See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/222714973

Spectral resolution and sampling issues in Fourier-transform spectral

interferometry

Article in Journal of the Optical Society of America B · October 2000

DOI: 10.1364/JOSAB.17.001795

CITATIONS READS
394 2,558

4 authors, including:

Christophe Dorrer Nadia Belabas

University of Rochester French National Centre for Scientific Research
242 PUBLICATIONS 5,818 CITATIONS 130 PUBLICATIONS 1,592 CITATIONS

SEE PROFILE SEE PROFILE

Manuel Joffre
École Polytechnique
174 PUBLICATIONS 5,344 CITATIONS

SEE PROFILE

All content following this page was uploaded by Nadia Belabas on 11 March 2014.

The user has requested enhancement of the downloaded file.

Dorrer et al. Vol. 17, No. 10 / October 2000 / J. Opt. Soc. Am. B 1795

Spectral resolution and sampling issues in

Fourier-transform spectral interferometry

Christophe Dorrer, Nadia Belabas, Jean-Pierre Likforman, and Manuel Joffre

Laboratoire d’Optique Appliquée, École Nationale Supérieure des Techniques Avancées, Ecole Polytechnique,
Centre National de la Recherche Scientifique, Unité Mixte de Recherche 7639, F-91761 Palaiseau Cedex, France

Received January 18, 2000; revised manuscript received May 12, 2000
We investigate experimental limitations in the accuracy of Fourier-transform spectral interferometry, a widely
used technique for determining the spectral phase difference between two light beams consisting of, for ex-
ample, femtosecond light pulses. We demonstrate that the spectrometer’s finite spectral resolution, pixel
aliasing, and frequency-interpolation error can play an important role, and we provide a new and more accu-
rate recipe for recovering the spectral phase from the experimental data. © 2000 Optical Society of America
[S0740-3224(00)00109-0]
OCIS codes: 320.7120, 320.7100, 120.3180, 120.5050

1. INTRODUCTION length calibration of the spectrometer.11 Indeed, it was

shown that a calibration accuracy better than one-tenth
Spectral interferometry, a technique relying on the use of
of the spacing between two pixels is often required to
frequency-domain interferences between two beams of
achieve the best possible accuracy in the measured spec-
different optical paths,1,2 has been shown in recent years
tral phase. In this paper we address two other experi-
to be of great use in femtosecond spectroscopy.3–11 In-
mental limitations affecting the reliability of spectral in-
deed, spectral interferometry allows the retrieval, in a
terferometry: spectral resolution and frequency
simple way, of the difference in spectral phase between
sampling. Both can result in phase measurement distor-
two time-delayed light pulses. This makes possible the
tion when not properly taken into account.
measurement of the complex transfer function of any lin-
In Section 2 we review one of the most common imple-
ear optical element by use of a broadband light source
mentations of spectral interferometry, also known as
such as a femtosecond laser or an incoherent white lamp. Fourier-transform spectral interferometry (FTSI), which
It also allows the full characterization of the electric field allows the retrieval of the spectral phase from a few Fou-
of an unknown pulse, assuming that a well-characterized rier transforms of the experimental interference spec-
reference pulse of appropriate spectrum is available. Be- trum. In Section 3 we address the problem of spectral
cause the measured quantity is linear in the electric field data sampling: In spectrometers, data are usually avail-
of the unknown pulse, this technique is much more sen- able as an array of points evenly spaced in the wave-
sitive than its nonlinear counterparts12–15 and can be length domain, while available fast Fourier transform
used for extremely weak pulses, which is one of the main (FFT) algorithms are most efficient when the data points
reasons for its widespread use in femtosecond spectros- are evenly spaced in the frequency domain. In Section 4
copy. A complete measurement of the electric field thus we discuss limitations arising from the finite spectral
allowed the transposition of two-dimensional nuclear resolution of the spectrometer, as well as from the use of a
magnetic resonance to the optical domain,16,17 as well as detector made of a finite number of pixels. We will show
the time-resolved measurement of photon-echo that such effects can be carefully characterized and in
emissions.18–21 Spectral interferometry has also been most cases corrected for. Finally, we propose in Section 5
used for measuring the linear dispersion of materials,22 an improved FTSI procedure, which relies on the same ex-
for characterizing the complex dielectric function of semi- perimental scheme but involves more careful data pro-
conductor nanostructures,23 and for discriminating be- cessing.
tween coherent and incoherent radiation in secondary
emission from semiconductor quantum wells.24,25 Fi-
nally, spectral interferometry is a key ingredient in a re- 2. FOURIER-TRANSFORM SPECTRAL
cent nonlinear pulse-measurement technique, known as INTERFEROMETRY
spectral phase interferometry for direct electric field re- In this section we describe how FTSI permits the retrieval
construction. This very efficient and noniterative tech- of the difference in spectral phase between two light
nique makes use of spectral interferences between two pulses from their interference spectrum. Let us call
frequency-sheared replicas of the unknown pulse.15,26–29 E 0 (t) and E(t) the time dependence of the two electric
Despite the widespread use of spectral interferometry, fields, E 0 ( ␻ ) and E( ␻ ) their Fourier transforms, and
there have been few detailed studies up to now on its re- ⌬ ␸ ( ␻ ) ⫽ arg关E(␻)兴 ⫺ arg关E0(␻)兴 the difference in spectral
liability, with the exception of a recent work demonstrat- phase that we intend to measure. In a typical spectral
ing the large sensitivity of the retrieved data on the wave- interferometry experiment, a relative time delay ␶ is in-

