Probing Strong Adsorption of Solute onto C18-Silica Gel by

Justin T. Cooper , Eric M. Peterson , and Joel M. Harris .... Wayland W. L. Cheng, Dee Ann Stults, Emily R. Wiebracht, John J. Kasianowicz, and Daniel...
0 downloads 0 Views 583KB Size
Download PDF
Anal. Chem. 2005, 77, 2303-2310

Articles

Probing Strong Adsorption of Solute onto C18-Silica Gel by Fluorescence Correlation Imaging and Single-Molecule Spectroscopy under RPLC Conditions Zhenming Zhong, Mark Lowry,† Gufeng Wang, and Lei Geng*

Department of Chemistry, Optical Science and Technology Center, and Center for Biocatalysis and Bioprocessing, University of Iowa, Iowa City, Iowa 52242

Understanding molecular adsorption at a chromatographic interface is of great interest for addressing the tailing problem in chemical separations. Single-molecule spectroscopy and confocal fluorescence correlation imaging are used to study the adsorption sites of C18 silica beads under RPLC chromatographic conditions. The experiments show that cationic molecule rhodamine 6G laterally diffuses through the chromatographic interface of a C18 hydrocarbon monolayer and acetonitrile with occasional reversible strong adsorptions. Fluorescence correlation imaging extracts the rare strong adsorption events from large data sets, revealing that the strong adsorption sites are randomly distributed throughout the silica beads. Virtually every imaging pixel of silica beads adsorbs molecules. Single-molecule spectroscopy of the 584 strong adsorption events observed indicates that the strong adsorptions persist on the time scales from several milliseconds to seconds, having an average desorption time of 61 ms. The strong adsorption events are rare, comprising 0.3% of the total observation time. The sizes of strong adsorption sites are within the optical resolution of confocal imaging.

ion-exchange interactions in RPLC.1-7 Ion-exchange interaction can be suppressed in several ways to minimize tailing. The most common process is end capping.1 End capping improves the quality of chemically modified stationary phase by bonding to residual silanols or rendering them sterically inaccessible. Other techniques include the use of mobile-phase additives, rehydroxylation of the silica surface, thermal treatment of silica before derivatization, and manipulation of the pH of the mobile phase.3,8 Wirth et al. achieved a dense, two-dimensionally cross-linked network on the chromatographic silica surface through the mixed self-assembly of C18 and C1 alkyl chains.9,10 The C18-C1 monolayer proved to be acid-stable and had better chromatographic performance for organic cations compared to a conventional C18 monolayer. Despite much effort and advancement, tailing has not been completely eliminated in chromatography. The physical origin and retention mechanism for tailing is still a widely debated topic. To achieve an improved understanding of tailing in and retention mechanism of RPLC, the organization of bonded phase and the interaction between solute and the stationary phase have been intensively investigated with a variety of spectroscopic methods, such as IR,6,7,11-13 Raman,14-16 NMR,17-19 and

Reversed-phase liquid chromatography (RPLC) has been one of the most important separation methods for the analysis and purification of chemical and biological mixtures for the past few decades. Chemically modified silica, for example, octadecylsilyl (C18)-modified silica gel, is the most commonly used stationary phase in RPLC. Despite decades of practical applications, the problems of band broadening and peak tailing are still persistent for the separation of basic compounds at high pH. Many published reports document the role of silanol groups in the chromatographic properties of silica and provide evidence that tailing correlates with the acidic residual silanol groups on exposed silica substrate after surface derivatization, which cause tailing through

(1) Cox, G. B. J. Chromatogr. 1993, 656, 353-367. (2) Nawrocki, J.Chromatographia 1991, 31, 177-192. (3) Nawrocki, J. J. Chromatogr., A 1997, 779, 29-71. (4) Rogers, S. D.; Dorsey, J. G. J. Chromatogr., A 2000, 892, 57-65. (5) Zhuravlev, L. T. Colloids Surf., A 2000, 173, 1-38. (6) Ko¨hler, J.; Chase, D. B.; Farlee, R. D.; Vega, A. J.; Kirkland, J. J. J. Chromatogr. 1986, 352, 275-305. (7) Ko ¨hler, J.; Kirkland, J. J. J. Chromatogr. 1987, 385, 125-150. (8) Sunseri, J. D.; Cooper, W. T.; Dorsey, J. G. J. Chromatogr., A 2003, 1011, 23-29. (9) Wirth, M. J.; Fairbank, R. W. P.; Fatunmbi, H. O. Science 1997, 275, 4447. (10) Fairbank, R. W. P.; Xiang, Y.; Wirth, M. J. Anal. Chem. 1995, 67, 38793885. (11) Sander, L. C.; Calls, J. B.; Field, L. R. Anal. Chem. 1983, 55, 10681075. (12) Rivera, D.; Poston, P. E.; Uibel, R. H.; Harris, J. M. Anal. Chem. 2000, 72, 1543-1554. (13) Rivera, D.; Harris, J. M. Langmuir 2001, 17, 5527-5536. (14) Ducey, M. W., Jr.; Orendorff, C. J.; Pemberton, J. E.; Sander, L. C. Anal. Chem. 2002, 74, 5576-5584.

