Supporting Information Selective Manipulation of Molecules by Electrostatic Force and Detection of Single Molecules in Aqueous Solution Zhongbo Yan, Ming Xia, Pei Zhang, and Ya-Hong Xie* Department of Materials Science and Engineering, University of California Los Angeles, Los Angeles, CA 90095-1595, USA Address correspondence to yhx@ucla.edu. S1
1. Fabrication of Graphene Au Nano-pyramid Hybrid SERS Substrate The Hybrid SERS substrate used in the present study is identical to that described in ref.31. Key steps of the process flow are given below. First, we generated a template using a single layer of self-assembled polystyrene balls. The near hexagonal periodicity of different pitch is then transferred to SiO 2 mask over a Si (001) wafer via plasma etching. Similar patterns of substantially smaller pitch on the order of tens of nanometers can be produced using self assembly of diblock copolymer. 37 These two methods produce nanometer scale 2-dimensional features of poorly defined shape. We employ an additional step of anisotropic etching of Si to transfer the fuzzy 2-D features into well defined 3-D inverted pyramids bounded by {111} facets on a (001) oriented Si wafer. 38 We further employ geometrical hindrance during thermal oxidation of Si to fine tune the sharpness of the apex of the inverted pyramids. 39 200 nm thick Au films are deposited over the pitted surface, bond to a handle substrate using epoxy, and then lifted off of the surface thereby completing the nano-casting process. Because of the way it is fabricated, the Au tipped surface has the unique features of in-plane anisotropy with wafer-scale coherency in the precise orientation and shape of individual pyramids. 2. Preparation and Transfer of Graphene 25 micron-thick copper foil is cut into a 2x2 inch square. The copper foil is placed at the center of a quartz chemical vapor deposition (CVD) tube of 15 cm diameter. It serves the purpose of catalyst during CVD growth. The furnace is heated up to ~1060 C under the flow of H 2 with 1 Torr total pressure. After 30 minute S2
annealing, the growth commences under 20 Torr total pressure in the flow of CH 4 (~20 sccm) and H 2 (~1000 sccm) for 15 minutes. Then the chamber was cooled down to room temperature within 10 hours. A ~500 nm-thick poly(methylmethacrylate) (PMMA) layer was spin-coated on the CVD graphene covered Cu foil to provide mechanical support to the monolayer graphene during the subsequent Cu etching step. The Cu foil was removed in an etching solution of FeCl 3 : H 2 O (1:5 vol.%). Then the floating PMMA-graphene structure was transferred onto the surface of de-ionized water. Then the sample was transferred onto a target substrate. In the final step the PMMA supporting layer was removed by acetone. Figure S1 shows the SEM image of graphene transferred on Au tip region. Figure S1. SEM of graphene on Au tips. 3. Raman Enhancement Factor Detail calculation steps of Raman enhancement factor are described in the ref. 31. Here we give the key steps of the calculation to show the enhancement factor is about 10 7. S3
Compared to conventional Raman spectra, The surface enhanced Raman signals comes predominantly from electromagnetic field concentration that is extremely localized to typically nanometer region. Here we introduce a simple definition of graphene Raman enhancement factor for graphene-plasmonic composite structures: I( patt ) / A( hotspot) ( ) ( ) F = (1) I unpatt / A excitation spot Where I(patt) is the graphene Raman peak intensity including G and 2D bands measured at hotspot on the patterned Au tip region. A(hotspot) is the area of a typical hotspot. The I(unpatt) graphene Raman peak intensity measured at Au film without patterned plasmonic features. The A(excitation spot) is the area of the excitation spot. For 633 nm laser source, the theoretical minimum value of the spot diameter is 633 nm. In the experiment, the spot diameter we observed is about 1 µm. As shown in Figure S2, the hotspots are highly localized region within area of 10 nm 2 scale (nanotip region) which is supported by EM wave simulation. Raman signal from tipped surface is composed of two parts: the non-resonant part from the entire illuminated area of 1 µm x 1 µm, and the resonant part from hot spot of the 5 nm x 5 nm area. 40 Part one is assumed to be the same as that from flat Au or SiO 2 region which contribute to 1/1000 of the signal we measured. The other 99.9% of the signal comes from an area of 5 nm x 5 nm. To calculate the enhancement factor, this signal should be divided by the corresponding non-resonant signal from the same 5 nm x 5 nm area, or alternatively the typical signal intensity multiplied by a factor of (5 nm x 5 nm)/(1 µm x 1 µm). Here is the additional 10 4 factor of enhancement. Together with the apparent intensity enhancement 10 3, we obtain 10 7 for total enhancement. Figure S4
S3 shows the performance of the SERS substrate. Figure S2. Graphene Raman spectra on Au tip region. The spectrum 2 is taken from a hotspot region in the Raman intensity map. Inset: Raman 2D band intensity mapping of graphene on Au tip region. 31 S5
Figure S3. Raman spectra of Lysozyme on graphene Au pyramids substrate. The graphene signal is marked by the D peak and G peak and the concentration of Lysozyme could achieve as low as 10-12 M. 31 4. Fabrication Process Flow of the PDMS Cells Key steps of fabrication process flow is shown in Figure S4. A mould of cylinder and square wall was placed on an ITO glass (50 mm x 50 mm, 1.1 mm thick). The conductive layer of the IT glass is 200 nm thick and the sheet resistance is about 10-15 Ω per square inch. PDMS was then poured into the mould and cured at 80 ºC for 2 hours (Fig.S4a and b). In the meantime we pasted SERS substrate onto another ITO glass using silver conductive epoxy, and transferred graphene onto it (Fig. S4c and d). Transfer method has been described in Part 2. In the last step these two pieces of ITO glass were adhered to each other to form a cell and the solution was ejected into the cell by syringes. S6
