Anal. Chem.
2001, 73, 2302-2309
Two-Dimensional Fluorescence Correlation
Spectroscopy with Modulated Excitation
Yan He, Gufeng Wang, Jason Cox, and Lei Geng*
Department of Chemistry, University of Iowa, Iowa City, Iowa 52242
Overlap of multiple states or multiple species in a chemi-
cal system often creates a congested fluorescence spec-
trum that is difficult to interpret. The resolution of
component spectra is essential for the understanding of
the structure and dynamics of such multicomponent
systems. In this paper, two-dimensional fluorescence
correlation spectroscopy (2D FCS) is presented for the
dissection of component spectra using the time correlation
function. In 2D FCS, the time response of fluorescence
intensity is collected at various wavelengths upon an
external perturbation. The time correlation function is
evaluated between wavelengths. A two-dimensional fluo-
rescence correlation spectrum, or a plot of the correlation
intensity as a function of two wavelength axes, resolves
the overall spectrum into component spectra. The char-
acteristics of the two-dimensional time correlation func-
tion are demonstrated in the frequency domain fluores-
cence spectroscopy in which the sinusoidally modulated
excitation provides the external perturbation. Using 2D
FCS, fine vibronic structures of the component fluores-
cence emission spectra were completely resolved from a
strongly overlapped one-dimensional mixture spectrum.
The existence of multiple microenvironments of a probe
molecule in a biological system is evidenced by nonzero
asynchronous correlation intensities. The corresponding
spectra are retrieved from correlation analysis. Unlike
traditional resolution methods in fluorescence spectros-
copy based on statistical fitting of fluorescence decays,
2D FCS can resolve species whose fluorescence decays
are linked by the rate constants in chemical reactions and
species displaying multiexponential decay kinetics.
Molecular fluorescence is widely used for probing molecular
structure and interactions because of its high signal-to-noise ratio
and its sensitivity to the physicochemical properties of the
surrounding microenvironments.1-5 In the gas phase, high-
resolution fluorescence spectroscopy allows detailed investigation
of molecular dynamics.1 In macromolecular systems, the fluores-
cence of the probe molecules, either intrinsic or extrinsic to the
molecular system, reveals the rigidity of and solvent access to
(1) Jacobson, M. P.; Field, R. W. J. Phys. Chem. A 2000, 104, 3073-3086.
(2) Eftink, M. R. Methods Enzymol. 1997, 278, 258-286.
(3) Millar, D. P. Curr. Opin. Struct. Biol. 1996, 6, 637-642.
(4) Dahms, T. E.; Szabo, A. G. Methods Enzymol. 1997, 278, 202-221.
(5) Millar, D. P. Curr. Opin. Struct. Biol. 1996, 6, 332-336.
2302 Analytical Chemistry, Vol. 73, No. 10, May 15, 2001
the local molecular structure at the site where the probe molecule
is located.2-6 Fluorescence spectroscopy has been used to study
the guest-host interaction in cyclodextrins,7,8 micelles,9,10 pro-
teins,11,12 DNA,13 polymers,14 modified silica surfaces,15-18 biogels,19
and dendrimers.20 In addition, the superior sensitivity of fluores-
cence makes it an excellent technique for environmental21,22 and
biomedical sensing.23
A challenge in fluorescence spectroscopy of complex multi-
component chemical and biological samples is to resolve the
component spectra from the overall fluorescence.1-5,21,23 The
components can be various states or isotopomers of a molecule
in the gas phase, a probe molecule in multiple microenvironments,
or different molecules in an environmental or biological sample.
Severe overlap between many states of gas phase molecules can
lead to highly congested fluorescence spectra that are difficult to
resolve. In condensed phases, inhomogeneous broadening results
in broad vibronic bands in fluorescence spectra for individual
components. These component spectra overlap each other to form
an overall sample spectrum that generally lacks detailed structural
features and is difficult to interpret. To investigate the molecular
environment of each probe molecule or a probe molecule at
different sites of the macromolecular structure, it is essential to
resolve the component spectra from the overall sample spectrum.
(6) Lakowicz, J. R. Principles of Fluorescence Spectroscopy; Kluwer Academic/
Plenum Publishers: New York, 1999.
(7) Dey, J.; Roberts, E. L.; Warner, I. M. J. Phys. Chem. A 1998, 102, 301-305.
(8) Agbaria, R. A.; Roberts, E.; Warner, I. M. J. Phys. Chem. 1995, 99, 10056-
10060.
(9) Freeman, K. S.; Tang, T. T.; Shah, R. D. E.; Kiserow, D. J.; McGown, L. B.
J. Phys. Chem. B 2000, 104, 9312-9316.
(10) Villegas, M. M.; Neal, S. L J. Phys. Chem. A 1997, 101, 6890-6896.
(11) Cuppett, C. M.; Doneski, L. J.; Wirth, M. J. Langmuir 2000, 16, 7279-
7284.
