How Fast Can Thiols Bind to the Gold Nanoparticle Surface?

Kinetics of gold nanoparticle surface modification with thiols can take more than one hour for completion. 7‐mercapto‐4‐methylcoumarin can be used to follow the process by fluorescence spectroscopy and serves as a convenient molecular probe to determine relative kinetics. SERS studies with aromatic thiols further support the slow surface modification kinetics observed by fluorescence spectroscopy. The formation of thiolate bonds is a relatively slow process; we recommend one to two hour wait for thiol binding to be essentially complete, while for disulfides, overnight incubation is suggested.


INTRODUCTION
While there are numerous ways to stabilize gold nanoparticles (AuNP), the most common ones involve the formation of the thiolate S-Au bond (~130 kJ mol À1 ), frequently by reaction of thiols with the gold surface (1). The modification of gold surfaces with thiol compounds has been extensively explored as one of the best gold surface passivation methods, as well as the best way to anchor different functional groups to the surface (1)(2)(3)(4)(5)(6). While the strategy has proven thermodynamically effective and efficient, the complex kinetics of the reaction have been frequently overlooked. The process is well understood when it involves the formation of self-assembled monolayers (SAM) on atomically flat gold surfaces, but to a lesser extent in the case of AuNP where polydispersity and irregularities on the surface and curvature variations make the binding process more complex and the conclusions more difficult to generalize (7). Surface enhanced Raman spectroscopy (SERS) has proven a useful tool to study structural effects and kinetics of thiol-AuNP interactions (7,8), although relatively large particle sizes (e.g. 50-200 nm) and aromatic thiols tend to be preferred to suit SERS optimal experimental conditions (9). Thus, we combine fluorescence and SERS as a way to generalize our results to different chemical structures and particle sizes.
While working on preparing thiolated stabilized AuNP, we have asked ourselves how long we had to wait until the derivatization was complete, so as to avoid working with nanostructures that were still undergoing significant change. A few literature studies (such as those mentioned above) address this question for specific systems, yet, much more frequent are literature contributions where the delay between thiolate derivatization and usage of the nanostructures is simply not mentioned. A study of the influence of thiolate derivatization on the catalytic reduction of 4-nitrophenol serves as a perfect example of the importance of thiolate coverage on the catalytic performance (10). The free catalytic site density determines the activity of the material. Interestingly, considerable catalytic activity is retained even when the coverage is 90%.
The motivation for this study is our interest in the delay between reagent mixing and completion of the S-Au derivatization process. In order to address this issue, we employed two strategies. One of these methodologies involves simple fluorescence spectroscopy. In our work, fluorescence spectroscopy is used as a tool that is suitable for different combinations of nanostructure and sensing molecules (Scheme 1). In the other approach, Raman spectroscopy is used to further support our results. While recognizing that our data contain kinetic and mechanistic information and do some simple kinetic analysis, we use a rather pragmatic approach by trying to answer this question. It is important to note that the type (particularly size) of nanoparticles and the organic structures required to optimize fluorescence and SERS measurements are different, and thus, the sections that follow dealing with fluorescence and Raman spectroscopy are not directly comparable. In this case, this is an asset, as it allows us to establish the generality of the conclusions we reach in this contribution. Scheme 1 shows the molecules examined in this contribution.

MATERIALS AND METHODS
Materials and instrumentation. All reagents have been purchased from Sigma-Aldrich (St. Louis, MO) and have been used without further purification unless otherwise stated. S-methyl-7-mercapto-4-methylcoumarin (C-SMe) was synthesized as previously described (11). AuNPs were synthesized using a reported method (12) via reduction with sodium citrate. Briefly, 230 mL of a 0.3 mM aqueous solution of HAuCl 4 was heated up to boiling point, and then, 20 mL of a 39 mM solution of sodium citrate was quickly added to the solution and boiled for 2 h. The ruby-red solution obtained was kept at room temperature and properly diluted before use (see Data S1 for the calculation of AuNP concentration). Transmission electron microscopy (TEM) images were collected using a JEM-2100F FETEM (JEOL) working at an acceleration voltage of 200 kV. Steady-state absorbance and fluorescence measurements were recorded on a Cary 100 spectrophotometer and a Photon Technology International (PTI) fluorimeter, respectively. Raman spectra were recorded in a Horiba Xplora microscope configured with 532 nm (at 24 lW) and 785 nm (at 50 lW) laser lines at 100% power. Data analyses were done using LabSpec 6 software. Fluorescence spectroscopy measurements. Initial fluorescence testing was run to determine appropriate concentrations of 7-mercapto-4methylcoumarin (dimethyl sulfoxide-DMSO-solution) for fluorescence testing. Kinetic studies were performed by monitoring the fluorescence of fixed-volume AuNP solutions (~12 nm average diameter, 1.2 nM) in the presence of varying amounts of 7-mercapto-4-methylcoumarin; kinetic runs were followed at 430 nm (excitation wavelength = 358 nm). In addition, fluorescence spectra were recorded by exciting 7-mercapto-4methylcoumarin solutions before and after 1.5 h of kinetic testing.
Surface enhanced Raman spectroscopy measurements. AuNP's surface functionalization was monitored utilizing two different thiols, namely thiophenol (PhSH) and 2-Naphthalenethiol (NaphSH). Stock solutions of PhSH were prepared daily and diluted in purified 18.2 MO water, while stock solutions of NaphSH were prepared in EtOH and stored at 4°C for a maximum of one week, followed by the appropriate daily dilution in a 1:1 EtOH/water mixture. Measurements were performed in solution upon addition of thiol into AuNP (~60 nm average diameter, 31 pM) solutions at given concentrations. SERS measurements were performed on AuNP solution in the presence of various thiol concentrations and recorded at the following acquisition conditions: 785 nm laser (25% power), 5 s integration time, 10 accumulations per spectrum and 60 s measurement interval time. The peak areas were calculated using LabSpec software (HORIBA).
All kinetic experiments were fitted using a user-defined fitting with Kaleidagraph or Origin software.

