Single-molecule studies of diffusion by oligomer-bound dyes in organically modified sol-gel-derived silicate films

Article abstract of DOI:10.1021/ac0491511

Single-molecule fluorescence spectroscopy is used to study dye diffusion within organically modified silicate (ORMOSIL) films. ORMOSIL films are prepared from sols containing tetraethoxysilane and isobutyltrimethoxysilane in 2:1 and 1:9 molar ratios. Nile

Full text of DOI:10.1021/ac0491511

Anal. Chem. 2005, 77, 486-494
Single-Molecule Studies of Diffusion by
Oligomer-Bound Dyes in Organically Modified
Sol-Gel-Derived Silicate Films
Skylar A. Martin-Brown, Yi Fu, Ginagunta Saroja, Maryanne M. Collinson,* and Daniel A. Higgins*
Department of Chemistry, Kansas State University, 111 Willard Hall, Manhattan, Kansas 66506
epoxys.4-6 The preparation of such organic/inorganic composites
provides a means to produce silicate materials with continuously
tunable chemical and physical properties by simply changing the
precursors employed, their molar ratios, or both. ORMOSIL
materials with tailored hydrophobicity, flexibility, reactivity, poros-
ity, and stability have previously been described,4-6 as have their
applications in the areas of ion exchange materials,7 chemical
sensors,8-10 catalysts,11-14 and stationary phases for chromato-
graphic separations.15,16
Of utmost importance to these applications is the rate at which
reagents and encapsulated guests move through the material of
choice. Diffusion in sol-gel-derived materials, however, can be
very complicated.17-21 Dopant-matrix interactions, matrix poros-
ity, and pore interconnectivity all influence diffusion in sol-gel-
derived matrixes.17-20 One powerful method for studying diffusion
of dyes in thin films or on their surfaces is fluorescence correlation
spectroscopy (FCS).21-26 FCS methods have been applied previ-
ously in studies of silicate materials. For example, Dai and co-
workers have studied the diffusion of rhodamine 6G in mesopo-
rous silicate glasses.23 Their results showed that diffusion through
these materials occurred on two distinct time scales. The fast
component was attributed to the motions of “free” molecules, while
Single-molecule fluorescence spectroscopy is used to
study dye diffusion within organically modified silicate
(ORMOSIL) films. ORMOSIL films are prepared from sols
containing tetraethoxysilane and isobutyltrimethoxysilane
in 2:1 and 1:9 molar ratios. Nile red and a new silanized
form of nile red that can be covalently attached to the
silicate matrix are used as fluorescent probe molecules.
The number and rate of single molecules diffusing through
these films increases dramatically with increasing film
organic content. Autocorrelation of the fluorescence im-
ages yields a quantitative measure of the relative popula-
tions of fixed and diffusing species. Surprisingly, both
“free” and silicate-bound nile red exhibit relatively facile
translational motions. Single-molecule/single-point fluo-
rescence correlation spectroscopy (FCS) is used to mea-
sure the dye diffusion coefficients in submicrometer-scale
film regions. The most common diffusion coefficients for
“free” and silicate-bound nile red molecules in the 1:9
films are 3.9 × 10-10 and 1.6 × 10-10 cm2/s, respec-
tively. The unexpectedly rapid diffusion of silicate-bound
nile red is attributed to the presence of liquidlike silicate
oligomers in the films. A lower bound for the molecular
weight of the oligomers is estimated at 2900. Bulk
solution-phase FCS experiments performed on “free” and
silicate-bound nile red species extracted into chloroform
solutions provide valuable support for these conclusions.
Comparison of the results derived from experimental and
simulated time transients indicates film heterogeneity
occurs on sub-100-nm-length scales and likely results
from the presence of inorganic- and organic-rich domains.
(4) Wen, J.; Wilkes, G. L. Chem. Mater. 1996, 8, 1667.
(5) Collinson, M. M. Trends Anal. Chem. 2002, 21, 30.
(6) Schmidt, H. J. Sol-Gel Sci. Technol. 1994, 1, 217.
(7) Wei, H.; Collinson, M. M. Anal. Chim. Acta 1999, 397, 113.
(8) Wolfbeis, O. S.; Reisfeld, R.; Oehme, I. Sol Gels and Chemical Sensors. In
Structure and Bonding; Springer-Verlag: Berlin, 1996; Vol. 85; p 51.
(9) Lin, J.; Brown, C. W. Trends Anal. Chem. 1997, 16, 200.
(10) Dave, B. C.; Dunn, B.; Valentine, J. S.; Zink, J. I. Anal. Chem. 1994, 66,
1120A.
(11) Blum, J.; Avnir, D.; Schumann, H. Chemtech 1999, 32.
(12) Gelman, F.; Avnir, D.; Schumann, H.; Blum, J. J. Mol. Catal. A 1999, 146,
123.
(13) Abu-Reziq, R.; Avnir, D.; Blum, J. J. Mol. Catal. A 2002, 187, 277.
(14) Sandee, A. J.; Reek, J. N. H.; Kamer, P. C. J.; van Leeuwen, P. W. N. M. J.
Am. Chem. Soc. 2001, 123, 8468.
(15) Rodriguez, S. A.; Colon, L. A. Chem. Mater. 1999, 11, 754.
(16) Wang, D.; Chong, S. L.; Malik, A. Anal. Chem. 1997, 69, 4566.
(17) Howells, A. R.; Zambrano, P. J.; Collinson, M. M. Anal. Chem. 2000, 72,
5265.
(18) Watson, J.; Zerda, T. W. Appl. Spectrosc. 1991, 45, 1360.
(19) Koone, N.; Shao, Y.; Zerda, T. W. J. Phys. Chem. 1995, 99, 16976.
(20) McCain, K. S.; Hanley, D. C.; Harris, J. M. Anal. Chem. 2003, 75, 4351.
(21) McCain, K. S.; Schluesche, P.; Harris, J. M. Anal. Chem. 2004, 76, 939.
(22) Elson, E. L.; Magde, D. Biopolymers 1974, 13, 1.
