In Situ Nanoscale Redox Mapping Using Tip-Enhanced RamanSpectroscopyGyeongwon Kang, Muwen Yang, Michael S. Mattei,† George C. Schatz,*and Richard P. Van Duyne*

Department of Chemistry, Northwestern University, Evanston, Illinois 60208, United States

*S Supporting Information

ABSTRACT: Electrochemical atomic force microscopy tip-enhanced Ramanspectroscopy (EC-AFM-TERS) was used for the first time to spatially resolvelocal heterogeneity in redox behavior on an electrode surface in situ and at thenanoscale. A structurally well-defined Au(111) nanoplate located on apolycrystalline ITO substrate was studied to examine nanoscale redoxcontrast across the two electrode materials. By monitoring the TERS intensityof adsorbed Nile Blue (NB) molecules on the electrode surface, TERS mapswere acquired with different applied potentials. The EC-TERS maps showeda spatial contrast in TERS intensity between Au and ITO. TERS line scansnear the edge of a 20 nm-thick Au nanoplate demonstrated a spatialresolution of 81 nm under an applied potential of −0.1 V vs Ag/AgCl. Theintensities from the TERS maps at various applied potentials followedNernstian behavior, and a formal potential (E0′) map was constructed byfitting the TERS intensity at each pixel to the Nernst equation. Clear nanoscale spatial contrast between the Au and ITOregions was observed in the E0′ map. In addition, statistical analysis of the E0′ map identified a statistically significant 4 mVdifference in E0′ on Au vs ITO. Electrochemical heterogeneity was also evident in the E0′ distribution, as a bimodal distributionwas observed in E0′ on polycrystalline ITO, but not on gold. A direct comparison between an AFM friction image and the E0′map resolved the electrochemical behavior of individual ITO grains with a spatial resolution of ∼40 nm. The variation in E0′was attributed to different local surface charges on the ITO grains. Such site-specific electrochemical information with nanoscalespatial and few mV voltage resolutions is not available using ensemble spectroelectrochemical methods. We expect that in situredox mapping at the nanoscale using EC-AFM-TERS will have a crucial impact on understanding the role of nanoscale surfacefeatures in applications such as electrocatalysis.

KEYWORDS: Tip-enhanced Raman spectroscopy (TERS), nanoscale electrochemical imaging, site-dependent electrochemistry,Nernst equation

Understanding nanoscale electrochemistry on heteroge-neous electrode surfaces has become an important area

of research due to recent developments in the nanofabricationof electrode materials1−4 for applications including sensors2

and electrocatalysis.4 In particular, nanoscale features on theelectrode surface play an important role in determining themechanism, kinetics, and thermodynamics of interfacial redoxreactions;5−8 however, the relationship between local redoxbehavior and electrode features at the nanoscale is poorlydefined in most work. The most common techniques to mapthis spatially varying electrochemistry are scanning electro-chemical microscopy (SECM)9,10 and scanning ion-conduc-tance microscopy (SICM).11 SECM measures electrodereactivity at the nanoscale via the current produced by aredox mediator cycling between a nanoelectrode tip and asubstrate electrode.9,12,13 SICM detects local variations insurface topography and reactivity by measuring the ioniccurrent through a nanopipette tip. The spatial resolution ofthese methods has been mainly limited by the tip geometryand diffusion. Recent efforts to fabricate structurally well-

defined nanoelectrodes2,3,14−16 have enabled nanoscale spatialresolution imaging of local electrochemical currents.6,17−22

However, a detailed chemical understanding of adsorbate−surface interactions cannot be readily obtained from theelectrochemical current alone.Electrochemical tip-enhanced Raman spectroscopy (EC-

TERS) based on atomic force microscopy (AFM)23,24 andscanning tunneling microscopy (STM)25−28 has recently beendeveloped as a powerful tool for selectively probing differentelectrode surface locations. TERS is known to provide detailedchemical information with subnanometer lateral resolutionunder ultrahigh vacuum conditions29−31 and a few nanometersunder ambient32,33 or liquid conditions.27 In EC-TERS,subdiffraction-limited spatial resolution can be achieved usingthe nanoscale resolution of STM and AFM,34−36 and this iscombined with the single-molecule sensitivity and richchemical information available from surface-enhanced Raman

Received: January 22, 2019Published: February 14, 2019

Letter

pubs.acs.org/NanoLettCite This: Nano Lett. XXXX, XXX, XXX−XXX

© XXXX American Chemical Society A DOI: 10.1021/acs.nanolett.9b00313Nano Lett. XXXX, XXX, XXX−XXX

Dow

nloa

ded

via

NO

RT

HW

EST

ER

N U

NIV

on

Febr

uary

27,

201

9 at

19:

41:0

9 (U

TC

).

See

http

s://p

ubs.

acs.

org/

shar

ingg

uide

lines

for

opt

ions

on

how

to le

gitim

atel

y sh

are

publ

ishe

d ar

ticle

s.