0740-3224/2000/101795-08$15.00 © 2000 Optical Society of America

1796 J. Opt. Soc. Am. B / Vol. 17, No. 10 / October 2000 Dorrer et al.

troduced between the two beams, which are then recom- direct measurement of the correlation function f(t) with
bined collinearly with a beam splitter. The total electric time-domain interferometry, also known as dispersive
field, E 0 (t) ⫹ E(t ⫺ ␶ ), is then spectrally resolved with a Fourier-transform spectroscopy.31 However, the advan-
spectrometer and a CCD detector. The total frequency tage of FTSI lies in the multichannel detection of the
spectrum thus reads whole data by use of CCD detectors, which makes the in-
terferometric requirements less difficult to fulfill and the
I 共 ␻ 兲 ⫽ 兩 E 0 共 ␻ 兲 ⫹ E 共 ␻ 兲 exp共 i ␻ ␶ 兲 兩 2
technique more practical than a scanning measurement
⫽ 兩 E 0 共 ␻ 兲 兩 2 ⫹ 兩 E 共 ␻ 兲 兩 2 ⫹ E 0* 共 ␻ 兲 E 共 ␻ 兲 of the correlation function.
However, actual detectors never provide directly the
⫻ exp共 i ␻ ␶ 兲 ⫹ c.c., (1) power spectrum I( ␻ ). Obviously, the signal is always
where c.c. holds for the complex conjugate of its preceding spoiled with some amount of electronic and photon noise,
term. The last two terms result in spectral interferences an effect of minor importance that is discussed in Appen-
through a term in cos关⌬␸ (␻) ⫹ ␻␶兴, causing a rapidly os- dix A. More important, the measured signal is not I( ␻ )
cillating frequency dependence. but an array of data points related to I( ␻ ) through the ap-
The interference pattern therefore strongly depends on paratus function. Taking this apparatus function into
the spectral phase difference, although an experimental account turns out to be of particular importance in the
measurement of the power spectrum yields only the phase case of spectral interferometry, as will be shown in Sec-
cosine. However, there are a number of ways for retriev- tion 4.
ing the phase from its cosine, e.g., by use of polarization To demonstrate experimentally the incidence of the ap-
multiplexing.6 We are here interested in the technique paratus function, we used a homemade Ti:sapphire oscil-
that uses Fourier transforms,6,7 or FTSI, which we briefly lator that delivers pulses of duration ranging between 20
review below. and 50 fs, depending on the operating conditions. A se-
Let us call f(t) ⫽ E 0* (⫺t) 丢 E(t) the correlation product quence of two nearly identical pulses is obtained with a
between the two fields. The power spectrum then reads balanced Michelson interferometer; the time delay be-
tween the two pulses is controlled with a step motor.
I 共 ␻ 兲 ⫽ 兩 E 0 共 ␻ 兲 兩 2 ⫹ 兩 E 共 ␻ 兲 兩 2 ⫹ f 共 ␻ 兲 exp共 i ␻ ␶ 兲 ⫹ c.c. (2) Therefore, in the following, E(t) ⫽ E 0 (t) and the mea-
sured spectral phase ⌬␸(␻) should reflect only the spec-
Note that f(␻) ⫽ F.T.f(t) ⫽ E 0* (␻)E(␻) ⫽ 兩 E 0* ( ␻ )E( ␻ ) 兩
tral dispersion of the interferometer. The interference
⫻ exp关i⌬␸(␻)兴 carries all the information on the spectral
spectra are recorded with a Jobin–Yvon HR-460 spec-
phase difference ⌬ ␸ ( ␻ ) ⫽ arg关 f (␻)兴. Therefore extract-
trometer followed by an EG&G 1024 ⫻ 256 CCD detector.
ing f( ␻ ) from the other terms in Eq. (2) will fulfill our
Note that in this particular set of experiments, which is
purpose. This can be achieved by Fourier transforming
intended only to demonstrate the limitations of spectral
Eq. (2):
interferometry rather than actually to measure the elec-
F.T.⫺1 I 共 ␻ 兲 ⫽ E 0* 共 ⫺t 兲 丢 E 0 共 t 兲 ⫹ E * 共 ⫺t 兲 丢 E共 t 兲 tric field, it was not required to characterize the spectral
phase of E 0 ( ␻ ), since it cancels out in the measured spec-
⫹ f 共 t ⫺ ␶ 兲 ⫹ f 共 ⫺t ⫺ ␶ 兲 * . (3) tral phase difference ⌬␸(␻). For the same reason, iden-
f(t⫺␶ ) is centered on t ⫽ ␶ , while the last term is cen- tical experimental results would have been obtained if an
tered on t ⫽ ⫺␶ . The first two terms, autocorrelation incoherent white lamp had been used instead of a femto-
functions of the individual fields, are centered at t ⫽ 0. second laser.
Therefore, for reasonably well-behaved pulses and for
large enough values of ␶, f(t) does not overlap with the
other terms in Eq. (3) and can be easily extracted from the 3. FREQUENCY SAMPLING
interference spectrum.30 Note that the first two terms In this section we address the issue of frequency sam-
can also be directly subtracted off in the frequency do- pling. We will neglect here the finite spectral resolution
main if two additional measurements are made while one of the spectrometer, as such issues will be discussed in
of the two beams is blocked; thus the noninterfering parts Section 4. Let us call x the spatial coordinate in the de-
are subtracted. This allows the use of smaller values of tector plane along which the spectrum dispersion occurs.
the time delay ␶, which will be shown in the next sections Although x is nearly proportional to wavelength in most
to be a desirable feature. spectrometers, this is never exactly the case, so we prefer
To summarize, FTSI relies on a few simple steps: An to use a general calibration function ␻ (x) that relates the
inverse Fourier transform of the interference spectrum, frequency ␻ to the spatial coordinate x. We develop this
followed by a selection of a finite time window so as to calibration law with respect to frequency around the laser
keep only the correlation product between the two fields. center frequency, ␻ 0 :
The time delay must be adjusted so that this truncation is
made possible. A Fourier transform back into the fre- ␻共 x 兲 ⫽ ␻ 0 ⫹ ␣ 1x ⫹ 1
␣ 2x 2 ⫹ 1
␣ 3x 3 ⫹ ¯ . (4)
2 6
quency domain then allows the retrieval of f( ␻ )
⫽ E 0* ( ␻ )E( ␻ ) and the spectral phase difference ⌬ ␸ ( ␻ ) As a result of this nonlinear dependence of ␻ versus x,
⫽ arg关 f (␻)兴. In cases in which E 0 ( ␻ ) has been indepen- the frequency values for which the signal is sampled, ␻ i ,
dently measured with nonlinear phase measurement are not evenly spaced, since the detector pixels are evenly
techniques, this allows the determination of E( ␻ ) and spaced in x. Because x is roughly proportional to 1/␻, the
hence E(t) after an inverse Fourier transform. Note that nonlinear terms in Eq. (4) are usually not small in femto-
the same information could have been obtained through a second experiments in which the spectral extent is quite
Dorrer et al. Vol. 17, No. 10 / October 2000 / J. Opt. Soc. Am. B 1797

