Geometrical and chemical effects on the electrochemistry of single-wall carbon nanotube (SWCNT) network electrodes

Single-wall carbon nanotube (SWCNT) network is a promising electrode material for bio detection. Unfortunately, the associations between their physical as well as chemical properties and observed electrochemical performance are not known. This hinders any systematic optimization of the network properties towards specific analytes. Here we present a consistent physicochemical and electrochemical characterization of differently treated SWCNT networks. The results unambiguously show that ( i ) even if the electrochemical properties of different electrodes are practically identical when assessed by surface insensitive outer sphere redox (OSR) probes their behavior with inner sphere redox (ISR) probes can be drastically different. Further, ( ii ) the choice of the modification method (structural, chemical, electrochemical) heavily depends on nature of the target analyte, which are typically ISR probes. Although, ( iii ) chemical changes in the carbon phase appeared to be minor, effects of different treatments on oxidation states of Fe appeared to have a strong effect on the electrochemical performance of the networks in the case of ISR probes.


Introduction
Since carbon nanotubes (CNTs) were discovered in 1991 [1], this class of nanomaterials has been extensively investigated for various applications [2] owing to their unique material properties [3][4][5].Different kind of configurations containing multi-walled (MWCNT) or single-walled carbon nanotubes (SWCNTs) have been realized, including also SWCNT networks [6].The latter materials are especially attractive from the electroanalytical point of view as it is possibly to heavily modify them both (i) structurally and (ii) chemically with different treatments.The feasible material properties of CNTs has led to their wide use particularly in the field of electroanalytical applications, where they have been extensively used in the detection of different (bio) molecules [7][8][9][10].However, despite the large number of applied research, the published electroanalytical studies frequently lack consistent characterization of physicochemical properties of CNTs [11][12][13].As CNTs performance heavily depend on the synthesis parameters as well as the possible purification/modification steps often used to remove residual metal catalyst particles and amorphous carbonaceous impurities [5,14], without in depth structural knowledge about them, discovering any structure-property associations for these materials becomes extremely difficult.As electrochemical performance of a carbon electrodes is affected by numerous factors including microstructure, surface cleanliness and surface chemistry [15], it can be expected that CNTs fabricated under different conditions can show a wide spectrum of electrochemical properties.This is especially true in detection of inner-sphere (ISR) analytes, which are sensitive to the reactive sites at the electrode surface such as surface oxides [16][17][18] and defects [19].These issues have been recently discussed in detail in a comprehensive investigation of structure property relationships in carbon electrochemistry [20].
Here, we use our previously extensively characterized system, namely SWCNT networks [20][21][22][23].We investigate the effects of (i) structural and (ii) chemical modifications on the electrochemical properties of these materials.To systematically assess the structure-property associations, we start by providing in depth analyses of the physical and chemical properties of the SWCNT networks and how these are affected by the different treatments.Then we proceed to investigate the electrochemical properties of the materials and especially how the different treatments have affected them.To have a consistent step by step approach to this complex problem, we start with Ru(NH 3 ) 6 2+/3+ and IrCl 6 3-/2− redox probes known to be true outer sphere redox (OSR) probes insensitive to any surface chemical features of the electrodes.In the next step, we use Fe 2+/3+ redox species known to be sensitive to oxygen functionalities present on the surface.Finally, we proceed to the most complex problem of true inner sphere redox probes and use dopamine (DA) as our probe molecule to assess the full complexity of chemical effects on the electrochemical kinetics.By using the above stated systematic approach we can show that (i) even rather drastic geometrical changes of the networks did not induce significant effects on the electrochemical properties.(ii) Electrochemical oxidation slightly increased the number of oxygen functionalities in the network and especially affected the oxidation state of Fe present in the SWCNT networks somewhat favoring the formation of more oxygen rich oxides.(iii) Nitric acid treatment on the other hand did not favor similar formation of oxygen rich Fe oxides, as its X-ray absorption spectroscopy (XAS) spectra resembled closely that of the pristine SWCNTs.Instead, some cleaning effect and subsequent increase in the electrochemically active surface area (ECAS) was observed in comparison to pristine SWCNTs based on the electrochemical data.(iv) These two different effects on type of Fe-oxides were highlighted in the opposite trends seen with two different ISR redox probes, namely Fe 2+/3+ and DA.In the case of the former probe, the electrochemically treated networks exhibited fastest kinetics whereas in the case of DA the HNO 3 treated, and the pristine ones were clearly faster.These results show that (i) even if the electrochemical properties of different electrodes are practically identical when assessed by OSR probes their behavior with ISR probes may be drastically different.Further, (ii) the choice of modification method heavily depends on the target analyte (ISR probe).Finally, (iii) even though chemical changes in the carbon phase after the various treatments were very minor, the changes in the oxidation state of iron nanoparticles appeared to have a strong effect on the electrochemical properties of the SWCNT networks.