* Corresponding author. E-mail: [email protected]. Phone: (319)-335-3167. Fax: (319)-335-1270. † Current address: Department of Chemistry, Louisianan State University, Baton Rouge, LA 70803. 10.1021/ac048290f CCC: $30.25 Published on Web 03/03/2005

© 2005 American Chemical Society

Analytical Chemistry, Vol. 77, No. 8, April 15, 2005 2303

fluorescence,20-27 as well as numerous chromatographic studies.28-31 Wang and Harris revealed the lateral diffusion of covalently bound organosiloxane on a silica surface using fluorescence spectroscopy.21 The migration proceeded through the hydrolysis of the siloxane and the reattachment. Rivera and Harris studied the adsorption of pyridine from n-heptane solutions onto both bare and cyanopropyl-derivatized silica surfaces using attenuated total internal reflection FT-IR.13 The results showed that there are two distinct adsorption sites on both underivatized and derivatized silica surfaces, but derivatization decreases the total number of adsorption sites by 43%. IR spectra indicated that surface-bonded water is the site of weaker adsorption. The stronger site of adsorption was found to be composed mainly of isolated silanols for underivatized surface and exclusively of weakly hydrogenbonded silanols for derivatized surface. Gritti and Guiochon used the expectation-maximization method to derive the adsorption energy distributions of caffeine and phenol on two C18 stationary phases with different solid supports, silanol-group-rich silica, and silica-methylsilane hybrid surface with less acidic free silanols.29 Comparing the adsorption energy distributions between caffeine and phenol, they concluded that the presence of silanol groups increases the heterogeneity of the structure of the C18-bonded layer and that the strong adsorption sites causing the severe peak tailing of caffeine are located within the C18-bonded layer, not on the bare surface of silica. Despite these and other findings, direct experimental observations of the strong adsorption sites responsible for tailing in chromatographic interface are desired. Recently, single-molecule spectroscopy has been widely used as a tool for exploring the behavior of individual molecules in heterogeneous chemical systems. Single-molecule fluorescence tracking revealed spatial variation in the molecular diffusion coefficient of a sol-gel film.32 Xu and Yeung used total internal reflection fluorescence imaging to observe electrostatic interactions between single protein molecules and fused-silica surfaces, which are spatially dependent.33Kang et al. studied the adsorption/ desorption dynamics of individual DNA and protein molecules under various pH and ionic strength conditions at liquid-solid interfaces, which are related to retention behavior in capillary (15) Ducey, M. W., Jr.; Orendorff, C. J.; Pemberton, J. E.; Sander, L. C. Anal. Chem. 2002, 74, 5585-5592. (16) Doyle, C. A.; Vickers, T. J.; Mann, C. K.; Dorsey, J. G. J. Chromatogr., A 2000, 877, 25-39. (17) Pursch, M.; Sander, L. C.; Egelhaaf, H. J.; Raitza, M.; Wise, S. A.; Oeikrug, D.; Albert, K. J. Am. Chem. Soc. 1999, 121, 3201-3213. (18) Pursch, M.; Sander, L. C. Albert, K. Anal. Chem. 1999, 71, 733A741A. (19) Kelusky, E. C.; Fyfe, C. A. J. Am. Chem. Soc. 1986, 108, 1746-1749. (20) Lochmu ¨ ller, C. H.; Colborn, A. S.; Hunnicutt, M. L.; Harris, J. M. J. Am. Chem. Soc. 1984, 106, 4077-4082. (21) Wang, H. B.; Harris, J. M. J. Am. Chem. Soc. 1994, 116, 5754-5761. (22) He, Y.; Geng, L. Anal. Chem. 2001, 73, 5564-5575. (23) He, Y.; Geng, L. Anal. Chem. 2002, 74, 1819-1823. (24) Burns, J. W.; Bialkowski, S. E.; Marshall, D. B. Anal. Chem. 1997, 69, 38613870. (25) Hansen, R. L.; Harris, J. M. Anal. Chem. 1998, 70, 4247-4256. (26) Montgomery, M. E., Jr.; Wirth, M. J. Anal. Chem. 1994, 66, 680-684. (27) Lowry, M.; He, Y.; Geng, L. Anal. Chem. 2002, 74, 1811-1818. (28) Gritti, F.; Guiochon, G. J. Chromatogr., A 2004, 1048, 1-15. (29) Gritti, F.; Guiochon, G. J. Chromatogr., A 2004, 1028, 75-88. (30) Quinones, I.; Cavazzini, A.; Guiochon, G. J. Chromatogr., A 2000, 877, 1-11. (31) Stanley, B. J.; Krance, J.; Roy, A. J. Chromatogr., A 1999, 865, 97-109. (32) McCain, K. S.; Hanley, D. C.; Harris, J. M. Anal. Chem. 2003, 75, 43514359. (33) Xu, N. X.; Yeung, E. S. Science 1998, 281, 1650-1653.