(12) Doody, M. A.; Baker, G. A.; Bright, F. V. Anal. Chem. 2000, 72, 227-233.
(13) Davidson, Y. Y.; Gunn, B. M.; Soper, S. A. Appl. Spectrosc. 1996, 50, 211-
221.
(14) Kane, M. A.; Baker, G. A.; Pandey, S.; Maziarz, E. P., III; Hoth, D. C.; Bright,
F. V. J. Phys. Chem. B 2000, 104, 8585-8591.
(15) Kovaleski, J. M.; Wirth, M. J. J. Phys. Chem. B 1997, 101, 5545-5548.
(16) Swinton, D. J.; Wirth, M. J. Anal. Chem. 2000, 72, 3725-3730.
(17) Hansen, R. L.; Harris, J. M. Anal. Chem. 1996, 68, 2879-2882.
(18) Hansen, R. L.; Harris, J. M. Anal. Chem. 1998, 70, 4247-4256.
(19) Doody, M. A.; Baker, G. A.; Pandey, S.; Bright, F. V. Chem. Mater. 2000,
12, 1142-1147.
(20) Wade, D. A.; Torres, P. A.; Tucker, S. A. Anal. Chim. Acta 1999, 397, 17-
31.
(21) Bloch, J.; Johnson, B.; Newbury, N.; Germaine, J.; Hemond, H.; Sinfield, J.
Appl. Spectrosc. 1998, 52, 1299-1304.
(22) Sharpless, C. M.; McGown, L. B. Environ. Sci. Technol. 1999, 33, 3264-
3270.
(23) Richards-Kortum, R.; Sevick-Muraca, E. Annu. Rev. Phys. Chem. 1996, 47,
555-606.
10.1021/ac001261u CCC: $20.00 © 2001 American Chemical Society
Published on Web 04/18/2001
In medical and environmental sensing, simultaneous detection of
many components without separation would greatly increase
sample throughput, reduce analysis time, and enhance portability
of sensing instruments.
Various methods have been investigated to enhance the
spectral resolution of multicomponent systems. A numerical
pattern-recognition method, extended cross correlation (XCC),24,25
has been introduced to resolve high-resolution dispersed fluores-
cence spectra of acetylene.1 Spectral classification in the tree
analysis method is based on the construction of a hierarchical
tree from spectra of different resolutions.26 Rank annihilation factor
analysis27,28 was developed to analyze fluorescence excitation-
emission matrixes.29 Besides statistical methods, the fluorescence
lifetimes have been used to enhance the spectral resolution. The
overall fluorescence spectrum of the molecular system is resolved
into component decay-associated spectra (DAS) through fitting
the globally linked fluorescence decay data at a range of
wavelengths by multiple exponential decays.30,31 The spectra of
various lifetime components are excellent reporters of the struc-
ture and dynamics of biological systems.32-34 The resolution of
emission spectra associated with individual decay times requires
that the fluorescence decay be single exponential for all molecules,
and that the decay time be distinctive for each molecule. In the
frequency domain fluorescence spectroscopy, spectral resolution
is enhanced by measurements at multiple excitation frequen-
cies.35,36 One-dimensional fluorescence correlation spectroscopy37-39
has been applied extensively in the studies of molecular dynamics,
kinetics, and diffusion in biological cells.40-42 In these experiments,
only the time dimension was used, and fluorescence intensity was
integrated over the entire emission spectrum. Spectral information
on energy levels contained in the wavelength dimension was not
utilized.
Two-dimensional correlation analysis can be used to resolve
component signals from the overall spectral response of a complex
sample. The principles of two-dimensional correlation were first
(24) Jacobson, M. P.; Coy, S. L.; Field, R. W. J. Chem. Phys. 1997, 107, 8349-
8356.
(25) Coy, S. L.; Jacobson, M. P.; Field, R. W. J. Chem. Phys. 1997, 107, 8357-
8369.
(26) Davis, M. J. J. Chem. Phys. 1993, 98, 2614-2641.
(27) Ho, C.-N.; Christian, G. D.; Davidson, E. R. Anal. Chem. 1980, 52, 1071-
1079.
(28) Sanchez, E.; Kowalski, B. R. Anal. Chem. 1986, 58, 499-501.
(29) Warner, I. M.; Christian, G. D.; Davidson, E. R.; Callis, J. B. Anal. Chem.
1977, 49, 564-573.
(30) Knorr, F. J.; Harris, J. M. Anal. Chem. 1981, 53, 272-276.
(31) Knutson, J. R.; Walbridge, D. G.; Brand, L. Biochem. 1982, 21, 4671-4679.
(32) Savikhin, S.; Xu, W.; Chitnis, P. R.; Struve, W. S. Biophys. J. 2000, 79, 1573-
1586.
(33) Lam, W. C.; Seifert, J. M.; Amberger, F.; Graf, C.; Auer, M.; Millar, D. P.