spectroscopy (SERS). Recently, Mattei et al. reported theobservation of single- and few-molecule electrochemistry ofNile Blue (NB) adsorbed on an indium tin oxide (ITO)electrode24 based on the AFM-TERS platform (EC-AFM-TERS).23,24 In that work, TERS intensity−potential curves, so-called TERS voltammograms, were obtained by acquiring TERspectra concurrently with surface cyclic voltammetry (CV).The acquired TERS voltammograms showed single or doublesteps, and the step-like results were quantitatively analyzedusing the Laviron model. Surface site heterogeneity across theITO surface was reflected in spatial variations in the formalpotential (E0′) extracted from the quantitative analysis.Although this work provided a description of how the localelectrochemistry is affected by molecule−surface interactions,the correlation of local variations in E0′ with electrodestructure was lacking. Herein, we present potential-dependentand nanoscale TERS maps of NB, acquired in tandem withAFM imaging, on a heterogeneous electrode surface. TheTERS intensity at each pixel in a series of images at differentapplied potentials was fit to the Nernst equation. The resultingsite-dependent E0′ map represents the first TERS study ofnanoscale electrochemical heterogeneity correlated with localelectrode structure.All EC-AFM-TERS experiments were performed on a home-

built TERS setup described previously.23,24 In order toproduce a well-defined electrode surface for correlated EC-TERS and AFM imaging, single crystal Au nanoplates(typically either hexagons or equilateral triangles, and someof the latter having truncated tips) were synthesized byreducing AuCl4

− with lemon grass extract following apreviously reported procedure.37,38 A 50 g amount oflemongrass was finely chopped and boiled in 250 mL ofwater for 5 min. A 5 mL aliquot of room temperature extractwas added to 45 mL of 1 mM HAuCl4 (Sigma-Aldrich) andleft to react for ∼24 h without stirring. The solution-containing

plates were purified by centrifuging 25 mL aliquots at 5000rpm for 10 min, pouring off the supernatant, and redispersingin 25 mL of Mili-Q water. The cleaning process was repeatedonce more. ITO coverslips (40 mm × 22 mm, 8−12 Ωresistivity, SPI Supplies) were cut into 20 mm × 22 mm pieces,sonicated in ethanol (Thermo Fischer Scientific) for 20 min,and dried under nitrogen. A 100 μL aliquot of cleaned Aunanoplate solution was drop-cast onto an ITO coverslip andwas dried under ambient conditions overnight. To removeweakly bound Au nanoplates and nanospheres (side products),the sample was sonicated in ethanol and water for 3 min each.The sample was then incubated in 10 μM Nile Blueperchlorate (Sigma-Aldrich) solution in ethanol overnight.The excess Nile Blue solution was removed by touching theside of ITO to a piece of Kimwipe and incubating the samplein 20 mL of ethanol for 20 min. After drying the sample,copper tape (Ted Pella) was attached to the edge of the ITOto make an electrical contact. The sample (working electrode)was then sealed in a custom-made electrochemical cell24 with asilicone O-ring. Pt and Ag/AgCl (Pt and Ag wires from AlfaAesar) wires were used as counter and quasi-referenceelectrodes, respectively, and fixed into the cell using Hysolepoxy (Loctite). A mixture of 50 mM Tris buffer at pH = 7.1and 50 mM NaCl was used as a supporting electrolyte. Aucoated contact mode silicon AFM tips with a cantileverfrequency of 15 kHz and a spring constant of 0.2 N/m wereused (NaugaNeedles). A closed-loop scanner with active XYfeedback was used, and the instrument is in a thermallyinsulated chamber to minimize drift. A drift rate of ∼3.4 nm/min was measured (Figure S1). A 633 nm continuous waveHe−Ne laser (Spectra Physics) was coupled into a single modefiber (Thorlabs) and directed into an isolation chamberhousing the AFM-TERS system. TER signals collected fromthe AFM-TERS setup were then passed through a custommultimode fiber bundle (50 μm core diameter, FiberTech

Figure 1. (A) Schematic of EC-AFM-TERS imaging experiment. Redox reaction mechanism of NB involving 2 e− and 1 H+ at pH = 7 is shown inthe top left corner. (B) Representative TER spectrum of NB (10 μM incubation concentration) acquired on a Au nanoplate using 500 μW of 633nm excitation. 110 TER spectra obtained with a 1 s acquisition time were averaged. The asterisk at 520 cm−1 indicates the Si signal from the AFMtip underneath the Au film. (C) AFM image of a Au nanoplate on ITO (3 μm × 3 μm). (D) Two AFM line profiles corresponding to the blue andred lines in (C).

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b00313Nano Lett. XXXX, XXX, XXX−XXX

B

Optical), which was directed to a spectrograph (Isoplane SCT-320, Princeton Instruments) equipped with a CCD camera(Pixis 400, Princeton Instruments).Figure 1A shows a schematic of the EC-AFM-TERS imaging

experiment. At pH = 7, the electronic resonance of NB at∼634 nm results in an extra enhancement of the TERSintensity with 633 nm laser excitation.23 At reducing potentials,NB undergoes a two-electron, one-proton transfer reaction.Because this process breaks the conjugation of the phenoxazinemoiety, the intensity of the TER spectrum decreases upon thereduction of the molecule. In order to investigate the redoxbehavior on a well-defined nanoscale electrode feature, Aunanoplates with a thickness of 15−25 nm were placed on anITO electrode substrate. As-synthesized, the Au nanoplates areknown to be single-crystalline with the (111) face predom-inantly exposed,37 in contrast to the polycrystalline ITOsurface. The ultrathin nanoplates allow enough transmission ofboth excitation and TERS photons through the electrodesurface for sufficient TERS signal to be collected. Someabsorption is still expected due to the finite thickness of thenanoplates, as will be discussed later in the paper. TER spectrawere acquired while raster scanning the Au-coated AFM tipover the border between the Au nanoplate and the ITOsubstrate. The average of 110 TER spectra of NB with the tiplocated on the nanoplate is shown in Figure 1B. The TERspectra of NB on ITO have been published previously,23,24 andno significant difference in Raman shift or relative intensity wasobserved on Au compared to ITO (Figure S2). An AFM imageof the electrode surface (Figure 1C) reveals that thepolycrystalline ITO surface has a significant number of grainswith a lateral size of 50−100 nm. In contrast, the Au nanoplatehas a well-defined triangular shaped surface with an ∼1.6 μmedge length. Two AFM line scans selected in Figure 1C areshown in Figure 1D. Both line scans show the well-definedtopography of an ∼20 nm thick Au nanoplate, with a fewprotrusions due to the roughness of ITO grains underneath theplate. An SEM image of a similar nanoplate supported on ITOalso shows clear morphological differences in the surface of theAu nanoplate and ITO (Figure S3).