2304

Analytical Chemistry, Vol. 77, No. 8, April 15, 2005

electrophoresis and capillary liquid chromatography.34,35 Wirth pioneered single-molecule probing of the strong adsorption sites on a chemically modified silica surface.36-40 The mixed-mode adsorption of 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) onto a fused-silica surface chemically modified with chlorodimethyloctadecylsilane was observed on the singlemolecule level. Three different strong adsorption processes were observed with desorption times on the time scales of 70 ms, 7 s and >2 min. Fluorescence imaging identified that the latter two extremely strong events happen at nanoscale topographical indentations of fused silica. The independence of pH for these adsorption events revealed that these adsorption sites are neutral, and AFM imaging of the adsorption sites showed that they have a topographical origin. These single-molecule experiments revealed remarkable details of adsorption/desorption processes that control peak tailing in chromatographic retention. Most of these studies were conducted on model systems: unmodified or C18-modified flat silica surface. Experiments have showed that differences in structure and chemistry may exist between alkyl-derivatized silica surfaces and derivatized porous silica beads.41 It is of interest to study the adsorption properties of a chromatographic stationary phase under RPLC separation conditions. In this paper, we study the adsorption events of a cationic dye, rhodamine 6G, at the interface of C18 on chromatographic silica gel and acetonitrile using single-molecule spectroscopy and fluorescence correlation imaging. C18 silica beads and acetonitrile are the most widely used stationary phase and solvent. The results presented here are directly relevant to RPLC separations. EXPERIMENTAL SECTION Chemicals and Materials. Rhodamine 6G was purchased from Aldrich (Milwaukee, WI) and used as received. HPLC grade acetonitrile and microscope coverslip (12-545M, size 60 × 24 mm, thickness 0.13-0.17 mm) were purchased from Fisher Scientific (Fair Lawn, NJ). The coverslip was cleaned in concentrated nitric acid and rinsed thoroughly with deionized water, which was purified with a MilliQ system (MilliQ-Plus, Millipore, Bedford, MA). The Luna 10-µm C18 silica beads were obtained from Phenomenex (Torrance, CA). According to the manufacturer, they were end-capped and had an average pore size of 100 Å, a carbon loading of 17.5%, a calculated bonded phase coverage of 3.00 µmol/ m2, and a surface area of 400 m2/g. Fluorescent impurities inside the C18 silica beads were minimized by multiple rinsings with fresh acetonitrile followed by centrifugal separation until few photon bursts were observed for blank C18 silica beads in the singlemolecule experiment. Mobile phase consisting of 100% acetonitrile was used throughout the experiments. Immersion oil with a refractive index of 1.51 was obtained from Cargille Laboratories Inc. (Cedar Grove, NJ). (34) Kang, S. H.; Shortreed, M. R.; Yeung, E. S. Anal. Chem. 2001, 73, 10911099. (35) Kang, S. H.; Yeung, E. S. Anal. Chem. 2002, 74, 6334-6339. (36) Wirth, M. J.; Swinton D. J.; Ludes, M. D. J. Phys. Chem. B 2003, 107, 62586268. (37) Ludes, M. D.; Wirth, M. J. Anal. Chem. 2002, 74, 386-393. (38) Swinton, D. J.; Wirth, M. J. Anal. Chem. 2000, 72, 3725-3730. (39) Wirth, M. J.; Ludes, M. D.; Swinton D. J. Anal. Chem. 1999, 71, 39113917. (40) Wirth, M. J.; Swinton D. J. Anal. Chem. 1998, 70, 5264-5271. (41) Hartner K. C.; Carr J. W.; Harris J. M. Appl. Spectrosc. 1989, 43, 81-87.