Biochem. 1998, 37, 1800-1809.
(34) Croce, R.; Dorra, D.; Holzwarth, A. R.; Jennings, R. C. Biochem. 2000, 39,
6341-6348.
(35) Millican, D. W.; McGown, L. B. Anal. Chem. 1990, 62, 2242-2247.
(36) Neal, S. L. Anal. Chem. 1997, 69, 5109-5120.
(37) Magde, D.; Elson, E.; Webb, W. W. Phys. Rev. Lett. 1972, 29, 705-708.
(38) Eigen, M.; Rigler, R. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 5740-5747.
(39) Maiti, S.; Haupts, U.; Webb, W. W. Proc. Natl. Acad. Sci. U.S.A. 1997, 94,
11753-11757.
(40) Haupts, U.; Maiti, S.; Schwille, P.; Webb, W. W. Proc. Natl. Acad. Sci. U.S.A.
1998, 95, 13573-13578.
(41) Meseth, U.; Wohland, T.; Rigler, R.; Vogel, H. Biophys. J. 1999, 76, 1619-
1631.
(42) Korlach, J.; Schwille, P.; Webb, W. W.; Feigenson, G. W. Proc. Natl. Acad.
Sci. U.S.A. 1999, 96, 8461-8466.
introduced in infrared spectroscopy by Noda.43,44 Since then,
two-dimensional vibrational (IR and Raman) correlation spectros-
copy has been used extensively to probe the conformation and
dynamics of molecular systems including peptides,45,46 proteins,47
and polymers.48 Two-dimensional fluorescence correlation spectros-
copy49-53 has been developed in recent years. In the first applica-
tion of 2D correlation analysis in fluorescence, Roselli et al. studied
the two binding sites of Yb3+ in transferrin using a series of
emission spectra collected at various excitation wavelengths
(excitation wavelength perturbation).49 Geng used a sinusoidally
modulated laser as the external perturbation to construct the two-
dimensional time correlation function, and demonstrated 2D FCS
with mixtures of polycyclic aromatic hydrocarbons (PAHs) and
ligand-protein binding systems.50 Cox and Geng demonstrated the
resolution power of 2D FCS by recovering the fine vibronic
structure in the heavily overlapped fluorescence spectra of
mixtures of pyrene and anthracene and investigated the utilities
of 2D FCS in tissue classification.51 Nakashima et al. achieved
similar resolution of PAH spectra with the external perturbation
introduced by varying the excitation wavelengths and concentra-
tion ratios.52 The two-dimensional fluorescence correlation prin-
ciples have also been applied to chemical separations. With 2D
FCS, peak resolution and species identification were substantially
improved in capillary electrophoresis.53
In this paper, we present the dynamic two-dimensional fluo-
rescence correlation spectroscopy, which utilizes the time cor-
relation function between wavelengths to enhance species reso-
lution. The method relies on the time response of fluorescence
intensity when an external perturbation is introduced. Coupling
both temporal and wavelength information, 2D FCS significantly
improves the capability of resolving overlapping components. The
high sensitivity of fluorescence makes the technique well-suited
for chemical and biomedical samples at low concentrations.
THEORY
The general principles of two-dimensional correlation analysis
have been developed by Noda. 44 In this section, we will discuss
the two-dimensional time correlation function in the fluorescence
spectroscopy, especially when the excitation source is sinusoidally
modulated in the frequency domain. In 2D FCS, an external field
is introduced to perturb the molecular system. Component
chemical species respond to the perturbation according to their
characteristic fluorescence relaxation times. The induced fluctua-
tion in fluorescence is monitored at multiple wavelengths covering
(43) Noda, I. J. Am. Chem. Soc. 1989, 111, 8116-8118.
(44) Noda, I. Appl. Spectrosc. 1993, 47, 1329-1335.
(45) Graff, D. K.; Pastrana-Rios, B.; Venyaminov, S. Yu.; Prendergast, F. G. J.
Am. Chem. Soc. 1997, 119, 11282-11294.
(46) Müller, M.; Buchet, R.; Fringeli, Urs P. J. Phys. Chem. 1996, 100, 10810-
10825.
(47) Wang, Y.; Murayama, K.; Myojo, Y.; Tsenkova, R.; Hayashi, N.; Ozaki, Y. J.
Phys. Chem. B 1998, 102, 6655-6662.
(48) Czarnecki, M. A.; Okretic, S.; Siesler, H. W. J. Phys. Chem. B 1997, 101,
374-380.
(49) Roselli, C.; Burie, J. R.; Mattioli, T.; Boussac, A. Biospectroscopy 1995, 1,
329-339.
(50) Geng, L. Presented at the 23rd Annual Conference of the Federation of the
Analytical Chemistry and Spectroscopy Societies (FACSS), September 1996.
(51) Cox, J.; Geng, L. Presented at the Pittsburgh Conference, March 1998.
