Binding Kinetics of Methylene Blue on Monolayer Graphene Investigated by Multiparameter Surface Plasmon Resonance.

In this paper, we study the interaction of a small dye molecule, namely, methylene blue (MB) with graphene surfaces using surface plasmon resonance (SPR). We show that by utilizing all of the parameters of the SPR angular dip and exploiting the fact that MB absorbs light at the operating wavelength, it is possible to detect the binding of small molecules that would otherwise not give a significant signal. The binding of MB to unmodified graphene is found to be stronger than that for gold. By studying the interaction at modified surfaces, we demonstrate that electrostatic effects play a dominant role in the binding of MB on to graphene. Furthermore, following the binding kinetics at various concentrations allows us to estimate apparent equilibrium binding and rate constants for the interaction of MB with graphene.

Introduction

The interaction of dyes, such as methylene blue (MB), with carbon-based materials has been a matter of widespread interest due to the redox properties of MB[1] and broad applicability in electrochemical (bio)sensors.[2−5] MB intercalates with double-stranded DNA, and the electroactivity of MB has been exploited in nucleic acid sensors.[6,7] Moreover, composites made of MB and graphene oxide are promising for active components of immunosensors.[2] Methylene blue was found to bind strongly with carbon nanotubes and reduced graphene oxide, and the interactions were noncovalent in nature attributed to electrostatic and π–π interactions.[8−10] A recent study, involving the interaction of another dye rhodamine 6G with graphene oxide, found that the dye may covalently attach to defect sites on graphene oxide derivatives.[11] There is little systematic information yet on the interaction of MB with monolayer graphene (e.g., using exfoliation or chemical vapor deposition (CVD)). In contrast to graphene oxide-based materials, CVD-graphene[12] is rather free of ionizable groups, which might have a considerable effect on the nature and mechanism of binding. Moreover, the kinetics of binding was not investigated in the previous studies.

Surface plasmon resonance (SPR) is ideally suited for studying the interaction of molecules on surfaces since the involved binding partners can be studied in their native forms without the need for labeling or complex protocols.[13−18] Moreover, the kinetics of binding can be followed in real time. Typically, a gold surface with immobilized ligands is used for this purpose. Here, we deposit a graphene monolayer on the gold surface to study the interaction of the dye directly with graphene. To be able to detect the binding of analyte molecules on the SPR sensor surface, in the simplest case, the analyte should induce a sufficient change in the local dielectric constant at the interface.[19,20] Although this is easily observed when using large biomolecules, it is a challenge to detect very low amount of small analytes binding to the sensor surface. The minimal changes in dielectric constant set a limitation on the sensitivity for the detection of low molecular weight (LMW) molecules (molecular weight < 300 Da, such as MB).[21] Understanding the interaction of LMW compounds is crucial in the design of new pharmaceuticals.[22] LMWs that absorb light in the wavelength used for SPR measurements allow a more sensitive detection.[21] Indeed, the ability of graphene to absorb light at visible wavelengths has motivated a series of theoretical[23] and experimental[12,24−28] investigations on using graphene on gold as a sensor surface when using SPR. For the detection of analytes with receptors immobilized on the graphene surface, graphene offers some key advantages. First of all, the adsorption of biomolecules increases as a result of high surface-to-volume ratio and π–π stacking interactions between graphene and receptor molecules.[26,27,29,30] Second, it has been proposed that a single sheet of graphene may enhance the SPR response due to its unique electronic and optical properties.[12] Related to this concept are experiments where dyes have been tested as a label to enhance the SPR response.[21,31] In SPR studies reported until now, graphene was used only to facilitate the binding or improve the quality of SPR signal. In contrast, here we focus on the direct interaction of a LMW molecule (MB) with an unmodified monolayer graphene surface.