Kinetics based on fluorescence spectroscopy
In our work, fluorescence spectroscopy was used as a suitable tool to study different combinations of AuNP and various sensing molecules (Scheme 1). While analyzing the obtained data to extract kinetic and mechanistic information, we use a pragmatic approach by trying to answer the question on hand: How fast can thiols bind to the gold nanoparticle surface?
We have found that 7-mercapto-4-methyl coumarin (C-SH) is an excellent substrate for fluorescence spectroscopy and commercially available. Interestingly, C-SH is weakly fluorescent as a result of non-radiative deactivation (13), attributed to the thione resonance structure contributions (Scheme 2) (11).
When the S-H bond is substituted, for example by a methyl group (C-SMe), the molecule becomes strongly fluorescent (Fig. 1), despite what is expected for its hydroxy counterpart (14). Notice that Fig. 1 shows that C-SMe is a much stronger emitter than C-SH, typically 20-100 times. We have explained this effect in an earlier contribution using Scheme 2 to rationalize the effect (11). Additionally, a modest fluorescence enhancement due to the presence of AuNP is detected, in contrast to typical plasmonic enhancements reported to be an order of magnitude larger (15). It is known that while proximity enhances the signal, surface contact results in emission quenching; the experimental observation depends on the balance of these effects, on the system as well as on the experimental conditions (15,16).
The addition of C-SH to the gold surface leads to the displacement of citrate and the formation of thiolate bonds (12), Scheme 3, with concomitant spectral changes in the AuNP region (~530 nm) and the appearance of the coumarin absorption band in the 360 nm region, Fig. 2. Thus, the S-H bond can be eliminated by coordination to the gold surface (Scheme 3), which is typically the key step in AuNP derivatization. With this in mind, we decided to design a simple way to determine how fast the interaction between thiols and AuNP surface can take place, using C-SH as a probe. Thus, we expect the thione-like resonance form of C-SH to be less favored as the thiol moiety is   engaged in binding to the AuNP surface, increasing the fluorescence emission of the coumarin. The increase in fluorescence signal can account then for the formal interaction between the thiol moiety and the Au surface and help reveal kinetic information about the process. While fluorescence quenching is expected, when fluorophores sit right on the surface of plasmonic materials such as AuNP, the experimental balance in this case is a moderate (but readily detectable) fluorescence enhancement (16). Thus, the use of fluorescence enhancement provides a novel approach to monitor reactivity with the gold plasmonic surface. The changes in Fig. 2 are attributed to modifications on the surface of the AuNPs, as well as changes in the dielectric media, as a consequence of the presence of C-SH. Changes produced by the addition of the different solvent were ruled out. Pure DMSO was added to AuNP under the same conditions, and no deviations in the absorption spectrum were detected.
AuNP used for fluorescence spectroscopy (Fig. 3) has an average size of 12 nm, significantly smaller than those used for Raman spectroscopy (~60 nm, vide infra). Particles were initially tested at two different AuNP concentrations: 2.4 and 1.2 nM (calculated as previously reported (16)). Figure 4 shows how the emission at 440 nm, corresponding to C-SH, increases after mixing with AuNP solution. Notice that while essentially the same rate constants are obtained, diluted AuNP solutions reach higher fluorescence intensities, more likely due to better light penetration and minimal light re-absorption. Figure 4 clearly shows that changes in the AuNP concentration led to unusual effects (such as more signal with less AuNP) that while readily explained on a qualitative basis pose serious challenges for quantification. Thus, most of the experiments that follow center on kinetic studies where thiol concentrations are changed while maintaining the AuNP concentration constant. Representative results are shown in Fig. 5, where C-SH was added in concentrations ranging from 1 to 6 lM and monitored for 2000 s. Note that only the emission growth component of the signal is displayed in Fig. 5, as C-SH, while weak is somewhat emissive (see Fig. 1). This emission is probably slightly enhanced by non-reactive initial interaction with the gold surface. An example of the uncorrected traces is shown as an inset in Fig. 5. A simple visual inspection of Fig. 5 shows that about 30 min is required for 90% of the reaction to take place.
The surface of each AuNP must present sites with different reactivity, further, the material size distribution is not monodisperse, as Fig. 3 shows. Under these conditions, it is remarkable that the plots of Fig. 5 can be reasonably fitted with a monoexponential growth, in other words, they follow excellent Langmuir-type first order kinetics. In fact, it was possible to use a kinetic global fitting approach to fit the data in Fig. 5 using Equation (1).
where k is the observed rate constant for the growth of A (enhanced Raman or emission signal) and is equal to the inverse of the lifetime (1/s). We obtain k as the value that minimizes the mean squared error between the fitted and the observed kinetic curves with different values of A 0 .
Global kinetic analysis of Fig. 5 yields a lifetime (s) of 735 s. Considering that three lifetimes corresponds to 95% completion, the visual estimate of 30 min mentioned above appears quite reasonable. Beyond a single lifetime or rate constant, global analysis also yields a projected plateau value for the signal (A 0 ). These values were recorded and plotted against concentration in Fig. 6. The plot reaches a plateau that implies the mercaptan and not the AuNP is the likely limiting reagent in this system. At the highest C-SH concentration, this corresponds to~3300 thiolate bonds per nanoparticle.
In order to prove the reproducibility of this methodology, a second batch of AuNP (~17 nm, see Figure S1) was prepared and the same type of kinetics was evaluated. Fig. 7 shows the kinetic traces obtained.
The lifetime values of 767, 638 and 660 s, derived from single curve analysis (Fig. 7) and global analysis (Fig. 5), illustrate the type of reproducibility that can be expected in these systems as the sample batch and/or type of analysis is changed.