slips. The slips were then tightly fixed on the xy piezoflexture stage to prevent solvent evaporation and movements of individual silica beads. By scanning the xy piezoflexture stage, the laser focal point was directed into specific silica beads and fluorescence signal was recorded. The measurements were performed without a driving flow. Typically, the single-molecule photon bursts were continuously monitored for over 4 min for each selected focal point for the construction of fluorescence correlation curves of that pixel. To acquire the fluorescence image of a silica bead, the continuous scanning of the laser focal point throughout the bead was achieved by scanning the xy piezoflexture stage with the dwell time set to 1 ms for each imaging pixel. Data Treatment. To construct fluorescence correlation curves of one specific pixel, the collected single-molecule data were evenly divided into 200 small data sets. For each data set, the timedependent fluorescence intensity F(t) was analyzed by calculating the temporal autocorrelation function of fluorescence fluctuations,

Figure 1. Schematic optical diagram for single-molecule experiment. The laser beam was focused into the chromatographic beads by a 100× oil immersion objective, and the probing of specific sites was achieved by scanning the piezoelectric stage at nanometer resolution.

Single-Molecule Spectroscopy. The system for both singlemolecule detection and confocal fluorescence imaging was constructed on an inverted microscope (Eclipse TE2000-U, Nikon) and is shown in Figure 1. The 514.5-nm excitation light was provided by an air-cooled argon ion laser. The power of the beam was reduced by neutral density filters to below 0.1 mW, at which no photobleaching of rhodamine 6G was observed in the duration of the experiments. The excitation beam was expanded to slightly overfill the back of an oil immersion objective (100×, 1.45 NA). A nearly Gaussian profile beam was introduced into the back port of the microscope and reflected by a dichroic mirror (Z514RDC, Chroma Technology Corp.). The objective focused the beam into a diffraction-limited spot. The three-dimensional movement of the focal point in and out of the C18 silica beads was achieved by the combination of a piezoelectric objective stepper and an xy piezoflexure stage (Physik Instrumente) with nanometer resolution. Fluorescence emitted by rhodamine 6G was collected by the same objective and passed through the dichroic mirror. An Omega Optical 560RDF55 band-pass filter (560 ( 27.5 nm) was used to transmit fluorescence and reject background signal. A 50-µm confocal pinhole was placed in the primary image plane to efficiently reject out-of-focus signal. The spatially filtered fluorescence signal was then passed through a notch filter (HNPF-514.51.0, Kaiser Optical Systems) and focused onto the active area of an avalanche photodiode (SPCM-AQR, PerkinElmer Optoelectronics). Time-dependent photoelectron pulses were counted by a multichannel scaler (MCS) that outputs up to 65 535 channels. In single-molecule measurement, 500 µL of 10 pM rhodamine 6G in acetonitrile was mixed with 500 µL of dilute suspension of C18 silica beads in acetonitrile. After thorough mixing, 20 µL of the mixture was deposited onto the surface of a microscope coverslip. A second coverslip was placed immediately on top of the solution to produce a sample layer between the two cover

G(τ) ) 〈δF(t)δF(t + τ)〉/〈F(t)〉2

(1)

where the angular brackets indicate time average, δF(t) is the fluctuations of the fluorescence signal defined as the deviations from the temporal average, and δF(t) ) F(t) - 〈F(t)〉. So 200 normalized autocorrelation curves were calculated for each imaging pixel. When molecules of one species undergo translational diffusion through the Gaussian-distributed probe volume, the autocorrelation function is given by,

G(τ) )

(

)(

1 4Dτ 1+ 2 N ω

-1

1+

)