Experimental
The SWCNT network fabrication, transmission electron microscopy (TEM) tomography, scanning electron microscopy (SEM), conductive atomic force microscopy (PeakForce Tunneling AFM, PF-TUNA) and Raman spectroscopy are described in more detail in [22] and thus, only brief description is provided here.

SWCNT network fabrication
Single-walled carbon nanotube networks were fabricated by Canatu Oy using a high-temperature floating catalyst chemical vapor deposition in laminar flow reactor.Growth was carried out in carbon monoxide atmosphere.Fe particles decomposed from a ferrocene precursor were used as a catalyst for the growth of SWCNTs.More detailed information about the fabrication process can be found from Kaskela et al. [6], Moisala et al. [24].
In this study, two different types of SWCNT networks were investigated.The networks were synthesized at different collection rates of 100% and 500%, respectively, by adjusting the amount of ferrocene in the gas phase.The amount of iron was regulated by changing the temperature of the ferrocene cartridge, which influenced the vapor pressure of ferrocene.Collection time was adjusted accordingly, based on in-situ optical rate absorbance monitoring, to obtain networks with the same thickness.The sample obtained with a slower collection rate was identified as Low-Rate SWCNT, while the one obtained at a rate of 500% was identified as High-Rate SWCNT.
For tomography, maximum entropy method (MEM) [25,26] was used for 3D reconstruction of aligned.alistacks.Before running tomography, aligned.alistacks were 2 x binned and ROI cropped with scripts initiated in the Aalto Nanomicroscopy Center laboratory.MEM-3D-reconstruction volumes produced were visualized and scrutinized, in stereo mode, using UCSF CHIMERA application.Programs were all run in Mac Pro, Cylinder model, Twelve Core (64 GB).

Scanning electron microscopy
Hitachi S-4700 SEM at 15 kV accelerating voltage was used for imaging self-standing SWCNT networks.Open-to-close ratios were computed from SEM images by gray-scale thresholding to binary color in Image J.

X-ray absorption spectroscopy
Soft XAS measurements were performed on the 33-pole wiggler beamline 10-1 at Stanford Synchrotron Radiation Lightsource (SSRL) using 55 • incidence angle (magic angle) of X-rays.A resolution of ~200 meV was achieved with a spherical grating monochromator which was operated using 40 × 40 μm slits.X-ray beam spot size was approximately 1 × 1 mm2 with total flux in order of 10^11 photons/sec.The X-ray energy for iron 2p edge was measured from 695 eV to 735 eV, whereas for carbon and oxygen 1 s edge were from 260 eV to 340 eV and 520 eV to 560 eV, respectively.A Keithley picoammeter was used for amplifying the drain current to collect all the data in total electron yield (TEY) mode, where the incoming flux was measured using a nickel grid coated with Au sputtered film.Here, a reference sample was used for energy calibration of the data prior to the data analysis.The C 1s spectra were confirmed to match their energy calibration by observing the core- exciton signature at 291.65 eV [26][27][28] after the reference sample energy alignment.The presented C 1s and Fe 2p spectra are normalized, where the Fe 2p spectra were energy aligned to match metallic iron signature at 707.36 eV [27].Beamline energy calibration is based on matching the Ni L3 in 2nd order at 426.35 eV prior to measurement for low energy samples (up to O1s) and Ni L3 at 852.7 eV for high energy samples (above O1s).The presented C 1s, Fe 2p and O 1s spectra in Fig. 2 shows the average value from three different locations.Furthermore, all the data treatment (background subtraction and energy correction) was done using IGOR Pro v. 8.02 software.