large. This causes changes in the frequency step, ␻ i⫹1 count the nonlinearity in the calibration law is different
⫺ ␻ i , by as much as ⫾20% over the spectral range of our from the actual temporal shape of f(t). This is especially
spectrometer. The noneven frequency sampling of the true for shorter pulses, for which the frequency-step
data might be thought to preclude the use of the ex- variation from one end of the spectrum to the other is
tremely efficient Cooley–Tukey FFT algorithm, thus mak- greater.
ing the FTSI spectral phase retrieval much more time
consuming. In the following, however, we will show that B. Discrete Fourier Transform
the FFT can still be used. One approach to account for the discrepancy reported in
Subsection 3.A is to use a Fourier-transform algorithm
A. Plain Fast Fourier Transform of the Data that can handle nonevenly spaced data points, such as the
Let us first consider what happens when we ignore the discrete Fourier transform. This will obviously yield the
nonlinear calibration law and simply proceed in comput- correct answer; however, none of these algorithms will be
ing the FFT of the experimental data array 兵 I( ␻ i ) 其 . We as efficient as the Cooley–Tukey FFT in terms of comput-
obtain an array 兵 I(k i ) 其 that actually corresponds to the ing time. We will therefore attempt to use other tech-
Fourier transform of 兵 I(x i ) 其 , where k is the spatial fre- niques in the following, in order to obtain the correct an-
quency, swer more efficiently.

I 共 k 兲 ⫽ F.T.⫺1 I 共 x 兲 ⫽ 冕 I 共 x 兲 exp共 ⫺ikx 兲 dx C. Data Interpolation

The most straightforward approach for using the Cooley–

⫽ N.I.T. ⫹ 冕 f 关 ␻ 共 x 兲兴 exp关 i ␻ 共 x 兲 ␶ 兴
Tukey FFT algorithm, despite an uneven spacing of the
data points, is first to interpolate the experimental data
so as to numerically generate an array of points regularly
⫻ exp共 ⫺ikx 兲 dx ⫹ c.c., (5)
where N.I.T. stands for noninterferometric terms, which
do not depend on ␶. The result is plotted in Fig. 1 as a
function of ␰ ⫽ k/ ␣ 1 for three different values of the time
delay between the two pulses. Note that if we were to
neglect the nonlinear terms in Eq. (4), ␰ would be the ex-
act Fourier conjugate of ␻, i.e., the time t. Indeed, we ob-
serve that the data shown in Fig. 1 peak at ␰ ⫽ ␶ and ␰
⫽ ⫺␶ . However, the correlation peak is not simply
translated in time as would be expected for f(t ⫺ ␶ ) but
also broadens when ␶ increases. This can be easily ex-
plained by taking into account the calibration law,

I 共 k 兲 ⫽ N.I.T. ⫹ 冕 f 关 ␻ 共 x 兲兴 exp关 i ␻ 共 x 兲 ␶ 兴

⫻ exp共 ⫺ikx 兲 dx ⫹ c.c.

Fig. 1. Magnitude of the fast Fourier transform of the experi-
⫽ N.I.T. ⫹ exp共 i ␻ 0 ␶ 兲 (F.T.⫺1 兵 f 关 ␻ 共 x 兲兴 mental interference spectrum, 兩 I(k) 兩 , plotted as a function of ␰
⫽ k/ ␣ 1 , for three different values of the time delay.
⫻ exp关 i⌽ ␶ 共 x 兲兴 其 )共 ␰ ⫺ ␶ 兲 ⫹ 关 c.c.兴共 ⫺␰ ⫺ ␶ 兲
⫽ N.I.T. ⫹ exp共 i ␻ 0 ␶ 兲 f ␶ 共 ␰ ⫺ ␶ 兲
⫹ exp共 ⫺i ␻ 0 ␶ 兲 f ␶ 共 ⫺␰ ⫺ ␶ 兲 * , (6)

where ⌽ ␶ (x) ⫽ 21 ␣ 2 ␶ x 2 ⫹ 61 ␣ 3 ␶ x 3 ⫹ ... is a phase factor

resulting from the nonlinear terms in the calibration law.
f ␶( ␰ ) is the inverse Fourier transform of
f 关 ␻ (x) 兴 exp关i⌽␶(x)兴. f 关 ␻ (x) 兴 with respect to x has a shape
similar to f( ␻ ) with respect to ␻ and does not significantly
change the Fourier transform time width. In contrast,
the phase factor ⌽ ␶ (x) gives the main contribution to the
broadening in f ␶ ( ␰ ) observed in Fig. 1. Keeping only the
quadratic term in ⌽ ␶ (x), proportional to ␣ 2 ␶ , we find that
the broadening can be essentially interpreted as an arti-
ficial linear chirp in the pulse. This yields the asymmet-
ric shape observed in Fig. 1, in which the spectral shape
of our laser pulses can be recognized for large values of Fig. 2. Fourier transform of the same experimental data as
those used in Fig. 1, except that a linear interpolation of the fre-
the time delay.32 quency axis has first been performed to provide the FFT proce-
Not surprisingly, we conclude that the result obtained dure with an array of evenly spaced data points in frequency do-
in the time domain when there is a failure to take into ac- main.
1798 J. Opt. Soc. Am. B / Vol. 17, No. 10 / October 2000 Dorrer et al.