Kinetics based on Raman spectroscopy
Surface enhanced Raman spectroscopy (SERS) is also an important and complementary tool in determining the reactivity of the S-H bond toward the gold surface. The best suited molecules for SERS differ from those preferred for fluorescent studies, and thus, non-fluorescent or weakly fluorescent aromatic molecules, such as PhSH and NaphSH (see Scheme 1), were the choices for these studies. AuNP concentration was also optimized in order to have greater Raman signal enhancement. Further, larger particles are optimal for SERS, and thus, citrate-covered 65 nm AuNP was preferred. The Raman spectra for PhSH and NaphSH are shown in Fig. 8; notice in the inset that in addition to polydispersity, the sample shows some polymorphism.
In the case of PhSH, Raman peaks at 408 (C-S), 1067 (C-H) and 1565 cm À1 (C=C) could be monitored to observe the enhancement that results from binding at the AuNP surface. In our case, we found the 1565 cm À1 peak to be the most convenient for PhSH (Fig. 9). Likewise, we followed the addition of NaphSH monitoring the area of the C=C peak at 1370 cm À1 (Fig. 10).

Competition binding by lipoic acid
The case of lipoic acid (LA) is particularly interesting, as its behavior should mimic disulfides, however-due to its cyclic structure-once LA binds to the gold surface produces more stable molecular arrangements because of the formation (Fig. 11 of two S-Au bonds. (17). We have designed two types of fluorescence experiments where LA and C-SH compete for the gold surface. In the first one, equimolar concentrations of both are added simultaneously to a AuNP solution. As shown in Fig. 12, the presence of LA has virtually no effect on the growth of the fluorescence of C-SH from the gold-bonded mercaptocoumarin. This result implies that for C-SH and LA, thiols are more reactive than disulfides, likely because S-S bond breaking is required   for LA. In the second type of experiments, the reagents are exactly the same, however, in this case the AuNP were pre-incubated at the same concentration of LA for 24 h, prior to addition of C-SH. These results are also presented in Fig. 12, and show a smaller and somewhat slower fluorescence growth, illustrating that LA has bonded to the gold surface thus reducing the number of active sites available for C-SH reaction.
Thus, the competition of C-SH with LA shows that thiols are far more reactive than disulfide, as adding LA as a competitive reagent does not affect the growth curve giving a growth lifetime of 560 AE 70 s. On the other hand, if AuNP is pre-incubated overnight with LA, the disulfide can bind to the surface (frequently described as two-footed binding) and protect it from the attack by thiols. This results in a lifetime of 620 AE 70 s, and the final fluorescence is only 78% of that achieved without pre-incubation. With pre-incubation, less sites are available, but those that remain available are within experimental errors just as reactive as those encountered on a fresh surface. Overall LA binding must be about ten times slower than thiol reaction, thus needing overnight incubation to make a real difference in C-SH binding.

CONCLUSION
These results prove the reactivity of several thiols occurs within the same timescale; however, disulfides (such as LA) react much more slowly. To answer the title question, we recommend one to two hour wait for thiol binding to be essentially complete, while for disulfides, overnight incubation is recommended. Any "ready-mix" strategy is bound to lead to results obtained while the formation of thiolate bonds was still in progress.