Raman spectroscopy
Raman spectroscopy was performed by a Horiba Jobin-Yvon Labram HR confocal Raman system with an 488 nm argon laser with 10 mW power on sample.Spot size of 1 μm was used with an Olympus 100× objective.Spectra were acquired in the range of 50 to 3000 cm − 1 with a 600 lines/inch diffraction grating, exposure time of 15 s, and accumulation averaging count of two.Spectroscopic calibration was performed on intrinsic Si wafer (Ultrasil).Spectra were fitted by one Lorentzian peak for D-band, and two Lorentzian peaks for the G-band (G + and G − ), to obtain the I D /I G peak intensity ratios, as explained in literature [28].

Electrochemical measurements
Cyclic voltammetry experiments were carried out with Gamry Reference 600+ potentiostat in a three-electrode cell.Pt wire (Goodfellow) was used as a counter electrode, Ag/AgCl (+0.199V vs. SHE, Radiometer analytical) as a reference electrode and SWCNT networks investigated here as a working electrode.Detailed description of SWCNT preparation can from Leppänen et al. [21].Briefly, SWCNT electrodes prepared by press-transferring a smaller piece of network onto a glass substrate (Thermo Scientific, ISO 8037/I) and densified with a drop of 99.5 wt-% ethanol (Altia).Electrical contact was painted with conductive silver paint (Electrolube).The working area of the SWCNT network was defined by covering the surface with inert PTFE-tape (Saint-Gobain Performance Plastics CHR 2255-2) with a 3 mm hole.Electrochemical measurements were conducted for three different types of Low-and High-Rate SWCNT networks: (i) pristine (as-prepared), (ii) HNO 3 treated and (iii) electrochemically (EC) treated equaling into total of six (6) different sample types.15 min HNO 3 treatment was carried out by drop-casting 10 M HNO 3 on top of the prepared SWCNT electrode, whereas electrochemical oxidation was conducted in 10 mM phosphate-buffered saline (PBS, pH 7.4) by holding the electrode at 1.3 V (vs.Ag/AgCl) for 5 min.

Table 3
The peak-to-peak separation, onset potential and peak current ration obtained in Fe 2+/3+ with scan rates of 100 and 1000 mV/s (N = 3).

Table 4
The results obtained from the cyclic voltammetry measurements in 100 µM dopamine in PBS with scan rates of 50 mV/s and 5000 mV/s (N = 3).