Results and Discussion

To study the interaction of methylene blue with CVD-graphene using SPR, we have devised a modified procedure to transfer graphene on to a gold surface. The gold-coated glass slides are initially functionalized with carboxyl groups to render the surface hydrophilic. This ensures a conformal transfer of CVD-graphene with minimal wrinkles and breakoff. The details of functionalization and the transfer process are presented in the Experimental Section. Figure a shows an optical image of the SPR chip with transferred graphene, whereas Figure b shows SPR angular spectra obtained in air at various stages during the transfer process. It can be seen that the SPR angular dip shifts to higher angles after surface modification and after graphene transfer, attesting the formation of a functional layer and the presence of graphene, respectively, as has been reported in previous works.[24,32] SPR angular spectra were also acquired in water and in buffer before and after transfer of graphene. These results are discussed in Figure S1 (Supporting Information).

Figure 1

(a) Optical image of graphene on a gold-coated glass slide. (b) SPR angular spectra of bare gold (Au), carboxylated gold (Au*), and graphene on Au* (Au*Gr) measured in air. The inset shows a zoomed-in region of the spectrum around the SPR dip.

Figure a presents partial SPR angular spectra in acetate buffer (pH 4.2) for the interaction of 1 μM methylene blue with graphene transferred on to the modified gold surface (the complete spectra can be found in Figure S2). It is apparent from Figure a that in the presence of methylene blue the angular dip is offset slightly to lower angles. In addition, the entire curve shifts upward with an increase in intensity at the minimum angle. Figure b shows spectra obtained on the carboxylated gold surface (in the absence of graphene) for the same interaction. In this case, the shift in the minimum angle and the intensity change at the dip are rather negligible. The absence of complete recovery in Figure a suggests that MB binds to the graphene surface during the interaction. Furthermore, the negligible shift in Figure b indicates that the interaction of MB with graphene is likely much stronger than the interaction with the gold surface.

Figure 2

SPR angular spectra measured with and without methylene blue (MB) in acetate buffer (pH 4.2): (a) graphene on carboxylated gold surface (Au*Gr) (b) carboxylated gold surface (Au*). Measurements were taken initially after 25 min of baseline in buffer (t = 25), after subsequent association for 15 min in the presence of MB (t = 40), and finally after 20 min dissociation time in buffer (t = 60). MB concentration is 1 μM. Flow rate: 130 ± 5 μL/min.

Typically, SPR investigations focus on the changes in the position of angular dip to investigate binding kinetics. However, an SPR angular spectrum contains more information than just the minimum angle. Specifically, the spectrum around the minimum angle can be approximated by a parabola and can be expressed as I(θ) = I0 + w(θ – θmin)2, where I is the measured intensity as a function of the angle θ, I0 is the intensity at the minimum angle θmin, and w the width of the parabola.[21,33] Shifts in minimum angle θmin are indicative mainly of changes in the real part of dielectric constant at the sensor-medium interface, whereas the width w and intensity I0 provide additional information about the imaginary part of the dielectric constant, for example, if the adsorbed material absorbs light.[33]Figure presents kinetic data in the form of SPR sensorgrams, where the position of the angular dip θmin as well as the intensity at minimum angle I0 and the width w are recorded simultaneously as a function of time during the interaction of MB with the gold surface in the presence and absence of graphene. For graphene (black curve), the θmin (Figure a) and I0 (Figure b) channels show clear changes during the association and dissociation phases. However, for the case of gold in the absence of graphene (red curve) a significant response is only observed in the intensity (I0) channel. The width channel (Figure c) does not show sizeable changes for both cases. The higher sensitivity of the I0 channel can be understood by considering that the adsorbed MB absorbs light at the laser wavelength (λexc = 633.2 nm) causing a shift in I0 toward higher values during the association phase. The change in the real part of the dielectric constant of bound MB is, on the other hand, not high enough to observe shifts in the θmin channel. Shifts in the intensity of the plasmon resonance dip have been observed in wavelength scanning mode earlier on absorptive films.[5]

Figure 3

Interaction of 1 μM MB on Au* and Au*Gr surfaces. Real-time evolution of the relative changes in three SPR parameters: (a) angular dip θmin, (b) intensity I0, and (c) width w on carboxylated gold surface without (red line) and with graphene (black line). For each measurement, the association time with MB was 15 min and the dissociation in acetate buffer was 20 min. The baseline in acetate buffer was measured for 25 min prior to the start of association. Flow rate: 130 ± 5 μL/min.

