Supporting information Influence of electrolyte composition on liquid-gated carbon-nanotube and graphene transistors By: Iddo Heller, Sohail Chatoor, Jaan Männik, Marcel A. G. Zevenbergen, Cees Dekker, and Serge G. Lemay A. Device fabrication Both graphene and carbon nanotube devices were prepared using standard procedures. Carbon nanotube devices were fabricated on silicon wafers with a 200 nm thick thermally grown oxide. Prior to nanotube growth, we cleaned the wafers with acetone, 2-propanol, and concentrated nitric acid. Marker grids were defined using e-beam lithography in a bilayer PMMA resist followed by Ti/Pt deposition and a lift-off process in warm acetone followed by rinsing with 2-propanol. Islands consisting of alumina supported iron-molybdenum catalyst were then lithographically defined, and nanotubes were grown from these islands through chemical vapor deposition at 900 o C using methane as carbon feedstock. 46 In a final lithography step, Cr/Au electrodes were defined to establish electrical contact to the carbon nanotubes. Figure S1a and b show AFM images of a typical individual carbon nanotube device. Graphene devices were fabricated on silicon wafers with a 285 nm thick thermally grown oxide. The wafers were cleaned and marker grids were defined as described above. We subsequently deposited graphene flakes through mechanical exfoliation of graphite (NGS Naturgraphit GmbH) using adhesive tape. 47 Adhesive residues were dissolved in tetrahydrofuran. After identification of graphene flakes through their optical contrast, Ti/Au contact electrodes were deposited through a second e-beam lithography step identical to the first. Figure S1c and d show optical microscope images of a graphene flake before and after fabricating contact electrodes. S1
a b c d Figure S1. Typical carbon nanotube and graphene devices. (a,b) Atomic force microscopy height (a) and amplitude (b) images of a carbon nanotube in between source and drain electrodes. Scale bar: 1 µm. (c,d) Optical microscopy images of a graphene flake before (c) and after (d) fabricating contact electrodes. Scale bar: 10 µm. Ishigami et al previously observed the presence of resist residues on devices processed with common lithographic methods. 30 In an attempt to remove these residues, we tried to anneal carbon nanotube and graphene devices in an argon hydrogen atmosphere as suggested in ref 30. Figure S2 shows AFM images of carbon nanotube and graphene devices before (cf. fig S2a,b and c,d) and after (cf. fig S2e,f and g,h) annealing for 1 hour in an argon hydrogen gas flow at 300 o C. Although the roughness on the SiO 2 and carbon surfaces is clearly reduced after the treatment, complete removal of the apparent residues was not established, nor was the electrostatic gating shift as a function of salt concentration drastically reduced. Moreover, we observed that the overall device conductivity was drastically reduced in some instances, and occasionally the treatment induced the complete electrical breakdown of devices. We have not applied this annealing procedure to devices presented in the manuscript, because the argon hydrogen annealing procedure appeared unsuccessful in removing possible residues, and in order to avoid damaging devices in the process,. S2
a b c d e f g h Figure S2. Annealing carbon nanotube and graphene devices in a hydrogen argon atmosphere. (a,b) and (c,d) show AFM height and amplitude images of as-fabricated carbon nanotube and graphene devices. (e,f) and (g,h) show AFM images of the same devices after annealing for 1 hour in an argon hydrogen gasflow in a tube oven at 300 o C. scale bar in (b) and (f): 500 nm. Scale bar in (d) and (h): 2 µm. B. Modeling of ion-specific electrostatic gating Specific adsorption of ionic species provides a possible mechanism for the observed ion specificity in electrostatic gating. Simulations and experiments indeed indicate that alkali ions can have a considerable interaction energy with conjugated-π systems, SWNTs, graphene, graphite, and defect sites therein. In general, the interaction energies follow the electronegativity trend, where Li + displays a larger interaction energy than K +. Introducing competitive adsorption of alkali cations to ionizable sites in the model of equation (1) leads to an alternate form, σ σ ( ψ ) = + σ max IG OHP m( pka ph ) m( pk x px ) offset 1+ 10 exp( mβ eψ ) + 10 exp( mβeψ ) OHP OHP. (S1) Here the new parameter pk x represents the association constant of the cation to the ionizable site, and px = log(c) represents the bulk concentration of the cation, c. The grey lines in Fig S3 show the resulting surface potential for values of pk x of 2 and 4. These values fall well within the range of reported association constants. For comparison, the curve without adsorption, or pk x =, is also S3
plotted. This model is in qualitative agreement with the data presented in Figure 3 for different cations using pk Li = 2, pk K = 4, and pk TEA =. 0.12 V shift (V) 0.1 0.08 0.06 0.04 pk X - pk X -4 pk X -2 10-2 10-1 10 0 c (mol/l) Figure S3. Comparison of experimentally observed ion-specific electrostatic gating with modeling of specific ion adsorption. Blue and magenta circles represent V shift as function of salt concentration at ph 7 for SWNTs for KCl and LiCl, respectively, as previously presented in Fig 2b. Dashed lines represent the surface potential ψ S calculated using the specific adsorption model of Equation (S1) and different values of the model parameter pk x, as indicated in the graph. Another potential source of ion specificity stems from differences in hydrated ionic radius; the latter is 2.53 Å and 3.31 Å for Li + and K +, respectively. 48 In our model, the hydrated radius directly relates to the distance of closest approach of the ions to the carbon surface, x OHP, and therefore to the difference between ψ OHP and ψ S. Figure S4 compares predictions based on the model using the different hydrated ionic radii for x OHP. Interestingly, the experimentally observed difference between Li + and K + agrees quite well with the calculcation based on the hydrated ionic radii. Therefore, both specific ion adsorption and ionic radius variations are possible mechanisms for the observed differences in electrostatic gating for K + and Li +. S4
V shift (V) 0.12 0.1 0.08 0.06 0.04 x OHP = 4.09 Å x OHP = 3.31 Å x OHP = 2.53 Å 10-2 10-1 10 0 c (mol/l) Figure S4. Comparison of experimentally observed ion-specific electrostatic gating effect with modeling of ionic radii. V shift as function of salt concentration and ion-type at ph 7 for SWNT devices as previously presented in Fig 2b. Blue and magenta circles represent KCl and LiCl, respectively. Dashed lines represent the surface potential ψ S - calculated using the model and parameters explained in the main text - for different values of x OHP, as indicated in the graph. Supporting information references 46. Kong, J.; Soh, H. T.; Cassell, A. M.; Quate, C. F.; Dai, H. J. Nature 1998, 395, 878. 47. Novoselov, K. S.; Geim, A. K.; Morozov, S. V.; Jiang, D.; Zhang, Y.; Dubonos, S. V.; Grigorieva, I. V.; Firsov, A. A. Science 2004, 306, 666. 48. Nightingale, E. R. J. Phys. Chem. 1959, 63, 1381-1387. S5