3314 OPTICS LETTERS / Vol. 30, No.

24 / December 15, 2005

Precision measurement of the refractive index of

air with frequency combs
J. Zhang, Z. H. Lu, and L. J. Wang
Institute of Optics, Information and Photonics, Max-Planck Research Group and University Erlangen-Nuremburg,
91058 Erlangen, Germany

Received July 29, 2005; accepted August 29, 2005

We report an experimental method for high-precision refractive index measurement using a Michelson in-
terferometer setup with a femtosecond optical frequency comb as the light source. We demonstrate this tech-
nique by measuring the refractive index of air, under different pressures, inside a 29 m multipass cell. Both
spectral and temporal interferograms are recorded. The method has a sensitivity of 6.2⫻ 10−9 and a stan-
dard error of 7.4⫻ 10−8 at fixed parameters (pressure and temperature). The possible variation of the den-
sity factor from the Lorentz–Lorenz equation is also discussed. © 2005 Optical Society of America
OCIS codes: 120.3180, 120.3940, 120.4530.

Probably the most exciting recent advance in preci-
sion frequency metrology is the development of the
optical frequency comb technique.1 It enables a direct
link between microwave and visible optical frequency
measurements with one single mode-locked laser.2 In
addition to its many applications in precision laser
spectroscopy,3–6 it is possible to extend the frequency
measurement technique to the measurement of the
material’s refractive index and dispersion with a
higher precision. As is well known, nowadays fre-
quency measurements can be made very precisely. By
using the frequency comb as a primary standard, we
expect to obtain a higher precision of the optical prop-
erties of various gaseous molecules such as air.
The accurate knowledge of the refractive index of
air is very important in GPS communications and li-
dar. For most applications the refractive index of air
is indirectly given by Edlén’s formula,7–10 with an ac-
curacy of approximately 5 ⫻ 10−8.11 In this Letter we
describe a new experimental technique to measure
the absolute refractive index of air, taking advantage
of the accuracy of a frequency- and phase-stabilized
comb system. Our method has the potential of
achieving a 10−9 accuracy. Another added advantage
of using a frequency comb is its broad bandwidth in
the visible and near infrared (NIR), which enables us
to characterize the refractive index of air at different
wavelengths simultaneously.
The experimental setup, based on an unbalanced
Michelson interferometer, is shown in Fig. 1. The
light source is a femtosecond, mode-locked, Ti:sap-
phire laser frequency comb (MenloSystems FC-8004)
whose repetition rate and carrier envelope phase are
both locked to a Rb frequency standard (SRS FS725).
The laser has a repetition rate of 200 MHz. The cen-
ter wavelength is approximately 806 nm, with a
bandwidth of approximately 50 nm. The pulse width
is measured to be approximately 30 fs by using an
autocorrelator. The laser light is spatially filtered be-
fore entering the Michelson interferometer. The
power after the spatial filter is approximately
18 mW. The beam is divided into two paths using a
pellicle beam splitter. The first beam goes through a
multipass cell (Toptica CMP-30) that is filled with
synthetic dry air [Air Liquide synthetic air N57, ⬍2
parts in 106 共ppm兲 H2O, ⬍0.1 ppm CO2]. The total
path length inside the multipass cell is approxi-
mately 29 m. The other beam goes through an adjust-
able reference arm. The two beams are recombined in
a second beam splitter with equal intensity, and one
of the outputs is sent into a fast photodiode as the
time domain signal, analyzed with an Agilent 83480A
communications analyzer. The other output is
coupled into an Anritsu MS9701B optical spectrum
analyzer through a short length single-mode fiber as
the frequency domain signal. Both the time and fre-
quency domain signals are recorded and processed by
a computer.
In the time domain the interference signal has the
form
I共␶兲 ⬀ 具E12共t − ␶兲典 + 具E22共t兲典 + 具E1共t − ␶兲E2*典
+ 具E1*共t − ␶兲E2典, 共1兲
in which the angle brackets indicate the time averag-
ing done by the detector and ␶ is the time delay be-
tween the two paths. The time delay can be varied by
either moving the stepper motor or, as pointed out by
Ye,12 by changing the repetition rate f of the laser.
The relationship between the time delay ␶ and the
repetition rate f is
Fig. 1. Experimental setup. SF, spatial filter; BS1, pellicle
beam splitter; BS2, cube beam splitter; MPC, multipass
cell; RR, retroreflector; L, lens; PD, fast photodiode; FC, fi-
ber coupler; OSA, optical spectrum analyzer.

