Investigation of charge interaction between fullerene derivatives and single-walled carbon nanotubes

The charge interaction and corresponding doping effect between single-walled carbon nanotubes (SWNTs) and various fullerene derivatives, namely, C 60 , phenyl-C 61 -butyric acid methyl ester (PC 61 BM), methano-indenefullerene


| INTRODUCTION
One-dimensional single-walled carbon nanotubes (SWNTs) show outstanding ballistic transport properties and unique geometry of an assembled network.In addition, SWNTs can possess different energy bands depending on their diameters and chirality, ranging from semiconducting to metallic. 1,27][8] There have been myriad methods reported for the p-or n-type chemical doping of SWNTs.5][46][47] The n-type dopants are generally reducing agents (= electron donors), namely, alkali metals (K- [48][49][50][51][52] or Na-based dopants 52 ), N 2 H 4 , 22,53,54 aniline, 11 polyaniline, 53 ethylene diamine, 11 NH 3 , 42 polymer PEI, 55 viologen, 56 3-(4-dimethylaminobenzylidenyl)-2-indolinone, 57,58 and decamethylcobaltocene. 59Because of the challenging nature of the doping, the n-type dopants are less common than the p-type dopants.][66][67][68] With the emergence of flexible and stretchable electronics, the number of reports on SWNT-based solar cells has increased dramatically in recent years. 1,4In one of the examples, SWNT films were used as both an anode and cathode, and the cathode SWNT film was soaked in phenyl-C 61butyric acid methyl ester (PC 61 BM). 66PC 61 BM played a crucial role in controlling the energy level of the SWNTs.However, the exact mechanism of charge interaction and the doping were not discussed.][75] Here, we selected representative fullerene derivatives, namely, C 60 , methano-indenefullerene (MIF), 1,4-bis(dimethylphenylsilylmethyl) [60]fullerene (SIMEF-1), dimethyl(orthoanisyl) silylmethyl(dimethylphenylsilylmethyl) [60]fullerene (SIMEF-2), 1 0 ,1 00 ,4 0 ,4 00 -tetrahydrodi [1,4]methanonaphthaleno [5,6]fullerene (ICBA), and PC 61 BM. [76][77][78] For the carbon nanotubes, aerosol-synthesized, free-standing SWNTs and horizontally aligned SWNTs (HA-SWNTs) were used for the analyses.For the charge interaction study, the Van der Pauw method of the four-probe measurement, Kelvin probe, photoelectron yield spectroscopy (PYS), Seebeck coefficient measurement, X-ray-induced photoelectron spectroscopy (XPS), Raman spectroscopy, visible-near infrared (Vis-NIR) absorbance measurements and FET characterizations were used.The results show that no one measurement can conclusively tell us about the charge interaction between the fullerene derivatives and SWNTs.There are many factors, such as intrinsically weak doping strength of the fullerenes on SWNTs and intermolecular interactions within the fullerenes, hindering the analysis.Only collective data from a variety of different measurements have to be used, which in this work demonstrated that MIF, SIMEF-1, SIMEF-2, and PC 61 BM induce p-doping on SWNTs, while ICBA and C 60 induce ndoping.Despite discrepancies in some measurements, there is overwhelming evidence that the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) levels of the fullerenes govern the doping characteristics.The higher the molecular orbital energy levels, the more likely they induce p-type doping.However, in spite of the n-doping effect by certain fullerenes, SWNTs in the atmosphere still remained slightly p-doped on account of the oxidation by oxygen and moisture in the air.This work uncovers the effect of fullerene derivatives on SWNT in terms of energetics and electronic properties.Although the doping effect is marginal, it is possible to tune the energetics of SWNTs using fullerenes, which will benefit greatly in designing highperformance SWNT-based electronics.

