Femtosecond broadband fluorescence upconversion spectroscopy:
Improved setup and photometric correction
X.-X. Zhang, C. Würth, L. Zhao, U. Resch-Genger, N. P. Ernsting et al.
Citation: Rev. Sci. Instrum. 82, 063108 (2011); doi: 10.1063/1.3597674
View online: http://dx.doi.org/10.1063/1.3597674
View Table of Contents: http://rsi.aip.org/resource/1/RSINAK/v82/i6
Published by the American Institute of Physics.
Related Articles
Compact metal probes: A solution for atomic force microscopy based tip-enhanced Raman spectroscopy
Rev. Sci. Instrum. 83, 123708 (2012)
Broadband photon-counting Raman spectroscopy in short optical waveguides
Appl. Phys. Lett. 101, 211110 (2012)
Antenna-enhanced infrared near-field nanospectroscopy of a polymer
Appl. Phys. Lett. 101, 193105 (2012)
Comparisons of 2D IR measured spectral diffusion in rotating frames using pulse shaping and in the stationary
frame using the standard method
J. Chem. Phys. 137, 184201 (2012)
Constant current etching of gold tips suitable for tip-enhanced Raman spectroscopy
Rev. Sci. Instrum. 83, 103708 (2012)
Additional information on Rev. Sci. Instrum.
Journal Homepage: http://rsi.aip.org
Journal Information: http://rsi.aip.org/about/about_the_journal
Top downloads: http://rsi.aip.org/features/most_downloaded
Information for Authors: http://rsi.aip.org/authors
Downloaded 11 Jan 2013 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
REVIEW OF SCIENTIFIC INSTRUMENTS 82, 063108 (2011)
Femtosecond broadband fluorescence upconversion spectroscopy:
Improved setup and photometric correction
X.-X. Zhang,1,3
C. Würth,2
L. Zhao,1
U. Resch-Genger,2
N. P. Ernsting,3,a)
and M. Sajadi3,b)
1
Photonics Center, College of Physical Science, Nankai University, Tianjin, China
2
Federal Institute for Materials Research and Testing, Berlin, Germany
3
Department of Chemistry, Humboldt Universität zu Berlin, Germany
(Received 2 March 2011; accepted 16 May 2011; published online 17 June 2011)
A setup for fluorescence upconversion spectroscopy (FLUPS) is described which has 80 fs temporal
response (fwhm) for emission in the spectral range 425–750 nm. Broadband phase matching is
achieved with tilted gate pulses at 1340 nm. Background from harmonics of the gate pulse is removed
and sensitivity increased compared to previous designs. Photometric calibration of the upconversion
process is performed with a set of fluorescent dyes. For Coumarin 153 in methanol the peak position,
bandwidth, and asymmetry depending on delay time are reported. © 2011 American Institute of
Physics. [doi:10.1063/1.3597674]
I. INTRODUCTION
Time-resolved fluorescence upconversion was first reported
by Mahr and Hirsch in 1975.1
A chromophore in solution
is excited with short optical pulses, and its emission at
wavelength λF is mixed with short gate pulses at λG in a nonlinear
optical (NLO) crystal. Sum-frequency generation occurs
at a well-defined crystal orientation when phase matching
conditions are met. In this case upconverted light at λU
is generated and detected. By plotting the intensity IU(t) as
function of time delay t between pump and gate pulses, a
“kinetic trace” of the fluorescence at the chosen wavelength is
obtained. This kind of measurement is well suited to measure
short excited-state lifetimes. In step with advances in ultrafast
laser technology, the accuracy improved steadily and the
spectral scope widened. Representative examples and some
milestones on this way are given in Refs. 2–27.
Spectral evolution of the fluorescence may indicate a
photochemical reaction or a photophysical process in the excited
singlet state; examples are cis/trans isomerization,4,20,27
proton-transfer10–14,23
or electron-transfer,8,21,24
and dynamic
solvation.3,5,9,16,17,22,37
To understand such processes one
must obtain not only an amplitude at various delay times
t, but also the shape of the entire fluorescence quantum
distribution F(t,λF). This is usually achieved by recording
10–20 kinetic traces IU(t), corresponding to different
λF which cover the emission region. The results are depicted
schematically in Fig. 1. Because the crystal is rotated
between recording kinetic traces, their relative amplitudes
are usually not defined during measurement. Instead, amplitude
scaling is performed afterwards with the help of
additional information such as the steady-state emission
spectrum of the compound under investigation. Spectral
reconstruction then yields the desired transient emission
spectra.9,28,29
a)Electronic mail: nernst@chemie.hu-berlin.de.
b)Electronic mail: sajadimo@chemie.hu-berlin.de.
For photometric accuracy it is better to record the contiguous
emission spectrum for any given delay time. The
order of measurement should therefore be inverted. Whereas
before, λF is kept fixed while t is scanned, now instead, for
every fixed delay the fluorescence wavelength is scanned. By
this procedure, intensities for adjacent emission wavelengths
are compared directly, allowing a more accurate correction of
the apparatus response which may strongly depend on wavelength.