4Dτ l2

-1/2

(2)

where N is the average number of diffusing molecules in the probe volume, D is the diffusion coefficient, and ω and l are the distance from and along the optical axis, respectively, both at which the laser intensity drops to 1/e2 of its value at the maximum. From the measured autocorrelation function of rhodamine 6G in water and its known diffusion coefficient value (2.8 × 10-6 cm2 s-1),42 ω and l were directly determined to be 0.22 and 1.2 µm, respectively, for the system setup, defining a probe volume of 0.4 fL. A concentration of 5 pM rhodamine 6G is selected to ensure singlemolecule observation. At this concentration in the mobile phase, the average number of rhodamine 6G molecules in this probe volume is 0.001. The probabilities of finding 0, 1, or 2 molecules in the probe volume are 99.9%, 0.1%, and below 1 × 10-6%, respectively, according to the Poisson distribution. The probability of observing more than one molecule simultaneously is negligible. In the stationary phase, the solute concentration is higher, determined by the partition coefficient between the mobile and stationary phases. The concentration, however, is still low enough to ensure the observation of single-molecule behavior. By fitting the calculated autocorrelation function to eq 2, diffusion coefficients for rhodamine 6G in mobile phase and chromatographic interface were determined. Image construction of C18 silica bead and the calculation of autocorrelation functions were performed with programs written in MatLab (Mathworks Inc., Natick, MA). (42) Rigler, R.; Mets, U.; Widengren, J.; Kask, P. Eur. Biophys. J. 1993, 22, 169175.

Analytical Chemistry, Vol. 77, No. 8, April 15, 2005

2305

RESULTS AND DISCUSSION In RPLC, most molecular interactions in the separation process occur at the interface in the network of pores inside the chromatographic beads. With the aid of optical sectioning capability of confocal microscopy, it is possible to probe the separation process deep in the stationary phase.27,43 Figure 2A shows the raw intensity data recorded by the multichannel scaler for imaging a C18 silica bead equilibrating with a mobile phase of 5 pM rhodamine 6G in 100% acetonitrile. The data were acquired by scanning the xy piezoflexure stage with the focal plane sectioning through the center of the bead, line by line. We collected 10 000 data points in a single scanning experiment, creating 10 000 imaging pixels with the dwell time set to 1 ms/pixel. Figure 2B shows the image constructed from the data presented in Figure 2A. The image size is 20 µm × 20 µm. The focal point of the laser is moved through the sample by 0.2 µm between pixels. Insets in Figure 2A are the expanded view of photon burst data for different

areas in the image. The first data trace (pixels 800-1200) corresponds to an image section in the mobile phase. The second trace (pixels 5600-6000) corresponds to an image section across the silica bead. These image sections are indicated in Figure 2B by the white boxes. Single-molecule photon bursts were observed in both the mobile and the stationary phases, but the intensity data obtained from the stationary phase display much higher background signal due to the strong scattering from the porous silica medium. The shape of the silica bead is clearly observed in Figure 2B because of this strong scattering background. The pixels in the mobile phase, with the low background signal, are much darker in comparison. Brighter pixels that rise above the background are observed for both the stationary and the mobile phases due to the presence of individual rhodamine 6G molecules in the probe volume. It has been previously shown that rhodamine 6G molecules preferentially partition into the stationary phase under reversedphase conditions.27 When the probe volume is focused inside the silica beads, it is reasonable to assume that there is negligible contribution to the fluorescence from rhodamine 6G in the mobile phase occupying the pores of the bead. So the time-dependent intensity profiles acquired by single-molecule spectroscopy inside the beads are characteristic of rhodamine 6G diffusion in the stationary phase. Panels A and B in Figure 3 show typical singlemolecule photon burst data obtained from the mobile phase and the stationary phase, respectively, with 20-µs integration time. Due to the different populations of rhodamine 6G molecules between mobile phase and stationary phase, more photon bursts were observed from the stationary phase than those from the mobile phase. The integration time per data point is shorter than the characteristic diffusion time of rhodamine 6G molecules across the probe volume under both stationary- and mobile-phase conditions, so panels A and B in Figure 3 have similar peak counts The data obtained from the mobile phase and those from the stationary phase without any observed strong adsorption events were respectively correlated according to eq 1. The calculated autocorrelation functions are plotted in Figure 3C. It shows that the lateral diffusion of rhodamine 6G at the chromatographic interface of C18/acetonitrile is much slower than that in acetonitrile. The molecules are diffusing in a confined volume of the nanometer pores. This diffusion is different from those in an open space in confocal fluorescence correlation spectroscopy of free solution.44,45 For quantitative comparison, we fitted both of them to the three-dimensional diffusion model as described by eq 2. The diffusion coefficient of rhodamine 6G is 1.13 × 10-6 cm2 s-1 at the C18/acetonitrile interface for the three-dimensional diffusion model and 6.78 × 10-6 cm2 s-1 in acetonitrile. This smaller diffusion coefficient of rhodamine 6G at the chromatographic interface is interpreted as a result of geometric restriction and weak adsorption interaction between solute and stationary phase that is the basis of retention in RPLC. Tailing in chromatography suggests that there exists at least one other distinct type of strong adsorption sites on the silica substrate. These results are very similar to those at C18 interface on a fused-silica surface.38 The method of observing single-molecule adsorption events has been established by Wirth.36 In single-molecule spectroscopy,