spaced in the frequency domain. Figure 2 shows the re- It is then straightforward to retrieve f( ␻ ) after we sub-
sult obtained with a linear interpolation of the same data tract the phase ␻ (x) ␶ . Note, however, that since this lat-
as those used in Fig. 1. Although a sharp peak is then ter term does not vary linearly with x, it is important to
observed, in contrast with Fig. 1, a superimposed back- take into account the exact calibration law ␻ (x). This
ground now appears whose magnitude dramatically in- approach does allow us to get rid of the background
creases with increasing values of the time delay (dashed shown in Fig. 2 that resulted from the interpolation
area). Although such a feature remains small, it does scheme.
significantly affect the quality of the spectral phase thus Figure 3 shows the spectral phase retrieved by use of
retrieved. As demonstrated in more detail in Appendix the various techniques discussed above for a time delay
B, the observed background is a direct consequence of the between the two pulses set to 5 ps. Curve (a), obtained
error resulting from linearly interpolating the experimen- by ignoring the nonlinear dependence of the calibration
tal data. This error is most important for large values of law, exhibits a large parabolic spectral phase, directly re-
the time delay, owing to the rapid frequency oscillations flecting the first nonlinear term in the calibration law.
of the spectral interferogram. This large quadratic phase is consistent with the broad-
It might be claimed that a more elaborate interpolation ening observed in Fig. 1. Curve (b) shows the result ob-
scheme would improve the result. However, aiming at tained by performing, prior to the FFT, a linear interpo-
pushing the technique to its limits, we would like to be lation in the frequency axis, as discussed in Subsection
able to use time delays as great as the Nyquist limit, as 3.C. Although the retrieved phase is more accurate, it
will be discussed in Section 4. This means that the oscil- exhibits strong oscillations that are due to the interpola-
lation period can be as small as two pixels. In such a tion error. Such oscillations around the exact value of
case, any local interpolation scheme such as cubic spline the phase are due to the fact that the error in the linear
is bound to fail and would not provide satisfactory results. interpolation of the cosine function between two points is
There is a global interpolation scheme that does work, dependent on their position. Indeed, let us consider the
however, known as zero filling. This technique consists interpolation on evenly spaced points in the frequency do-
in first performing a FFT of the data to ␰ space, then in- main of the function cos关␻␶ ⫹ ⌬␸ (␻)兴 recorded on points
creasing the ␰-window size, e.g., to 4N or 8N, where N is roughly evenly spaced in the wavelength domain. A
the number of detector pixels, filling the new data points negative, zero, or positive error is obtained, thus giving a
with zeroes. A FFT back into x space yields an array periodic-like structure. The local period is varying be-
with a finer sampling, now making possible a proper in- cause the wavelength interval associated to a fixed spec-
terpolation of the data. Although this scheme works and tral interval depends on the wavelength. In contrast,
uses only FFT’s, it requires larger arrays to handle. We this oscillating noise is totally absent in curve (c), which
will show in Subsection 3.D that similar results can be ob- has been obtained with the approach discussed in this
tained with only arrays of the same size as the number of subsection. This result is exactly identical to that of the
pixels on the detector. zero-filling method (d), despite the smaller number of
points used in the calculation. Note that the residual
D. Retrieving the Spectral Phase in a First Step spectral phase observed here results from the dispersion
The approach we propose here consists of retrieving the of the interferometer used in these experiments.
spectral phase with the ␰ domain instead of the time do- Finally, to retrieve the electric field in the time domain,
main. We will show below that such a method is possible we need to perform a Fourier transform toward the true
and that once the spectral phase is retrieved, data inter- time domain t, instead of ␰. Fortunately, in most cases
polation will be made easier, allowing the retrieval, as a the amplitude and phase of the unknown electric field
last step, of the electric field as a function of time. vary slowly with frequency, unlike the spectral interfero-
Let us first note the similarity between Eq. (2) and Eq.
(6). In both cases, we have a sum of a few terms centered
on 0 and ⫾␶, either in t space or in ␰ space. As is evident
in Fig. 1, although there is a broadening, the relevant
term can still be extracted in ␰ space. Indeed, the broad-
ening mentioned in Subsection 3.A can be explained by
the fact that a given value of ␶ does not yield a unique ␰
for all frequency components, as d␻ /dx is equal to ␣ 1 only
at the center of the spectrum, ␻ 0 . Therefore this broad-
ening cannot exceed a fixed fraction of ␶, namely, the rela-
tive variation of the frequency spacing over the spectrum.
As a consequence, such a broadening cannot cause the
overlap between components separated by ␶. Thus,
choosing a value of ␶ so that the relevant term can be ex-
tracted, we obtain, after a FFT back to x space:

Fig. 3. Spectral phase obtained from the interference spectrum

exp共 i ␻ 0 ␶ 兲 f 关 ␻ 共 x 兲兴 exp关 i⌽ 共 x 兲兴 exp共 i ␣ 1 x ␶ 兲 between two pulses separated by 5 ps. The phase-retrieval tech-
niques used are (a) plain FFT, (b) linear interpolation, (c) the
technique described in Subsection 3.D, and (d) zero-filling inter-
⫽ f 关 ␻ 共 x 兲兴 exp关 i ␻ 共 x 兲 ␶ 兴 . (7) polation. The curves have been vertically shifted for clarity.
Dorrer et al. Vol. 17, No. 10 / October 2000 / J. Opt. Soc. Am. B 1799

of the time delay ␶. Furthermore, we will show that such

effects can be accounted for after the spectrometer’s appa-
ratus function has been carefully measured, a task for
which interference spectra have been shown to be particu-
larly useful.33,34
We assume that the spectrometer response can be ap-
proximated to a convolution with a response function
R(x), so that the spatial dependence of the intensity in
the detector plane is R(x) 丢 I 关 ␻ (x) 兴 . R(x) depends on
the spectrometer characteristics, such as focal length, dif-
fraction grating, and entrance-slit width. Furthermore,
when the laser beam is not highly diffracted by the en-
trance slit, such as when we deal with low-energy pulses
for which no loss can be afforded, R(x) may also depend
on the laser spatial profile within the slit area. This light
Fig. 4. Time-domain determination of the correlation function, intensity in the detector plane is then integrated over the
f(t ⫺ ␶ ), by use of (a) a plain FFT of the data, (b) a linear inter- pixel area, yielding the following expression for the sig-
polation before the FFT, (c) the technique described in Subsec- nal, S i , collected on a given pixel at position x i ,
tion 3.D, and (d) the zero-filling method. (c) and (d) cannot be dis-
tinguished because the difference between the two curves is
within the line thickness.

grams we started from. This is true in many cases such

Si ⫽ 冕 x i ⫹a/2

x i ⫺a/2
R共 x 兲 丢 I 关 ␻ 共 x 兲兴 dx

as for a short pulse, a highly chirped pulse for which the

phase variation is dominated by lower-order terms, four-
wave mixing emission, etc. In such cases, it is straight-
⫽ 冕⫺⬁
⫹⬁
P共 x ⫺ x i 兲兵R共 x 兲 丢 I 关 ␻ 共 x 兲兴 其 dx

forward to interpolate linearly 兩 E( ␻ ) 兩 and ␸(␻) so that a ⫽ 兵P共 x 兲 丢 R共 x 兲 丢 I 关 ␻ 共 x 兲兴 其 共 x i 兲 , (8)