Coverage of the fluorophore’s emission region has
been realized in various ways. The methods are: (i) rotation
of the NLO crystal while synchronously scanning the detection
monochromator,17,19,30
(ii) wobbling the crystal while
detecting with a spectrograph,31
and (iii) simultaneous phase
matching across a broad spectral range followed by multiplex
detection.32–34
Broadband fluorescence upconversion spectroscopy
(FLUPS), developed systematically in our laboratory along
(iii),32–34
has similarities with ultrafast pulse generation by
nonlinear-optical parametric amplification. Indeed, parametric
amplification offers an alternative route to time-resolved
fluorescence spectra,35,36
but so far the band shape could not
be recorded accurately with this method. This renders FLUPS
more promising. For gating we use near-infrared pulses at
λG ≈ 1340nm. Tilting of the gate pulses34
turned out to be
crucial to achieve a time resolution (fwhm of instrument
response) of 80 fs. However, two deficiencies remained: first,
background due to direct fluorescence, and second, signal
from the third-harmonic of the gate pulse which blocked
the range λF > 600 nm. To overcome these drawbacks, we
designed an improved setup which is easier to align, more
sensitive, and – most importantly – background free. This
enables photometric correction of broadband sum-frequency
for emission wavelengths λF = 425 − 750 nm.
In addition to instrument characterization, the increasing
collaboration between laboratories triggers the need for
a dynamic fluorescence standard. The dynamics of methanol
is a case in point. For such measurements one observes the
dynamic solvation of a polar molecular probe, for exam-
0034-6748/2011/82(6)/063108/8/$30.00 © 2011 American Institute of Physics82, 063108-1
Downloaded 11 Jan 2013 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
063108-2 Zhang et al. Rev. Sci. Instrum. 82, 063108 (2011)
FIG. 1. With conventional fluorescence-upconversion, the emission wavelength
λF is kept fixed while the time delay t is scanned, resulting in a kinetic
trace. Relative amplitudes are obtained with the help of the stationary fluorescence
spectrum. Transient spectra are constructed from 10–20 traces.
ple, Coumarin 153 (C153). Figure 2 shows its absorption
and stationary emission spectra. An estimate of the initial
fluorescence band, after vibrational relaxation but before
solvation,38
is shown as a dashed line. The methanol dynamics
is assumed to be reflected by the spectral position ˜ν(t) of
the fluorescence band (peak position or first moment) as a
function of time. Polar solvation of C153 has been the subject
of numerous experimental9,17,34,37–40
and theoretical41,42
studies. In Fig. 3 we compare the ˜ν(t) curves for methanol
reported by different groups.9,17,37,40,43
To facilitate the comparison,
curves C(t) ∝ ˜ν(t) − ˜ν(∞) are shown that are scaled
to a common value9
at t = 0.20 ps. At this time the fluorescence
spectrum should have been fully time-resolved in all
.qb
.bb
FIG. 2. Absorption and fluorescence spectra (quantum distributions over energy)
of Coumarin 153 (inset) in methanol. (Dashed) Time-zero estimate,38
(gray) transient fluorescence spectrum at 0.1 ps of this work. The timedependent
Stokes shift of fluorescence is indicated by a horizontal arrow.
FIG. 3. (Color online) Reports of the dynamic Stokes shift of C153 fluorescence
in methanol: The “spectral relaxation functions” C(t) are set to a
common value 0.704 at t = 0.20 ps10 when spectra were fully time-resolved.
The relative variance at 10 ps, for example, is ∼30%, limiting the accuracy
of C(t) measurements. Reports of the time-dependent fluorescence bandwidth
and asymmetry10,35,40 vary more strongly.
cases. The variances between these results are notable, considering
that C153 is the best-studied polarity probe. Similar
discrepancies must be expected when results from different
laboratories are combined to support a microscopic model for
the spectral relaxation, thus limiting the significance of the
model.
Why are the C(t) curves of Fig. 3 so different? When
measurements of transient absorption are used37,43
the main
problem is the extraction of the band for stimulated emission.
Conventional upconversion methodology,9,40
whereby spectra
are reconstructed from kinetic traces, suffers from a lack
of stationary calibration light on the blue side, in combination
with (relatively) sparse spectral coverage. The latter problem
is solved by synchronous wavelength scanning of the crystal
angle and detection monochromator, before the next delay
time is accessed.17
In this case geometrical instabilities might
play a role.
Broadband upconversion with a fixed crystal angle, by
comparison, simultaneously monitors a range of frequencies
and therefore needs less integration time to achieve the same
signal/noise; also it allows better control of the geometry.
Using this method, we quantify the C153 fluorescence in
methanol in terms of peak position, bandwidth, and asymmetry
depending on delay time.
