam. This process is
illustrated in Fig. 5(a). Of course, the zero-order beam cannot be frequency-modulated.
However, it is observed that the first-order output beam carries residual amplitude
modulation (RAM) at fFM at the level of 2 10-3 in addition to frequency modulation [20].
Therefore, the zero-order beam is also amplitude modulated at the frequency f FM leading to
interference detection. The addition of the modulator AOM#2 shifting the frequency of the
probe beam (f2 = 40 MHz) cancels this spurious signal by transposing the beat note
frequency around 40 MHz Fig. 5(b). The beat note between the zero-orders of both AOM
outputs is negligible.

Voltage (V)
Voltage (u.a.)
(a) 0.35 (c)
0,30 Df = 90

0.25

0.20

0.15

0.10
8%
0.05

0.0 Df = 40
Time (hour)
-0.05
Frequency 0 2 4 6 8 10

0.3
(b)
Voltage (V)

0.02 V
0.2

0.1

with PBS
0.0

-0.1
0.5 V

with circulator
-0.2

Time (s)
-0.3
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1

Fig. 4. Characterization of interference fringes in the setup: (a) Detected error signal without the
frequency shift introduced by AOM#2; (b) Detected interference fringes with PBS#1 (grey curve) and
with a circulator (black curve); (c) Detected interference fringes according to the detection phase
(black curve: minimum amplitude - red curve: maximum amplitude).

f2 = 40 MHz

Laser Laser

Probe Probe
beam beam

fFM fFM fFM fFM

Saturating Saturating
beam beam

f1 = 25 MHz f1 = 25 MHz
Zero-order First-order Zero-order First-order
(AM) (FM + AM) (AM) (FM + AM)

Fig. 5. Interference due to the zero-order of the AOM: (a) The probe beam is at the same frequency
as the residual zero-order of the AOM that shifts and modulate the saturating beam frequency. The
interference results from the detection at fFM of this zero-order that is amplitude modulated. (b) By
shifting the frequency of the probe beam by 40 MHz, the beat note between the probe beam and
the zero-order of the AOM is rejected at 40 MHz and therefore no interference is detected, the beat
note between AOM zero-orders being negligible.
With AOM#2 in the setup, the residual interference fringes result from the direct detection
of reflections from the saturating beam itself, namely the first-order output beam of
AOM#1. It is frequency modulated at fFM = 200 kHz and carries a slight RAM as seen above.
Fig. 4(b) shows the interference fringes detected when the probe beam is switch off
(AOM#2 RF input OFF). The two traces correspond to the interference fringes recorded
when PBS 1 is used to separate and route the optical beams, as in Fig. 1, and when PBS#1 is
replaced by an optical circulator. In this latter case, a high level of interference is observed.
The introduction of PBS#1 in the setup leads to an efficient rejection (by a factor of 25) of
interference by filtering the saturating beam reflections according to its polarization.
Nevertheless, it is seen that this rejection is limited by the polarization maintaining fiber
(1m) at the PBS output presenting a polarization isolation of less than 30 dB between the
two principal propagation axes.
Fig. 4(c) shows a ten-hour record of detected interference. The acquisition is performed
with PBS#1, without the probe beam and when the saturating beam is replaced by the
metrological reference at ref. Although the optical frequency of the source is perfectly
constant, we observe a regular scrolling of the interference fringes during the recording. It
causes the periodic shift of the zero crossing of the error signal for the laser frequency lock,
leading to the bumps observed on the stability curves in Fig. 3. The detected fringes are
strongly sensitive to the detection phase of the error signal. Their amplitude is minimized
for a detection phase of 40° (black trace of Fig. 4(c)). The maximum interference amplitude
is obtained for a detection phase of 90° (red curve in Fig. 4(c)) for which the detection of a
slowly variable background is also observed.
We have seen above that the AOM output beam is both frequency and amplitude
modulated (RAM). In addition, the RAM is strongly sensitive to the experimental
environment. This leads to a variable background detection. Note that the phases
associated with both modulations are different. For a detection phase that cancels the RAM,
the interference fringes resulting only from the detection of the FM signal are obtained. This
is the case of the black curve in Fig. 4(c) without background. For another detection phase,
the interference fringes resulting from the RAM detection appear associated with a variable
background, as for the red curve in Fig. 4(c).
This effect can be modeled simply by computing the detected signal resulting from the
reflection by a low finesse cavity of a beam modulated simultaneously in frequency and
amplitude. For the parameters of our experiments (FM index = 2, modulation frequency =
200 kHz, RAM = 2 10-6), we found that the maximum amplitude of interference fringes due
to RAM detection is 2.7 times greater than that due to the detection of the FM modulation
alone. In addition, the interference resulting from RAM detection is superimposed to a
continuous level while interference resulting from FM detection is centered at zero, which is
consistent with the experimental results in Fig. 4(c).
The phase shift between FM and RAM generated by the AOM makes it possible to find a
detection phase that cancels the signal due to RAM detection, and thus to minimize both
interference and background while preserving an error signal of sufficient amplitude.
Consequently, RAM rejection and beam intensity control seem to be essential in such an
experiment.
5- Sensitivity to temperature
The RAM generated by AOM#1 is temperature sensitive. This leads to a time-varying shift of
the detection phase associated with the minimum of interference fringes (black curve in Fig.
4(c)) and thus a variation of the fringe offset. This is illustrated by Fig. 6 which shows the
recording of the frequency fluctuations of the reference (Fig. 6(a)) and of the temperature in
the thermal box (Fig. 6(b)) during a night with air conditioning in the lab corresponding to a
regulation cycle of 1800 s (blue curves) and without air conditioning (red curves). In both
cases, the detection phase is initially adjusted to minimize the interference amplitude.
With air conditioning in the lab (blue curves), the temperature is sufficiently stable to
maintain the initial adjustment leading to a constant average value of the frequency. The
corresponding stability is given by the blue curve in Fig. 3. The periodic variation of the
detected frequency (period 2200 s) is reflected by the multiple bumps in Allan's variance as
noted out above. Without particular drifts, we expect that for long integration times, the  -
1/2
trend reaches the stability level of 10-14.
When the air conditioning is turned off (red curves), the room temperature is less stable
(variation of 3.5 °C during the night) and the frequency is observed to be unstable in the
long term. A significant degradation of the stability of the frequency reference is observed
(red curve in Fig. 3) leading to a trend in  for integration time between 30 and 200 s and in
 1/2 beyond.
These measurements were performed with active beam intensity stabilization. The
consequences of beam intensity variations are discussed in the next section.
6- Sensitivity to beam power
Another limitation of long-term stability of the reference is the sensitivity of the detected
line center to the power of the probe beam and the saturating beam. We find a sensitivity
of 65.7 Hz/mW for the probe beam and 40.7 Hz/mW for the saturating beam. The origin of
instability of the intensity of the beams in the cell is the polarization fluctuations in fibers
induced by the thermal variations. The PBS’s convert them into intensity variations. This
effect is compensated by controlling the intensity of both beams with VOA’s. Residual
fluctuations are much lower than 1 mW, which makes this effect negligible.
In order to quantify the sensitivity of our reference to beam power variations, we have
measured the reference frequency with a stable temperature in the lab, but without
correction of the power of the probe and saturating beams. The resulting stability is given
by the black curve in Fig. 3. The period of interference fringes is modified as compared to
the blue curve obtained with the beam power correction and the interference bump
corresponding to the thermal sensitivity of the AOM is cancelled.
This result shows that the observed long-term instability is due to both beam power and
RAM fluctuations, both due to thermal effects. In our setup, the control of the beam power
and the external temperature rejects the long-term instabilities up to the limit given by the
scrolling of interference fringes superimposed on the detected signal (blue curves of the
Fig. 3).