Physical and chemical characterization
Two SWCNT networks of the same optical transmittance were collected at widely different rates by varying the ferrocene cartridge temperature as discussed in more detail in [22].The I D /I G ratios as fitted from the Raman spectra given in Fig. 1A, showed a small increase in the number of defects when moving from the low-rate to the high-rate SWCNTs.However, these values were still very low, indicating highly defect free tubes in both cases [28].Further, the two samples exhibited similar sheet resistance values as well as comparable PF-TUNA conductivities [22] indicating that the electrical properties of the two types of structurally different networks were alike.As a note, the sheet resistance technique measures the resistance between opposite sides of a square of a material, whereas the PF-TUNA mainly probes the z-axis conductivity through the network.From the electrochemical application point of view, the latter property is much more important.
Low magnification SEM images showing the overall network structure are exhibited in Fig. 1B and C. Open-to-close -ratios of the CNT networks were calculated by grayscale to binary conversion from SEM images to assess the network density.The obtained values are included in Table 1.
The three-dimensional reconstructions for both pristine SWCNT networks based on the TEM tomography technique are presented in Fig. 1D and E. From the figures it is clearly seen that in both networks the SWCNTs exist mainly in bundles and not as an individual tubes.Further observations include the following: (i) high-rate SWCNT sample network appears overall less dense and (ii) the average size of the bundles seems to be significantly larger in the case of high-rate sample than in the low-rate one.Quantitative analyses from the XPS revealed that the content of Fe was identical in both networks [22].
Analysis of the surface chemistry of the geometrically different samples was carried out by measuring C 1s, Fe 2p and O 1s spectra with XAS (Fig. 2).No marked differences were observed between the pristine SWCNT samples.However, based on the XAS Fe 2p spectra a small but clear difference was observed in the composition of iron between the two types of networks.With pristine Low-Rate network the intensity of peak ~709 eV is higher in comparison to High-Rate which indicated that the former contains slightly more oxidized phases of iron, presumably, Fe 3 O 4 and Fe 2 O 4 [29][30][31].These oxides have considerably higher intensity than the metallic iron at 707.36 eV [27,32].However, the relative intensities of the two peaks around 720-723 eV indicate that there could also be some contribution from lower iron oxides (FeO) in this case [31,33].As metallic iron may be masked by oxidized iron, it is very likely that the signal is combination of all these phases.
Next the surface chemistry of electrochemically and nitric acid treated networks was assessed.The following main results were obtained.Based on the C1s and O1s spectra neither the electrochemical nor nitric acid treatment induced marked changes in the surface chemistry of carbon phase.Due to the high reactivity of SWCNT networks towards atmospheric impurities such as oxygen and hydrogen, the age of the sample can have a slight impact on the observed surface chemistry.This phenomenon is also evident in the subtle variations observed between Low-Rate and High-Rate SWCNT networks.Therefore, it is important to acknowledge that minor changes in the surface chemistries of carbonaceous samples should not be given significant consideration unless the sample ages are identical.
On the Fe 2p spectra, small but still clear differences were seen in the chemical states of Fe after the two treatments.Intensity of the iron oxide peak enhanced significantly when the SWCNTs were electrochemically pretreated.Based on the extensive literature survey [27,[29][30][31]34,35], Fe 2p spectrum of the EC treated is similar to Fe 2 O 3 with separate and well-defined peak at lower energy.We suggest that Fe 2p spectrum for the EC treated arise from phases of metallic Fe, Fe 3 O 4 and Fe 2 O 3 .Additionally, it is very likely that some carbon is dissolved into the iron as well [32,36].In the double peak region (around 722 eV) one can now observe that the intensities of the two peaks are different when compared to pristine sample, which indicates less contribution from oxygen poor oxides relative to the oxygen rich oxides [31,33].The HNO 3 treated sample resembles chemically the pristine samples more than the electrochemically treated ones.Especially, within the double peak region (722 eV) the ordering of the intensities is similar to pristine SWCNT.Thus, based on the XAS the main differences between the different SWCNT networks are in the nature of the iron oxidesmore oxygen rich Fe-oxides are present in electrochemically treated SWCNTs.This highlights that the main effects of the treatments are not found in the carbon phase, but instead in the oxidation states of Fe.Finally, we note that in all spectra (pristine, electrochemically and HNO 3 treated) the sp 2 π* peak at 285.2 eV and the exciton peak at 291.65 eV stays well defined providing evidence that CNT long range ordering is well preserved in all cases.