| RESULTS AND DISCUSSION
1][82] It is important to differentiate the metallic SWNTs (m-SWNTs) from the semiconducting SWNTs (s-SWNTs) as SWNTs can have mixtures of m-SWNTs and s-SWNTs. 83The Schottky barriers between m-SWNTs and s-SWNTs have to be considered when investigating the doping effect. 84,85heet resistance measurement is one of the methods used for assessing the electrical property change upon doping.Figure 2A shows relative sheet resistance changes of SWNT films after a 10 mg mL −1 solution of fullerene was spin-coated (Tables S1 and S2).Application of fullerene derivatives showcases small decreases in the relative sheet resistance in all cases except C 60 .Application of C 60 actually increased the relative sheet resistance, indicating ndoping.The relative decrease in the sheet resistance is the greatest for MIF, implying the strongest p-doping.We can infer that neither electron affinity nor crystallinity affected the relative sheet resistance change because there are no strongly electronegative atoms in the fullerene derivatives.In addition, MIF and C 60 , both of which are small fullerenes, showed a totally different trend, whereas SIMEF-1 and SIMEF-2, which show different packing properties, showed similar doping effects.
The energy levels of a pristine SWNT film and fullereneapplied SWNT films were measured using the Kelvin probe and PYS (Figure 2B).For semiconducting materials, the Kelvin probe and PYS measurement values represent Fermi levels and valence levels, respectively, whereas in the case of conductive materials, both the Kelvin probe and PYS are supposed to give the same value as the Fermi level lies within either the conduction band or the valence band.The pristine SWNTs show the Kelvin probe and PYS values of −5.0 eV and −5.05 eV, respectively.The difference of 0.05 eV is negligible, and those values can be interpreted as the same Fermi level.By considering the difference of the Kelvin probe and PYS values within 0.1 eV to be negligible, we can deduce that, except ICBA, the fullerene-applied SWNTs all manifest p-type conductors.Among the samples, the MIF-applied SWNTs show the strongest p-doping effect, which is in agreement with the sheet resistance measurement.The data of the ICBA-applied SWNTs are difficult to interpret as the PYS measurement value (ca.−4.93 eV) is much higher than the Kelvin probe measurement value (ca.−5.12 eV).
7][88] Figure 3 shows that the Seebeck slopes of the SWNTs are positive, manifesting a p-type characteristic due to the oxidation by water and oxygen.From the gradient change of the slopes, we can infer the type of doping effect and the degree of doping the fullerene induces.It is clear that, except ICBA and C 60 , the other fullerenes induced p-type doping.Again, MIF induced the greatest pdoping as evidenced by the slope gradient.ICBA and C 60 decreased the Seebeck coefficients, implying n-type doping.
Raman spectroscopy was used to identify the doping effect from the G and 2D bands of the fullerene-applied SWNT samples.Figures S1 and S2 show the G and 2D bands, respectively, of the fullerene-applied SWNTs.The F I G U R E 1 Band energy diagrams of the m-single-walled carbon nanotubes (SWNTs), s-SWNTs with a diameter of 1.5 nm, methanoindenefullerene, 1,4-bis(dimethylphenylsilylmethyl) [60]fullerene, dimethyl(orthoanisyl) silylmethyl(dimethylphenylsilylmethyl) [60]  fullerene, 1 0 ,1 00 ,4 0 ,4 00 -tetrahydrodi [1,4]methanonaphthaleno [5,6]  fullerene, PC 61 BM (PCBM), and C 60 F I G U R E 2 A, Relative sheet resistance changes of single-walled carbon nanotube (SWNT) films after application of the fullerene solutions.B, Kelvin probe and photoelectron yield spectroscopy measurement of fullerene-applied SWNT films initial measurement (using 10 mg mL −1 ) did not show clear trends among the samples.All the peaks were shifted to lower Raman shifts, ascribed to the overlap of the G and 2D bands from fullerenes.In fact, this explains the seemingly ntype doping of PC 61 BM on SWNT in our previous report, which is not accurate in hindsight. 66To minimize the influence of the fullerenes, we lowered the fullerene concentration by 10 times (1 mg mL −1 ).The G and 2D bands of low concentration of fullerenes-applied SWNTs displayed much clearer trends.From the quasilinearity plot ([Δω2D/ ΔωG] n hole = 0.75 ± 0.04) of the two fullerene concentrations, we can see that the low concentration and high concentration do not align, indicating that lowering the concentration does not weaken the doping effect (Figure S3). 89Therefore, a meaningful conclusion could only be drawn from the G and 2D bands of the low concentration of fullerenes-applied SWNTs.p-doping shifts the band to the right (higher Raman shift), and n-doping shifts the band to the left (lower Raman shift).Figure 4A shows that the G bands of the low concentration of the fullerenesapplied SWNTs are shifted in the order of MIF > SIMEF-1> > SIMEF-2 > PCBM> ICBA ≈ C 60 .The 2D bands are trickier to interpret as the difference is finer, and all the peaks are left-shifted because of the influence of the fullerene core.The 2D band shift trend is SIMEF-1 = MIF = SIMEF-2 > PCBM> ICBA ≈ C 60 (Figure 4B).Despite the slight discrepancy, the general trend is similar to those of the previous measurements.XPS measurement can provide binding energy information of carbon atoms in SWNTs.Brønsted acid p-doping of SWNTs has been reported to decrease the energy of the C1s because the Fermi level lies further from the valence band edge on p-doping. 10,90We ran XPS on the fullerene derivative-applied SWNT films (Figure 5A).However, all the peaks exhibited higher binding energy than that of the pristine SWNTs.This is because XPS was used to measure C1s of the fullerene derivatives. 91Figure 5B shows the XPS C1s peaks of only the fullerene derivatives.It is clear that the C1s peaks of the fullerene derivatives possess relatively high binding energy.We estimated the C1s peak positions of the fullerene-applied SWNTs by subtracting the corresponding fullerene binding energy.Table S3 shows the calibrated C1s peak positions of the fullerene-applied SWNTs.Although the degree of doping does not agree with the previous outcome because of the inaccuracy of the calibration, the binding energy difference between the calibrated binding energies from that of the pristine SWNTs clearly shows that ICBA and C 60 are the only two fullerenes inducing n-doping.
Vis-NIR absorbance spectroscopy of SWNT is a useful method for observing doping effects as p-type doping is accompanied by the suppression of the Van Hove transitions.We prepared samples with high and low concentrations of the and M 11 (m-SWNTs) transition peaks in either concentration (Figure S4).We ascribed this to the mild doping effect of the fullerene derivatives.