To gather further support that we are mainly observing bound MB and not effects due to changes in the background medium (e.g., due to MB in solution), we have characterized the samples using UV–vis spectroscopy before and after the association phase. After association, the sample is gently rinsed in water before the measurement. Figure a presents the absorption spectra (measured in air) for graphene on quartz before and after incubation with MB. The absorption spectrum of bare graphene (black curve) shows a clear Fano resonance at shorter wavelengths and a rather flat absorption close to the theoretical value of πα (∼2.3%) at higher wavelengths in accordance with previous observations.[34] After incubation with MB, there is a slight overall increase in absorption and the occurrence of an additional band at around 650 nm (red curve). The absorption spectrum can be further improved by using another graphene on quartz sample, as a reference. In this case, we can directly obtain the absorption only due to the adsorbed MB layer on graphene. Such a spectrum is shown in Figure b (black line), together with the absorption spectrum of 1 μM MB in solution (red line). It is apparent that both the absorption bands (around 290 and 650 nm) characteristic of MB are clearly discernible on the graphene sample confirming that MB indeed adsorbs strongly on to the graphene surface. Moreover, it is clear from these spectra that there is a sizeable absorption of light at the SPR laser wavelength (λexc = 633.2 nm) explaining the changes in the intensity at minimum (I0), as discussed earlier. In comparison to free MB in solution, the bound MB shows a broader absorption band around 630 nm, which may be due to aggregation of several molecular layers of MB on the graphene surface.[35]

Figure 4

(a) Absorption spectrum of graphene on quartz (black line) and graphene with bound MB (red line), using a quartz slide as blank. (b) Red curve: absorption spectrum of 1 μM MB in acetate buffer (blank: acetate buffer); black curve: absorption spectrum of bound MB on graphene/quartz, with graphene/quartz used as blank. In this case, the absorption is only due to the MB layer on graphene.

The downward shift of minimum angle θmin upon MB binding (Figure a) is consistent with thicker layers of MB depicting a lower dielectric constant (real part) for wavelengths below 665 nm.[5] Another possibility is that the bound MB introduces some kind of charge transfer on to graphene, which may modify the optical properties of graphene. However, in this case typically an upward shift of the angular dip has been reported.[11] Hence, we believe that the dominant factor causing the observed changes is due to the absorptive nature of the bound MB layer. In contrast to the case of rhodamine 6G,[11] we do not have any indication of covalent bonding, which is confirmed by Raman spectroscopy (Figure S3, Supporting Information) and due to the fact that we can recover the free graphene surface after flushing with buffer. On the basis of these considerations, we conclude that methylene blue binds (noncovalently) to graphene through electrostatic and hydrophobic interactions similar to interactions between DNA and MB.[36]