II. OPTIMIZED OPTICAL SETUP FOR FS BROADBAND
FLUORESCENCE UPCONVERSION
A. Gate pulse compression and tilting
The fs fluorometer is shown in Fig. 4. A Ti:sapphire laser
(Femtolasers sPro) provides 30 fs, 500 μJ pulses at 800 nm
(500 Hz repetition rate). This beam is split with a 6:1 ratio.
Pulses of 430 μJ drive a traveling-wave optical parametric
Downloaded 11 Jan 2013 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
063108-3 Zhang et al. Rev. Sci. Instrum. 82, 063108 (2011)
gate, 1340 nm
pump, 400 nm
1 J, 40 fsμ
KDP
time delay
sample
cell
P2
Spectrograph
P1
M2
M1
M3
L4
L5
M4
Pr3 Pr2
Pr1
X
L1
fiber bundle
L3 L2
100 mm
λ/2
F D
spectrograph
triple
mirror
60 J, 40 fsμ
Ue
G
e
Fo
α′
FIG. 4. (Color online) Setup for broadband fluorescence upconversion spectroscopy. Pr – prisms; M 1–4 – spherical mirrors; L – lenses; P – polarizers.
amplifier of superfluorescence (TOPAS, Lightconversion)
which delivers 60 μJ, 1340 nm pulses in horizontal polarization,
to be used for optical gating. The beam is widened
and collimated by a telescope L2/L3 with −50-mm/+200mm
lenses. All lenses in the gate path have antireflective
coating at 1300 nm. For compression and subsequent tilting
a combination of three prisms Pr1–3 (Schott SF50, apex
angle 55.5◦
) is used. The fourth prism of a normal compressor
would be located at point X.44
The tilt angle
= 3.55◦
of the pulse front at X (due to the action of prism
Pr3) is calculated from Eq. (1) of Ref. 34. X is imaged by
a thin lens L4 (f = 140 mm) onto the potassium dihydrogen
phosphate (KDP) crystal, whereby the tilt angle is leveraged
to 20◦
at the crystal. Fine adjustment of the crystal-lens
distance, and hence the tilt, is done simply by observing the
generated sum frequency (from the pump laser) on a white
piece of paper. When aligned properly, the upconverted light
appears and disappears at the same location on the paper when
the time difference between pump and gate pulses is varied,
otherwise it moves across the wider upconversion pattern.
B. Fluorescence excitation, collection, and imaging
For optical pumping, the rest of the fundamental light is
frequency doubled to 400 nm pulses which have 40 fs fwhm
after compression. Their polarization is set with a λ/2 plate.
After passing a triple mirror on a variable delay stage, attenuated
pump pulses (∼1 μJ) are focused by a thin lens L1 (f
= 200 mm, fused silica, f.s.) onto the sample cell, to a spot
diameter ≤ 0.1 mm. To obtain an optical image with the subsequent
collection optics, the cell is made as thin as possible
for a given solution (f.s. windows, 0.2 mm thick, are typically
spaced 0.2–0.3 mm apart). The sample solution is pumped
through the cell which can also be oscillated perpendicular to
the beam direction.
FIG. 5. (Color online) Off-axis Schwarzschild objective for fluorescence collection, allowing to oscillate the sample cell perpendicular to the pump beam.
Downloaded 11 Jan 2013 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
063108-4 Zhang et al. Rev. Sci. Instrum. 82, 063108 (2011)
An off-axis Schwarzschild objective (Fig. 5) refocuses
the generated fluorescence with sevenfold magnification at
intermediate position D. This arrangement provides more
degrees-of-freedom to the sample cell compared to the axial
geometry.34
Fluorescence emitted from source A into the full
angles 18◦
and 40◦
in horizontal and vertical planes, respectively,
is collected by concave mirror M1, consistent with the
acceptance angles for sum-frequency generation. The effective
waist diameter at D is 0.6 mm, allowing to select vertical
polarization by a miniature Glan polarizer here.45,46
For fluorescence
wavelengths λF ≥ 420nm a wire-grid polarizer P1
(Moxtec PPL04C, on 700 μm of Corning 1737F glass) is used
instead. Pump scatter and low-frequency Raman signal are reduced
with a long-pass edge filter (Schott GG420, 2 mm). The
fluorescence from D is relayed 1:1 onto the KDP crystal by
concave mirror M3 (R3 = −400 mm).
C. Sum-frequency generation
For sum-frequency generation we use a KDP crystal cut
at θN = 65◦
, with the crystal axis in the horizontal plane. At
first sight BBO might be considered superior due to its higher
nonlinear susceptibility, but this advantage is more than cancelled
by increased dispersion which limits broadband operation.