The results of the analyses carried out up to this point are summarized in Table 1.Despite some discrepancies among the analytical techniques, the general trend of the doping follows that the molecular orbital energy levels of the fullerene derivatives are a dominant factor in inducing the doping effect.
Horizontally aligned SWNTs between two gold electrodes on a 100 nm-thick insulator-coated p-type silicon substrate (HA-SWNT FETs) were used to investigate the doping effect further (Figure 6A).In this setup, the measurements were made in air by applying a gate voltage (V G ) from −10 V to +10 V and then from +10 V to −10 V while measuring the drain current (I DS ) with a fixed drain voltage (V DS ) of −1 V.The negative V G values mean that the SWNTs are positively charged by holes, whereas the positive V G values mean that the SWNTs are negatively charged by electrons.HA-SWNT FETs can provide data of mixed SWNTs (Figure 6B) and s-SWNTs only by selectively burning off m-SWNTs using high voltage for the latter (Figure 6C).Previously, researchers used SWNT FETs to investigate charge transfer between fullerene and SWNTs.][94][95][96][97][98][99][100] Park et al reported thermally deposited C 60 on SWNTs resulting in ambipolar SWNT FETs and a left shift of the threshold voltage, which became stronger with the increasing thickness of the deposited C 60 . 101Here, we demonstrate HA-SWNT FET analysis when various fullerene derivatives are solution-coated in air and the interaction with the mixed Note: Blue with "+" represents a p-doping trend, and red with "−" represents an n-doping trend.Abbreviations: ICBA, 1 0 ,1 00 ,4 0 ,4 00 -tetrahydrodi [1,4]methanonaphthaleno [5,6]fullerene; MIF, methano-indenefullerene; SIMEF-1, 1,4-bis(dimethylphenylsilylmethyl) [60]  fullerene; SIMEF-2, dimethyl(orthoanisyl) silylmethyl(dimethylphenylsilylmethyl) [60]fullerene; XPS, X-ray-induced photoelectron spectroscopy.
SWNTs and the s-SWNTs by a simple process of m-SWNT removal using high current, both of which have not been reported to date.Each threshold gate voltage V TH of HA-SWNT-FETs was evaluated by extrapolation of the linear region method using Equation (1) 102 : where V Ge is the gate voltage obtained by the linear extrapolation as shown in Figure S5.The HA-SWNT FET data show strong hysteresis arising from the trapped charge (Figure S6 and Table S4) attributed to the sp 3 defects on the SWNT walls, 103 which come from atmospheric H 2 O, [60][61][62] O 2 , 62,63 and hydroxyl groups on the dielectric surface of SWNTs. 104Thus, the obtained V TH values are different depending on the V G sweep directions, which include hole conduction for the increasing V G and electron conduction for the decreasing V G .Therefore, the doping effect was interpreted by looking at the V TH shifts before and after the fullerene application, with the consideration of both increasing and decreasing V G (Figure 6B).The integrated results point to the same conclusion: ICBA and C 60 induce n-doping, while the other fullerenes induce p-doping on SWNTs (Table 2).Purely semiconducting HA-SWNT FETs can give data with a high ON-OFF ratio, which translates to a more accurate doping effect observation.6][107] Figure S7 and Table S5 show that the FET transfer characteristics demonstrate unipolar transport properties.Contrary to those from the mixed SWNTs, right shifts were dominant for the semiconducting HA-SWNT FETs after the fullerene application (Figure S7).By observing the V TH changes, we can infer that all of the fullerene derivatives except C 60 induce the p-doping effect.
The charge carrier mobility of a single string of SWNT can be determined from FET transfer characteristics in the linear region using Equation (2). 108 where L is the length of the SWNT channel (10 μm); g m = ∂I DS ∂V GS is the transconductance at the threshold gate voltage; W is the width of the SWNT channel, which is approximately 10 μm from the scanning electron microscopy (SEM) image (Figure S8); and C i is the capacitance, which is Equation (3). 108 where Λ 0 is the average distance between each nanotube (ca.500 nm from Figure S8), C Q is the quantum capacitance value (4.0 10 −10 F m −1 ), 109 ε is the dielectric constant (1 F m −1 in air), R is the radius of an SWNT (~0.7 nm), and After calculation, C i ≈ 8.00 10 −6 F is found.Therefore, Equation ( 2) can be calculated as the following: The calculated mobilities of SWNTs before and after the fullerene application are shown in Figure S9.Surprisingly, there was no significant trend in the mobility among the different fullerene derivatives for both the mixed HA-SWNT FETs and semiconducting HA-SWNT FETs (Table S6).
Figure 7 illustrates the charge transfer mechanism in the SWNTs.SWNTs in air are inevitably p-doped by oxygen T A B L E 2 Doping trends of fullerene derivatives-applied single-walled carbon nanotubes (SWNTs) from the V TH data of HA-SWNT-FETs (mixed SWNTs and s-SWNTs) performed in this study Threshold voltage, V TH
F I G U R E 7 Illustration of charge transfer A, in single-walled carbon nanotubes (SWNTs) in air, B, in C 60 -applied SWNTs, C, in SWNTs on a perovskite film under illumination, and D, in C 60 -applied SWNTs on a perovskite layer under illumination and water (Figure 7A).With the presence of fullerenes, charge transfer can occur between the fullerenes and SWNTs.The direction and degree of the charge transfer depend on the HOMO and LUMO levels of the fullerene (Figure 1).For example, in the case of C 60 , electrons will transfer from C 60 to SWNTs as C 60 induces n-doping (Figure 7B).However, SWNTs still remain p-doped because fullerene-induced doping effects are found to be mild.When these fullerene-applied SWNTs are used as electrodes in devices, such as perovskite solar cells, excitons will be generated, and either electrons or holes will transfer to the SWNT electrodes (Figure 7C).Being a p-type conductor, SWNTs are more inclined to accept holes than electrons, functioning as an anode.From the energy diagram, which includes CH 3 NH 3 PbI 3 (MAPbI 3 ) and HC(NH 2 ) 2 PbI 3 (FAPbI 3 ), we can deduce what kind of charge transfer may occur (Figure S10).Fullerene derivatives with strong pdoping ability, such as MIF, would shift the Fermi level of SWNTs down.In addition, LUMO of MIF (−5.5 eV) aligns well with the valence level of the perovskite materials (−5.3 eV for MAPbI 3 and −5.6 eV for FAPbI 3 ), which implies that the application of MIFs will enhance the performance of SWNTs as an anode.On the other hand, the application of C 60 will make SWNTs cathodes by bringing the Fermi level closer to the HOMO of C 60 (−3.95 eV) and the conduction band of the perovskite layers (−3.95 eV for MAPbI 3 and −4.20 eV for FAPbI 3 ) (Figure 7D).