FFT can be performed on the interpolated points, which
are now evenly sampled. The result is plotted in Fig. 4 where P(x) is a rectangle function taking the value 1 for
and compared with the other techniques. It appears that 兩 x 兩 ⬍ a/2, a being the pixel width. If we now take into
this approach performs much better than the linear- account the fact that the signal is sampled for discrete
interpolation technique, as the background is reduced by values of the spatial coordinate, x ⫽ x i , we find that the
one order of magnitude. Furthermore, on these experi- actual function we can experimentally access is
mental data, the result of this technique cannot be distin-
guished from that of the a priori more exact zero-filling S共 x 兲 ⫽ ⌸共 x 兲兵P共 x 兲 丢 R共 x 兲 丢 I 关 ␻ 共 x 兲兴 其 , (9)
method.
We must mention that there are some pulses for which where ⌸(x) is a Dirac comb of period ␦ x, the pixel spac-
the technique described in this section would fail to yield ing. Note that ␦ x ⭓ a. The Fourier transform in ␰
the same accuracy as the zero-filling method (for example, space reads
when the unknown pulse is a sequence of two pulses sepa-
rated by several picoseconds). Then the spectral ampli- S共 ␰ 兲 ⫽ ⌸共 ␰ 兲 丢 (P 共 ␰ 兲 R 共 ␰ 兲 T.F.⫺1兵 I 关 ␻ 共 x 兲兴 其 ), (10)
tude itself oscillates rapidly with frequency, so that the
interpolation required at the latest stage results in sig-
nificant errors. In such a case, one would have to resort
to the zero-filling technique, as described at the end of
Subsection 3.C.

4. SPECTRAL RESOLUTION AND

ALIASING
In spectral interferometry, larger values of the time delay
result in a smaller fringe spacing, hence in a reduced
fringe contrast that is due to the finite spectral resolution
of the spectrometer. This is illustrated, for example, in
Fig. 5(a) in which the spectral fringes almost vanish for a
time delay of 8 ps. One must therefore compromise when
choosing ␶, which must be small enough so that this effect
is not too important but large enough so as to make pos-
sible the extraction of the correlation product from the
Fig. 5. (a) Blow-up of a particular spectral region of the inter-
Fourier transform of the interferogram. In this section
ference spectra obtained for ␶ ⫽ 3 ps (lower curve) and ␶ ⫽ 9 ps
we discuss the incidence of the spectrometer’s finite spec- (upper curve). (b) Amplitude of the FFT of the above data, plot-
tral resolution and of the detector’s finite number of pix- ted as a function of ␰. The curve corresponding to ␶ ⫽ 9 ps has
els, which will be obviously most evident for large values been multiplied by a factor of 10.
1800 J. Opt. Soc. Am. B / Vol. 17, No. 10 / October 2000 Dorrer et al.

where ⌸(␰) is a Dirac comb of period T ⫽ 2␲/( ␣ 1 ␦ x), or 16

ps for our setup. The convolution with this Dirac comb
results in a folding within the Nyquist window, a phe-
nomenon also known as aliasing. This is illustrated in
Fig. 5(a), which shows the spectral interferograms for two
values of the time delay, ␶ ⫽ 3 ps and ␶ ⫽ 9 ps. In the
former case, in which the delay is significantly smaller
than T/2, the fringes are properly sampled. In contrast,
the latter case corresponds to a time delay of the order of
T/2, which yields a spectrum highly undersampled. Fig-
ure 5(b) shows the FFT of these spectra in ␰ space. For
␶ ⬇ T/2, half of the broadened pulse is actually folded in
the Nyquist window, i.e., shifted by T. The part of
f ␶ (⫺␰ ⫺ ␶ ) * , where t ⬍ ⫺T therefore interferes with the
part of f ␶ ( ␰ ⫺ ␶ ) for which t ⬍ T, resulting in the ob- Fig. 6. (a) FFT computed from a series of experimental interfer-
served time-domain fringes (b) and also in a characteris- ence spectra obtained for different values of ␶. (b) FFT com-
tic beating in the frequency domain (a). puted from a series of numerically computed interference spectra
A situation in which ␶ ⬎ T/2 should therefore be obtained for different values of ␶ and by use of the experimental
laser spectrum.
avoided. More precisely, the condition max(␦␻␶) ⬍ 1/2
must be fulfilled to avoid the occurrence of any aliasing,
where ␦␻ is the (nonconstant) frequency separation be-
tween two adjacent pixels. However, even when this is
the case, the signal will still be distorted through the mul-
tiplication by P( ␰ )R( ␰ ). This term, characterizing the
spectral resolution of our setup, must now be measured.
For our purpose, one may use either an atomic narrow
spectral line or spectral interferometry itself. In the first
case, I( ␻ ) can be approximated to a Dirac distribution, so
that the FFT of the spectrum yields P( ␰ )R( ␰ ), or rather
its aliased version, ⌸( ␰ ) 丢 关 P( ␰ )R( ␰ ) 兴 . In the second ap-
proach, we record a series of interference spectra for dif-
ferent values of the time delays.35 The spectral resolu-
tion can clearly be deduced from the decrease in the
fringe contrast. More precisely, we plot on Fig. 6(a) the Fig. 7. ␰-domain apparatus function of the spectrometer ob-
FFT of the data, which shows a decrease of the signal as ␶ tained with a narrow spectral line (thin solid curve) or with spec-
increases. However, as was mentioned in Section 3, a tral interferences (dashed curve). The thick solid curve shows
large part of this decay results from the uneven sampling the aliased apparatus function deduced after periodization (i.e.,
after adding the dotted curve), thus simulating the convolution
of the data. More quantitatively, the signal at the peak
product with P( ␰ ). The apparatus function that should be used
for a given value of ␶ reads P( ␶ )R( ␶ )f ␶ (0), where f ␶ (0) for correction of interference spectra is the dashed curve.
⬍ f(0) owing to the broadening resulting from ⌽ ␶ (x).
This can be taken into account by numerically generating
interference spectra from the experimental laser spec- 5. IMPROVED FOURIER-TRANSFORM
trum, i.e., multiplying the experimental 兩 E 0 关 ␻ (x) 兴 兩 2 by SPECTRAL INTERFEROMETRY
cos关␻(x)␶兴 and computing the Fourier transform for the SCHEME AND CONCLUSION
experimental values of ␶. The decay thus observed in To summarize, we have investigated some instrumental
Fig. 6(b) is now entirely due to f ␶ (0), since the calculation limitations in FTSI, which, to our knowledge, have not
was not limited by the spectrometer resolution. By divid- been reported up to now. Our results lead us to propose
ing the maxima of Fig. 6(a) by those of Fig. 6(b), we obtain an improved FTSI procedure that allows a partial com-
P( ␶ )R( ␶ ) for several values of ␶, from which we can de- pensation for instrumental limitations. First, the spec-
duce the entire response function through interpolation, tral calibration must be performed with great care, fol-
as this function is slowly varying with ␰. lowing the technique reported previously.11 The spectral
To compare these two approaches, we plot in Fig. 7 the resolution of the spectrometer should then be character-
function P( ␰ )R( ␰ ) obtained with this technique, which we ized with the technique described in Section 4, yielding
compare with the data derived from a narrow spectral the product P( ␰ )R( ␰ ). After data acquisition, the spec-
line. Provided that we add the contribution from the in- tral interferograms are Fourier transformed into ␰ space
tervals 关 ⫺3T/2,⫺T/2兴 and 关 T/2, 3T/2兴 to the data ob- by use of a Cooley–Tukey FFT, where the data can be di-
tained with spectral interferometry, we reach a good vided by P( ␰ )R( ␰ ). After truncation, a FFT back into the
agreement between the two techniques, except for small wavelength domain allows recovery of the spectral phase
values of ␰ for which spectral interferometry is not valid. difference between the two pulses. Finally, the data can
However, only the second approach yields the unaliased be obtained in time domain after proper interpolation so
P( ␰ )R( ␰ ), which can then be directly used to correct the as to generate an array of evenly spaced frequencies, as
experimental interference spectra. described in Subsection 3.D.
Dorrer et al. Vol. 17, No. 10 / October 2000 / J. Opt. Soc. Am. B 1801