(23) Mahurin, S. M.; Dai, S.; Barnes, M. D. J. Phys. Chem. B 2003, 107, 13336.
(24) Wirth, M. J.; Swinton, D. J. Anal. Chem. 1998, 70, 5264.
(25) Wirth, M. J.; Swinton, D. J. J. Phys. Chem. B 2001, 105, 1472.
(26) Wirth, M. J.; Ludes, M. D.; Swinton, D. J. Appl. Spectrosc. 2001, 55, 663.
The sol-gel process provides a convenient approach to prepare
technologically important silicate materials.1-3 Relatively simple
inorganic silicate matrixes can be made via the acid-catalyzed
hydrolysis and condensation of tetraethoxysilane (TEOS).1-3 More
complex organically modified silicate materials (ORMOSILs) can
be prepared from mixtures of organoalkoxysilanes RSi(OR′)3 and
alkoxysilanes, where R represents different organic functional
groups such as alkyl chains, aromatic rings, amines, thiols, and
* Corresponding authors. E-mail: mmc@ksu.edu; higgins@ksu.edu.
(1) Brinker, J.; Scherer, G. Sol-Gel Science; Academic Press: New York, 1989.
(2) Avnir, D. Acc. Chem. Res. 1995, 28, 328.
(3) Collinson, M. M. Structure, Chemistry, and Applications of Sol-Gel Derived
Materials. In Chalcogenide Glasses and Sol-Gel Materials; Nalwa, H. S., Ed.;
Academic Press: New York, 2001; Vol. 5; p 163.
486 Analytical Chemistry, Vol. 77, No. 2, January 15, 2005
10.1021/ac0491511 CCC: $30.25 © 2005 American Chemical Society
Published on Web 12/07/2004

slow diffusion was assigned to molecules adsorbing/desorbing
from the surface or becoming entrapped in pores. Diffusion in
silicate films prepared from silica particles has been studied by
McCain and Harris, using total internal reflection FCS21 and single-
molecule tracking.20 Their results point to micrometer-scale
variations in the diffusion coefficients observed for dye molecules
and dye-labeled dendrimers. Importantly, the level of heterogene-
ity observed varied inversely with the size of the diffusing species.
Swinton and Wirth have also used FCS methods to examine lateral
diffusion and rare strong adsorption in C18 phases fused onto silica
surfaces in contact with aqueous solutions.24,25 Their results point
to the presence of reactive sites on the silica surfaces, to which
the molecules can periodically bind.
As demonstrated by some of the aforementioned examples, it
is now possible to study diffusion (e.g., mass transport) at the
single-molecule level.27-29 Single-molecule methods offer several
unique advantages over more common bulk physical methods for
studying diffusion. Most importantly, single-molecule methods
eliminate the averaging associated with bulk methods. As a result,
spatial variations in the properties of heterogeneous materials can
be directly probed. Since sol-gel-derived hybrid organic/
inorganic composites are known to be heterogeneous,30 single-
molecule methods may be able to help unravel some of the
complexities associated with dopant diffusion.
Figure 1. Chemical structures of nile red (NR) and the silanized
form of nile red (NR-Si).
nile red (NR-Si) that covalently binds to the silicate matrix when
cohydrolyzed and cocondensed with the precursor silanes. Figure
1 shows the chemical structures of these two dyes. Diffusion is
studied in two classes of film: one derived from sols of low (33
mol %) organic content and the other from sols of high (90 mol
%) organic content. These two films are representative of the
extremes observed in our previous studies;31,32 no apparent
diffusion is observed in the first, while facile diffusion occurs in
the second. The results of these studies indicate that liquidlike
silicate oligomers are present in the organic/inorganic hybrids
and that these oligomers allow for diffusion of even the silicate-
bound dye. The results of bulk diffusion studies in which the “free”
and silicate-bound forms of NR are extracted from the films into
chloroform solution provide strong support for these conclusions.
In this paper, FCS at the single-molecule/single-point level is
used to examine diffusion in organic/inorganic hybrid silicate films
prepared by the sol-gel process. Such films likely incorporate
phase-separated domains of inorganic- and organic-rich silicate
materials.31-36 In earlier studies from our research program, it was
shown that dye molecules entrapped in hybrid silicates of high
organic content exhibited unexpectedly facile translational diffu-
sion.31,32 At lower organic content, the dye molecules appeared to
be entrapped at fixed sites within the films. The percentage of
molecules that appeared to diffuse was observed to increase with
increasing film organic content. However, the diffusion of these
molecules was not quantitatively explored in these previous
studies, which focused primarily on film polarity properties and
the possibility of phase separation of the inorganic and organic
film components.
The main objectives of the present studies include a much
more quantitative understanding of the mechanism and rates of
probe diffusion through these materials, as well as more complete
information on the level of film heterogeneity as it relates to probe
diffusion. To this end, two different dye molecules are employed
as single-molecule probes of the diffusion process. Nile red (NR)
is employed as a probe of “free” diffusion through the films. The
results are compared to those obtained from a silanized form of
EXPERIMENTAL SECTION
Sample Preparation. Nile Red Samples. ORMOSIL films were
prepared from sols containing TEOS (Aldrich 99+%) and isobu-
tyltrimethoxysilane (BTMOS, Fluka 95%). The precursor sols were
composed of either 67 mol % TEOS and 33 mol % BTMOS (2:1
molar ratio) or 10 mol % TEOS and 90 mol % BTMOS (1:9 molar
ratio). In both cases, the sols were prepared in 1:4:5:0.006 (total
silane/ethanol/water/HCl) mole ratios. Each sol was stirred for
∼30 min after addition of all ingredients. The sols were then
allowed to stand for ∼48 h at 33% humidity. For the single-
molecule experiments, a methanolic solution of NR (Aldrich) was
then added to each to yield a total dye concentration of 0.5 nM.
Immediately after addition of the dye, 100 µL of the sol was spin
cast under nitrogen onto a cleaned glass microscope coverslip
(Fisher Premium). Films of 350-550-nm thickness were ob-
tained.31 The films were dried in a desiccator overnight prior to
use. All single-molecule experiments were conducted under dry
air (20% relative humidity) at room temperature. Experiments were
performed on each sample for ∼6-8 h.