To assess the nature of these interactions further, we have investigated the effect of the concentration of MB on the SPR response on gold surfaces with and without graphene. The I0 sensorgrams for the interaction of MB at various concentrations are depicted in Figure for both surfaces. On gold surfaces, it is difficult to see a significant response below 1 μM. However, on the graphene surface a clear response is observable down to 10 nM. This is also apparent from the equilibrium plots shown in Figure , where it is clear that it is possible to detect MB at concentrations much lower than that on gold. From such plots, we have extracted an apparent equilibrium binding constant (KD) for the interaction of MB with the surface by assuming a 1:1 site binding model.[37] For the sample in Figure , we obtain KD values of 772 ± 303 nM for MB–graphene binding and 4.61 ± 2.21 μM for MB–gold binding. Similar values were observed on other samples (statistics of responses on other samples are shown in Figure S5, Supporting Information). In addition to single-step method, kinetic data can also be obtained using the kinetic titration method,[38] which avoids the need for regenerating the surface for every concentration step and allows the recording of kinetic data in a much shorter time interval. An example of such a dataset is shown in Figure , depicting a similar response as for the single-step method. From such datasets, we have estimated the kinetic rate constants by performing a global fit using the 1:1 site binding model. The kinetic parameters for the data in Figure are as follows: ka = 3.48 ± 0.2 × 103 M–1s–1 (association rate constant) and kd = 2.25 ± 0.25 × 10–3 s–1 (dissociation rate constant). The modeled kinetic binding curve using these parameters is shown as the red curve in Figure .

Figure 5

Concentration dependence of the kinetic response of relative changes in I0 for MB concentrations from 0.01 to 10 μM on a modified gold surface with ((a) Au*Gr) and without ((b) Au*) graphene. After each measurement, 0.2 M HCl was injected for 15 min to regenerate the surface for the next concentration. See Figure S4 for the time evolution of θmin and w.

Figure 6

Equilibrium curves (black lines) obtained by plotting the relative change in I0 at t = 8 min as a function of concentration for the data in Figure : (a) graphene on carboxylated gold, (b) carboxylated gold. The red lines are curves obtained by fitting a 1:1 binding model to the data, allowing us to estimate an apparent equilibrium binding constant KD.

Figure 7

Kinetic titration curve for the interaction of MB with graphene, where the surface is exposed to increasing concentrations of MB without regenerating the surface. The black arrows indicate the injection of MB at the concentration indicated, while the gray arrows indicate the transition to buffer. The red line is a fitting curve obtained by using 1:1 binding model, used to extract the kinetic parameters.

A higher KD for graphene suggests that the binding of MB on to the graphene surface is stronger than the carboxyl-terminated gold surface. It is worth mentioning that the binding is weak also on an unmodified gold surface (i.e., without carboxylation). These data are shown in Figures S6 and S7 (Supporting Information). At a pH of 4.2, the carboxylated gold surface is rather neutral (pKa of propionic acid is around 4.9) and hydrophilic, whereas the unmodified gold surface is hydrophobic and devoid of a significant charge density. In contrast, monolayer graphene on SiO2 was found to exhibit an isoelectric point less than 3.3, rendering the surface negative at all pH values greater than 3.3.[39] It is possible that graphene on gold also exhibits a significant negative charge density, which induces strong electrostatic interactions with the positively charged MB ions, explaining the stronger binding on graphene. To evaluate this hypothesis systematically, we have performed two separate sets of measurements. In a first set, we modify the surface of graphene with aromatic amino groups by electropolymerization of 4-aminobenzylamine (ABA). This renders the surface positively charged at pH 4.2 (as we have shown earlier),[39] while maintaining a similar level of hydrophobicity as the bare graphene surface. Figure a shows the kinetic response after modification of the graphene surface with 1 mM ABA, where it is apparent that the response is weaker than that of graphene. The density of the amino groups on the surface can be increased by using a higher concentration of the precursor during electropolymerization. By increasing the density of positive charges (with 10 mM ABA concentration), we observe that the response is even lower, as shown in Figure b. Figure c shows the equilibrium responses for these two cases together with that of bare graphene. It is clear that the binding of the cationic MB to the positively charged surface is significantly hampered in comparison to the bare graphene surface, as exemplified by the higher binding constant values for these two cases. In a second set of experiments, we evaluated the role played by hydrophobic interactions. For this purpose, we render the gold surface hydrophobic by functionalizing the surface with a long chain alkanethiol, namely, dodecylmercaptan (DM). Also for this case, the binding of MB on to the surface is considerably weaker, as is apparent from the kinetic response and the equilibrium plot shown in Figure . On the basis of these two observations, we conclude that electrostatic interactions between the cationic MB and the negatively charged graphene surface predominantly dictate the binding kinetics of MB on to graphene at the micromolar concentrations investigated here.