(52) Nakashima, K.; Yashuda, S.; Ozaki, Y.; Noda, I. J. Phys. Chem. A 2000,
104, 9113-9120.
(53) Wang, G.; Geng, L. Anal. Chem. 2000, 72, 4531-4542.
Analytical Chemistry, Vol. 73, No. 10, May 15, 2001 2303
the spectral range of the mixture. The cross correlation between
fluorescence intensity at wavelength λ1, A(t;λ1), and intensity at
wavelength λ2, A(t;λ2), at time t, is evaluated as
1
∫
T
C(t; λ1, λ2) ) lim -T
A*(τ; λ1)A(t + τ; λ2)dτ, (1)
Tf∞ 2T
where A* denotes the complex conjugate of the fluorescence
intensity response. If the fluorescence at both wavelengths is
dominated by the same chemical species, a strong correlation will
be observed between λ1 and λ2. Two wavelengths will display zero
cross correlation if the two species responsible for fluorescence
at these wavelengths have very different response times.
The correlation function can be collected in the frequency
domain. When the external perturbation is sinusoidally modulated,
the fluorescence response is also sinusoidally modulated at the
same frequency as the perturbation, with a phase delay φ that is
determined by the fluorescence relaxation time of the molecule
A(t; λ) ) A0(λ)ei(ωt-φ) (2)
where ω is the angular modulation frequency, and A0(λ) is the
average fluorescence intensity at wavelength λ. When the rate of
fluorescence relaxation is very fast, as compared to the modulation
frequency, the phase angle is close to zero. When the relaxation
is very slow, on the other hand, the phase delay approaches 90
degrees. With a sinusoidal perturbation, the correlation function,
as defined in eq 1, is simplified to
C(ω; λ1, λ2) ) [Φ(ω; λ1, λ2) + iΨ(ω; λ1, λ2)]eiωt (3)
At a certain modulation frequency, ω, the real part of the
correlation function
Φ(ω; λ1, λ2) ) A(λ1)A(λ2) cos(φλ1) cos(φλ2) +
A(λ1)A(λ2) sin(φλ1) sin(φλ2) (4)
describes the similarity between fluorescence fluctuations at
wavelengths λ1 and λ2. The imaginary part
Ψ(ω; λ1, λ2) ) A(λ1)A(λ2) cos(φλ1) sin(φλ2) -
A(λ1)A(λ2) sin(φλ1) cos(φλ2) (5)
describes the difference between the two wavelengths. The real
and imaginary parts are termed the synchronous and asynchro-
nous correlation intensities at frequency ω.43 A(λ1) and A(λ2) are
the intensity averaged over time at wavelength λ1 and λ2,
respectively, multiplied by the modulation factor. φλ1 and φλ2 are
the phase delays at the respective wavelengths. The synchronous
and asynchronous correlation intensities are calculated from the
experimentally measured in-phase spectrum
AR(λ) ) A(λ) cos φ (6)
2304 Analytical Chemistry, Vol. 73, No. 10, May 15, 2001
and quadrature spectrum
AI(λ) ) A(λ) sin φ (7)
The in-phase and quadrature spectra are the real and imaginary
components of the fluorescence response function upon a sinu-
soidally modulated excitation. An exponential decay is represented
as a vector in the complex plane in the frequency domain, where
AR and AI are the projections of the vector to the real and the
imaginary axes, respectively.53,54 The synchronous and asynchro-
nous correlations are the dot product and the cross product
between the vectors representing fluorescence responses at the
two wavelengths
Φ(ω; λ1, λ2) ) A
B(λ1)‚A
B(λ2) ) A(λ1)‚A(λ2) cos ∆φ (8)
Ψ(ω; λ1, λ2) ) [A
B(λ1) × A
B(λ2)]‚k̂ ) A(λ1)‚A(λ2) sin ∆φ
(9)
where
∆φ ) φλ1 - φλ2 (10)
is the phase difference between the two wavelengths. k̂ is the
unit vector perpendicular to the complex plane containing the
fluorescence vectors.
The external perturbation that induces the fluorescence
fluctuation can be quite general in nature. It can be an electric
field or a mechanical perturbation applied to a polymer system to
reorient the side chains, a polarization modulation to induce dipole
orientation in a molecular solution, or chemical modulation that
varies the population ratios of components (such as H/D ratio in
high-resolution spectroscopy). The relaxation times of fluores-
cence fluctuation are related to the rates of polymer reorientation,
rotational correlation times, and rates of chemical reactions,
respectively. In this paper, we will demonstrate the enhanced
spectral resolution in two-dimensional fluorescence spectroscopy
with a light field that is sinusoidally modulated in intensity as the
external perturbation. The relaxation time used in time correlation
analysis, in this case, is the characteristic decay time of fluores-
cence.