| CONCLUSION
We investigated the charge transfer between SWNTs and the fullerene derivatives using various analyses.Due to the complexity of the system, no single measurement could provide a conclusive outcome.Collective evidence had to be put together to derive a valid conclusion.It was found that the fullerene derivatives induce charge transfer depending on their HOMO and LUMO levels.Fullerene derivatives with high HOMO and LUMO levels induce stronger p-doping than those with low HOMO and LUMO molecular energy levels, while mild n-doping was observed in the case of ICBA, and C 60 .If we look at the p-doping trend, it is MIF > SIMEF-1 > SIMEF-2 > PC 61 BM < ICBA > C 60 .
From the trend of HOMO, it is MIF > SIMEF-1 = SIMEF-2 > ICBA = PC 61 BM > C 60 .This means that the HOMO trend follows the doping trend, which makes sense as the hole movement from fullerene to SWNT can be speculated to be hindered by the energetic barrier between the HOMO of fullerene and the energy level of the SWNTs.The similar HOMO levels between SIMEF-1 and SIMEF-2, and between ICBA and PC 61 BM, beg a different explanation.By looking at the trend of the LUMO level, we can hypothesize that the LUMO also influences the strength of the doping effect.The different doping effect between SIMEF-1 and SIMEF-2 can be thought to arise from the difference in the LUMO levels.The strong electron affinity of fullerene is thought to p-dope SWNT.However, having a high energy barrier between the LUMO of fullerene and the energy level of SWNTs can hinder the electron transfer, which leads to a reduced doping effect.The same is true for the difference between ICBA and PC 61 BM, in which the higher-lying LUMO of ICBA is suspected to weaken the p-doping.Therefore, we can say that the HOMO and LUMO levels of fullerene are thought to govern the doping behavior.Other factors, such as crystallinity, size, and electronegativity of fullerenes, did not play an important role in the charge interactions.This finding demonstrates that fullerenes can be used as mild dopants and energy level tuners for the carbon electrodes.It serves as a guideline for systems, in which carbon electrodes, such as SWNTs, and fullerenes are used together by providing an insight into the charge interaction and corresponding doping effect.