Silanized Nile Red Samples. Sols were prepared as described
above. However, for these samples, a methanolic solution of NR-
Si was added immediately after the initial stirring of the sol. NR-
Si was synthesized as described below. The dye-doped sol was
allowed to stand for 48 h, spin cast, and dried in a manner identical
to the NR-doped films.
Samples for Bulk Diffusion Studies. Thin films were generally
prepared as described above, from 90 mol % BTMOS sols.
However, more dye (NR or NR-Si) was added to each sol to yield
a final dye concentration of 10 nM in each case. Once the films
had been cast and dried, the entrapped dye molecules were
extracted into chloroform solution. This was accomplished by
transferring each sample to a small vial, to which 100 µL of
chloroform was subsequently added. The vials were then capped
and shaken vigorously for several minutes. This procedure was
(27) Dickson, R. M.; Norris, D. J.; Tzeng, Y.-L.; Moerner, W. E. Science 1996,
274, 966.
(28) Schmidt, T.; Schu¨tz, G. J.; Baumgartner, W.; Gruber, H. J.; Schindler, H.
Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 2926.
(29) Schu¨tz, G. J.; Schindler, H.; Schmidt, T. Biophys. J. 1997, 73, 1073.
(30) Dunn, B.; Zink, J. I. Chem. Mater. 1997, 9, 2280.
(31) Bardo, A. M.; Collinson, M. M.; Higgins, D. A. Chem. Mater. 2001, 13,
2713.
(32) Higgins, D. A.; Collinson, M. M.; Saroja, G.; Bardo, A. M. Chem. Mater.
2002, 14, 3734.
(33) Brennan, J. D.; Hartman, J. S.; Ilnicki, E. I.; Rakic, M. Chem. Mater. 1999,
11, 1853.
(34) Keeling-Tucker, T.; Brennan, J. D. Chem. Mater. 2001, 13, 3331.
(35) Frenkel-Mullerad, H.; Avnir, D. Chem. Mater. 2000, 12, 3754.
(36) Rottman, C.; Grader, G.; Avnir, D. Chem. Mater. 2001, 13, 3631.
Analytical Chemistry, Vol. 77, No. 2, January 15, 2005 487

Scheme 1. Synthesis of 9-Diethylamino-2-(triethoxysilyl-3-propyloxy)-5H-benzo[r]phenoxazin-5-one
(NR-Si)
repeated three times for each vial, using three films of each type.
A small quantity of the chloroform solution containing the
extracted dye was then deposited onto a microscope cover glass.
A second cover glass was placed on top to spread the solution
and to prevent its evaporation. The sample was then immediately
transferred to the microscope for characterization.
recorded in alternating (i.e., left-to-right and right-to-left) direc-
tions. The fluorescence signal was integrated for 40 ms/pixel in
the 100 × 100 pixel images. Single-point time transients were
obtained by positioning selected sample regions in the laser focus
and recording the spectrally integrated fluorescence in time. These
latter experiments employed a dwell time of 10 ms. The total
length of the time transients recorded was 670 s in the case of
NR and 1340 s for NR-Si. The results of computer simulations
described below indicated these transient lengths were sufficient
for accurate measurement of the relevant diffusion coefficients.
Bulk Solution-Phase Fluorescence Correlation Spectroscopy.
Fluorescence time transients from bulk solutions of the dyes were
recorded using the confocal microscope. For these measurements,
however, a fast multichannel scaler (Perkin-Elmer) was employed
to record the transients with much better (50 µs) time resolution.
In addition, the incident laser power into the microscope was
increased to ∼200-400 µW to obtain sufficient signal from the
dye molecules as they diffused through the focal volume of the
microscope. Finally, to ensure the data obtained were from
diffusion in bulk solution (avoiding surface effects), the microscope
focus was positioned between the two glass/solution interfaces,
which were separated from each other by distances ranging from
10 to 40 µm.
Instrumentation. Single-Molecule Imaging and Diffusion Stud-
ies. All single-molecule studies were conducted on a sample
scanning confocal microscope described previously.31,32 Briefly,
the system was built around an inverted microscope (Nikon TE-
200). Light from a green helium-neon laser (543 nm) was used
as the excitation source. The excitation beam first passed through
a 543-nm band-pass filter (CVI, 10-nm pass-band), polarization
control optics, and a telescope, before being directed into the epi-
illumination port of the microscope. A dichroic beam splitter
(Chroma 565DCLP) was used to reflect the excitation light into
the back aperture of the oil immersion objective (Nikon Plan
Fluor, 1.3 numerical aperture, 100×). This system produces a
nearly diffraction limited Gaussian spot of width s ≈ 200 nm in
the sample. The incident laser power was always maintained in
the 200-400-nW range. The sample was mounted above the
objective on a piezoelectric scanning stage (Queensgate) employ-
ing closed-loop feedback in X and Y. The software used to control
the stage and collect images and time transients was written in-
house.
The fluorescence emitted by single molecules was collected
by the same oil immersion objective. Emission to the red of 565
nm passed through the dichroic beam splitter and was directed
into the detection path. To further isolate single-molecule emission
and reduce background, a 100-nm-wide band-pass filter centered
at 650 nm was also employed. A single photon counting avalanche
photodiode was used as the detector. Images were recorded by
raster scanning the sample above the focused spot of the incident
laser. The fast-scan direction is oriented horizontally on all images
reported herein, while the slow-scan direction corresponds to
vertical on the images. Raster scanning was performed in a
bidirectional fashion so that the individual line scans were
Synthesis of Silanized Nile Red. A four-step procedure was
required to obtain NR-Si. The entire synthesis is outlined in
Scheme 1.37
Synthesis of 5-(Diethylamino)-2-nitrosophenol (2). m-Diethyl-
aminophenol (6 g, 0.0363 mol, Aldrich, 1) was dissolved in a
mixture of 13 mL of HCl and 8 mL of water. The solution was
cooled to 0 °C while stirring. To this cold solution, a solution of
NaNO2 (2.5 g, 0.036 mol) in 18 mL of water was added dropwise
over a period of 30 min, while maintaining a temperature of 0-5
°C. A vigorous exothermic reaction occurred. Once the addition
was complete, the reaction mixture was stirred at 0-5 °C for an
additional 4 h. The crude HCl salt was collected by filtration and
(37) Briggs, M. S. J.; Bruce, I.; Miller, J. N.; Moody, C. J.; Simmonds, A. C.;
Swann, E. J. Chem. Soc., Perkin Trans. 1 1997, 1051.