The KDP crystal of thickness L = 0.3 mm has antireflection
(AR) coating at 520 nm and 1300 nm on the input face,
and at 360 nm on the output face (EKSMA). The central rays
of the fluorescence and the gate beam form an external gating
angle α ≈ 20◦
at the crystal (internal angle α ≈ 13◦
). Type
II phase matching is chosen because it provides the broadest
spectral window. Vertically polarized fluorescence at λF
(o − ordinary in terms of crystal axes) interacts with
horizontally polarized gate pulses at λG = 1340 nm (e − extraordinary),
to generate horizontally polarized upconverted
light at λU (e). Phase matching requires minimizing the wave
vector mismatch k = keU(θU) − {koF + keG(θG)} over a
wide range of λF, for example 400–800 nm. The quantum
efficiency of the upconversion process is47
ηNLO =
NU(z = L)
NF(z = 0)
=
27
π3
IG(z = 0)
c3
×
ωUωFd2
ef f
ne(ωU)no(ωF)ne(ωG)
sin( kL/2)
kL/2
2
L2
.
(1)
Here, N is the photon flux, is the intensity of the gate light, L
is the crystal thickness, neo are the pertinent refractive indices,
def f is the effective nonlinear susceptibility, and ωFGU are
the angular frequencies. Figure 6 shows mismatch curves
k(λF) for different cut angles θN (Note that Eq. (11) of
Ref. 34 contains a sign error which has been corrected in
Fig. 6(a)). The efficiency curve η(λF) depends sensitively on
thickness L (Fig. 6(b)) and external beam angle α (Fig. 6(c)).
A range of angles contributes to the upconverted signal,
resulting in the measured efficiency curve (Fig. 6(d)) which
was obtained upon photometric correction. Calculating the
group velocity mismatch (between gate and fluorescence
λ
μ
FIG. 6. (Color online) (a) Phase matching curve for type II sum-frequency
generation, for three cut angles θN of a KDP crystal with gate pulses at
1340 nm. Gate and fluorescence beams have external angle α = 19.5◦
between
them. (b) Relative conversion efficiency sin( kL/2)/( kL/2) L2
for different crystal thicknesses (θN = 65◦, α = 19.5◦
). (c) Efficiency when
varying the angle α , for a crystal of 0.3 mm thickness cut at 65◦, with gate
wavelength of 1340 nm. (d) Experimental NLO efficiency, i.e., 1/¯c(p) of
Fig. 8, divided by the quantum efficiency of the UV detection system.
light, in the fluorescence direction) for these angles, we find
that their passage times differ by < 20 fs.
D. UV collection, dispersion, and detection
The upconverted signal is collimated by a meniscus lens
L5 (f = +50 mm, f.s., AR-coated for the UV). It is then
passed through a Glan polarizer P2 (10 × 10 × 10 mm3
,
Halle Nachfl.) to suppress direct fluorescence, pump
scatter, and gate harmonics. The concave mirror M4 (R4
= −70 mm) collects and focuses the signal light on the
entrance of a fiber bundle. The bundle (CeramOptec) consists
of 30 f.s. fibers of 100 μm diameter, compacted into a round
shape (0.6 mm diameter ) at the input and into a slit (0.1
× 3.3 mm) at the spectrograph. This UV geometry allows suppressing
the background signal which hampered the previous
setup for λF > 600 nm due to better use of the Glan polarizer.
An important aspect of the geometry is the large input
angle α .
Downloaded 11 Jan 2013 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
063108-5 Zhang et al. Rev. Sci. Instrum. 82, 063108 (2011)
The spectrograph employs a ruled plane grating
(Newport 53–080R) in a Czerny–Turner mount. Photon loss
to unused orders is reduced, and the UV efficiency is maximized
by the choice of blaze angle. The linear wavelength
calibration λU(p) as function of pixel number p is obtained
with a mercury lamp. The spectral range λU = 307 − 478 nm
(corresponding to λF = 400 − 750nm) is mapped onto a spectrum
21 mm wide. It is recorded with a CCD camera (Andor
DV420-BU, 25 μm pixel width). For example, photons at
λU = 374 nm (λF = 522 nm) are binned into a 14.5 cm−1
energy
segment. The resolution of the spectrograph is 50 cm−1
(fwhm of spectral instrument response).
Using the dye C153 in methanol as described below, we
measure a count rate of ∼100 s−1
at long delay times. (The
sensitivity is ∼1.5-fold higher than in Ref. 34 because of the
new collection optics, but signal/noise is increased by a factor
3 because background has been removed.) The integration
time is 1 s. For a time scan, the background is first recorded
four times at negative delay, averaged, and smoothed. This
background is subsequently subtracted from each recorded
spectrum while the delay is scanned. Results from 16 time
scans are individually corrected for cosmic spikes and then
averaged.
III. CORRECTION PROCEDURES
A. Photometric correction
The correction is performed with the “standard” dyes
BBOT (2,5-bis[5-tert-butylbenzoxazolyl(2)]thiophene, CAS
7128–64-5), Coumarin 6H (58336–35-9), Coumarin 153
(53518–18-6), and DCM (51325–91-8) in methanol. The
absorbance at the pump wavelength is set to 0.4 in order
to limit the inner filter effect. (Steady-state absorption
spectra are shown in the supplementary material.53
Note
that also excited-state absorption affects the fluorescence
signal33
.) The λF regions which were used to correct the
spectral responsivity of the setup are 425–630, 470–690, and
600–750 nm for the last three dyes.