0146-9592/05/243314-3/$15.00 © 2005 Optical Society of America

 December 15, 2005 / Vol. 30, No. 24 / OPTICS LETTERS
␶ = 共N/f2兲␦f, 共2兲
in which ␦f is the change of the repetition rate. N is
an integer such that the initial pulse from the multi-
pass cell arm overlaps with the Nth pulse from the
reference arm. N can be calculated by varying the
repetition rate and measuring the relative time shift
of the two pulses when they are not overlapping by
using a fast oscilloscope. In our experiment N is mea-
sured to be 19, corresponding to a relative path-
length difference of approximately 29 m. The resolu-
tion of our stepper motor is 0.114 ␮m. Although the
laser repetition rate can be controlled to the 10 mHz
accuracy, the limiting factor is still the mechanical
stability of the interferometer. In our case, the real-
istic resolution is limited by the mechanical drift dur-
ing the data-acquisition time to approximately
0.1 ␮m (Ref. 13).
In the experiment we first evacuate the multipass
cell to vacuum. We adjust the stepper motor so that
the pulse from the multipass cell arm almost over-
laps with the 19th pulse from the reference arm. The
laser repetition rate is then scanned at a 0.4 Hz step
size at approximately 201.12 MHz while the laser re-
mains frequency and phase locked. An interferogram
is obtained. Then we fill the multipass cell with air to
a preset pressure of approximately 1 bar 共1
⫻ 105 Pa兲. This time we only scan the laser repetition
rate at a 0.4 Hz step size at approximately
201.07 MHz. The result is shown in Fig. 2(a). The
repetition rate difference ␦f from the peaks of the en-
velopes of the two interferograms is measured, and
with Eq. (2) we can calculate the group delay from
the multipass cell ␶g. Since
关ng共␭0兲 − 1兴lcell
␶g = , 共3兲
c
we can then calculate the total length of the multi-
pass cell lcell. Here ng共␭兲 = n共␭兲 − ␭dn共␭兲 / d␭ is the
Fig. 2. (a) Interferograms in the time domain versus laser
repetition rate when the multipass cell is filled with stan-
dard air or pumped out to vacuum. (b) Interferograms in
the frequency domain. The solid curve shows spectral
modulation, and the dotted curve shows spectrum Ĩ1共␻兲
+ Ĩ2共␻兲.
3315
group index of air, and ␭0 = 806 nm is the center
wavelength of the laser pulse. In the first-order ap-
proximation we can directly calculate the group in-
dex by using Edlén’s formula,7,8,10 from which we can
obtain lcell. The uncertainty of lcell associated with the
frequency measurement is approximately 0.1 mm. A
more serious problem is the thermal expansion of the
multipass cell. The laboratory temperature is con-
trolled to within 0.5 ° C, and real-time measurements
show a fluctuation of the cell length up to 7 mm; this
contributes an uncertainty of 6.8⫻ 10−8 to the refrac-
tive index measurement accuracy. In the future, a
temperature-stabilized cell will be used.
Frequency domain interferograms are simulta-
neously taken to obtain the refractive index. In the
frequency domain the interference signal has the
form
冑
Ĩ共␻兲 = Ĩ1共␻兲 + Ĩ2共␻兲 + 2 Ĩ1共␻兲Ĩ2共␻兲 cos ⌽共␻兲. 共4兲
A spectral modulation is be observed at the output of
the optical spectrum analyzer when the two arms are
unbalanced. The periodicity is inversely proportional
to the delay of the two arms. When the delay is too
large, such that the spectral modulation is washed
out, we get only Ĩ1共␻兲 + Ĩ2共␻兲. This result is shown in
Fig. 2(b), where the solid curve corresponds to spec-
tral modulation Ĩ共␻兲 and the dotted curve corre-
sponds to no spectral modulation Ĩ1共␻兲 + Ĩ2共␻兲. The
phase ⌽共␻兲 can be measured with the same tech-
nique used by Dogariu et al.13 When the multipass
cell is filled with air, the phase is written as
␻共nlcell + n⬘⌬l兲
⌽air共␻兲 = + ␾⬘共␻兲, 共5兲
c
in which n⬘ is the refractive index of air outside the
multipass cell, which might be different from the re-
fractive index of air n inside the multipass cell. ⌬l is
the path-length difference of the two arms excluding
the contributions from the multipass cell and cell
window. ␾⬘共␻兲 represents a common phase shift in-
cluding those introduced by the window and the mul-
tireflecting mirrors. Similarly when the multipass
cell is pumped to vacuum the phase can be written as
␻共lcell + n⬘⌬l兲
⌽vac共␻兲 = + ␾⬘共␻兲. 共6兲
c
Accurate knowledge of ␾⬘共␻兲 is not required since we
are interested only in the phase difference
共n − 1兲lcell␻
⌬⌽共␻兲 = ⌽air共␻兲 − ⌽vac共␻兲 = . 共7兲
c
From Eq. (7) we can calculate the refractive index
of air to be