488 Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

Figure 2. Fluorescence excitation and emission spectra of NR (s)
and NR-Si (- - -). All spectra were taken in spectrophotometric grade
ethanol with a dye concentration of 1 µM.
dried under reduced pressure for 8 h at 50 °C. The product was
dissolved in 100-200 mL of boiling ethanol and then cooled to
40°C. Diethyl ether was slowly added until crystallization began.
The mixture was cooled to 5 °C and kept refrigerated for 1 day.
Finally, the product was collected by filtration and dried to yield
4.9 g (70%) of the title compound.
Synthesis of 9-Diethylamino-2-hydroxy-5H-benzo[a]phenoxazin-
5-one (3). 1,6-Dihydroxynaphthalene (0.725 g, 4.53 mmol) and 2
(1 g, 4.53 mmol) were heated under reflux in DMF (100 mL) for
4 h. The solvent was removed by rotary evaporation. The crude
product was purified by column chromatography using 5:2 ethyl
acetate/2-propanol to yield 1.5 g (60%) of the title compound as a
green solid: 1H NMR δH (400 MHz, DMSO-d6) 10.46 (1H, br, OH),
7.95 (1H, d, 4-H), 7.88 (1H, d, 1-H), 7.58 (1H, d, 11-H), 7.10 (1H,
dd, 3-H), 6.80 (1H, dd, 10-H), 6.64 (1H, d, 8-H), 6.15 (1H, s, 6-H),
3.49 (4H, q, N(CH2CH3)2) and 1.20 (6H, t, N(CH2CH3)2).
Synthesis of 9-Diethylamino-2-allyloxy-5H-benzo[a]phenoxazin-
5-one (4). Allyl bromide (0.5 mL, 5 mmol) and 3 (500 mg, 1.5
mmol) were dissolved in anhydrous methanol to which excess
K2CO3 was added. The reaction mixture was refluxed for 48 h.
After evaporation of the solvent, the crude product was purified
by column chromatography using 3:1 hexane/ethyl acetate to give
the title compound as a purple solid: 1H NMR δH (400 MHz,
CDCl3) 8.21 (1H, d, 4-H), 8.06 (1H, d, 1-H), 7.59 (1H, d, 11-H),
7.17 (1H, dd, 3-H), 6.65 (1H, dd, 10-H), 6.45 (1H, d, 8-H), 6.29
(1H, s, 6-H), 6.1 (1H, m, OCH2CHdCH2), 5.49 (1H, d, OCH2CHd
CH2), 5.33 (1H, d, OCH2CHdCH2), 4.73 (2H, d, OCH2CHdCH2),
3.45 (4H, q, N(CH2CH3)2), and 1.25 (6H, t, N(CH2CH3)2).
Synthesis of 9-Diethylamino-2-(triethoxysilyl-3-propyloxy)-5H-ben-
zo[a]phenoxazin-5-one (5, NR-Si).38 Compound 4 (80 mg, 0.21
mmol) was dissolved in 20 mL of anhydrous methanol in a round-
bottom flask. The flask was fitted with a condenser and drying
tube and heated to reflux. To the warm solution, 15 mg of H2PtCl6
dissolved in a few drops of 2-propanol was added, followed by
HSi(OCH2CH3)3 (0.7 g, 4.3 mmol) 10 min later. The solution
quickly changed from purple to brown upon silane addition. At
this point, heating was terminated and the flask was wrapped in
aluminum foil. The solution was subsequently stirred under dry
nitrogen for 36 h. The solvent was removed by rotary evaporation
to give the title compound. The product was stored in a drybox
or desiccator: 1H NMR δH (400 MHz, CDCl3) 8.23 (1H, d, 4-H),
8.07 (1H, d, 1-H), 7.63 (1H, d, 11-H), 7.18 (1H, dd, 3-H), 6.68 (1H,
dd, 10-H), 6.48 (1H, d, 8-H), 6.32 (1H, s, 6-H), 4.16 (2H, t,
OCH2CH2CH2Si), 3.74 (6H, q, Si(OCH2CH3)3), 3.49 (4H, q,
Figure 3. Fluorescence images of (A) a 33% BTMOS film doped
with NR and (B) a 90% BTMOS film doped with NR-Si. Both images
are of 10 × 10 µm regions. (C, D) Autocorrelations along the
horizontal (circles) and vertical (triangles) directions for the images
shown in (A) and (B), respectively. Fits to the autocorrelation functions
in each case are depicted by the solid lines. The vertical autocorre-
lation data have been offset in the y-direction. Average values for
the amplitudes of the Gaussian and diffusional decay components
(Ag and Ad, respectively) can be found in the main text.
N(CH2CH3)2), 1.90 (2H, m, OCH2CH2CH2Si), 1.33 (9H, t, Si-
(OCH2CH3)3), 1.25 (6H, t, N(CH2CH3)2), 1.11 (2H, t, OCH2-
CH2CH2Si).
RESULTS AND DISCUSSION
Fluorescence Spectra for NR and NR-Si. Figure 2 depicts
the fluorescence spectra obtained from NR and NR-Si in bulk
ethanol solution. Aside from a subtle blue shift for NR-Si, the two
dyes exhibit nearly identical fluorescence excitation and emission
spectra. Each is most efficiently excited near 550 nm, and each
emits most strongly near 625 nm. These results indicate that the
two dyes will exhibit similar fluorescence characteristics when
(38) Leventis, N.; Chen, M. Chem. Mater. 1997, 9, 2621.