The development in Ref. 34 is summarized here for
completeness. Consider an upconversion experiment with a
standard dye whereby a total number (U)
of counts in the
UV is registered. The time delay is chosen to be 250 ps
so that nuclear relaxation in methanol has ceased. Actually
one records the “technical” signal- or count-distribution
over pixels p, s(U)
(p) ≡ (∂ (U)
/∂p)(p). This is to be compared
with the “molecular” stationary distribution of (F)
fluorescence
photons over wavenumbers in the visible, f(F)
(˜νF)
≡ ∂ (F)
/∂ ˜νF (˜νF). The latter is conveniently provided as a
sum of lognormal functions:48
log norm(˜νF)=h exp
⎧
⎨
⎩
−ln 2
ln 1 + 2γ (˜νF − ˜ν0)/
γ
2
⎫
⎬
⎭
.
(2)
Optimal values for peak wavenumber ˜ν0 [cm−1
], width
[cm−1
], asymmetry γ [−], and amplitude h from conventional
fluorescence measurements are collected in Table I. In
TABLE I. Lognormal parameters (Eq. (2)) for the fluorescence quantum
distributions over wavenumbers, f(F)(˜νF), in methanol.
˜ν0(cm−1
) (cm−1) γ h
BBOT 24 457 952 0.0303 0.4153
23 859 2126 0.001 0.1797
22 975 1517 0.0564 0.8736
21 532 1599 −0.3423 0.6292
19 826 1989 −0.567 0.1546
18 600 9937 −0.7929 0.0158
C6H 23 200 886 0 0.0255
21 000 774 0.087 0.0203
20 717 2950 −0.308 0.9774
15 500 3186 0 0.0499
C153 20 481 1601 −0.218 0.0724
18 519 2947 −0.402 0.9913
16 537 2884 −0.781 0.0904
DCM 15 829 2388 −0.056 1.0000
addition to the spectrograph calibration λU(p) we also need
the relationship
λF(p) =
λGλU(p)
λG − λU(p)
, (3)
which involves the gate wavelength λG. Figure 7 shows several
upconversion spectra which are obtained when dye fluorescence
is passed through narrowband interference filters
as indicated. For any such combination, the peak wavelength
λF max of the transmitted fluorescence light is known from stationary
measurements, and the corresponding pixel number
pmax is obtained from the figure. Optimization gives λG =
1340 ± 4nm (see supplementary material).53
Next, the stationary distribution f(F)
(˜νF) is transformed
into an equivalent distribution of “upconverted” fluorescence
λ
λ
FIG. 7. Experimental UV-count distributions s(U)(p) over pixels (black lines)
from fluorescence upconversion measurements of the dyes C6H, C153, and
DCM in methanol at 250 ps delay time (for BBOT see Fig. 9). Also shown are
the count distributions when the fluorescence has been passed through interference
filters which are centered at the indicated wavelengths. For comparison,
the expected UV-photon distributions f(U)(p) (gray lines) were inferred
from stationary fluorescence measurements.
Downloaded 11 Jan 2013 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
063108-6 Zhang et al. Rev. Sci. Instrum. 82, 063108 (2011)
λ
FIG. 8. (Color online) Photometric correction curves ci(p) (see text for definition
and use) obtained by comparing the count distributions in Fig. 7 with
the expected photon distributions. The black line ¯c(p) is a polynomial fit.
photons over pixels:
f(U)
(p) ≡
∂ (U)
∂p
(p) ∝ λ−2
U f(F)
(˜νF) ·
∂λU
∂p
with λU(p) and ˜νF(p). (4)
The two types of spectra, s(U)
(p) and f(U)
(p), are shown
in Fig. 7 (black and gray lines, respectively) for each standard
dye. The correction function is then
c(p) ∝ f(U)
(p)/s(U)
(p). (5)
Figure 8 shows the c(p) segments which correspond to
the dyes in Fig. 7. They are scaled so that their rms differences
are minimized, and a smooth fit provides the final correction
curve ¯c(p).
The correction function ¯c(p) is used as follows. Pointwise
multiplication of a measured, technical count distribution
s(U)
(p) with ¯c(p) gives the distribution f(U)
(p) of upconverted
UV photons. To obtain the fluorescence distribution over
fluorescence wavenumbers, f(U)
(p)λ2
U(p) is plotted against
107
/λF(p). To obtain the fluorescence distribution over fluorescence
wavelengths, f(U)
(p)λ2
U(p)/λ2
F(p) is plotted against
λF(p). (Here it was assumed that λU is linear in p.)