Figure 2A shows an AFM image of the corner of a Aunanoplate on ITO. We chose this well-defined electrodefeature for TERS mapping experiments. TERS maps wereconstructed by acquiring TER spectra pixel by pixel whileraster scanning the AFM tip over the surface, followed byintegrating the NB 591 cm−1 mode. In our previous EC-AFM-TERS work regarding single- or few-molecule electrochemistry,the redox behavior of NB was extracted from the integratedTERS intensity while concurrently sweeping the potential.24

Conversely, in this work, the electrode was held at a constantpotential for the duration of each image. It is known that thekinetics of the NB redox reaction (k0 = 10 s−1) are ∼10 timesfaster than the acquisition time in our TERS measurements (1spectrum/s).24 Thus, the kinetics of the reaction are obscured,and only the thermodynamics (E0′) will be discussed. TERSintensity maps at applied potentials of −0.1, −0.3, −0.4, −0.45,−0.5, and −0.6 V vs Ag/AgCl, respectively, are shown inFigure 2B−H. The TERS intensity is generally stronger on Authan on ITO at potentials positive of E0′ (Figures 2B−D) butis similar after the reduction (Figures 2E−H). No edge effectwas observed in the TERS intensity maps, unlike the work byEl-Khoury et al. in which a stronger TERS signal of 4-mercaptobenzonitrile at the Au(111) step edge under ambientconditions was reported.39 The possibility of lower molecularcoverage on the nanoplate edge can contribute to this result,but a major factor is likely the coarse spatial resolution of thepresent measurements compared to El-Khoury. In addition,resonant NB is less sensitive to the local electric field intensityand polarization than 4-mercaptobenzonitrile which is non-resonant and forms a well-defined self-assembled monolayer.As expected from the surface CV measurements (details insections S1 and S2 and Figure S4 in the SupportingInformation) and the potential-sweep TERS voltammogramcurves,24 the overall TERS intensities in the TERS mapssteeply decrease around the E0′ (∼ −0.4 V vs Ag/AgCl) of NB.Also, the calculated surface coverages of NB on ITO and Auare 0.005 and 0.035 monolayer, respectively, which corre-sponds to 2 and 14 molecules in a 20 nm × 20 nm pixel area(details in section S3). Since there is likely at least one

Figure 2. (A) AFM image of the corner of a Au nanoplate on ITO (320 nm × 320 nm). (B−H) TERS intensity maps showing the variations of 591cm−1 peak area of NB TERS acquired from the electrode surface shown in (A) with a 1 s acquisition time per pixel. Potentials were held at −0.1,−0.3, −0.4, −0.45, −0.5, −0.6, and −0.8 V vs Ag/AgCl, respectively. White dotted lines represent the borderline of the Au nanoplate. Each TERSpixel size is 20 nm × 20 nm. (I) Selected zoomed-in TER spectra around the 591 cm−1 NB TERS peak with a 1 s acquisition time obtained on Au(black) and ITO (red) pixels under each potential. Tip-retracted spectrum after the imaging experiment is shown as a blue spectrum.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b00313Nano Lett. XXXX, XXX, XXX−XXX

C

molecule residing in each pixel, nonzero TERS intensity isobserved everywhere in TERS maps. Multiple cycles of surfaceCVs confirmed that the coverage of NB on both Au(111)single crystal and ITO does not change on the time scale of ourimaging experiments (Figure S5). We observe no significantchange in the average TERS intensity during the course of eachimaging experiment, verifying the chemical stability of ourmeasurements (Figure S6). A tip-retracted spectrum after theimaging experiment was acquired to confirm if the tip apex isnot contaminated (Figure 2I). There seems to be a small NBpeak originating from the few molecules adsorbed on the upperpart of the tip that is electrochemically inaccessible. Thisbackground peak is not critical in our experiment due to its lowsignal-to-noise-ratio (Table S1). Further, in the TERS imagesdiscussed below, we still observe electrochemical contrast inspite of this small amount of tip contamination, confirmingthat it has no significant impact on our results. RepresentativeTER spectra at the marked pixels in the intensity maps shownin Figure 2I indicate that TERS intensities on both Au andITO decrease as the potential becomes more negative due tothe reduction of NB.The spatial resolution of our TERS images was determined

by plotting TERS line scans across the border of the Aunanoplate. Figure 3A is the same AFM image shown in Figure2A, but with a red horizontal line indicating the trace on whichan AFM topography profile was obtained. Figure 3B shows