3000 2200 s

2000
with Air Conditioning
1000
f - f0 (Hz)

-1000

-2000

-3000

-4000 without Air Conditioning

0 10000 20000 30000

Time (s)

2.0

1.5
Temperature Variation (°C)

3.5 °C
1.0

0.5
without Air Conditioning
0.0
0.9 °C

-0.5

-1.0

-1.5 with Air Conditioning

1800 s
-2.0

-2.5
0 10000 20000 30000
Time (s)

Fig. 6. Effect of temperature on the detected interference: (a) Frequency fluctuations of the beat-
note between our reference and the REFIMEVE+ signal. The interference fringes scroll with a period
of 2200s; (b) Corresponding temperature in the thermal box with and without air conditioning
system in the lab (cycle period of 1800s).

7- Conclusion
We have described an optical frequency reference at 1.5 µm, based on standard fiber
components. Frequency stabilization is achieved by locking the laser frequency to a
saturated absorption signal in acetylene. The signal detection is performed in a cell by a
modulation transfer method. Stability in the range of 10-13 between 1 and 104 s is
demonstrated.
We have investigated the origin of interference fringes and estimated their impact on the
frequency stability of the reference. We have also highlighted the long-term effect of
temperature and power fluctuations.
Setups based on fiber components have many advantages (compactness, lightness,
robustness of optical alignments) which make them interesting, especially for the realization
of transportable devices. However, even if some improvements could still be made to the
setup (different powers of the pump and probe beams, replacement of connectors by
splicing, choice of components), this work shows that interference constitutes an essential
technological limitation of the fiber-based setups, limiting their stability in the 10-13 range.
Moreover, we demonstrate that accurate temperature and power control may considerably
limit RAM fluctuations and pave the way to the 10-14 range for long integration times.

Acknowledgments. We are grateful to H. Mouhamad and L. Malinge for their electronics

support. This work was supported by the LABEX Cluster of Excellence FIRST-TF (ANR-10-
LABX-48-01), within the Program " nvestissements d'Avenir” operated by the rench
National Research Agency (ANR), and by PN-GRAM (Gravitation, Références, Astronomie,
Métrologie). We acknowledge RENATER (Réseau National pour l’Enseignement et la
Recherche), partner of the REFIMEVE+ project (ANR Equipex REFIMEVE+ ANR-11-
EQPX-0039).