For some implementations of FTSI (for example, in traction of the phase, as discussed in Subsection 3.C. In
spectral phase interferometry for direct electric field re- this case, the interferometric part in the measured spec-
construction), the extent of the time-dependent correla- trum reads
tion product is sufficiently small so that the response of
g 共 ␻ 兲 ⫽ 兩 E 0 共 ␻ 兲 兩兩 E 共 ␻ 兲 兩 cos关 ␻ ␶ ⫹ ⌬ ␸ 共 ␻ 兲兴 . (B3)
the spectrometer does not significantly modify this quan-
tity. In such cases, there is no need to correct data in the We assume here that the variation of g( ␻ ) is due only
␰ domain, which saves one step in the above procedure. to the cosine term in the interval in which interpolation is
performed, i.e., that the spectrum does not change signifi-
cantly in this small interval. In this case, the second de-
APPENDIX A: INFLUENCE OF rivative of g( ␻ ) reads
EXPERIMENTAL NOISE
⳵ 2g ⳵ 2 cos关 ␻ ␶ ⫹ ⌬ ␸ 共 ␻ 兲兴
In this appendix we discuss the influence of experimental ⬇ 兩 E 0 共 ␻ 兲 兩兩 E 共 ␻ 兲 兩
noise on the spectral phase retrieved with FTSI. We as- ⳵␻ 2 ⳵␻ 2

再冉 冊
sume that a frequency-dependent noise, N( ␻ ), is added to
⳵␸ 2
the spectral intensity, I( ␻ ), so that the total detected sig- ⫽ ⫺兩 E 0 共 ␻ 兲 兩兩 E 共 ␻ 兲 兩 ⫹␶ cos关 ␻ ␶ ⫹ ⌬ ␸ 共 ␻ 兲兴
nal reads ⳵␻
I 共 ␻ 兲 ⫹ N 共 ␻ 兲 ⫽ 兩 E 0共 ␻ 兲兩 2 ⫹ 兩 E 共 ␻ 兲兩 2 ⫹ f 共 ␻ 兲
⫻ exp共 i ␻ ␶ 兲 ⫹ c.c. ⫹ N 共 ␻ 兲 . (A1)
⫹
⳵ 2␸
⳵␻ 2
sin关 ␻ ␶ ⫹ ⌬ ␸ 共 ␻ 兲兴 . 冎 (B4)

Applying FTSI, as defined in Section 2, we compute the Owing to the large value of the delay ␶, the first term in
inverse Fourier transform of the above expression and the above sum is dominant. This shows that on a point ␻
multiply by a window function, H(t), to extract f(t ⫺ ␶ ). at which the interpolated interferogram is obtained be-
Assuming that H(t) does not overlap with the nonrel- tween points ␻ a and ␻ b , the error is
evant terms and that it is equal to 1 when f(t ⫺ ␶ ) is non- N 共 ␻ 兲 ⫽ ⫺兩 E 0 共 ␻ 兲 兩兩 E 共 ␻ 兲 兩 cos关 ␻ ␶ ⫹ ⌬ ␸ 共 ␻ 兲兴
zero, we obtain f(t ⫺ ␶ ) ⫹ H(t)N(t). Finally, a Fourier
transform yields f( ␻ ) ⫹ F.T.关 H(t ⫹ ␶ )N(t ⫹ ␶ ) 兴 . For a
given value of the frequency ␻, let us write this expres-
sion as a exp(i⌬␸) ⫹ b exp(i␪). The error on the extracted
⫻ 冉 ⳵␸
⳵␻
⫹␶ 冊 2
k共 ␻ 兲. (B5)