that the thickness of the imaged Au nanoplate is 20 nm, with a70 nm wide red shaded region indicating the edge of thenanoplate. The AFM topography at the nanoplate edge slowlydecreases rather than showing a vertical drop. However, theedge of the nanoplate shows a vertical drop in the SEM image(Figure S3). The gradual drop in the AFM topography is anartifact due to the steep drop at the edge of the nanoplate andthe radius of the AFM tip.40,41 At this length scale, we havereached the resolution limit of our system. In Figure 3C, TERSline scans at −0.1 and −0.2 V vs Ag/AgCl are shown to obtainthe spatial resolution of our measurements. At each acquisitionpoint indicated in Figure 3A, a 2 × 2 array of TERS imagepixels was averaged to obtain a smoothed TERS line scan (40nm × 40 nm acquisition area for each point; raw TERS linescans are shown in Figure S7). The averaging was performeddue to signal fluctuations commonly observed in the fewmolecule TERS measurements.42,43 The pixels in each linescan are divided into two categories by the nanoplate edge (redshaded) region, where the left four pixels and the right threepixels represent TERS intensities from NB on Au and ITO,respectively. At both potentials, the averaged TERS intensityon Au is 21% higher on Au than on ITO. TERS line scans werefit to the normal cumulative distribution function with thenonlinear curve-fitting function (lsqcurvef it function) imple-mented in Matlab to determine the spatial resolution of eachline scan (details in section S4). The fits yield spatial resolutionvalues of 81 and 90 nm at −0.1 and −0.2 V vs Ag/AgCl,respectively. A recent study reported that the tip-broadeningeffect results in a lower lateral resolution in AFM-TERSmeasurements due to the interaction between the tip shaft andthe side of an imaged object.44 Therefore, the origin of themeasured spatial resolution is likely limited by the AFMresolution which is mainly determined by the radius ofcurvature and the geometry of the tip.The origin of the observed intensity contrast needs to be

discussed since the contrast plays an important role indetermining the spatial resolution of our measurements.Considering the ∼8 times higher coverage of NB on Aucompared to ITO obtained from surface CVs (Figure S4), theobserved intensity contrast is smaller than expected. One of thefactors contributing to this unexpected weaker intensitycontrast is due to the absorption of light by the Au nanoplate.Deckert-Gaudig et al.45 reported Au nanoplates as idealsubstrates for TERS due to their optical transparency.However, a significant power loss is still expected throughthe nanoplate due to the finite thickness of Au. The intensity ofan electromagnetic field transmitted through a material with athickness of z is described by the Beer−Lambert law:

= α−I z I e( ) z0 (1)

where I0 and I(z) are the intensities before and aftertransmission, respectively, and α is the absorption coefficientof the material. A penetration depth (δp) is defined to be theinverse of α which can be described with respect to thecomplex index of refraction, Im(n(λ)), where λ is thewavelength of the laser:

αδ

πλ

λ= = Im n1 4

( ( ))p (2)

Then δp can be described with dielectric constants andexcitation wavelengths as described in the following equation:

Figure 3. (A) AFM image shown in Figure 2A with an AFM line scan(red horizontal line) and TERS line scan pixels marked (red verticallines). (B) AFM topography line scan along the horizontal line in (A),across the Au nanoplate and ITO (Nanoplate thickness = 20 nm).(C) TERS line scans (dots) acquired at pixels marked in (A) andfitting curves at potentials of −0.1 (red) and −0.2 (blue) V vs Ag/AgCl. The lateral resolutions are 81 and 90 nm, respectively. In (B)and (C), the red shaded region indicates the border range dividing Au(left) and ITO (right) in the AFM topography. Calculated (D)penetration depth profile and (E) Raman collection efficiency througha 20 nm Au nanoplate using dielectric constants of Au. Red verticalline indicates the excitation laser wavelength (633 nm) used in theexperiment. (F) Simulated gap-mode enhancement factor profile,(EAU/EITO)

4, along the axis normal to the surface and the tip apexwith a tip−surface distance of 5 nm at 633 nm excitation wavelength.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b00313Nano Lett. XXXX, XXX, XXX−XXX

D

δ λπ λ

=Im n4 ( ( ))p

(3)

Figure 3D shows a plot of δp with varying laser wavelengthscalculated using the above equation with the experimentallymeasured dielectric function of Au.46 At 633 nm, thepenetration depth of Au is calculated to be 15 nm. Followingeq 1, the intensity after the 633 nm laser is transmitted througha 20 nm-thick Au plate is 26% relative to the initial intensity. Ina bottom illumination geometry, the intensity of the scatteredlight is also attenuated by absorption by the nanoplate.Assuming equal attenuation of the incident and scattered light,the collected Raman signal is proportional to the square of thetransmission efficiency. In Figure 3E, the attenuation of thecollected Raman signal for a bottom illumination optical setupthrough a 20-nm-thick Au plate with varying laser wavelengthsis shown. This plot indicates the Raman collection efficiencythrough the plate at 633 nm is 7% of the collection efficiencyon ITO. After taking the coverage difference of NB on Au andITO in addition to the attenuation effect into account, theRaman collection efficiency through the Au nanoplate is stillexpected to be 54% of the efficiency on ITO. This value is∼2.3 times smaller than the observed TERS intensity contrast.The extra enhancement on Au is attributed to the gap-modeenhancement from the coupling between the plasmonic tip andthe substrate. To confirm the coupling between the tip and thesubstrate, a boundary element method (BEM) based on theMNPBEM Matlab toolkit47 was carried out (details in sectionS5). On the ITO surface, higher electric field intensity isobserved in proximity to the tip apex which is in line with theplasmonic nature of the Au nanostructure. On the Au surface,the plasmon mode of the tip-nanoplate assembly is dominatedby that of the Au nanoplate (Figure S8). The calculated gap-mode enhancement factor ((Eau/EITO)

4) within the tip−surface junction ranges from 2.6 to 15.7 which is in goodagreement with our experimental result (Figure 3F). Thus, weclaim the gap-mode enhancement is the major factor for theobserved spatial contrast in our TERS intensity maps.In order to quantitatively investigate the site dependence of