EXPERIMENTAL SECTION
Anthracene and bovine serum albumin were purchased from
Sigma (St. Louis, MO). Pyrene and benzo[a]pyrene were obtained
from Aldrich (Milwaukee, WI). Glycogen was acquired from J. T.
Baker (Phillipsburg, NJ). 6-Propionyl-2-(dimethylamino)naphtha-
lene (PRODAN) was purchased from Molecular Probes (Eugene,
OR). The pyrene, benzo[a]pyrene, and anthracene solutions were
prepared in ethanol through serial dilutions from stock solutions
at the millimolar concentration range. The solutions were soni-
cated to break up possible microcrystals in the stock solutions.
The PRODAN-bovine serum albumin solutions were prepared
in deionized water (MilliQ-Plus, MilliPore Corp., Bedford, MA).
Benzo[a]pyrene is carcinogenic and should be handled with
extreme care.
(54) He, Y.; Geng, L. Anal. Chem. 2001, 73, 943-950.
Figure 1. Fluorescence emission spectrum of a sample containing
glycogen and benzo[a]pyrene.

The one-dimensional steady state and the phase-resolved

fluorescence spectra were collected with a spectrofluorometer
(SLM48000MHF, Jobin Yvon, Edison, NJ). To provide sufficient
SNR, a spectral bandwidth of 4 nm was maintained for the
emission in all experiments. To measure the in-phase and
quadrature spectra, the exciting light was sinusoidally modulated
using a Pockels cell. The detector, a photomultiplier tube, was
modulated at a frequency that is the sum of the excitation
frequency and a low frequency between a few hertz and a few
tens of hertz in the heterodyne detection. The detector phase
angle was calibrated using scattered exciting light of a glycogen
solution. The detector phase angle was tuned until the integrated
phase-resolved intensity of the scattered light was zero. The
detector-phase angle at this point was calibrated to be 90°. With
the detector set in-phase with and 90° phase-delayed from the
excitation, the fluorescence intensity was integrated at each
wavelength to generate the in-phase spectrum AR(λ) and the
quadrature spectrum AI(λ). The synchronous and asynchronous
fluorescence correlation spectra were calculated according to eqs
4 and 5 in MatLab (Mathworks, Inc., Natick, MA).

RESULTS AND DISCUSSION

Two-Dimensional Fluorescence Correlation Spectra. The
characteristics of two-dimensional fluorescence correlation spectra
are illustrated in Figures 1 and 2, by a sample containing glycogen
and benzo[a]pyrene. The conventional one-dimensional fluores-
cence emission spectrum of the sample (Figure 1) shows five
vibronic transitions of benzo[a]pyrene centered at 386, 395, 404,
427, and 454 nm, corresponding to the 0-0, 0-590 cm-1, 0-1154
cm-1, 0-2488 cm-1, and 0-3880 cm-1 transitions. The exciting
light scattered by glycogen appears at 364 nm. The scattered light
behaves as a molecular component with a lifetime of zero. The
two-dimensional fluorescence correlation spectra are plots of the
synchronous and asynchronous correlation intensity, or the real
and imaginary parts of the time correlation function, versus the
two axes of fluorescence emission wavelengths λ1 and λ2. The
synchronous correlation spectrum is symmetric, as shown in
Figure 2A, displaying cross-correlation peaks that are symmetric
with respect to the diagonal of the spectrum in addition to
autocorrelation peaks on the diagonal. Vibronic features of the
same species show strong cross-correlation because the fluores-
cence relaxes at an identical characteristic rate at all peaks.
Different species, however, respond to the external perturbation
Figure 2. Two-dimensional synchronous (A) and asynchronous (B)
fluorescence correlation spectra of the sample in Figure 1, in 3D and
contour plots. The conventional one-dimensional fluorescence spec-
trum is plotted on the sidebars of the contour plots to demonstrate
the correlation between bands. Both of the wavelength axes for the
2D correlation spectra are wavelengths of fluorescence emission.
Excitation wavelength, 364 nm; modulation frequency, 20 MHz.
with different relaxation times. In an ideal situation in which the
fluorescence relaxation rate is very fast for one species and very
slow for the other, the synchronous correlation between the
vibronic features of the two species should vanish. The phase
difference between such completely different species would be
90°. The vibronic features would be clearly separated into two
groups in a synchronous correlation spectrum for such completely
Analytical Chemistry, Vol. 73, No. 10, May 15, 2001 2305
different species. The 0° phase delay of glycogen scattering
represents a situation as close to the ideal situation as possible
by generating the largest phase difference between the two
component species in the sample. Significant off-diagonal cross-
correlation intensity was observed, however, as a result of the
non-90° phase difference, as illustrated in the contour plot of
Figure 2A. In fact, at the modulation frequency of 20 MHz, benzo-
[a]pyrene, whose fluorescence lifetime is 13.7 ns in ethanol, is
delayed in-phase by 59.8°, as compared to glycogen. The syn-
chronous correlation intensity, or the dot product between the
glycogen and the benzo[a]pyrene vectors, shows 50% of the
projection of the benzo[a]pyrene to the glycogen vector (eq 8).