Figure 8

Concentration-dependent kinetic response of relative changes in I0 for the interaction of MB with Au*Gr (in buffer, pH 4.2) electrochemically modified with (a) 1 mM ABA and (b) 10 mM ABA (ABA: 4-aminobenzylamine). The association and dissociation times are the same as in Figure . (c) Equilibrium curves for the binding of MB to unmodified Gr (black line) and graphene modified with 1 mM ABA (red line), 10 mM ABA (blue line). As the density of positive charges on the graphene surface increases (with ABA modification), the KD values are found to increase.

Figure 9

(a) Concentration-dependent kinetic response of relative changes in I0 for the interaction of MB with a gold surface modified with dodecylmercaptan (AuDM). This modification makes the gold surface highly hydrophobic. (b) Equilibrium curves for the modified surface AuDM (black line) in comparison to the graphene surface Au*Gr (red line). With a hydrophobic surface, such as AuDM, KD is found to increase.

To demonstrate the applicability of our technique, we have also investigated the interaction of three other small molecules with the graphene surface. All these three molecules have a sizeable extinction coefficient at the SPR laser wavelength. A summary of these observations is shown in Figure . With nile blue A (Figure a), which has a chemical structure close to that of MB, a similar binding behavior in the same concentration range like that for MB is observed. This is consistent with our finding that electrostatic interactions play a dominant role, since nile blue A is completely dissociated and is found as a cation in solution (similar to MB). Figure b shows the kinetic response for the interaction of a neutral compound calmagite with the graphene surface. In this case, we observe that there is no binding observable up to a concentration of 10 μM, which further confirms the role played by electrostatic interactions. Finally, Figure c presents data obtained for the interaction of a rather hydrophobic species perylene-3,4,9,10-tetracarboxylic dianhydride (PTCDA) with graphene, where we do not see any response even up to 100 μM. PTCDA has often been used in studies for deposition from the vapor phase in ultrahigh vacuum and was observed to form self-assembled layers on epitaxial graphene through π–π stacking.[40,41] We do not see such an interaction in solution, indicating that π–π stacking probably plays a less important role under the aqueous conditions experimented here. These examples demonstrate that the method (multiparameter absorptive SPR) is applicable for studying the interaction of a range of LMW species on nanostructures, with the only requirement that the analyte molecule must have a sizeable extinction at the SPR wavelength used.

Figure 10

Screening the interaction of three different LMW molecules with graphene in buffer (pH 4.2). Concentration-dependent kinetic response of relative changes in I0 for (a) nile Blue A, (b) calmagite (Cal), and (c) perylene-3,4,9,10-tetracarboxylic dianhydride (PTCDA) on Au*Gr. After each measurement, 0.2 M HCl was injected for 15 min to regenerate the surface for the next concentration. The association and dissociation times are the same as in Figure .

Conclusions

In summary, we have quantified the interaction of a small molecule, namely, methylene blue with CVD-graphene using multiparameter SPR. By exploiting the optical absorption of the dye at the working wavelength, we observe that MB binds strongly to the graphene surface without forming a covalent bond, as exemplified from equilibrium SPR measurements and spectroscopy. The binding of the dye occurs at a much lower concentration on graphene than that on gold. We attribute the larger affinity of MB to graphene to predominant electrostatic effects corroborated by experiments using modified surfaces. MB is a widely used redox couple and hence the strong interaction of this dye with graphene will have important consequences in electrochemistry and electroanalysis. On the other hand, our results demonstrate that multiparameter SPR can be efficiently used to study interactions not only on absorbing surfaces, such as graphene, but also involving analyte molecules that absorb at the measurement wavelength.