Analytical Chemistry, Vol. 77, No. 2, January 15, 2005 489

probed in silicate films. Any differences observed between NR
and NR-Si single molecules may therefore be attributed to
differences in the interactions of the dye with the silicate matrix.
The NR-Si dye was prepared specifically for covalent attachment
to the matrix. Numerous studies have shown that similar orga-
noalkoxysilanes can be chemically attached to the matrix by
cohydrolysis and cocondensation with another silicon alkoxide.14,38-43
Imaging and Spatial Autocorrelations. Figure 3A shows a
representative fluorescence image of a 33% BTMOS film doped
with NR, while Figure 3B depicts such an image for a NR-Si-doped
90% BTMOS film. NR-Si-doped 33% BTMOS films and NR-doped
90% BTMOS films yield images that are nearly indistinguishable
by eye from those shown in Figure 3A and B, respectively.
Examples are given in the Supporting Information (Figure S-1).
These images are consistent with those obtained in previous
studies of a series of NR-doped hybrid silicate films prepared from
sols having a range of TEOS/BTMOS ratios.31,32 Regardless of
the specific form of NR employed, images obtained from films of
low BTMOS content are composed of bright round fluorescent
spots, consistent with the excitation and emission of single
molecules entrapped at fixed positions within the films. In contrast,
images obtained from films of high organic content (i.e., >66%
BTMOS)31 exhibit dramatic line-by-line variations in single-
molecule emission, whether they are doped with NR or NR-Si.
Previous studies have shown that translational diffusion is a major
contributor to the “streaky” appearance of these images.31,32 It is
most surprising that the NR-Si molecules exhibit what is appar-
ently facile translational diffusion (see Figure 3B), since they are
expected to be bound to the silicate matrix.
The “streaks” in such images result from a combination of
spatial and temporal contrast (referred to below as “spatiotem-
poral” contrast) associated with the mobile molecules. In these
experiments, the images are recorded by sample-scanning meth-
ods. The horizontal direction on each image represents the “fast-
scan” direction. Images are constructed by recording a series of
individual line scans in the fast-scan direction for a series of
positions along the vertical direction. The vertical direction on an
image represents the “slow-scan” direction. Individual image pixels
along the horizontal direction are recorded at 40-ms intervals,
while the image pixels along the vertical direction (in alternating
horizontal lines) are separated by 8 s (the latter complexity arises
from the bidirectional nature of the fast scan axis). Hence, the
data in the horizontal and vertical image directions incorporate
temporal information on two distinct time scales. Molecules may
appear in a fixed position for a single horizontal line scan and
subsequently move to a new position prior to the recording of
the next line scan. Images incorporating narrow horizontal
fluorescent streaks result, reflecting temporal variations in the
positions of mobile molecules. The horizontal widths of these
streaks are often governed by diffraction-limited excitation of the
molecules, while their vertical widths reflect the rates of temporal
variation in their positions.
Autocorrelation of these fluorescence images44 can then be
used to determine the relative contributions of fixed and mobile
molecules to the signal fluctuations (“spatiotemporal” contrast)
observed. Representative autocorrelations are shown in Figure
3, along with the images from which they were derived. These
image autocorrelations were calculated as follows:44
C(ê) ) i(x)i(x + ê) / i(x)2
(1)
where i(x) represents an intensity trace along the horizontal (x)
direction in an image. A similar expression defines the autocor-
relation in the vertical (y) direction. Again, the x and y coordinates
in these image autocorrelations contain both spatial and temporal
information. The brackets (< >) indicate the average over all
points in each trace is taken. The autocorrelation functions plotted
in Figure 3 represent average autocorrelation functions, Ch(ê),
obtained by averaging all autocorrelation functions in the hori-
zontal and vertical directions for each image. As noted above, the
images were recorded by raster scanning the sample in a
bidirectional fashion. Therefore, to maintain a constant temporal
separation (8 s) between the data points in the vertical image
direction, alternate pixels were employed to produce two separate
autocorrelations along the vertical direction for each image. This
process results in a 2-fold reduction in the data density. All such
autocorrelations were then averaged together for use in Figure
3.
To obtain a measure of the relative contributions of fixed and
mobile molecules to the signal fluctuations in the images, the
autocorrelations calculated above were fit with an approximate
decay function. For this model, variations in the signal were
assumed to arise from fixed molecules, whose intensities fall off
spatially with a Gaussian (i.e., diffraction-limited) profile, and
diffusing molecules, whose intensities fluctuate more rapidly
because of their time-dependent motions. Hence, the fitting
expression combines both Gaussian and diffusional decays as
follows:23,25,45
Ad
2
ê
2s
Ch(ê) )
+ Ag exp -
+ B
(2)
(
(
) )
(1 + D′|ê|)
In eq 2, Ad and Ag are the amplitudes of the decays associated
with diffusing and fixed species, respectively. D′ is related to the
diffusion coefficient of the molecule, the scan rate, and the beam
variance but is treated here simply as a fitting parameter (i.e., it
was not used to determine the diffusion coefficient), s2 is the
variance of the focused laser beam used to excite the molecules,
and B is a constant. The value used for s (1.9 × 10-5 cm) was
determined by fitting the 33% BTMOS image autocorrelations to
eq 2. In fitting the 90% BTMOS data, its value was held fixed to
within 25% to allow for variations in the microscope focus. The
relative contributions of fixed and diffusing molecules to the spatial
signal fluctuations were then obtained from the relative amplitudes
(39) Pandey, S.; Baker, G. A.; Kane, M. A.; Bonzagni, N. J.; Bright, F. V. Chem.
Mater. 2000, 12, 3547.
(40) Chambers, R. C.; Haruvy, Y.; Fox, M. A. Chem. Mater. 1994, 6, 1351.
(41) Suratwala, T.; Gardlund, Z.; Davidson, K.; Uhlmann, D. R.; Watson, J.;
Peyghambarian, N. Chem. Mater. 1998, 10, 190.