B. Group delay correction
Before gating, fluorescence passes the exit window of the
sample cell, polarizer, and glass filter. These optical elements
cause group velocity dispersion, which is measured by gating
a white-light continuum. The latter is generated by focusing
10 μJ pump pulses with a thin lens (f = 100 mm f.s.) into the
cell containing pure methanol. For each pixel the transient signal
is fitted with a temporal Gaussian function, and the dispersion
curve is obtained by taking the peak position of the fit as
a function of pixel number. The temporal response is given by
the fwhm ≈ 80 fs of the Gaussian fits. To check the accuracy
of the time correction obtained from continuum measurement,
FIG. 9. (Color online) Left: The curve for time-zero as function of emission
wavelength is plotted on top of the transient fluorescence of BBOT in
methanol. The band ±40 fs indicates the width (fwhm) of the temporal apparatus
function. Following red-edge excitation at 400 nm, spectral broadening
is observed during the first 0.2 ps. Right: The measured count distribution
over wavelengths at late delay time (black line), compared to the stationary
photon distribution (gray).
FIG. 10. Transient fluorescence spectra of C153 in methanol at 0.1, 0.3, 1.2,
5, 15, and 50 ps (Bottom panel). Shown are the relative quantum distributions
over wavelength, i.e., after photometric and time corrections, for perpendicular
polarization between gate pulse and fluorescence. The data were obtained
by averaging over 16 scans. The sharp peak at blue side of the spectrum at
0.1 ps is the tail end of pump scatter. At TH, the third harmonic of the gate
pulse rendered measurement impossible in previous designs.34 Top panel: the
time trace of residual pump scatter shows the instrument response function.
Downloaded 11 Jan 2013 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
063108-7 Zhang et al. Rev. Sci. Instrum. 82, 063108 (2011)
we measure the transient fluorescence of BBOT in methanol.
The raw signal (λF, t) is shown as contour plot in Fig. 9, together
with the time-zero function. With a time resolution of
80 fs, small changes of the vibronic fluorescence structure are
observed before 200 fs. In the right panel the late raw signal
(1 ps, black line) is compared to the fluorescence quantum
distribution from stationary measurements (gray). Such comparisons
were made above for photometric calibration. In the
present case, the vibronic emission band of BBOT at 437 nm
is reduced by the yellow glass filter.
IV. C153 IN METHANOL: A REFERENCE FOR
TRANSIENT FLUORESCENCE BAND-SHAPE
MEASUREMENTS
C153 has been studied extensively as a probe in timeresolved
Stokes shift measurements9,17,34,37–40,43
and has become
a standard in this field. Sets of scans were taken from
negative delays up to 4 ps in 20 fs steps, to 40 ps in 200 fs
steps, and to 100 ps in 500 fs steps. Transient upconverted
spectra corrected for group velocity dispersion, and photometrically
as discussed above, are shown in Fig. 10. Fluorescence
polarized perpendicularly to the pump polarization is
recorded in order to minimize contributions from Raman scattering
of the solvent. As a consequence, the amplitude rises
on the ps time scale due to electronic change and rotational
diffusion of the emitting chromophore. We find biexponential
behavior of the anisotropy decay, as in Ref. 52 but with
faster initial component (0.3 ps, 38%). However, the fact that
FIG. 11. (Color online) The reproducibility of the upconversion experiment
is seen by measurements which were recorded two years apart. Shown is the
peak frequency as function of time, for methyl quinolone (a)49 and 4-amino
phthalimide (b)50 in methanol.
Δ
γ
τ
τ
τ
τ
τ
τ
τ
τ
τ
τ
τ
FIG. 12. Evolution of lognormal parameters (Eq. (2)) for the fluorescence
distribution f(F)(˜νF) from C153 in methanol. Gray bands outline the 95%
confidence interval of the lognorm fits. For the time range 0.1–80 ps, the evolution
is described by ai exp {−t/τi} + a0 (black lines) with the parameters
given as insets. A weak (2%) satellite pulse is responsible for a kink in (b) at
3.4 ps. The evolution of the average emission frequency can be derived.51
the amplitude rises does not influence the determination of the
spectral position and band shape. The sharp peak near the blue
edge of the spectrum at 100 fs is due to tailing pump scatter.
After transformation to photon distributions over wavenumbers,
the spectra are fitted with a lognormal function. Thus,
for every delay time t, we obtain the fit parameters together
with their standard error (SE), comprising the final result.
The reproducibility of dynamic fluorescence measurements
with the new setup is demonstrated in Fig. 11. Here the
evolution of peak frequency, in methanol, is shown for methyl
quinolone (MQ, a) and for 4-aminophthalimide (4AP,b). The
two measurements in each case were performed two years
apart.
Let us return to C153 in methanol. Figure 12 shows
the evolution of peak position ˜ν0(t), bandwidth parameter
(t), and asymmetry parameter γ (t), in bands of ±2SE for
95% confidence. Here the confidence interval refers to the
lognormal fit only and may be subject to other experimental
problems. For example, a weak satellite of the pump pulse
is responsible for a sudden increase of bandwidth at 3.4 ps
(panel b). This perturbation can be removed by data analysis.