(43) Wang, G. F.; Lowry, M.; Zhong, Z. M.; Geng, L. J. Chromatogr., A 2005, 1062, 275-283.

(44) Mahurin, S. M.; Dai, S.; Barnes, M. D. J. Phys. Chem. B 2003, 107, 1333613340. (45) Gennerich, A.; Schild, D. Biophys. J. 2000, 79, 3294-3306.

Figure 2. (A) Raw intensity data acquired for confocal fluorescence imaging of a C18 silica bead equilibrating with mobile phase of 5 pM rhodamine 6G in acetonitrile, by scanning the laser focal point through the sample. (B) Image constructed from the data presented in (A). Image size 20 µm × 20 µm. Pixel time 1 ms. Insets in (A) are the fluorescence intensity traces for image sections in the mobile phase and across the silica bead, as shown in (B).

The fitting of calculated autocorrelation functions was processed in Microcal Origin (OriginLab, Northhampton, MA).

2306 Analytical Chemistry, Vol. 77, No. 8, April 15, 2005

Figure 3. (A) Single-molecule burst data from a pixel in the mobile phase. (B) Single-molecule burst data from a pixel inside the C18 silica bead. Dwell time 20 µs. (C) Comparison of normalized experimental autocorrelation functions for the mobile phase (0) and the stationary phase (O) and the corresponding fits (-) to eq 2.

photon bursts of various amplitudes and durations are produced when individual rhodamine 6G molecules diffuse through the Gaussian laser beam with different diffusion trajectories. The autocorrelation function of fluorescence fluctuation reveals the average diffusion time of molecules across the probe volume. If a molecule were strongly adsorbed at the chromatographic interface, it would stop diffusing and reside at a fixed position inside the probe volume for some time, producing constant fluorescence signals within the shot noise before it diffuses away. This is the basic principle for single-molecule spectroscopy to distinguish the adsorbed molecules from diffusing molecules. On the other hand, strong adsorption renders the adsorbed molecules spending more time in the probe volume, prolonging the overall diffusion time of molecules across the probe volume. In other words, the autocorrelation function would decay at a relatively slower rate in the presence of strong adsorption events.

In recognition of the fact that the strong adsorption events would probably be rare, we divided the total data collected from a single pixel evenly into 200 small data sets to enhance the contribution from strong adsorption to the autocorrelation function. For each data set, an autocorrelation function was calculated. Figure 4A shows the fluorescence correlation curves of a pixel in the mobile phase. Since all the molecules entering the probe volume take a random walk, the 200 autocorrelation curves calculated from the 200 data sets have the same decay time within error. Figure 4B shows the fluorescence correlation curves of a pixel selected from the stationary phase that exhibits rare strong adsorption events. Two autocorrelation curves, corresponding to the 17th and 73rd data sets, decay much slower than the other 198 autocorrelation curves. An examination of the raw photonburst data for these two autocorrelation curves revealed that a strong adsorption event lasting for ∼12 ms happens for each data set as shown in Figure 4C and D. The photon-burst acquisition time is 1310 ms for calculating each autocorrelation curve in Figure 4. It is clearly seen that the presence of a 12-ms adsorption process during molecular diffusion causes the autocorrelation curve to decay significantly slower than those calculated from data sets without any strong adsorption events. The effect of strong adsorption on the decay time of the autocorrelation curve is attributed to the small beam radius of 0.22 µm employed in the experiment, which makes the contribution from diffusing molecules to decay very quickly and then augments the contribution from adsorbed molecules to the overall average diffusion time. The fluorescence correlation analysis is ideal to characterize the adsorption properties for each imaging pixel and distinguish the strong adsorption events from singlemolecule photon bursts, especially when the strong adsorption events are considerably rare and long data acquisition time is required. To characterize the adsorption properties of a C18 stationary phase, fluorescence correlation curves of different pixels inside different C18 silica beads were studied. For each pixel, 200 singlemolecule data sets, thus 200 autocorrelation curves were acquired. These correlation curves reveal that virtually all the pixels exhibit strong adsorption events. Figure 5 shows results from 21 pixels inside 7 randomly selected beads. The location classification for each pixel is summarized in Table 1. To plot the results in one graph for comparison, we integrated each autocorrelation curve and, therefore, displayed 200 integrated values for each pixel in Figure 5. If there were strong adsorption events in a raw data set, the calculated autocorrelation curve should decay much slower than that calculated from the raw data set without any strong adsorption events, producing a higher integrated value corresponding to that autocorrelation curve. In other words, a peak in Figure 5 represents at least one strong adsorption event. For comparison of integrated values between stationary phase and mobile phase, one pixel from mobile phase of 5 pM rhodamine 6G in acetonitrile is also shown in Figure 5 as pixel 1. The 200 integrated values are almost constant in the mobile phase due to the absence of any strong adsorption events. On the contrary, all the studied pixels inside the stationary phase display peaks with different magnitudes. Higher integrated value signifies strong adsorption events with longer adsorption time in that data set from which the autocorrelation curve is calculated. As shown in Figure Analytical Chemistry, Vol. 77, No. 8, April 15, 2005