The asynchronous correlation spectrum (Figure 2B) is antisym-
metric. For each peak of positive correlation, a negative peak exists
at the reflection point with respect to the diagonal. Most
importantly, the asynchronous correlation intensity becomes zero
between vibronic transitions of the same fluorescent species,
including both the autocorrelation and cross-correlation, as
predicted by Equation 9. This forms the basis of component
resolution in 2D FCS of multicomponent molecular systems. The
overall fluorescence spectrum of the mixture is resolved into
individual spectra of glycogen and benzo[a]pyrene by drawing
cross sections at 395 and 364 nm, respectively. The one-
dimensional slice at 395 nm of the two-dimensional asynchronous
correlation spectrum shows a single peak at 364 nm, the spectrum
of glycogen. The five vibronic bands of benzo[a]pyrene are
obtained by drawing a cross section at 364 nm in the 2D
correlation spectrum.
Figure 3. Fluorescence emission spectra of pyrene (curve i in 3A),
Spectral Resolution with Two-Dimensional Fluorescence anthracene (curve ii in 3A), and a mixture of anthracene and pyrene
Correlation Spectroscopy. The advantages of 2D FCS over (3B).
conventional one-dimensional fluorescence are illustrated by a
mixture of anthracene and pyrene. Anthracene and pyrene are 2 individual spectra of anthracene and pyrene. Specifically, the cross
of the 16 Environmental Protection Agency priority polycyclic section at 373 nm shows fluorescence transitions of anthracene
aromatic hydrocarbon (PAH) pollutants that appear in ground- at 379, 400, 423, and 450 nm, and the cross section at 400 nm
water, aerosols, and soil. The analysis of environmental samples reveals pyrene emission bands at 373, 384, 394, and 414 nm. The
frequently involves such a mixture.21 Conventional steady-state fluorescence emission spectra of pyrene and anthracene were
(one-dimensional) fluorescence spectra of anthracene and pyrene extracted from the completely overlapped mixture spectrum that
overlap strongly, as shown in Figure 3A. Eight vibronic transitions does not show any pyrene vibronic transitions. This demonstrates
from the two molecules exist between 360 and 460 nm, which the superior ability of 2D FCS to enhance spectral resolution for
makes this a very congested spectral range. The one-dimensional multicomponent molecular systems.
fluorescence spectrum of the mixture is depicted in Figure 3B. Two-Dimensional Fluorescence Correlation Spectroscopy
The spectrum shows prominent fluorescence vibronic features of of an Equilibrium System. In equilibrium systems such as a
anthracene, but pyrene vibronic bands are absent. The only ligand bound to a biological macromolecule, it is often necessary
indications of pyrene in the mixture are the broadening of peak to resolve the various forms of the ligand in order to fully
widths of anthracene transitions and a small shoulder peak of the characterize the macromolecule. The ligand is distributed into
384 nm band. many microenvironments, both the free unbound form and various
The two-dimensional synchronous and asynchronous fluores- binding forms. In a simple solution, it is possible to vary
cence correlation spectra of the mixture (Figure 4) consist of many experimental conditions, such as the concentration ratio of the
peaks of correlation. The synchronous spectrum (Figure 4A) ligand to the macromolecule, to enrich a certain form of the ligand.
shows strong autocorrelation on the diagonal, but the asynchro- The spectral and structural properties of this form can thus be
nous spectrum (Figure 4B) displays zero autocorrelation intensity. probed. In an in vivo experiment, such as a measurement in a
As the result of spectral congestion, the synchronous correlation biological cell, however, it can be very difficult to enrich various
spectrum does not provide resolution enhancement over a one- forms of the ligand. A technique that resolves the free and bound
dimensional spectrum. Upon close examination, it is evident that forms without any a priori information is desirable.
correlation peaks in the asynchronous spectrum form two rect- The two-dimensional fluorescence correlation spectra of
angular grids, which indicates the existence of two fluorescent PRODAN in bovine serum albumin are shown in Figure 5. The
species in the solution. More importantly, the cross sections in one-dimensional fluorescence of the binding system, depicted as
the asynchronous spectrum recovered fluorescence bands of the sidebars of the 2D spectra, displays a single broad peak. It is not
2306 Analytical Chemistry, Vol. 73, No. 10, May 15, 2001

Figure 4. Two-dimensional synchronous (A) and asynchronous (B)
fluorescence correlation spectra of the mixture in Figure 3. A one-
dimensional fluorescence emission spectrum of the mixture (curve i)
is plotted on the sidebars of the correlation spectra to demonstrate
cross-correlation between bands. Vibronic bands in the fluorescence
of pyrene (curve ii) and anthracene (curve iii) are recovered from the
two-dimensional asynchronous correlation spectrum by drawing cross
sections, as illustrated in B. Excitation wavelength, 325 nm; modula-
tion frequency, 17 MHz.