NB electrochemical behavior, we determined E0′ values at eachpixel in our potential-dependent TERS images. In theequilibrium regime, the TERS intensity at a single imagepixel will vary with potential according to the Nernst equation:

= ′ −[ ][ ]

E ERTnF

lnNBNB

0 red

ox (4)

= ′ −[ ] − [ ]

[ ]E E

RTnF

lnNB NB

NB0 ox 0 ox

ox (5)

where [NBox] and [NBred] refer to the coverages of oxidizedand reduced forms of NB, respectively, and [NBox]0 is theinitial coverage of oxidized NB. The number of electronstransferred through the process is assumed to be 2 at pH = 7(n = 2). Assuming that the TERS intensity is proportional tothe coverage of the oxidized form of NB ([NBox]), thecoverage expression in eq 5 can be converted into anexpression that depends on the TERS intensity:

= −−

E ERTnF

I I

Iln0 TERS,max TERS

TERS (6)

where ITERS is the TERS intensity at the potential, E, andITERS,max is the maximum TERS intensity over the potentialwindow. ITERS can be now written in terms of potential E:

=+ − −

+′

ÄÇÅÅÅÅÅÅ

ÉÖÑÑÑÑÑÑ

II

1 E EI

exp ( )nFRT

TERSTERS,max

0 TERS,min

(7)

Here, ITERS is shifted by ITERS,min to compensate for the weakbackground signal detected regardless of the potential. TheTERS intensity averaged over the whole surface, and separatelyover the Au and ITO regions at each potential, was fit to eq 7and shown to follow Nernstian behavior (Figure S9). It isimportant to note that in our TERS imaging experiments, E0′represents an average value for the forward and reversereactions. In our previous potential sweep experiments, E0′values for the forward and reverse reactions were consideredseparately due to unusual quasi-reversibility, and would moreaccurately be called half-wave potentials.24

To further investigate spatial variations in the redox behaviorof NB, a distribution of E0′ over the imaging window wasobtained by fitting the TERS intensities in Figure 2 over thepotentials at each TERS pixel to the Nernst equation. Figure4A shows the resulting E0′ map, and it is evident that there is aclear contrast between Au (left) and ITO (right) where E0′values are generally more negative on Au than on ITO. Thiscomparison is in good agreement with the ensemble E0′ valuesobtained from the surface CVs on Au(111) single crystal andITO, where E0′ on Au is 16 mV more negative than on ITO(Figure S4). Figure 4B is the reconstructed E0′ map comprisedof 40 nm × 40 nm size pixels by averaging the intensities of 4neighboring pixels (a 2 × 2 grid of 20 nm × 20 nm pixels) inTERS maps followed by fitting TERS intensities at each pixelto the Nernst equation. This pixel conversion was performedsince the aforementioned spatial resolution was based on a 40nm × 40 nm pixel size. Even with a coarser binning of the map,the spatial contrast in the E0′ map is still observed.To statistically analyze the E0′ values and study their site

dependence, the pixels in Figure 4A were categorized using thewhite lines that determine the Au nanoplate border. Figure 4Cshows a histogram of the distribution of E0′ values on the twodifferent substrates overlaid with unimodal Gaussian fittingcurves. Two E0′ distributions were confirmed to be statisticallydistinguishable by performing a two-sample t test (p-value: 6 ×10−11 and 5% significance level). Therefore, the distributionsshow that the E0′ values on Au are more negative than on ITO.Parameters from the Gaussian fits using the unimodal Gaussiancurves are given in Table 1. The mean value of the E0′distribution is 4 mV more negative on Au, with comparablevariance magnitudes. Although a single Gaussian curve fits theAu distribution well, the ITO E0′ distribution clearly deviatesfrom a unimodal fit. The well-defined surface morphology ofthe Au nanoplate likely leads to the unimodal character of theE0′ distribution. In contrast, Figure 4D reveals that while asingle Gaussian curve is sufficient to describe the Audistribution, a bimodal distribution appears to more accuratelyfit the ITO distribution. The parameters resulting fromunimodal and bimodal fits for the Au and ITO distributionsare reported in Table 2. It is noticeable that two Gaussians areevenly mixed to fit the ITO distribution with a ratio of 56:44whereas one of the Gaussians dominates for the Audistribution. The goodness of fits for unimodal vs bimodalGaussian curves were compared using a χ2 test for the Au andITO distributions. The results from the χ2 test clearly indicate

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b00313Nano Lett. XXXX, XXX, XXX−XXX

E

that a bimodal distribution fits the ITO E0′ distributionsignificantly better, whereas the Au E0′ distribution fits betterto the unimodal Gaussian fit (Table S2). In addition, one ofthe mean formal potentials for the ITO distribution (μ =−0.391 vs Ag/AgCl) becomes far more positive than the meanformal potential of the dominating Au distribution by 6 mV.