Electrochemical characterization
Effects of structural differences and chemical treatments on the charge transfer properties were characterized by using Ru(NH 3 ) 6 2+/3+ and IrCl 6 3-/2-species.Cyclic voltammograms are shown in Fig. 3.These two species are known true outer sphere redox probes that can be utilized to probe the electronic properties (surface density of states (DOS)) of the electrode materials in the absence of additional chemical complications.Note also that the two probes have very different charges and formal potentials enabling one to assess possible electrostatic effects as well.No changes in the peak positions that would indicate significant differences in surface DOS were seen as expected based on the practically identical electrical properties (Table 2).In addition, the geometrical features of the networks did not induce any notable thin liquid layer effects within the scan range region investigated except the small drop in the peak-to-peak separation, ΔE p, values with the electrochemically treated networks in the scan rate range of 100-400 mV/s with IrCl 6 3-/2− probe (see Fig. 4).The absence of notable thin liquid layer formation is most likely caused by the mismatch in the diffusion layer thickness and the network dimensions [37].
When the oxidation peak currents for the two reactions are compared, it is clear that Ru(NH 3 ) 6 2+/3+ shows in all cases higher currents than IrCl 6 3-/2− despite the equal concentrations and number of electrons (n = 1 in both cases) transferred.This can be rationalized by considering how different the electrode surfaces are around the formal potentials of the two probes.(i) Network will be negatively charged around the formal potential of the Ru(NH 3 ) 6 2+/3+ reaction whereas the opposite is true for IrCl 6 3-/2− reaction as the point of zero total charge (pztc) is between the two formal potentials [20].In addition, the pztc of the SWCNT network is closer to the formal potential of Ru(NH 3 ) 6 than that of IrCl 6 3-/2− [20] resulting into a higher charge density on the surface in the latter case.This leads to (ii) the presence of less ordered water at a potential range for the formal potential of the Ru(NH 3 ) 6 reaction and a relatively more compact water layer for the potential range corresponding to the IrCl 3-/2− reaction.Further, since the carbon phase of the SWCNT network is positively charged around the formal potential of IrCl 6 3-/2− , (ii) stronger adsorption of ions (anions) can shift the outer Helmholtz plane further into the solution at these adsorption sites.This will further lead to (iii) lowering of the analyte concentration within the effective electron transfer distance from the surface, as some of the IrCl 6 3-/2− probes will stay further away from the electrode surface, therefore having weaker electronic coupling and thus lower probability for electron transfer [38].The overall result from (i) -(iii) would be (iv) a decrease in the measured peak current for the IrCl 6 3-/2-reaction in comparison to the Ru(NH 3 ) 6 2+/3+ reaction, as observed.
Interestingly, the electrochemical oxidation of the SWCNT network seems to further increase the difference between the currents measured for the two probes.In fact, the IrCl 6 3-/2-reaction on the electrochemically treated electrodes is the most reversible, but still shows the lowest E. Leppänen et al. currents, whereas the pristine and HNO 3 treated electrodes appear to behave practically identically.These differences can be caused by the fact that there is higher number of oxygen rich Fe-oxide species on the surface of the electrodes after the electrochemical treatment in comparison to other SWCNT networks, as this is known to affect the location of the pztc [43].However, further investigations are needed to clarify this issue.Oxidation to reduction peak current ratios were observed to be slightly less than one with Ru(NH 3 ) 6 2+/3+ , which may indicate weak physisorption of the oxidized species.On the other hand, redox reactions of IrCl 6 3-/2− did not show any indication of adsorption interactions as the current ratio was very close to 1 with all the electrodes.Next, we moved on to a slightly more complex redox probe Fe 2+/3+ of which reaction kinetics is known to be sensitive to oxygen functionalities present on the electrode surface [16,18].Results are presented in Fig. 5 and Table 3.As expected, the EC-oxidation treatment improved the kinetics of the Fe 2+/3+ reaction, which is consistent with the spectroscopy results showing that there were more oxygen rich Fe-oxide species on its surface in comparison to the other types of samples.The HNO 3 treated and the pristine SWCNT networks were electrochemically practically the same (peak potentials were almost identical), but the HNO 3 treated network had a bit higher peak current.This suggests that the HNO 3 treatment somewhat increased the ECAS.Additionally, Raman spectroscopy data revealed an increase in the intensity of the G peak, providing further evidence to support the claim that HNO 3 treatments increase electrochemically active surface area (see supporting information, Figure S1).Low-rate networks showed in general faster kinetics than the high-rate ones, despite the exact nature of the treatment utilized.One reason for this difference is anticipated to be denser network structure and smaller bundle size (as shown in Fig. 1D and E) that should further increase the ECAS inside the SWCNT network.
Finally, we moved on to a true inner sphere redox system dopamine to investigate the surface chemical effects on the reaction kinetics.The results are shown in Fig. 6 and Table 4.At low scan rates the reaction rates appeared to be almost the same with HNO 3 treated and pristine ones being the fastest.However, as a function of increasing scan rate, the differences started to magnify, HNO 3 being the fastest, then the pristine network almost as fast and finally the electrochemically treated network clearly as the slowest.The difference between the two treatments (HNO 3 and electrochemical oxidation) can be rationalized in the following way.Based on the XAS results it appears that the Fe oxides are less oxygen rich in the case of HNO 3 treated and pristine SWCNT than in electrochemically treated network.It is known that when one moves to more oxygen rich metal oxides [44], the IP and thus pKa decrease and subsequently the acidity of the oxide surfaces increase.Therefore, one can conclude that the surface of the HNO 3 treated network (as well as pristine) should be a better hydrogen bond acceptor (less acidic) than the electrochemically treated one with its oxygen rich more acidic oxides [45].This could facilitate the proton transfer reaction associated with the DA oxidation reaction [46] on the surface of the HNO 3 treated networks with respect to other types of networks and thus increase the overall rate of reaction.The influence of the oxidation state of Fe is also supported by the results of Raman spectroscopy (see Figure S1 in the supplementary material), which indicated an increase in the number of defects in the EC-treated network, compared to both the HNO3-treated and pristine networks.However, even with the increase in defects, the EC network still exhibited the slowest overall reaction rate.
From the structural point of view, the low iron SWCNTs appeared to be slightly better than high iron ones in all cases.This is likely caused by (i) higher number of Fe (oxide) particles with smaller size in the case of low rate SWCNTs and (ii) slightly denser network leading into higher reactive site density.Peak potentials for the oxidation of low Fe were also consistently slightly less anodic than those of high Fe.Also, the peak currents were consistently somewhat higher.However, evidence of marked adsorption was not seen with any of the electrode types.