(42) Suratwala, T.; Gardlund, Z.; Davidson, K.; Uhlmann, D. R.; Watson, J.; Bonilla,
(44) Petersen, N. O.; Hoddelius, P. L.; Wiseman, P. W.; Seger, O.; Magnusson,
K.-E. Biophys. J. 1993, 65, 1135.
S.; Peyghambarian, N. Chem. Mater. 1998, 10, 199.
(43) Rottman, C.; Turniansky, A.; Avnir, D. J. Sol-Gel Sci. Technol. 1998, 13, 17.
(45) Srivastava, M.; Petersen, N. O. Biophys. Chem. 1998, 75, 201.
490 Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

of the two decay components as Ag/(Ad + Ag) and Ad/(Ad + Ag),
respectively.
To properly fit the image autocorrelations, the contributions
of uncorrelated shot noise to the first point, Ch(ê ) 0), in each
was subtracted. Shot noise was removed by simply subtracting
the average signal in each image from Ch(ê ) 0) and renormalizing
the corrected autocorrelation. This process improves the auto-
correlation fits dramatically, but the contributions of uncorrelated
noise from other sources are still present and limit the accuracy
to some degree. While such corrections are inexact, they are
necessitated by the limited data density in the image autocorre-
lation functions.
For 33% BTMOS films doped with either NR or NR-Si, the
autocorrelation functions could be fit well with Ag ) 1 and Ad )
0 in every instance, indicating that the dyes remain in “fixed”
locations on the time scale of the experiment. For data obtained
from 90% BTMOS films, the relative contributions of fixed and
mobile molecules depend on the autocorrelation direction (see
Figure 3D), as well as the specific dye molecule used as the
dopant. For NR in the horizontal (fast scan) direction, average
values of Ag ) 0.35 ( 0.15 and Ad ) 0.65 ( 0.14 were obtained
from several images. The same images yielded averages of Ag)
0.16 ( 0.05 and Ad ) 0.84 ( 0.05 for the vertical (slow scan)
direction, for those images having a nonzero Ag value. Ap-
proximately 43% of the images yielded Ag ≈ 0 for autocorrelations
in the vertical direction. For NR-Si in the horizontal direction,
average values of Ag ) 0.71 ( 0.28 and Ad ) 0.29 ( 0.28 were
obtained from several images, while average values of Ag ) 0.17
( 0.02 and Ad ) 0.83 ( 0.02 were obtained in the vertical direction
for those images yielding nonzero Ag values. The values obtained
for Ag (and Ad) in the horizontal direction differ for NR and
NR-Si at greater than the 90% confidence level. These results
indicate that a significantly smaller fraction of the NR-Si molecules
is “mobile” on the time scale of the horizontal line scans in 90%
BTMOS samples, as expected for silicate-bound molecules.
Fluorescence Time Transients and Temporal Autocorre-
lations. Information on the rate of single-molecule diffusion was
obtained by recording single-point fluorescence time transients.
In these studies, selected points in each sample were positioned
above the focused excitation spot and the integrated (total) sample
fluorescence recorded as a function of time, with 10-ms resolution.
These time transients were then autocorrelated and the autocor-
relation functions fit to appropriate expressions, as defined below.
Figure 4 shows representative results of these experiments for
NR in a 33% BTMOS film (Figure 4A,B) and for NR-Si in a 90%
BTMOS film (Figure 4C,D).
Figure 4. Representative single-point fluorescence time transients
and their autocorrelations. (A, B) For NR in a 33% BTMOS film. (C,
D) for NR-Si in a 90% BTMOS film.
NR and NR-Si in 90% BTMOS films yielded time transients
dramatically different from the single molecules entrapped in 33%
BTMOS films (see Figure 4C,D). In 90% BTMOS films, the
fluorescence time transients exhibited large-amplitude, rapid signal
fluctuations over much longer periods of time (the longest time
transients recorded for NR-Si were 22.3 min in length). No clear
evidence of single-molecule photobleaching was observed for
either NR or NR-Si, even for these extremely long transients. Such
an observation represents strong evidence for relatively facile
diffusion of different single molecules in to and out of the
excitation region of the microscope, even in the case of the silicate-
bound NR-Si.
The fluorescence time transients recorded for NR and NR-Si
single molecules in the 33% BTMOS films and for single locations
in the 90% BTMOS films were used to calculate autocorrelation
functions as follows:
The signal in Figure 4A arises from a single molecule (a single
round fluorescent spot) entrapped in the film at a fixed location.
The shutter blocking the excitation beam was opened early in
the transient, leading to the initial abrupt increase in fluorescence.
Emission from this molecule remained relatively stable for a period
of ∼70 s. At that time, an irreversible transition to a dark state
was observed, likely due to photodestruction of the chromophore.
This particular molecule survived longer than most molecules in
these materials, which typically last no more than several seconds.