clear from the 1D spectrum whether multiple microenvironments
exist for the ligand PRODAN. Similarly, the 2D synchronous
correlation spectrum exhibits a single, broad feature. The two
antisymmetric peaks in the asynchronous correlation spectrum,
however, clearly suggested the existence of two microenviron-
ments. These correspond to the equilibrium between two forms
of PRODAN, the free and bound forms. The fluorescence emission
spectra of PRODAN molecules in the aqueous solution and bound
to bovine serum albumin are recovered from drawing cross
sections in the two-dimensional asynchronous correlation spec-
trum. The strong overlap between the fluorescence spectra of the
free and bound PRODAN resulted in an equilibrium fluorescence
spectrum and a synchronous correlation spectrum consisting of
a single broad peak devoid of any spectral details. It is the zero
asynchronous correlation intensities between wavelengths for a
species that made the spectral resolution possible.
Figure 5. Synchronous (A) and the asynchronous (B) correlation
spectra of 10 µM PRODAN in 7.4 mg/mL bovine serum albumin.
Excitation wavelength, 365 nm; modulation frequency, 30 MHz.
Figure 6. Asynchronous correlation spectrum of 4.5 mM pyrene in
ethanol. Excitation wavelength, 325 nm; modulation frequency, 5
MHz.
Two-Dimensional Fluorescence Correlation Spectroscopy
of a Reacting System. The two-dimensional asynchronous
fluorescence correlation spectrum of 4.5 mM pyrene in ethanol
is shown in Figure 6. At high concentrations, an excited-state
pyrene molecule reacts with a ground-state molecule to form an
excited-state dimer (excimer). In contrast to the monomer
fluorescence, which displays narrow vibronic bands, the excimer
fluoresces at a longer wavelength with a broad structureless peak.
The fluorescence emission of both the monomer and the excimer
displays double exponential decays. The decay times are functions
of the fluorescence lifetimes of the monomer and the excimer
and the rate constants of excimer formation. Specifically, the
Analytical Chemistry, Vol. 73, No. 10, May 15, 2001 2307

Figure 7. One-dimensional fluorescence spectrum of (A) 4.5 mM
pyrene and the fluorescence spectra of (B) the monomer and (C)
the excimer of pyrene resolved from the two-dimensional asynchro-
nous fluorescence correlation spectrum.
fluorescence decay in the monomer wavelength region follows a
double exponential decay law55
[ ]
kFM(Γ2 - X) X - Γ1 -Γ2t
IM(t) ) e-Γ1t + e (11)
(Γ2 - Γ1) Γ2 - X
but the excimer fluorescence decays according to
kFDkDMCM -Γ1t
ID(t) ) [e - e-Γ2t]
(Γ2 - Γ1)
CM is the concentration of pyrene. The two apparent decay rates
Γ1 and Γ2, as well as X and Y, are dependent on the radiative
fluorescence decay rates of the monomer, kFM, and the excimer,
kFD; the total relaxation rates of the monomer, kM, and the excimer,
kD; the bimolecular rate constant of excimer formation, kDM; and
the rate constant of excimer dissociation kMD
X ) kM + kDMCM
Y ) kD + kMD
1 optical probes will be used to direct the exciting laser beam into
Γ2,1 ) {X + Y ( [(Y - X)2 + 4kMDkDMCM]1/2
2
The fluorescence decay of this system possesses two interesting
features. First, both the monomer and the excimer deactivate to
the ground state by a double-exponential process. Second, the
monomer and the excimer share two identical decay rates. The
two-dimensional asynchronous correlation spectrum shows two
correlation peaks having opposite signs, which indicates the
correlation between the monomer and the excimer fluorescence.
The fluorescence emission spectra of the monomer and the
excimer are obtained from one-dimensional slices in the 2D
spectrum. The recovered spectra are illustrated in Figure 7, along
with the conventional 1D spectrum.
(55) Birks, J. B. Photophysics of Aromatic Molecules; Wiley-Interscience: New
York, 1970.
2308 Analytical Chemistry, Vol. 73, No. 10, May 15, 2001
The capability of resolving individual spectra of components
that display multiexponential decays is a unique feature of the
two-dimensional fluorescence correlation spectroscopy. In other
techniques of resolution, such as the decay-associated spectra
(DAS), fluorescence decays are collected at various wavelengths.
All of the decay curves are globally linked and fit with nonlinear
least-squares method (NLLS). The individual fluorescence spectra
of each fluorescence decay time are calculated from the fractional
intensity contributions at all wavelengths. Molecules whose
fluorescence decays cannot be described by single exponential
decays thus pose a problem for these conventional methods of
spectral resolution. In a reactive system such as the excimer
formation, the fluorescence decays of the reactants and the
products are associated with each other by the reaction kinetics.