On the other hand, the other mean formal potential for theITO distribution (μ = −0.396 vs Ag/AgCl) is very close to theAu formal potential. This bimodal behavior for the ITOdistribution is attributed to the polycrystalline nature of ITO,as observed in the AFM and SEM images in Figure 1C andFigure S3. Thus, we have demonstrated the power of EC-TERS imaging by detecting a small (∼4 mV) but statisticallysignificant difference in E0′ on Au and ITO.The bimodal distribution in the ITO distribution reveals

that the polycrystalline ITO surface affects the equilibriumnature of the NB redox reaction. In Figure 5, a direct

comparison between the friction image and the E0′ map isshown to further investigate the origin of bimodal behavior inthe ITO E0′ distribution. The friction image was acquiredsimultaneously with the topography image during the AFMimaging (Figure 5A). The friction image depicts the lateralforce experienced by the tip and contains contributions fromthe topography and the physical and chemical properties of thesurface (e.g., stiffness or surface charge).48 This image providesa better spatial contrast in the surface structure of ITO. Figure5B shows a comparison between the AFM friction image andthe E0′ map in which 13 ITO grains are assigned based on theAFM friction image. This comparison reveals a strongcorrelation between specific grains which appear dark in thefriction image and regions with more negative E0′. We proposethat the negative shift in E0′ on these grains and the largerforce evident in the friction image are due to a more negativelocal surface charge on these particular grains.49−51 A morenegative surface charge would result in less stable binding ofthe neutral reduced form of NB compared to the cationicoxidized form, favoring a more negative E0′. Further, a more

Figure 4. Formal potential (E0′) maps with (A) 20 nm × 20 nm and(B) 40 nm × 40 nm pixel sizes. (C and D) Histograms of the E0′distribution from Au (purple) and ITO (cyan) pixels from (A) with(C) unimodal and (D) bimodal Gaussian fits. The histograms wereconstructed from the E0′ map with 20 nm × 20 nm pixels. Theunimodal Gaussian fits are shown as solid curves in (C). Two bimodalGaussian curves are shown as dashed curves and the sums of the twocurves are plotted as solid curves in (D). (Gray and black curves: Au.Red and orange curves: ITO. Inset in (D) is a zoomed-in plot around−0.41 V vs Ag/AgCl to show a Gaussian curve that is not dominant inthe Au distribution.)

Table 1. Fit Parameters from the Unimodal Gaussian Fit ofE0′ Distribution

Mean potential, μ (V vs Ag/AgCl) Standard deviation, σ (mV)

Au −0.397 4.95ITO −0.393 4.51

Table 2. Fit Parameters from the Bimodal Gaussian Fit ofE0′ Distribution

Mean potential,μ (V vs Ag/AgCl)

Standarddeviation,σ (mV)

Gaussianmixing

proportion

Au (Gaussian 1) −0.397 4.53 0.97Au (Gaussian 2) −0.408 6.34 0.03ITO (Gaussian 1) −0.391 4.13 0.56ITO (Gaussian 2) −0.396 3.02 0.44

Figure 5. (A) Friction image of the same surface area shown in Figure2A. (B) The E0′ map presented in Figure 4A is overlaid with thefriction image with white crosses indicating the location of 13 ITOgrains. The transparency of the two images was adjusted for a bettercomparison. (C) AFM friction line scan and (D) E0′ line scan alongthe white dashed line in (A). The locations of three ITO grainsmarked in (B) (grains 1, 2, and 3) are indicated as correspondingnumbers in (C).

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b00313Nano Lett. XXXX, XXX, XXX−XXX

F

negative surface charge would result in more repulsion of thetip, which is negatively charged under the negative appliedpotential, and therefore greater lateral force.The difference in E0′ of an adsorbed species occurs due to

the differential binding of the two redox forms.52,53 The surfacecharge is one of the dominant contributing factors to a changein differential binding free energy and thereby a change in E0′.A shift in E0′ of 5−10 mV has been previously attributed tovariations in surface charge.53 The observed shift in E0′ in ourwork is within this range. Thus, variation in local surfacecharge can account for some of the observed bimodal behaviorin the distribution of E0′ on ITO. A line scan containing threeITO grains in the AFM friction image (dashed line in Figure5A) was selected (Figure 5C) and directly compared to thecorresponding E0′ line scan (Figure 5D) to determine thespatial resolution of the E0′ map. This comparison clearlyshows the E0′ value on each assigned grain is more negativethan the neighboring pixels. We therefore determine the spatialresolution of the E0′ line scan to be a peak-to-peak distance(∼40 nm) that is enough to resolve individual surface featuresand their corresponding E0′ values on ITO. Unlike thenanoplate edge, there is no steep topographic drop on ITO;thus, the ITO grains are detectable without 3D effects of thetip.44 For a single crystal surface like the Au nanoplates, STM-TERS is likely a better method for distinguishing the chemistryon the terrace versus, for example, the step edge. However, ourEC-AFM-TERS system is better suited for resolving differenceson the rough ITO surface. To conclude, our EC-TERSmeasurements have spatially resolved distinct regimes ofelectrochemical behavior on a polycrystalline ITO surface atthe nanoscale and revealed the origin of bimodality in the E0′distribution of ITO.In summary, we have successfully acquired TERS intensity

maps representing the spatially dependent redox behavior ofNB on a Au nanoplate on an ITO electrode using EC-AFM-TERS at the nanoscale for the first time. TERS intensity mapsdemonstrated a spatial resolution of 81 nm for distinguishingthe border of the Au nanoplate and ITO electrode. A gap-mode enhancement with a factor of 2.3 difference between thegold nanoplate and the ITO was attributed to the origin of theobserved spatial contrast. The site dependent E0′ of NB wasobtained by fitting the TERS intensities to the Nernstequation, and a 4 mV voltage difference in E0′ was resolvedby statistical analysis of the E0′ distribution. The E0′distribution also measures the heterogeneity of the electrodesurface since a bimodal distribution was observed for ITOwhile Au showed unimodal behavior. The observed bimodalitywas directly correlated to the surface heterogeneity of ITOwith a spatial resolution of ∼40 nm. Our EC-AFM-TERSimaging experiment with such spatial and voltage resolutionshas therefore provided new insight into our knowledge of therole of local electrode structure. Further technical improve-ments to achieve better spatial resolution will have a significantimpact on the understanding of detailed chemical mechanismsat the molecular level, with profound implications forelectrocatalysis research.