However, the “constant” emission observed prior to photobleach-
ing is representative of typical single molecules in films of low
BTMOS content.31
C(τ) ) i(t)i(t + τ) / i(t)2
(3)
where i(t) represents the fluorescence time transient. Autocor-
relations obtained from fixed molecules in the 33% BTMOS films
were usually fit to exponential decays in order to obtain a measure
of the characteristic time scales for their signal fluctuations.31 Of
primary interest here, however, are the diffusion coefficients
associated with single molecules diffusing in to and out of the
excitation region in the 90% BTMOS films. In this case, the
autocorrelations were fit with the following equation:23-26
Analytical Chemistry, Vol. 77, No. 2, January 15, 2005 491

A1
A2
C(τ) )
+
+ B
(4)
(1 + D1τ/s2) (1 + D2τ/s2)
Here, D1 and D2 are diffusion coefficients associated with two
separate diffusional decays, A1 and A2 are the amplitudes of these
two decays, and s2 is the variance of the excitation region in the
microscope (s2 ) 3.7 × 10-10 cm2). In fitting these temporal
autocorrelations, the first point in each, C(τ ) 0), was simply
excluded from the fit, rather than attempting a correction as in
the case of the image autocorrelations. Also, only the first 5% of
each autocorrelation was fit, since insufficient information on
longer time-scale fluctuations exists in these data.46 A majority of
the autocorrelations (67% in the case of NR and 50% in the case
of NR-Si) analyzed herein required two separate diffusional
components to properly fit the observed decays, as shown in eq
4.23 A minority of the autocorrelations could, however, be fit with
a single component (i.e., A2 ) 0 in eq 4). Previously, a combination
of diffusional and exponential decays have been employed to fit
transients that exhibited both diffusion and strong adsorption of
single molecules on similar surfaces.24-26 However, the transients
recorded for 90% BTMOS films do not show evidence of long-
term adsorption, and hence, the present data have been inter-
preted to reflect diffusional processes alone. In reality, the signal-
to-noise ratio in most autocorrelations is too small to distinguish
between these two models based solely on the quality of the fit.23
Autocorrelations from 123 sites in NR-doped 90% BTMOS
samples and 112 sites in NR-Si-doped samples were fit to obtain
measurements of their respective diffusion coefficients. Figure 5
depicts histograms showing the full inhomogeneous distributions
for both the fast- and slow-diffusion components (D1 and D2, from
eq 4) observed in the data. The histograms given in Figure 5A,B
(fast diffusion component) were fit to Gaussian functions to
determine the peak of each distribution and, hence, the most
common diffusion coefficient observed in each sample. NR-doped
90% BTMOS samples yielded Dpeak ) 3.9 × 10-10 cm2/s and the
NR-Si-doped samples Dpeak ) 1.6 × 10-10 cm2/s. The average D
values obtained by numerically averaging all values obtained for
each sample were 1.2 × 10-9 and 2.9 × 10-10 cm2/s, respectively.
Hence, “free” NR diffusion through the films is only ∼3-4 times
faster than silicate-bound NR-Si. Similarly subtle differences in
the rotational mobilities of silicate-bound and unbound pyrene
molecules have been reported previously by Fox and co-workers.40
The small difference observed here suggests that the 90% BTMOS
films are best described as a “fluid” matrix, possibly incorporating
liquidlike organosilicate oligomers. These oligomers likely result
from incomplete condensation of the hydrolyzed organoalkoxysi-
lanes and would facilitate the translational motions of “free” NR.
In the case of silicate-bound NR-Si, it is hypothesized that the
diffusing species is actually a silicate-oligomer-bound NR-Si.
Bulk solution-phase FCS experiments were undertaken to test
the latter hypothesis. In these studies, time transients for bulk
chloroform solutions of both dyes were recorded and autocorre-
lated in order to obtain a measure of their solution-phase diffusion
coefficients. Figure S-2 in the Supporting Information presents
example results. Two different types of solution-phase samples
were characterized. In the first set of samples, NR and NR-Si
Figure 5. Histograms of the diffusion coefficients derived from fits
of more than 100 autocorrelations for each 90% BTMOS sample. (A,
B) Diffusion coefficient distributions for the fast components (D1) of
the autocorrelation decays for NR and NR-Si, respectively. (C, D)
Diffusion coefficients derived from the slow-decay components (D2)
for NR and NR-Si, respectively.
solutions prepared using fresh dye were characterized. In the
second, NR and NR-Si were extracted into chloroform solution
from previously prepared 90% BTMOS films. The diffusion
coefficients of these “liberated” dye molecules were then mea-
sured. Time transients (50-µs resolution, 3.3-s duration) for these
samples were obtained by focusing into the bulk of each solution.
The transients were autocorrelated as defined in eq 3 and fit to a
modified form of eq 4,23 assuming a focus depth of 2s. In all cases,
the bulk autocorrelation decays could be fit well to a single-
component diffusional decay (i.e., A2 ) 0 in eq 4).
Fresh NR in chloroform yielded D ) (1.9 ( 0.3) × 10-6 cm2/
s, while fresh NR-Si yielded D ) (1.8 ( 0.3) × 10-6 cm2/s. The
similarity in diffusion coefficients for fresh NR and NR-Si is
expected since the addition of the alkyltriethoxysilane moiety to
the NR structure does not significantly change the size of the
molecule. In contrast, the diffusion coefficients for the NR and
NR-Si species “liberated” from the silicate (90% BTMOS) films
were significantly different, exactly as observed in the single-
molecule/single-point measurements on the films. Specifically, the
“liberated” NR and NR-Si yielded D ) (2.2 ( 0.2) × 10-6 cm2/s
and D ) (1.0 ( 0.1) × 10-6 cm2/s, respectively. The absolute
values of these diffusion coefficients may include some systematic
error as the exact size of the detection area in these experiments
is not known. However, the relative magnitudes of the diffusion
coefficients is obtained accurately. As in the silicate film data
presented above, the diffusion coefficients of the “liberated” NR
and NR-Si species differ from each other by a factor of ∼2.2,
indicating that the “liberated” NR-Si species differs in size from
NR. This size difference is attributed here to the binding of
NR-Si to silicate oligomers.
(46) Schenter, G. K.; Lu, H. P.; Xie, X. S. J. Phys. Chem. A 1999, 103, 10477.
492 Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

kept constant by having molecules that diffuse out of one edge of
the region “wrap around” to reenter on the opposite side. At each
time slice, ∆t, each molecule was displaced from its previous
position by a step of random size. The step size probability
distribution was a Gaussian with a mean-square width of 2D∆t in
each direction (two dimensions). The excitation spot was posi-
tioned at the center of the 30 × 30 µm region and had a variance,
s2, of 4 × 10-10 cm2. The maximum signal from each single
molecule was set to 200 counts/10 ms. A background signal
equivalent to 20% of the signal expected from a single molecule
was added to each transient. Random variations were added to
the signal and background counts to simulate the effects of shot
noise. Two separate simulations were performed. In one, each
molecule was assigned a diffusion coefficient of 4 × 10-10 cm2/s
(Figure 6A), similar to the measured D1 for NR. In the other, the
diffusion coefficients were all set to 1.5 × 10-10 cm2/s (Figure
6B), similar to the measured D1 for NR-Si. A total of 125 transients
were simulated in each case. The simulated transients were
exactly the same lengths as those acquired experimentally (i.e.,
670 s for NR and 1340 s for NR-Si).