Conclusions
We show that (i) even rather drastic structural changes of the networks did not induce significant modifications of the electrochemical properties, except the slight differences in the case of ISR probe DA.(ii) Electrochemical oxidation somewhat affected the number of oxygen functionalities in the network, but especially it had an effect on the oxidation state of Fe present in the SWCNT networks favoring slightly formation of more oxygen rich oxides.(iii) Nitric acid treatment on the other hand did not favor similar formation of oxygen rich Fe oxides, as its XAS spectra resembled closely that of the pristine SWCNTs.Instead, some cleaning effect and subsequent increase in the electrochemically active surface area was observed in comparison to the pristine SWCNTs based on the electrochemical data.(iv) These two different effects were highlighted in the opposite trends seen with two different redox probes, namely Fe 2+/3+ and DA.In the case of the former probe, the electrochemically treated networks exhibited fastest kinetics whereas in the case of DA the HNO 3 treated ones were clearly the fastest closely followed by the pristine SWCNTs.The reason for this difference was anticipated to be the presence of less oxygen rich oxides on HNO 3 treated surface facilitating the proton transfer reaction associated with DA oxidation.These results clearly show that (i) even if the electrochemical properties of different electrodes are practically identical when assessed with the OSR probes, their behavior with the ISR probes may be drastically different.Further, (ii) the choice of modification method (structural, chemical, electrochemical) heavily depends on the target analyte (ISR probe).(iii) Although, chemical changes in the carbon phase very rather minor, the corresponding changes in the oxidation states of Fe appeared to have a strong effect of ISR redox probe kinetics, highlighting once more the importance of metallic catalyst particles in electrochemical properties of carbonaceous nanomaterials.These preliminary results open a new research avenue towards the application specific modification of the SWCNT networks for targeted analytes.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.A) Raman spectra normalized to G peak intensity and offset for clarity.SEM micrographs from non-densified self-standing films and the TEM 3D reconstructions illustrating the structure of pristine B,D) Low-Rate and C,E) High-Rate SWCNT networks.Reprinted from Leppänen et al. [22].

Fig. 2 .
Fig. 2. XAS C1s, O1s and Fe2p spectra of investigated SWCNT networks.C1s and Fe2p spectra are normalized, whereas O1s presents the absolute intensity.The inset in Fe2p spectra shows a magnification of the double peak at higher energy for HNO 3 and EC treated SWCNTs.

Fig. 6 .
Fig. 6.Cyclic voltammograms of the Low-and High-Rate networks in 100 µM dopamine in PBS with scan rates A) 50 mV/s, B) 1000 mV/s and B) 5000 mV/s.

Table 2
[40]lts of the cyclic voltammetry measurements in Ru(NH The standard heterogenous rate constant k 0 was calculated with Nicholson method[39], whereas the kinetic parameter Λ with method proposed by Matsuda and Ayabe[40].Electrochemically active surface area was obtained with Randles-Ševčík [41,42]e taken from[41,42].For all the tabulated values N = 3.