■ ASSOCIATED CONTENT

*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.nano-lett.9b00313.

Details for sample preparation and procedure for surfaceCV measurements; details for surface excess calculation;procedure for determining the TERS line scanresolution; BEM simulation details; drift rate measure-ment of the instrument; TER spectra on Au nanoplateand ITO; SEM image of a synthesized Au nanoplate onan ITO substrate; surface CVs at different scan rates onITO and Au(111) single crystal; coverage of NB on ITOand Au(111) single crystal; TERS intensity change ineach TERS maps; raw TERS line scans; simulatedelectric field distribution at the tip−sample junction;Nernst fit of TERS intensity; peak current density plotfrom CVs; signal-to-noise-ratio table; χ2 test result table(PDF)

■ AUTHOR INFORMATIONCorresponding Authors*E-mail: g-schatz@northwestern.edu (G.C.S.).*E-mail: vanduyne@northwestern.edu (R.P.V.D.).ORCIDGyeongwon Kang: 0000-0002-8219-2717Michael S. Mattei: 0000-0002-8276-5562George C. Schatz: 0000-0001-5837-4740Richard P. Van Duyne: 0000-0001-8861-2228Present Address†Department of Chemistry, University of WisconsinMadison, Madison, Wisconsin 53706.NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThe authors acknowledge financial support from the Air ForceOffice of Scientific Research MURI (FA9550-14-1-0003) andthe National Science Foundation Center for ChemicalInnovation dedicated to Chemistry at the Space-Time Limit(CaSTL) (CHE-1414466). The authors thank Dr. VitorBrasilience and Dr. Charles Cherqui for helpful discussions.

■ REFERENCES(1) Murray, R. W. Chem. Rev. 2008, 108 (7), 2688−2720.(2) Lin, Y.; Lu, F.; Tu, Y.; Ren, Z. Nano Lett. 2004, 4 (2), 191−195.(3) Menon, V. P.; Martin, C. R. Anal. Chem. 1995, 67 (13), 1920−1928.(4) Zhan, D.; Velmurugan, J.; Mirkin, M. V. J. Am. Chem. Soc. 2009,131 (41), 14756−14760.(5) Robinson, R. S.; Sternitzke, K.; McDermott, M. T.; McCreery, R.L. J. Electrochem. Soc. 1991, 138 (8), 2412−2418.(6) Basame, S. B.; White, H. S. J. Phys. Chem. 1995, 99 (44), 16430−16435.(7) Markovic, N. M.; Grgur, B. N.; Ross, P. N. J. Phys. Chem. B 1997,101 (27), 5405−5413.(8) Guell, A. G.; Cuharuc, A. S.; Kim, Y.-R.; Zhang, G.; Tan, S.-y.;Ebejer, N.; Unwin, P. R. ACS Nano 2015, 9 (4), 3558−3571.(9) Bard, A. J.; Fan, F. R. F.; Kwak, J.; Lev, O. Anal. Chem. 1989, 61(2), 132−138.(10) Kwak, J.; Bard, A. J. Anal. Chem. 1989, 61 (11), 1221−1227.(11) Hansma, P. K.; Drake, B.; Marti, O.; Gould, S. A. C.; Prater, C.B. Science 1989, 243 (4891), 641−643.(12) Fan, F.-R. F.; Bard, A. J. Science 1995, 267 (5199), 871−874.(13) Fan, F.-R. F.; Kwak, J.; Bard, A. J. J. Am. Chem. Soc. 1996, 118(40), 9669−9675.(14) Morris, R. B.; Franta, D. J.; White, H. S. J. Phys. Chem. 1987, 91(13), 3559−3564.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b00313Nano Lett. XXXX, XXX, XXX−XXX