The similarities in the widths and shapes of the experimental
and simulated histograms shown in Figure 5A,B and Figure 6,
respectively, indicate that the widths of the fast diffusion coefficient
distributions (D1) depicted in Figure 5A,B are limited by uncer-
tainties associated with the autocorrelation and fitting procedures.
That is, the distribution widths do not reflect sample heterogeneity
in this case.
Distinct differences between the experimental and simulated
data do exist, however. Most notably, the majority of experimental
autocorrelations required two diffusional decay components for
proper fitting (see eq 4), while far fewer of the simulated time
transients required such accommodations. Specifically, only 20%
of the simulated data for D ) 4 × 10-10 cm2/s required inclusion
of the slow diffusional component, compared to 67% of the
experimental transients. Likewise, only 14% of the simulated data
for D ) 1.5 × 10-10 cm2/s required this second, slow component,
compared to 50% of the experimental transients.
Figure 6. Histograms of diffusion coefficients (D1) derived from
simulated fluorescence time transients. Results are reported for
simulations employing diffusion coefficients of (A) D ) 4.0 × 10-10
cm2/s and (B) D ) 1.5 × 10-10 cm2/s.
The ratios of the diffusion coefficients obtained from the NR
and NR-Si-doped films (and from the bulk solutions) can be used
to estimate the size and mass of the silicate oligomers to which
the NR-Si molecules are bound. As a first approximation, this can
be accomplished via the Stokes-Einstein relation, which states
that the diffusion coefficient of a spherical object is inversely
related to its radius. An approximately 3-fold reduction in D for
the NR-Si molecules (relative to NR) indicates the silicate
oligomers may be 3 times larger (in radius) than NR itself. The
particle mass then scales as the cube of the radius, yielding an
upper bound to the mass of 27 times that of NR, or 8600 g/mol.
However, it is not likely the silicate oligomers are spherical in
shape. While their exact shape is unknown, D is often taken to
be inversely proportional to the square root of the mass in
empirical relationships describing diffusion of organic polymers.47
Hence, a lower bound for the mass may be estimated at 9 times
that of NR or 2900 g/mol. This latter model assumes the oligomers
take on a more extended-chain geometry.
Evaluation of Sample Heterogeneity. Single-molecule spec-
troscopy can also provide valuable information on the heterogene-
ity of the present samples. Evidence for the formation of inorganic-
and organic-rich domains in silicate films derived from mixtures
of TEOS and various organoalkoxysilanes has been reported
previously.31-36 As a result, heterogeneity in the rate of molecular
diffusion20 through the present films is expected. The broad
distribution shown in Figure 5A for NR in the 90% BTMOS films
might be reflective of such heterogeneity. This same distribution
also exhibits a profound tail to higher diffusion coefficients.
Interestingly, the NR-Si-doped samples (Figure 5B) yield a
distribution that is much narrower.
The appearance of the two decay components in the experi-
mental data is attributed here to sample heterogeneity on length
scales similar to or smaller than the size of the excitation spot.
Such heterogeneity is manifested as temporal variations in the
diffusion coefficient within each transient but not necessarily as
spatial variations between individual transients. The distributions
shown in Figure 5C,D reflect an ∼100-fold reduction in the rate
of dye diffusion (i.e., D2 ≈ D1/100). The two diffusional compo-
nents likely result from the phase separation of organic- and
inorganic-rich film regions.31-36 The dye likely moves more slowly
or may become briefly immobilized in the inorganic-rich regions.23
Diffusion of the molecules through (or with) liquidlike silicate
oligomers present in organic-rich film regions would yield
relatively more rapid diffusion.
Simulated time transients26 were employed to help interpret
the NR-Si and NR histograms in terms of sample heterogeneity.
In these studies, time transients for perfectly homogeneous
samples were numerically generated, autocorrelated and fit using
eq 4 in a manner identical to the analysis of the experimental data.
The results provide a view of the variations expected in the
experimentally determined diffusion coefficients from uncertainties
in the autocorrelations and their fits. Figure 6 depicts the
distribution of D values (fast component) obtained.
CONCLUSIONS
In summary, single-molecule spectroscopic methods have been
used to study diffusion in thin ORMOSIL films prepared by the
sol-gel process. Results for the diffusion of “free” and silicate-
bound dye molecules were obtained. Surprisingly, both molecules
showed relatively facile diffusion with the average diffusion
For these simulations, 180 molecules were positioned randomly
in a 30 × 30 µm region. The number of molecules present was
(47) Sun, S. F. Physical Chemistry of Macromolecules. Basic Principles and Issues;
John Wiley & Sons: New York, 1994.
Analytical Chemistry, Vol. 77, No. 2, January 15, 2005 493

coefficient associated with fast diffusion of the silicate-bound
molecules only ∼3-4 times smaller than that of the “free” dye.
These results were taken as clear evidence for the presence of
silicate oligomers in the films. Bulk solution-phase FCS studies
of dye molecules “liberated” from the silicate films provided
important support for these conclusions. Heterogeneity in the films
was manifested through the observation of two distinctly different
time scales for single-molecule diffusion. The origin of the two
components was attributed to the presence of inorganic- (slow
diffusion) and organic-rich (fast diffusion) domains. Such results
are important to those developing similar materials for use in
catalysts and chemical sensors. The presence of liquidlike oligo-
mers facilitates diffusion of reactants/analytes through these
materials and may therefore enhance the activity/response of
devices based upon ORMOSIL thin films.
ACKNOWLEDGMENT
The authors thank the National Science Foundation (CHE-
0316466), the Army Research Office, and the State of Kansas for
support of this work.
SUPPORTING INFORMATION AVAILABLE
Additional images and image autocorrelations to augment the
discussion associated with Figure 3 and examples of the bulk
solution-phase time transients and autocorrelations for the “liber-
ated” dyes. This material is available free of charge via the Internet
at http://pubs.acs.org.
Received for review June 9, 2004. Accepted October 20,
2004.
AC0491511
494 Analytical Chemistry, Vol. 77, No. 2, January 15, 2005