spectral phase is then arg关a exp(i⌬␸) ⫹ b exp(i␪)兴 It is the product of the interferometric part by the func-
⫺ arg关a exp(i⌬␸)兴, which simplifies to arg兵1 ⫹ (b/a) tion L( ␻ ) ⫽ ( ⳵␸ / ⳵␻ ⫹ ␶ ) 2 k( ␻ ). As a result, the correla-
⫻ exp关i(␪ ⫺ ⌬␸)兴其 and then to arctan兵(b/a)sin(␪ ⫺ ⌬␸)/ tion product in the temporal domain is the sum of the
关1 ⫹ (b/a)cos(␪ ⫺ ⌬␸)兴其. For a large value of the signal- error-free product and the convolution of this product by
to-noise ratio 兩 a/b 兩 , the phase error can be written as the function L(t). This explains why the superimposed
(b/a)sin(␪ ⫺ ⌬␸), whose magnitude is always smaller background that is due to the interpolation of the inter-
than 兩 b/a 兩 ⫽ 兩 H( ␻ ) 丢 N( ␻ ) 兩 / 兩 E 0 ( ␻ )E( ␻ ) 兩 . ferogram is moving with the correlation product when the
To summarize, we find that the noise in spectral phase delay is varied, as observed in Fig. 2. A direct conse-
is simply equal to the ratio of the filtered noise to the quence is that this noise cannot be filtered in the tempo-
product of the electric field spectral amplitudes. ral domain when the interferometric component is ex-
tracted. The level of this noise is roughly proportional to
␶ 2 . There is thus a dramatic parasitic effect as the delay
APPENDIX B: PHASE ERROR RESULTING between the two interfering pulses is increased. This er-
FROM LINEAR INTERPOLATION ror then leads to an error in the retrieved spectral phase,
In this appendix we compute the error in spectral phase whose magnitude can be deduced by applying Appendix A
resulting from linear interpolation in frequency of the ex- to the value of N( ␻ ) obtained in Eq. (B5). This noise is
perimental data. Let us first recall that for a function thus directly related to the function L(␻), and its level is
g( ␻ ) linearly interpolated on point ␻ between points ␻ a also roughly proportional to ␶ 2 .
and ␻ b , the interpolated value is In contrast, when the interpolation is performed on the
spectral phase according to the procedure discussed in
g inter共 ␻ 兲 ⫽ g 共 ␻ a 兲 ⫹ 关 g 共 ␻ b 兲 ⫺ g 共 ␻ a 兲兴 Subsection 3.D, the interpolation noise takes much
smaller values. Indeed, the error resulting from linearly
⫻ 共 ␻ ⫺ ␻a兲/共 ␻b ⫺ ␻a兲. (B1) interpolating the extracted spectral phase ␻ ␶ ⫹ ␸ ( ␻ ) is
The interpolation error then reads proportional to ⳵ 2 ␸ / ⳵␻ 2 k( ␻ ), according to relation (B2).
It is then independent of delay, since the linear interpo-
1 ⳵ 2g lation of a straight line does not introduce any error.
g inter共 ␻ 兲 ⫺ g 共 ␻ 兲 ⬇ ⫺ 共 ␻ ⫺ ␻ a 兲共 ␻ ⫺ ␻ b 兲
2 ⳵␻ 2
The corresponding author, M. Joffre, can be reached at
⳵ 2g the address on the title page, by fax at 33 1 69 31 99 96, or
⫽ k共 ␻ 兲 , (B2) by e-mail at manuel.joffre@polytechnique.fr.
⳵␻ 2
Note added in proof. It recently came to our attention
where k( ␻ ) ⫽ ⫺( ␻ ⫺ ␻ a )( ␻ ⫺ ␻ b )/2. that the use of filtering in the ␰ domain as developed in
Let us first consider the procedure consisting of the lin- Subsection 3.D of this paper has been also reported by Jo-
ear interpolation of the interferogram followed by the ex- nas and coworkers.36
 1802 J. Opt. Soc. Am. B / Vol. 17, No. 10 / October 2000 Dorrer et al.

REFERENCES AND NOTES M. Jonas, ‘‘Two-dimensional electronic spectroscopy,’’