␭⌬⌽
n共␭兲 = 1 + . 共8兲
2␲lcell
With the uncertainties associated with ␭共0.015 nm兲,
⌬⌽共0.8 rad兲, and lcell共0.1 mm兲, we can estimate that
3316 OPTICS LETTERS / Vol. 30, No. 24 / December 15, 2005
Fig. 3. Wavelength dependence of the refractive index of
air. The solid curve shows the refractive index of air de-
rived from Edlén’s equation. The dashed curve shows the
nonlinear least-squares fitting of the measured refractive
index. The standard error of the 800 nm data point is 4.1
⫻ 10−7.
Fig. 4. Refractive index difference between our results
and Edlén’s formula at 800 nm under different pressures.
Each data point is an average of five measurements. A typi-
cal error bar is 7.4⫻ 10−8.
our experiment has a sensitivity of 6.2⫻ 10−9. We
took 14 measurements with pressures close to 1 bar
and temperatures of approximately 21.5 ° C. The re-
sults are first converted into standard dry air condi-
tions (20 ° C, 1 bar, 400 ppm CO2).10 The average is
least-squares fitted with the standard formula 共n
− 1兲 ⫻ 108 = A0 + A1 / 共B1 − 1 / ␭2兲 + A2 / 共B2 − 1 / ␭2兲.8–10 The
result is compared with Edlén’s formula and shown
in Fig. 3. As can be seen, our result is very close to
Edlén’s result. At the wavelength of 800 nm, our
measured refractive index of air is nfit = 1.000266921.
Compared with Edlén’s result of ntheo = 1.000266791,
the difference is 1.3⫻ 10−7 with a standard error of
4.1⫻ 10−7. The error represents the distribution over
a wide range of pressure 共3000 Pa兲. Using the accu-
racy of temperature measurement 共0.1 ° C兲 and pres-
sure measurement 共250 Pa兲, an uncertainty limit of
6.7⫻ 10−7 is obtained.
We have also measured the effect of the air pres-
sure to density factor by converting different pres-
sure measurement results into standard conditions
by using the same density factor formula.10 The re-
sult is shown in Fig. 4. The preliminary results show
that the absolute error increases when the pressure
is farther away from 1 bar pressure. It could be due
to the deviation from the Lorentz–Lorenz equation
that is used to derive the density factor. Further ex-
periments are under way.
In conclusion, we have proposed and demonstrated
a method of measuring the refractive index of air us-
ing a frequency comb. A sensitivity of 6.2⫻ 10−9 was
demonstrated. The accuracy of our experiments
reaches a level of 1.3⫻ 10−7 compared with Edlén’s
formula at 800 nm. A standard error of 7.4⫻ 10−8 at
fixed parameters (pressure and temperature) is
reached. Our technique has its origin from white-
light interferometry,14,15 and by changing the laser
repetition rate instead of the stepper motor,12 we can
achieve a better resolution and accuracy. The prelimi-
nary results can be further improved if one chooses a
smaller repetition rate scanning step and improved
temperature and pressure reading of the air inside
the multipass cell.
Z. H. Lu acknowledges fellowship support from the
Alexander von Humboldt Foundation. We thank
Arthur Dogariu for helpful discussions on spectral
phase measurement. Z. H. Lu’s e-mail address is
zlu@optik.uni-erlangen.de.
References
1. T. Udem, J. Reichert, R. Holzwarth, and T. W. Hänsch,
Phys. Rev. Lett. 82, 3568 (1999).
2. S. A. Diddams, D. J. Jones, J. Ye, T. Cundiff, J. L. Hall,
J. K. Ranka, R. S. Windeler, R. Holzwarth, T. Udem,
and T. W. Hänsch, Phys. Rev. Lett. 84, 5102 (2000).
3. T. Udem, S. A. Diddams, K. R. Vogel, C. W. Oates, E. A.
Curtis, W. D. Lee, W. M. Itano, R. E. Drullinger, J. C.
Bergquist, and L. Hollberg, Phys. Rev. Lett. 86, 4996
(2001).
4. J. Stenger, C. Tamm, N. Haverkamp, S. Weyers, and
H. R. Telle, Opt. Lett. 26, 1589 (2001).
5. R. J. Jones, W. Y. Cheng, K. W. Holman, L.-S. Chen, J.
L. Hall, and J. Ye, Appl. Phys. B 74, 597 (2002).
6. P. C. Pastor, G. Giusfredi, P. de Natale, G. Hagel, C. de
Mauro, and M. Inguscio, Phys. Rev. Lett. 92, 023001
(2004).
7. B. Edlén, Metrologia 2, 71 (1966).
8. K. P. Birch and M. J. Downs, Metrologia 30, 155 (1993).
9. K. P. Birch and M. J. Downs, Metrologia 31, 315 (1994).
10. G. Bönsch and E. Potulski, Metrologia 35, 133 (1998).
11. P. Schelleknens, G. Wilkening, F. Reinboth, M. J.
Downs, K. P. Birch, and J. Spronck, Metrologia 22, 279
(1986).
12. J. Ye, Opt. Lett. 29, 1153 (2004).
13. A. Dogariu, A. Nahata, R. A. Linke, L. J. Wang, and R.
Trebino, Appl. Phys. B 74, S69 (2002).
14. V. N. Kumar and D. N. Rao, J. Opt. Soc. Am. B 12, 1559
(1995).
15. L. Lepetit, G. Chériaux, and M. Joffre, J. Opt. Soc. Am.
B 12, 2467 (1995).