2307

Figure 4. Fluorescence autocorrelation curves demonstrating the ability to recognize rare strong adsorption events from a large pool of burst data. (A) Fluorescence correlation curves of a single pixel in the mobile phase of 5 pM rhodamine 6G in acetonitrile. (B) Fluorescence correlation curves of a single pixel in C18 silica beads that exhibits rare strong adsorption events. (C) Photon burst data corresponding to the 17th FCS curve in (B). (D) Photon burst data corresponding to the 73rd FCS curve in (B). Insets in (C) and (D) are the expanded view of the adsorption events. Table 1. Summary of Pixel Locations pixel no. 1 2-6 7-13 14-17 18-22

Figure 5. Fluorescence correlation curves of 21 pixels from 7 randomly selected C18 silica beads equilibrating with 5 pM rhodamine 6G in acetonitrile. For each pixel, the integrated values for each FCS curve are plotted as a function of the number of FCS curves. For comparison, a pixel from mobile phase is also plotted and shown as pixel 1.

5, within the observation time of 4 min, strong adsorption events were observed for every pixel, even though some pixels exhibit more frequent strong adsorption events than the others. The result indicates that strong adsorption sites are randomly distributed throughout the silica beads. This is consistent with the random distribution of residual silanol groups on the surface of silica substrate because isolated silanol groups are the origin of strong adsorption sites.3 2308 Analytical Chemistry, Vol. 77, No. 8, April 15, 2005

randomly selected from the mobile phase inside bead 1 with a spacing of 0.2 µm/pixel inside bead 1 with a spacing of 0.4 µm/pixel inside bead 2 with a spacing of 0.4 µm/pixel randomly selected pixels inside beads 3-7, respectively

For a single pixel, the observed adsorption events have different levels of photon counts, reflecting that rhodamine 6G molecules are adsorbed onto different sites inside the probe volume of the Gaussian beam. Figure 6 shows two consecutive strong adsorption events observed in a data set. The first adsorption event starts at 41 ms and ends at 48 ms. The second one persists from 57.5 to 63.5 ms. The drop of intensity to the baseline and the presence of a spike at 55 ms between these two events indicate that they are two distinct adsorption events. Their different fluorescence signals, 497 and 318 counts/0.5 ms, respectively, suggest that they are from different adsorption sites within the probe volume. The first site is closer to the center of the Gaussian beam than the second one if these two sites are on the same cross section of the laser beam. This result reveals an important characteristic of the adsorption sites: the size of adsorption sites is beyond the optical resolution, on the nanometer level. Figure 7 displays several strong adsorption events persisting on different time scales. In Figure 7A, a strong adsorption event of only 6 ms is observed. The capability of distinguishing such

Figure 6. Single-molecule burst data showing two consecutive adsorption events occurred at two specific sites existing in the probe volume. The dotted lines show the (2σ levels of the shot noise for each adsorption events.