G

(15) Gray, N. J.; Unwin, P. R. Analyst 2000, 125 (5), 889−893.(16) Slevin, C. J.; Gray, N. J.; Macpherson, J. V.; Webb, M. A.;Unwin, P. R. Electrochem. Commun. 1999, 1 (7), 282−288.(17) Shevchuk, A. I.; Frolenkov, G. I.; Sanchez, D.; James, P. S.;Freedman, N.; Lab, M. J.; Jones, R.; Klenerman, D.; Korchev, Y. E.Angew. Chem. 2006, 118 (14), 2270−2274.(18) Laforge, F. O.; Velmurugan, J.; Wang, Y.; Mirkin, M. V. Anal.Chem. 2009, 81 (8), 3143−3150.(19) Comstock, D. J.; Elam, J. W.; Pellin, M. J.; Hersam, M. C. Anal.Chem. 2010, 82 (4), 1270−1276.(20) O’Connell, M. A.; Wain, A. J. Anal. Chem. 2014, 86 (24),12100−12107.(21) Kim, J.; Renault, C.; Nioradze, N.; Arroyo-Curras, N.; Leonard,K. C.; Bard, A. J. J. Am. Chem. Soc. 2016, 138 (27), 8560−8568.(22) Takahashi, Y.; Shevchuk, A. I.; Novak, P.; Babakinejad, B.;Macpherson, J.; Unwin, P. R.; Shiku, H.; Gorelik, J.; Klenerman, D.;Korchev, Y. E.; Matsue, T. Proc. Natl. Acad. Sci. U. S. A. 2012, 109(29), 11540.(23) Kurouski, D.; Mattei, M.; Van Duyne, R. P. Nano Lett. 2015, 15(12), 7956−7962.(24) Mattei, M.; Kang, G.; Goubert, G.; Chulhai, D. V.; Schatz, G.C.; Jensen, L.; Van Duyne, R. P. Nano Lett. 2017, 17, 590.(25) Zeng, Z.-C.; Huang, S.-C.; Wu, D.-Y.; Meng, L.-Y.; Li, M.-H.;Huang, T.-X.; Zhong, J.-H.; Wang, X.; Yang, Z.-L.; Ren, B. J. Am.Chem. Soc. 2015, 137 (37), 11928−11931.(26) Chen, X.; Goubert, G.; Jiang, S.; Van Duyne, R. P. J. Phys.Chem. C 2018, 122 (21), 11586−11590.(27) Martín Sabanes, N.; Ohto, T.; Andrienko, D.; Nagata, Y.;Domke, K. F. Angew. Chem., Int. Ed. 2017, 56 (33), 9796−9801.(28) Goubert, G.; Chen, X.; Jiang, S.; Van Duyne, R. P. J. Phys.Chem. Lett. 2018, 9 (14), 3825−3828.(29) Zhang, R.; Zhang, Y.; Dong, Z. C.; Jiang, S.; Zhang, C.; Chen,L. G.; Zhang, L.; Liao, Y.; Aizpurua, J.; Luo, Y.; Yang, J. L.; Hou, J. G.Nature 2013, 498, 82.(30) Chiang, N.; Chen, X.; Goubert, G.; Chulhai, D. V.; Chen, X.;Pozzi, E. A.; Jiang, N.; Hersam, M. C.; Seideman, T.; Jensen, L.; VanDuyne, R. P. Nano Lett. 2016, 16 (12), 7774−7778.(31) Liao, M.; Jiang, S.; Hu, C.; Zhang, R.; Kuang, Y.; Zhu, J.;Zhang, Y.; Dong, Z. Nano Lett. 2016, 16 (7), 4040−4046.(32) Stadler, J.; Schmid, T.; Zenobi, R. Nano Lett. 2010, 10 (11),4514−4520.(33) Chen, C.; Hayazawa, N.; Kawata, S. Nat. Commun. 2014, 5,3312.(34) Anderson, M. S. Appl. Phys. Lett. 2000, 76 (21), 3130−3132.(35) Hayazawa, N.; Inouye, Y.; Sekkat, Z.; Kawata, S. Opt. Commun.2000, 183 (1−4), 333−336.(36) Zaleski, S.; Wilson, A. J.; Mattei, M.; Chen, X.; Goubert, G.;Cardinal, M. F.; Willets, K. A.; Van Duyne, R. P. Acc. Chem. Res. 2016,49 (9), 2023−2030.(37) Shankar, S. S.; Rai, A.; Ankamwar, B.; Singh, A.; Ahmad, A.;Sastry, M. Nat. Mater. 2004, 3, 482.(38) Andoy, N. M.; Zhou, X.; Choudhary, E.; Shen, H.; Liu, G.;Chen, P. J. Am. Chem. Soc. 2013, 135 (5), 1845−1852.(39) Bhattarai, A.; Joly, A. G.; Hess, W. P.; El-Khoury, P. Z. NanoLett. 2017, 17 (11), 7131−7137.(40) Tranchida, D.; Piccarolo, S.; Deblieck, R. A. C. Meas. Sci.Technol. 2006, 17 (10), 2630.(41) Wang, Y.; Chen, X. Ultramicroscopy 2007, 107 (4−5), 293−298.(42) Sonntag, M. D.; Chulhai, D.; Seideman, T.; Jensen, L.; VanDuyne, R. P. J. Am. Chem. Soc. 2013, 135 (45), 17187−17192.(43) Park, K.-D.; Muller, E. A.; Kravtsov, V.; Sass, P. M.; Dreyer, J.;Atkin, J. M.; Raschke, M. B. Nano Lett. 2016, 16 (1), 479−487.(44) Wang, R.; Kurouski, D. J. Phys. Chem. C 2018, 122, 24334−24340.(45) Deckert-Gaudig, T.; Deckert, V. Small 2009, 5 (4), 432−436.(46) Olmon, R. L.; Slovick, B.; Johnson, T. W.; Shelton, D.; Oh, S.-H.; Boreman, G. D.; Raschke, M. B. Phys. Rev. B: Condens. MatterMater. Phys. 2012, 86 (23), 235147.

(47) Hohenester, U.; Trugler, A. Comput. Phys. Commun. 2012, 183(2), 370−381.(48) Grafstrom, S.; Neitzert, M.; Hagen, T.; Ackermann, J.;Neumann, R.; Probst, O.; Wortge, M. Nanotechnology 1993, 4 (3),143.(49) Frank, G.; Kostlin, H. Appl. Phys. A: Solids Surf. 1982, 27 (4),197−206.(50) Walsh, A.; Catlow, C. R. A. J. Mater. Chem. 2010, 20 (46),10438−10444.(51) Morales, E. H.; Diebold, U. Appl. Phys. Lett. 2009, 95 (25),253105.(52) Bard, A. J.; Faulkner, L. R.; Leddy, J.; Zoski, C. G.Electrochemical methods: fundamentals and applications; Wiley: NewYork, 1980; Vol. 2.(53) Petrovic, J.; Clark, R. A.; Yue, H.; Waldeck, D. H.; Bowden, E.F. Langmuir 2005, 21 (14), 6308−6316.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.9b00313Nano Lett. XXXX, XXX, XXX−XXX

H