Chem. Phys. Lett. 297, 307–313 (1998).
1. C. Froehly, A. Lacourt, and J. C. Vienot, ‘‘Notions de 18. J.-P. Likforman, M. Joffre, and V. Thierry-Mieg, ‘‘Measure-
réponse impulsionnelle et de fonction de transfert ment of photon echoes by use of Fourier-transform spectral
temporelles des pupilles optiques, justifications expérimen- interferometry,’’ Opt. Lett. 22, 1104–1106 (1997).
tales et applications,’’ J. Opt. (Paris) 4, 183–196 19. M. F. Emde, W. P. de Boeij, M. S. Pshenichnikov, and D. A.
(1973). Wiersma, ‘‘Spectral interferometry as an alternative to
2. J. Piasecki, B. Colombeau, M. Vampouille, C. Froehly, and time-domain heterodyning,’’ Opt. Lett. 22, 1338–1340
J. A. Arnaud, ‘‘Nouvelle méthode de mesure de la réponse (1997).
impulsionnelle des fibres optiques,’’ Appl. Opt. 19, 3749– 20. X. Chen, W. J. Walecki, O. Buccafusca, D. N. Fittinghoff,
3755 (1980). and A. L. Smirl, ‘‘Temporally and spectrally resolved ampli-
3. F. Reynaud, F. Salin, and A. Barthelemy, ‘‘Measurement of tude and phase of coherent four-wave-mixing emission from
phase shifts introduced by nonlinear optical phenomena on GaAs quantum wells,’’ Phys. Rev. B 56, 9738–9743 (1997).
subpicosecond pulses,’’ Opt. Lett. 14, 275–277 (1989). 21. A. L. Smirl, X. Chen, and O. Buccafusca, ‘‘Ultrafast time-
4. E. Tokunaga, A. Terasaki, and T. Kobayashi, ‘‘Induced resolved quantum beats in the polarization state of coher-
phase modulation of chirped continuum pulses studied with ent emission from quantum wells,’’ Opt. Lett. 23, 1120–
a femtosecond frequency-domain interferometer,’’ Opt. Lett. 1122 (1998).
18, 370–372 (1993). 22. V. N. Kumar and D. N. Rao, ‘‘Using interference in the fre-
5. J.-P. Geindre, P. Audebert, A. Rousse, F. Falliès, J. C. quency domain for precise determination of thickness and
Gauthier, A. Mysyrowicz, A. Dos Santos, G. Hamoniaux, refractive indices of normal dispersive materials,’’ J. Opt.
and A. Antonetti, ‘‘Frequency-domain interferometer for Soc. Am. B 12, 1559–1563 (1995).
measuring the phase and amplitude of a femtosecond pulse 23. J. Tignon, M. V. Marquezini, T. Hasche, and D. S. Chemla,
probing a laser-produced plasma,’’ Opt. Lett. 19, 1997–1999 ‘‘Spectral interferometry of semiconductor nanostructures,’’
(1994). IEEE J. Quantum Electron. 35, 510–522 (1999).
6. L. Lepetit, G. Chériaux, and M. Joffre, ‘‘Linear techniques 24. D. Birkedal and J. Shah, ‘‘Femtosecond spectral interferom-
of phase measurement by femtosecond spectral interferom- etry of resonant secondary emission from quantum wells:
etry for applications in spectroscopy,’’ J. Opt. Soc. Am. B 12, resonance Rayleigh scattering in the nonergodic regime,’’
2467–2474 (1995). Phys. Rev. Lett. 81, 2372–2375 (1998).
7. D. N. Fittinghoff, J. L. Bowie, J. N. Sweetser, R. T. Jen- 25. S. Haacke, S. Schaer, B. Deveaud, and V. Savona, ‘‘Inter-
nings, M. A. Krumbugel, K. W. DeLong, R. Trebino, and I. ferometric analysis of resonant Rayleigh scattering from
A. Walmsley, ‘‘Measurement of the intensity and phase of two-dimensional excitons,’’ Phys. Rev. B 61, R5109–R5112
ultraweak, ultrashort laser pulses,’’ Opt. Lett. 21, 884–886 (2000).
(1996); ‘‘Erratum,’’ 21, 1313 (1996). 26. C. Iaconis and I. A. Walmsley, ‘‘Self-referencing spectral in-
8. W. J. Walecki, D. N. Fittinghoff, A. L. Smirl, and R. terferometry for measuring ultrashort optical pulses,’’
Trebino, ‘‘Characterization of the polarization state of IEEE J. Quantum Electron. 35, 501–509 (1999).
weak ultrashort coherent signals by dual-channel 27. C. Dorrer, ‘‘Implementation of spectral phase interferom-
spectral interferometry,’’ Opt. Lett. 22, 81–83 etry for direct electric-field reconstruction using a simulta-
(1997). neously recorded reference interferogram,’’ Opt. Lett. 24,
9. S. M. Gallagher, A. W. Albrecht, J. D. Hybl, B. L. Landin, B. 1532–1534 (1999).
Rajaram, and D. M. Jonas, ‘‘Heterodyne detection of the 28. C. Dorrer, B. de Beauvoir, C. Le Blanc, S. Ranc, J.-P. Rous-
complete electric field of femtosecond four-wave mixing sig- seau, P. Rousseau, J. P. Chambaret, and F. Salin, ‘‘Single-
nals,’’ J. Opt. Soc. Am. B 15, 2338–2345 (1998). shot real-time characterization of chirped-pulse amplifica-
10. C. Dorrer and F. Salin, ‘‘Characterization of spectral phase tion systems using spectral phase interferometry for direct
modulation using classical and polarization spectral inter- electric-field reconstruction,’’ Opt. Lett. 24, 1644–1646
ferometry,’’ J. Opt. Soc. Am. B 15, 2331–2337 (1998). (1999).
11. C. Dorrer, ‘‘Influence of the calibration of the detector on 29. L. Gallmann, D. H. Sutter, N. Matuschek, G. Steinmeyer,
spectral interferometry,’’ J. Opt. Soc. Am. B 16, 1160–1168 U. Keller, C. Iaconis, and I. A. Walmsley, ‘‘Characterization
(1999). of sub-6-fs optical pulses with spectral phase interferom-
12. D. J. Kane and R. Trebino, ‘‘Single-shot measurement of the etry for direct electric-field reconstruction,’’ Opt. Lett. 24,
intensity and phase of an arbitrary ultrashort pulse by us- 1314–1316 (1999).
ing frequency-resolved optical gating,’’ Opt. Lett. 18, 823– 30. In cases such as four-wave mixing in which causality plays
825 (1993); R. Trebino and D. J. Kane, ‘‘Using phase re- a role, the emitted field is asymmetric in time, so that it is
trieval to measure the intensity and phase of ultrashort usually better to use positive values of ␶.
pulses: frequency-resolved optical gating,’’ J. Opt. Soc. 31. P. R. Griffiths and J. A. de Haseth, Fourier Transform In-
Am. A 10, 1101–1111 (1993). frared Spectrometry, Vol. 83 of Chemical Analysis (Wiley,
13. J. Paye, M. Ramaswamy, J. G. Fujimoto, and E. P. Ippen, New York, 1986).
‘‘Measurement of the amplitude and phase of ultrashort 32. Note that a similar explanation can be attributed to the
light pulses from spectrally resolved autocorrelation,’’ Opt. strong dependence of the retrieved phase on the exact
Lett. 18, 1946–1948 (1993); J. Paye, ‘‘How to measure the calibration.11 Even a small ␣ 2 in the calibration law will
amplitude and phase of an ultrashort light pulse with an result in a chirp proportional to time delay.
autocorrelator and a spectrometer,’’ IEEE J. Quantum 33. V. Coates and H. Hausdorff, ‘‘Interferometric methods of
Electron. 30, 2693–2697 (1994). measuring the spectral slit width of spectrometers,’’ J. Opt.
14. K. W. DeLong, R. Trebino, J. Hunter, and W. E. White, Soc. Am. 45, 425–430 (1955).
‘‘Frequency-resolved optical gating with the use of second- 34. V. Kumar and D. Rao, ‘‘Interferometric measurement of the
harmonic generation,’’ J. Opt. Soc. Am. B 11, 2206–2215 modulation transfer function of a spectrometer by using
(1994). spectral modulations,’’ Appl. Opt. 38, 660–665 (1999).
15. C. Iaconis and I. A. Walmsley, ‘‘Spectral phase interferom- 35. For the sake of clarity, we assume that the interferometer
etry for direct electric-field reconstruction of ultrashort op- is balanced, so that E(t) ⫽ E 0 (t).
tical pulses,’’ Opt. Lett. 23, 792–794 (1998). 36. A. W. Albrecht, J. D. Hybl, S. M. Gallagher, and D. M.
16. L. Lepetit and M. Joffre, ‘‘Two-dimensional nonlinear optics Jonas, ‘‘Experimental distinction between phase shifts and
using Fourier-transform spectral interferometry,’’ Opt. time delays: implications for femtosecond spectroscopy
Lett. 21, 564–566 (1996). and coherent control of chemical reactions,’’ J. Chem. Phys.
17. J. D. Hybl, A. W. Albrecht, S. M. Gallagher Faeder, and D. 111, 10934–10956 (1999).

View publication stats