For example, pyrene monomer and excimer share exactly the
same two decay constants. Because DAS resolves the individual
spectra of lifetime components, it cannot handle reactive systems.
Two-dimensional fluorescence correlation spectroscopy resolves
the individual spectra by the overall decay characteristics of the
components. The decay can be multiexponential. The pyrene
excimer decay presents a quite complex example in which the
temporal profile of fluorescence contains an exponential increase
and an exponential decay. Two-dimensional fluorescence correla-
tion spectroscopy can be applied to these complex decay profiles.
The capability of analyzing reactive systems opens a new field in
which the traditional methods are ineffective.
In addition to the capabilities of analyzing fluorescent mol-
ecules that display multiexponential decays, systems undergoing
(12) chemical reactions, and in vivo measurements without a priori
knowledge of the sample, 2D FCS affords faster and simplified
measurements. To calculate the correlation spectra, the in-phase
and quadrature fluorescence spectra are collected experimentally.
The integrative nature of phase-resolved fluorescence results in
a higher signal-to-noise ratio than that available in either the time-
domain or the frequency-domain fluorescence lifetime measure-
ments.54 The direct consequence of the improved S/N is the
reduced experimental time, which can be very beneficial for
biological samples and in vivo measurements. Biological samples
(13)
are susceptible to photodegradation, and prolonged illumination
by the light source can alter the properties of the sample. In in
(14)
vivo measurements such as optical diagnosis of diseases,23 fiber
(15) patients for optical measurements. High-speed of measurement
is one of the most important requirements in theses situations.
The separation of component spectra is accomplished by
coupling spectral and temporal resolution in 2D FCS. In chemical
and biological samples, these components can be different
fluorescent molecules in a multicomponent sample, the free and
bound ligands, a fluorescent probe at various local structures in
a polymer or a biomolecule, and different states or isotopomers
of a gas-phase molecule. Resolution provided by 2D FCS can help
to elucidate the physicochemical properties of these systems.
The two-dimensional correlation principle has been applied to
2D IR in the studies of a multitude of molecular systems. When
compared to IR, fluorescence spectroscopy has a much higher
sensitivity, which makes 2D FCS a suitable technique to investi-
gate chemical and biological samples at low concentrations. 2D
FCS provides the much-desired enhancement in spectral resolu-
tion in fluorescence, in which the inhomogeneous broadening and
spectral overlap are much more severe than in infrared spectros-
copy. An added advantage in fluorescence is the existence of both
the excitation and the emission axes. Two-dimensional correlation
analysis between the in-phase and quadrature excitation spectra
will allow the separation of individual excitation spectra of mixture
components. The correlation between the excitation and the
emission spectra can facilitate the recovery of all of the energy
levels for both the ground and the excited states of individual
components.
One-dimensional fluorescence correlation spectroscopy (FCS)
has been shown to be a powerful technique for chemical physics
and biophysics.37-39 In general, it does not provide species
resolution. Recent work has shown that resolution of two species
by one-dimensional FCS is possible in favorable conditions,
namely, when the diffusion constants of the two species are
significantly different.40,41 The addition of the spectral dimension
into FCS will help spectral resolution that is desirable in biophysi-
cal investigations.
The two-dimensional correlation analysis shown in this paper
is based on the fluorescence lifetimes. Photon excitation provides
the external perturbation. The principle can be extended to
techniques of fluorescence anisotropy decay and fluorescence
recovery after photobleaching, to probe the kinetics of rotational
reorientation and lateral diffusion. Through two-dimensional
fluorescence correlation analysis, the translational and rotational
diffusion of different fluorescent components can be investigated.
Perturbation sources other than photons can also be used. For
example, electric fields can be applied to reorient polymers. The
relaxation rates of different segments or branches of the polymers
can be used as the basis of component resolution in 2D FCS.
CONCLUSION
By coupling spectral and time responses, two-dimensional
fluorescence correlation spectroscopy provides significantly im-
proved resolution in fluorescence spectra of systems containing
multiple components. The technique does not require a priori
assumption of models for individual component spectra or the
decay kinetics. It is based on physical properties of the molecular
system: energy levels of individual components and their relax-
ation times responding to the external perturbation, rather than
post-experiment statistical analysis, as in many existing methods
of spectral resolution.
ACKNOWLEDGMENT
This work was supported by the National Institute of Health
(CA82706), Spectronics Instruments, Inc., and the University of
Iowa (Central Investment Fund for Research Enhancement). We
thank Dr. Robert Fugate and Dr. Kevin Mathias for lending an
instrument for the initial experiments of the project. We acknowl-
edge the Center for Biocatalysis and Bioprocessing of the
University of Iowa for a fellowship awarded to G.W.
Received for review October 26, 2000. Accepted February
18, 2001.
AC001261U
Analytical Chemistry, Vol. 73, No. 10, May 15, 2001 2309