References
1. J.C. Petersen and J. Henningsen, “Molecules as absolute standards for optical
telecommunication. Re uirements and characterisation,” paper 1777, XXV th General
Assembly of the International Union of Radio Science (URSI 2002), Maastricht, the
Netherlands (17-24 August 2002).
2. https://www.bipm.org/documents/20126/41590985/M-e-P_13C2H2_1.54.pdf/623e90e2-e816-
36f7-2fd3-93b2e741ee5c

3. F.-L. Hong, A. Onae, J. Jiang, R. Guo, H. Inaba, K. Minoshima, T. R. Schibli, H. Matsumoto,

and K. Nakagawa, “Absolute fre uency measurement of an acetylene-stabilized laser at
154 nm,” Opt. Lett. 8 ( 3), 3 4 ( 003).
4. P. Balling, M. ischer, P. Kubina, and R. Holzwarth, “Absolute fre uency measurement of
wavelength standard at 154 nm: acetylene stabilized D B laser,” Opt. Express 13 ( 3), 9196
(2005).
5. J. Hald, L. Nielsen, J. C. Petersen, P. Varming, and J. E. Pedersen, “ iber laser optical
fre uency standard at 1.54 μm,” Optic Express 19 (3), 05 ( 011).
6. M. de Labachelerie, K. Nakagawa, Y. Awaji, and M. Ohtsu, “High-frequency-stability laser
at 1.5 mm using Doppler-free molecular lines,” Opt. Lett. 0 (6), 57 (1995).
7. A. Czajkowski, A.A. Madej, P. Dubé, “Development and study of a 1.5 lm optical
fre uency standard referenced to the P(16) saturated absorption line in the (ν1 + ν3)
overtone band of 13C H ,” Opt. Commun. 34, 59 ( 004).
8. H. S. Moon, W.-K. Lee, and H.-S. Suh, “Absolute-Frequency Measurement of an
Acetylene-Stabilized Laser Locked to the P(16) Transition of 13C2H2 Using an Optical-
re uency Comb,” EEE Trans. nstrum. Meas. 56 (2), 509 (2007).
9. C. S. Edwards, H. S. Margolis, G. P. Barwood, S. N. Lea, P. Gill, G. Huang, and W. R. C.
Rowley, “Absolute fre uency measurement of a 1.5µm acetylene standard by use of a
combined fre uency chain and femtosecond comb,” Opt. Lett 9 (6), 566 (2004).
10. V. Ahtee, M. Merimaa, and K. Nyholm, “ iber-Based Acetylene-Stabilized Laser,” EEE
Trans. Instrum. Meas. 58 (4), 1211 (2009).
11. . Benabid, . Couny, J. C. Knight, T. A. Birks, and P. S. J. Russell, “Compact, stable and
efficient all-fibre gas cells using hollow-core photonic crystal fibres,” Nature 434, 490 ( 005).
12. C. Wang, N. V. Wheeler, C. Fourcade-Dutin, M. Grogan, T. D. Bradley, B. R. Washburn, F.
Benabid, and K. L. Corwin, “Acetylene fre uency references in gas-filled hollow optical fiber
and photonic microcells”, Applied optics 5 ( ), 5430 ( 013).
13. J. Hald, J. C. Petersen, and J. Henningsen, “Saturated Optical Absorption by Slow
Molecules in Hollow-Core Photonic Band-Gap ibers,” PRL 98, 1390 ( 007).
14. K. Knabe et al., “10 kHz accuracy of an optical fre uency reference based on 1 C H -
filled large-core kagome photonic crystal fibers,” Opt. Express 17 (18), 16017 ( 009).
15. https://www.refimeve.fr/index.php/en/
16. X. Xie, R. Bouchand, D. Nicolodi, M. Lours, C. Alexandre, and Y. Le Co , “Phase noise
characterization of sub-hertz linewidth lasers via digital cross correlation,” Opt. Lett. 4 (7),
1217 (2017).
17. D. Xu, W.-K. Lee, F. Stefani, O. Lopez, A. Amy-Klein, et P.-E. Pottie, “Studying the
fundamental limit of optical fiber links to the 10- 1 level,” Opt. Express 6 (8), 9515 ( 018).
18. E. Cantin, M. Tønnes, R. L. Targat, A. Amy-Klein, O. Lopez, and P.-E. Pottie, “An accurate
and robust metrological network for coherent optical fre uency dissemination,” New J.
Phys. 23 (5), 053027 (2021).
19. J. Rutman, “Characterization of phase and fre uency instabilities in precision fre uency
sources: ifteen years of progress,” Proc. EEE 66 (9), 1048 (1978).
20. F. Du-Burck, O. Lopez and A. El Basri, “Narrow-band correction of the residual amplitude
modulation in frequency-modulation spectroscopy,” EEE Trans. on nstrumentation and
Measurement 52(2), 288 (2003).