| SWNTs preparation for conductive films
SWNTs were synthesized by an aerosol (floating catalyst) CVD method based on ferrocene vapor decomposition in a CO atmosphere. 83The catalyst precursor was vaporized by passing room temperature CO through a cartridge filled with ferrocene powder.Ferrocene vapor was then introduced into the high-temperature zone of a ceramic tube reactor through a water-cooled probe and mixed with additional CO.To obtain stable growth of SWNTs, a controlled amount of CO 2 was mixed with the CO carbon source.SWNTs were directly collected at the downstream of the reactor by filtering the flow through a nitrocellulose membrane filter (Millipore Corp; HAWP, 0.45 μm pore diameter).These films were supplied on a large nitrocellulose membrane with 80% transparency (evaluated at 550 nm wavelength).SWNTs on a nitrocellulose membrane were drytransferred onto glass substrates, followed by ethanol application.Then, the samples were annealed at 100 C for 10 minutes and left in a vacuum for at least 12 hours.

| SWNTs growth for HA-SWNT-FETs
R-cut quartz was used for aligned growth.The quartz substrates were annealed at 900 C in air for 12 hours. 110An Fe catalyst (0.2 nm thick) was photolithographically patterned into parallel stripes using thermal evaporation, followed by a liftoff and heating at 500 C in air.SWNTs were grown at 800 C using ethanol as a feedstock gas.