Experimental Section

Graphene Transfer on to Gold-Coated Glass Slides

First, the gold surface was carboxylated to reduce the hydrophobicity of the surface. For this purpose, gold-coated glass slides were incubated overnight in 3-mercaptopropionic acid in ethanol. Then, the samples were washed with ethanol three times for 20 s and dried with N2 gas. After modification, the samples were cleaned with acetone and isopropanol. CVD-graphene (Graphenea) was cut into square pieces (typically 1 cm × 1 cm), and a solution of polystyrene (PS) (50 mg/mL in toluene) was spotted over the copper foil (PS/graphene/copper/graphene) and dried at 75 °C for 15 min. After the deposition of PS, the underlying copper was removed by etching in a solution of hydrochloric acid with added hydrogen peroxide (15% v/v 37% HCl + 5% v/v 30% H2O2). Then, graphene was transferred to modified gold samples and baked in the oven at 75 °C for 15 min before removal of PS using toluene. To modify the graphene surface, we use electropolymerization of a precursor, namely, 4-aminobenzylamine (ABA), whose preparation details have been reported earlier.[39] Hydrophobic gold surfaces are prepared by incubating the glass chip (coated with gold) overnight in a 1 M solution of dodecylmercaptan in ethanol.

Preparation of Methylene Blue Solutions

Solutions of methylene blue (MB), from 0.01 to 10 μM, were prepared in acetate buffer solution (10 mM, pH = 4.2, with 0.1 M KCl) freshly before each measurement.

Binding Assay

Equilibrium measurements were performed at room temperature using the classical single-step method.[42] Kinetic data were collected using either the single-step method or using kinetic titration.[38] In the classical single-step method, the association time for MB interaction with graphene was 15 min (tassoc) after obtaining a baseline. The bound MB was let to dissociate for 20 min (tdissoc) in the blank buffer. HCl (200 mM) was used to regenerate the surface for 10 min. In most concentrations, the equilibrium was attained within 15 min. In the kinetic titration method, the lowest concentration of MB was injected for 15 min (tassoc) followed by dissociation for 15 min (tdissoc). Without regeneration, this association–dissociation cycle was repeated for the remaining concentrations. The equilibrium data were fitted using a self-written Mathematica program using the Nonlinearmodelfit routine with the NMinimize method.

SPR Instrumentation

SPR measurements were performed using a RT2005 spectrometer from RES-TEC GmbH. The instrument operates in Kretschmann configuration using a HeNe laser (λexc: 632.8 nm) in angle scanning mode. The samples are glass slides (N-BK7, n = 1.515) coated with 48 nm gold and a 2 nm chromium adhesion layer. The liquid is delivered on to the chip surface using a flow channel controlled by a Fluigent 8-channel Microfluidics Flow Control System. All of the measurements were performed at a pressure of 50 mbar, which corresponds to a flow rate of 130 ± 5 μL/min.

Spectroscopy

Absorbance measurements were performed using a PerkinElmer Lambda 950/1050 spectrometer. A big graphene piece (2 cm × 1 cm) was transferred on to quartz substrate using the wet transfer method. Then, the graphene sample was placed in MB solution in acetate buffer, for 15 min, followed by rinsing in water and drying. For the thin films on quartz substrate, the PE-Labsphere accessory (150 mm Spectralon) was used to measure the transmittance (T%) and reflectance (R%). The absorbance (A) was calculated using the formula: A = (100 – T% – R%)/100. Auto zero was done with respect to quartz for Figure a, with respect to graphene on quartz for Figure b. Raman spectra were obtained on a JASCO NRS-4100 Raman spectrometer equipped with a 1024 × 256 charge-coupled device detector (Andor; air/Peltier cooled, operating temperature: −60 °C), a 400 1/mm grating, a frequency-doubled Nd:YAG diode-pumped solid state laser with an excitation of 532 nm, and a 100× (NA 0.90) objective and a power of 5.6 mW. The Raman spectra were recorded and processed using Spectra Manager and Origin 9.1.