Note that a multiexponential temporal fit of ˜ν0(t) may
Downloaded 11 Jan 2013 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
063108-8 Zhang et al. Rev. Sci. Instrum. 82, 063108 (2011)
have different errors (slightly smaller in this case) because
correlated changes over a large spectral range are measured.
The Stokes shift from t = 0.1 ps to time-infinity is determined
as 2020 cm−1
. This value is smaller than the total shift
(2190 cm−1
) from our stationary estimates9
(see Fig. 2). We
do not see the oscillations of ˜ν(t) which were reported and
analyzed recently.40
Parameters for multiexponential fits (red
lines) are given as insets. Such data could be used to develop
this and other femtosecond fluorescence spectrometers
consistently.53
ACKNOWLEDGMENTS
X.-X.Z. is grateful for the support by the Chinese Scholarship
Council. N.P.E. thanks the Deutsche Forschungsgemeinschaft
(Grant No. ER 154/9–1).
1H. Mahr and M. D. Hirsch, Opt. Comm. 13, 96 (1975).
2M. A. Kahlow, W. Jarzeba, T. P. DuBruil, and P. F. Barbara, Rev. Sci. Instrum.
59, 1098 (1988).
3W. Jarzeba, G. C. Walker, A. E. Johnson, M. A. Kahlow, and P. F. Barbara,
J. Phys. Chem. 92, 7039 (1988).
4D. C. Todd, J. M. Jean, S. J. Rosenthal, A. J. Ruggiero, D. Yang, and G. R.
Fleming, J. Chem. Phys. 93, 8658 (1990).
5R. Jimenez, G. R. Fleming, P. V. Kumar, and M. Maroncelli, Nature
(London) 369, 471 (1994).
6S. E. Bradforth, R. Jimenez, F. V. Mourik, R. V. Grondelle, and G. R.
Fleming, J. Phys. Chem. 99, 16179 (1995).
7B. P. Krueger, G. D. Scholes, R. Jimenez, and G. R. Fleming, J. Phys.
Chem. B 102, 2284 (1998).
8I. V. Rubtsov and K. Yoshihara, J. Phys. Chem. A 103, 10202 (1999).
9M. L. Horng, J. A. Gardecki, A. Papazyan, and M. Maroncelli, J. Phys.
Chem. 99, 17311 (1995).
10D. Marks, P. Prosposito, H. Zhang, and M. Glasbeek, Chem. Phys. Lett.
289, 535 (1998).
11M. Chattoraj, B. A. King, G. U. Bublitz, and S. G. Boxer, Proc. Natl. Acad.
Sci. U.S.A. 93, 8362 (1996).
12X. H. Shi, P. Abbyad, X. K. Shu, K. Kallio, P. Kanchanawong, W. Childs,
S. J. Remington, and S. G. Boxer, Biochemistry 46, 12014 (2007).
13S. Takeuchi and T. Tahara, J. Phys. Chem. A 102, 7740 (1998).
14T. Fiebig, M. Chachisvillis, M. Manger, A. H. Zeweil, A. Douhal,
I. Garcia-Ochoa, and A. D. L. H. Ayuso, J. Phys. Chem. A 103, 7419
(1999).
15S. K. Pal, J. Peon, and A. H. Zewail, Chem. Phys. Lett. 363, 57 (2002).
16S. K. Pal, L. Zhang, and A. H. Zewail, Proc. Natl. Acad. Sci. U.S.A. 100,
8113 (2003).
17T. Gustavsson, L. Cassara, V. Gulbinas, G. Gurzadyan, J.-C.,Mialocq, S.
Pommeret, M. Sorgius, and P. V. D. Meulen, J. Phys. Chem. A 102, 4229
(1998).
18T. Gustavsson, N. Sarkar, E. Lazzarotto, D. Markovitsi, and R. Improta,
Chem. Phys. Lett. 429, 551 (2006).
19H. S. Cho, H. Rhee, J. K. Song, C.-K. Min, M. Takase, N. Aratani, S. Cho,
A. Osuka, T. Joo, and D. Kim, J. Am. Chem. Soc. 125, 5849 (2003).
20A. Fürstenberg, M. D. Julliard, T. G. Deligeorgiev, N. I. Gadjev, A. A.
Vasilev, and E. Vauthey, J. Am. Chem. Soc. 128, 7661 (2006).
21Y.-T. Kao, C. Saxena, T.-F. He, L. J. Guo, L. J. Wang, A. Sancar, and D. P.
Zhong, J. Am. Chem. Soc. 130, 13132 (2008).
22L. Y. Zhang, Y. Yang, Y.-T. Kao, L. J. Wang, and D. P. Zhong, J. Am. Chem.
Soc. 131, 10677 (2009).
23P.-T. Chou, Y.-C. Chen, W.-C. Chen, W.-S. Yu, Y.-H. Chou, C.-Y. Wei, and
Y.-M. Cheng, J. Phys. Chem. A 105, 1731 (2001).