| Electrode patterning for HA-SWNT-FETs
Electrodes (Ti/Pt, typically 2 nm/23 nm) were photolithographically patterned on as-delivered silicon substrates (SUMCO Corp., highly p-doped Si with 100 nm-thick oxide layer) by sputtering. 105A photoresist layer and thin metal layers were used to obtain electrodes with smooth edges.Here, p-doped silicon substrates functioned as a global back gate.

| SWNTs transfer for HA-SWNT-FETs
A 4 wt% of poly(methyl methacrylate) (PMMA) in anisole was spin-coated at 2000 rpm on quartz substrates, on which SWNTs were grown, followed by heating at 170 C in air for 15 minutes. 105After scratching the edges using a diamond knife, the substrates were immersed in aqueous KOH (1 mol L −1 ) and heated at 100 C for 10 minutes.The PMMA thin films were spontaneously peeled from the substrate when placed in cold distilled water, which were then picked up by the target substrates with the prepatterned electrodes.After drying in air at room temperature, the substrates were heated at 170 C for 30 minutes.Finally, the PMMA films were dissolved in acetone, followed by annealing at 350 C in a vacuum for 3 hours.

| m-SWNT removal by Ohmic heating on HA-SWNT-FETs
[107] 4.6 | Fullerene solution preparation and application C 60 , PC 61 BM, and ICBA were purchased from Frontier Carbon.MIF, SIMEF-1, and SIMEF-2 were synthesized according to our previous reports. 76,81The fullerenes were dissolved in chlorobenzene at a concentration of 10 mg mL −1 or 1 mg mL −1 for the low-concentration doping analysis.The fullerene derivatives were spin-coated at 3000 rpm for 30 seconds on SWNT films or HA-SWNT-FETs, followed by annealing in air at 100 C for 10 minutes.

| Measurements of electronics properties
Sheet resistance of SWNT films was measured by using a Van der Paw method, which involves using four indium contacts pressed on the corners of SWNT films.Four-probe measurements were taken using an Analyzer Agilent 4156C, and the sheet resistance was determined using the Van der Paw equation.Kelvin probe force microscopy (Riken Keiki FAC-2) and PYS (Riken Keiki AC3) were used for energy level measurement.Absorbance measurements were performed using Shimadzu UV-3150.XPS measurements were carried out using a ULVAC PHI 5000 VersaProbe machine in high vacuum.Seebeck measurements were taken by a home-made system where SWNT films on glass were positioned to connect two copper sheets with carbon paste.Resistive heaters were attached on the hot side copper sheet, and T-type thermocouple was attached between hot and cold side copper sheets to measure the temperature gradient ΔT.Thermoelectric voltage (ΔV) was measured between the two copper plates.The measurements of HA-SWNT-FETs were performed with three probes (for Source, Drain and Gate) using an Analyzer Agilent 4156C.SEM measurement was carried out on an S-4800 (Hitachi).

60 F I G U R E 3
Seebeck measurement of single-walled carbon nanotube films with fullerene derivatives F I G U R E 4 Magnified Raman spectroscopy; A, G and B, 2D bands of the fullerene derivatives (low concentration)-applied singlewalled carbon nanotube samples fullerene derivatives.Nonetheless, it was difficult to observe the suppression difference of the typical S 11 , S 22 (s-SWNTs) Illustration of the HA-single-walled carbon nanotube (SWNT) field-effect transistors (FETs), where S represents the source, D represents the drain, and G represents the gate.B, V TH of the HA-SWNT-FETs (mixed SWNTs) before (black line) and after (red line) the fullerene derivative application.C, V TH of the semiconducting HA-SWNT-FETs (only s-SWNTs) before (black line) and after (red line) the fullerene derivative application d is the thickness of dielectric gate (100 nm of SiO 2 layer).
Summary of the doping analyses of the fullerene derivatives-applied single-walled carbon nanotubes conducted by various X-ray-induced photoelectron spectroscopy measurement of C1s peak of A, single-walled carbon nanotube films coated with fullerene derivatives solutions and B) fullerene derivatives alone T A B L E 1