relatively short strong adsorption events from the photon bursts due to single-molecule diffusion is attributed to the narrow Gaussian beam employed in the experiment. Based on the measured beam radius of 0.22 µm and diffusion coefficient of 1.13 × 10-6 cm2 s-1 at the studied C18/acetonitrile interface, the characteristic diffusion time of rhodamine 6G molecules across the probe volume is calculated to be ∼0.11 ms. This small diffusion time facilitates the observation of specific strong adsorptions lasting on the order of milliseconds, avoiding the underestimation of the population of short strong adsorption events that could also contribute to peak tailing. Figure 7B shows a strong adsorption event where the fluorescence signal dropped all the way down to the baseline several times during the adsorption event. A spike at ∼210 ms is due to

another molecule diffusing into the beam. The recovery of the fluorescence signal to the same initial intensity level suggests that the molecule is adsorbed at the same radial position and that the intensity drop is not due to photobleaching of the adsorbed molecule. This blinking phenomenon is caused by the excitedstate molecule transiently passing from the singlet state into the triplet state, where no fluorescence is emitted. This is similar to the blinking of DiI molecule adsorbed to C18 modified fused silica observed by Wirth.40 The blinking observed here is on the submillisecond to millisecond time scales, which could slightly impede the direct observation of short strong adsorption events on the time scale of several milliseconds, but it has negligible effect on the observation of longer strong adsorption events, such as shown in Figure 7B. Panels C and D in Figures 7 show two specific long strong adsorption events persisting for 310 and 1025 ms, respectively, which are significantly longer than the characteristic diffusion time of molecules across the beam, so many other molecules are diffusing through the beam during the strong adsorptions. Figure 8 shows the histogram of duration times for the observed 584 strong adsorption events. It reveals that the strong adsorption events persist on quite broad time scales, ranging from several milliseconds to several seconds. The observed strong adsorption events are rare, comprising 0.3% of the total observation time. The average duration time for these observed strong adsorption events is 61 ms. Three different strong adsorption sites have been reported on C18 modified fused silica with distinctly different desorption times of 70 ms, 7 s, and >2min.36 We also observed stronger adsorption events with desorption times of several seconds on C18 chromatographic silica gel, but the relative population of these stronger adsorption events is small compared

Figure 7. Selected adsorption events persisting on the time scales from milliseconds to seconds. Desorption time: (A) 6, (B) 50, (C) 310, and (D) 1025 ms.

Analytical Chemistry, Vol. 77, No. 8, April 15, 2005

2309

the desorption time of 70 ms the primary contributor. The relatively long time resolution of switching flow streams limited the study whether the strong adsorption site with desorption time of 70 ms found on fused silica also exists on chromatographic silica gel. The observed average desorption time of 61 ms reported in this work directly proves that this strong adsorption site exists on chromatographic silica gel and contributes significantly to the tailing zone.

Figure 8. Distribution of the duration of adsorption events. To show the broad range of desorption times, they are plotted in the logarithmic scale.

to that of adsorption events with millisecond desorption times. Longer observation times or higher fluorophore concentrations are needed for the observation of longer adsorption events in our experiment. One would expect that the slower desorption processes could contribute more significantly to the zone tailing than the faster desorption processes. Ludes et al. studied the desorption kinetics of DiI from fused silica versus silica gel, revealing that the two strong adsorption sites with desorption times of 7 s and >2 min observed on fused silica also exist on chromatographic silica gel.46 However, the very slow kinetics associated with the slow desorption processes for these sites weaken the contribution of these sites to the tailing zone, making the strong adsorption site having (46) Ludes, M. D.; Anthony, S. R.; Wirth, M. J. Anal. Chem. 2003, 75, 30733078.

2310

Analytical Chemistry, Vol. 77, No. 8, April 15, 2005

CONCLUSIONS Single-molecule spectroscopy and confocal fluorescence correlation imaging provide direct evidence of strong adsorption of cationic molecule onto C18 silica gel under separation conditions, which could cause peak tailing in RPLC. The strong adsorption sites are randomly distributed throughout the silica beads, and virtually every imaging pixel of silica beads absorbs molecules. Analysis of the observed 584 strong adsorption events indicates that the adsorptions persist on the time scales from several milliseconds to seconds, having an average desorption time of 61 ms. The strong adsorption events are rare, comprising 0.3% of the total observation time. These single-molecule results lend insights into the adsorption/desorption processes and the distribution of strong adsorption sites in chromatographic stationary phase. ACKNOWLEDGMENT We acknowledge the University of Iowa (MPSFP grant), the National Institutes of Health (CA100741), and a predoctoral fellowship awarded to G.W. by the Center for Biocatalysis and Bioprocessing of the University of Iowa for supporting this work. Received for review November 18, 2004. Accepted January 23, 2005. AC048290F