(f) Figure S4. Fabrication flow and setup of the PDMS cells 5. Simulations Using COMSOL Multiphysics COMSOL Multiphysics was employed to simulate the motion of molecules. Figure S5 shows the simulation model. In order to make it more close to the real experiment environment, the built in material is set as water, the substrate material is set as Au, and the particle model (including mass and radius) is based on R6G molecules. As observed in the experiment, the laser spot diameter is about 1 µm. As shown in Figure S5b, it could roughly cover seven Au pyramids. Therefore in our simulation, we monitored the variation of particle quantity within the region of seven pyramids. Considering the computational capability, we set the total side length of the simulation region as 3 µm, which is twice the length of the monitored region. S7
(a) (b) Figure S5. Simulation model by Comsol Multiphysics (a) 3D view, (b) xy plane First we simulated the motion of molecules at low concentration. When the solution concentration is about 10-9 M, the number of molecules in the volume 3 µm x 3 µm x 3 µm is about 16. The electric field intensity in this volume is uniformly distributed with the value of 10 2 V/m. In the simulation model, the boundary condition must be chosen very carefully. Figure S6a shows the initial state when the simulation started. For the boundary marked by A, B, C, D and E in Figure S6 (a) we applied the diffuse scattering condition. It means when particles arrive at these boundaries they are bounced back with a random velocity direction. The reason of choosing this boundary condition is that during the entire process the solution concentration should keep constant, so once some particles move out of the monitored region there should be about the same number of particles moving into this region. The diffuse scattering condition meets this requirement. Also the diffuse scattered particles could be seen as newcomers so they should have random velocity. Therefore by applying this boundary condition, the molecules that attracted to the hot spots and contributed to the SERS spectra do not have to be those molecules initially within the detection region. In other words, the molecules that attracted to the hot spots are very possibly traveled S8
from remote places. For the Au pyramid region, when the particles move onto it they would stick to the region so that we could count the amount of these particles. Figure S6. Simulation for the 16 particles motion under electrostatic field, xz-plane.(a) 0 s, (b) 30 s, (c) 60 s, (d) 90 s, and (e) 120 s. During the molecular motion, three driving force are involved in the process includes: (1) electrostatic force, (2) Brownian motion force and (3) drag force. For the electrostatic force: r r r r F = qe+ E E electrostatic ( εα ) (2) where q is the charge of molecules, E r is electrical field intensity, ε is the dielectric constant, α is the polarizability of molecules. The first item represents the electric force on point charge and the second item represents the electric force on dipoles. For the Brownian motion force and drag force, S9
r F brownian r =ζ 12π k µ Tr B P t (3) r 1 r r 2ρ r p F = m ( u v), τ = drag P P τ 9µ P 2 p (4) where ζ r is a Gaussian random number with zero mean and unit variance, µ is viscosity, T is temperature, r is the radius of particles, m P p is the mass of particles, ρ is the density of particles. u r and v r are the velocity of the medium and particles p respectively (in most cases u r =0). In the simulation we adopted these equations. The simulation result is shown in Figure S6. Within 120 s, there are 2-5 particles attracted to the substrate on average, which is in accordance with the experiment results. We also calculated the drag force and electrostatic force on one particle (Figure S7a). It shows that the drag force quickly increased at the very beginning and then approached to the same level of electrostatic force. The qualitative analysis is shown in Section 6. In order to understand the trend of molecular motion at higher concentration, we also simulated the motion of 5000 particles (~10-7 mol/l) in the volume of 3 µm x 3 µm x 10 µm. The variation of particle quantity in the monitor region with time and voltage bias is shown in Figure S7b. We could see the larger the voltage bias is the faster the molecules move to the pyramids region, which is correspondence with our expectation. At zero bias, the change of molecular distribution also exists even though it is much slower than other cases. S10
(a) (b) Figure S7. (a) The change of drag force and electrostatic force with time. (b) The change of particle number in the monitor region with time and voltage bias. 6. Qualitative Analysis of Molecular Motion There are three forces involved in the molecular motion. Based on Stokes' law, r r r F = 6πµ r u v, under the electrostatic force and drag force, the particles would drag p ( ) quickly reach a constant drift velocity: r v = r ee π µ 6 rp (4) Plugging in the parameters, we could get that the velocity is about 10-5 m/s, mp which is consistent with the simulation results (Figure 3). The time constant ( 6 πµ r ) of reaching equilibrium is decided by the molecular mass and radius. It is on the order of nanosecond p for both R6G and 4-MBA (the time constant of R6G is 2.5 larger than that of 4-MBA). The Brownian force leads to random motion, which obeys the diffusion equation: x ~ 2 Dt. D is the diffusivity with D~ kt. The diffusion distance in 1000 s is 6 π rµ p about 1 mm. Comparing with the drift distance of 1 cm in 1000 s under the bias voltage employed in our experiments, the diffusion length is one order of magnitude S11
smaller thus being a minor effect. Based on the above arguments, it is obvious that the molecular motion is dominated by the electrostatic force and drag force with the Brownian force playing a minor role. As such, the molecules are expected to undergo clear directional movement. The drift velocity is highly dependent on the electric field and the particle radius. S12