24C.-Y. Chen, C.-T. Cheng, J.-K. Yu, S.-C. Pu, Y.-M. Cheng, and P.-T. Chou,
J. Phys. Chem. B 108, 10687 (2004).
25T. Pancur, N. K. Schwalb, F. Renth, and F. Temps, Chem. Phys. 313, 199
(2005).
26N. K. Schwalb and F. Temps, Phys. Chem. Chem. Phys. 8, 5229 (2006).
27G. Zgrabli´c, S. Haacke, and M. Chergui, J. Phys. Chem. B 113, 4384
(2009).
28D. V. O’Connor and D. Phillips, Time-Correlated Single Photon Counting,
1st ed. (Academic, London, 1984).
29J. A. Gardecki and M. Maroncelli, J. Phys. Chem. A 103, 1187 (1999).
30T. Gustavsson, G. Baldcchino, J.-C. Mialocq, and S. Pommeret, Chem.
Phys. Lett. 236, 587 (1995).
31A. Cannizzo, O. Bräm, G. Zgrablic, A. Tortschanoff, A. AjdarzadehOskouei,
F. van Mourik, and M. Chergui, Opt. Lett. 32, 3555 (2007).
32R. Schanz, S. A. Kovalenko, V. Kharlanov, and N. P. Ernsting, Appl. Phys.
Lett. 79, 566 (2001).
33S. A. Kovalenko, R. Schanz, T. Senyushkina, and N. P. Ernsting, Phys.
Chem. Chem. Phys. 4, 703 (2002).
34L. Zhao, J. L. Pérez-Lustres, V. Farztdinov, and N. P. Ernsting, Phys. Chem.
Chem. Phys. 7, 1716 (2005).
35P. Fita, Y. Stepanenko, and C. Radzewicz, Appl. Phys. Lett. 86, 021909
(2005).
36X.-F. Han, X.-H. Chen, Y.-X. Weng, and J.-Y. Zhang, J. Opt. Soc. Am. B
24, 1633 (2007).
37P. Changenet-Barret, C. T. Choma, E. F. Gooding, W. F. DeGrado, and R.
M. Hochstrasser, J. Phys. Chem. B 104, 9322 (2000).
38R. S. Fee and M. Maroncelli, Chem. Phys. 183, 235 (1994).
39S. J. Rosenthal, R. Jimenez, G. R. Fleming, P. V. Kumar, and M.
Maroncelli, J. Mol. Liq. 60, 25 (1994).
40I. Eom and T. Joo, J. Chem. Phys. 131, 244507 (2009).
41D. V. Matyushov and M. D. Newton, J. Phys. Chem. A 105, 8516 (2001).
42R. Improta, V. Barone, and F. Santoro, J. Phys. Chem. B 111, 14080
(2007).
43J. L. Perez-Lustres, F. Rodriguez-Prieto, M. Mosquera, T. A.
Senyushkina, N. P. Ernsting, and S. A. Kovalenko, J. Am. Chem.
Soc. 129, 5408 (2007).
44Y. Nabekawa and K. Midorikawa, Opt. Express 11, 324 (2003).
45N. P. Ernsting, J. Breffke, D. Y. Vorobyev, D. A. Duncan, and I. Pfeffer,
Phys. Chem. Chem. Phys. 10, 2043 (2008); U.S. and European patents
pending DE 10 2006 039 425; U.S. Application patent 11/843,546.
46M. Sajadi, A. L. Dobryakov, E. Garbin, N. P. Ernsting, and S. A.
Kovalenko, Chem. Phys. Lett. 489, 44(2010).
47F. Zernike and J. E. Midwinter, Applied Nonlinear Optics (Wiley, New
York, 1973).
48Lognormal Distributions: Theory and Applications, edited by E. L. Crow
and K. Schimizu, (Marcel Dekker, New York, 1988).
49M. Sajadi, Y. Ajaj, I. Ioffe, H. Weingärtner, and N. P. Ernsting, Angew.
Chem. Int. Ed. 49, 454 (2009).
50M. Sajadi, T. Obernhuber, S. A. Kovalenko, M. Mosquera, B. Dick, and N.
P. Ernsting, J. Phys. Chem. A. 113, 44 (2008).
51From Eq. (3.11) of Ref. [9], a multiexponential fit of the average wavenumber
curve so obtained gives a1 = 1546 cm−1, τ1 = 0.091 ps, a2
= 599 cm−1, τ2 = 0.737 ps, a3 = −164.3 cm−1, τ3 = 21.9 ps, a4
= 1132 cm−1, τ4 = 14.6 ps, and a0 = 17744 cm−1.
52M. L. Horng, J. A. Gardecki, and M. Maroncelli, J. Phys. Chem. A. 101,
1030 (1997).
53See supplementary material at http://dx.doi.org/10.1063/1.3597674 for the
absorption spectra of calibration dyes, details of gate energy determination,
and parameters for solvation dynamic fits of C153 in methanol reported
previously by other groups.
Downloaded 11 Jan 2013 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions