Polymer-Free Electronic-Grade Aligned Semiconducting Carbon

Jul 31, 2017 - We further quantitatively describe the effect of polymer removal on FET device performance using short channel devices to analyze the c...
1 downloads 11 Views 5MB Size
Download PDF
Research Article www.acsami.org

Polymer-Free Electronic-Grade Aligned Semiconducting Carbon Nanotube Array Yongho Joo,† Gerald J. Brady,† Catherine Kanimozhi,† Jaehyoung Ko, Matthew J. Shea, Michael T. Strand, Michael S. Arnold, and Padma Gopalan* Department of Materials Science and Engineering, University of Wisconsin-Madison, Madison, Wisconsin 53706, United States S Supporting Information *

ABSTRACT: Conjugated polymers are used commonly to selectively sort semiconducting carbon nanotubes (S-CNTs) from their metallic counterparts in organic solvents. The polymer-wrapped S-CNTs can be easily processed from organic solvents into arrays of CNTs for scalable device fabrication. Though the conjugated polymers are essential for sorting and device fabrication, it is highly desirable to remove them completely as they limit the electronic properties of the device. Here, we use a commercially available polymer, namely, poly[(9,9-dioctylfluorenyl-2,7-diyl)-alt-co-(6,6′-(2,2′-bipyridine))] (PFO-BPy), to sort large-diameter S-CNTs with ultrahigh selectivity and fabricate CNT-array-based field effect transistors (FETs) via a floating evaporative selfassembly (FESA) process. We report quantitative removal of the polymer wrapper from the FESA aligned S-CNT arrays using a metal-chelation-assisted polymer removal (McAPR) process. The implementation of this process on FESA films requires the selective thermal degradation of the polymer into oligomers, combined with optimization of the solvent type and temperature of the metal complexation reaction. Resulting S-CNT array FET devices show that the electronic properties of pristine CNT are preserved through this process. Optical microscopy, UV−vis spectroscopy, and X-ray photoelectron spectroscopy (XPS) were used to characterize the quantitative polymer removal. We quantitatively describe the FET devices to analyze the fundamental characteristics of FETs (mobility (μ), on-conductance (Gon), and contact resistance (2Rc)) by comparing before and after polymer removal. The ability to completely remove the polymer wrapper in aligned CNT arrays without adversely affecting the device properties opens up applications beyond FETs into photovoltaics and biosensing. KEYWORDS: single-walled carbon nanotubes, polymer removal, aligned CNT array, FET, metal complexation

■

INTRODUCTION In the past few years, single-walled carbon nanotubes (CNTs), which posssess outstanding electronic properties, have led to significant breakthroughs in field effect transistors (FETs)1−4 and in low-power and high-speed electronic semiconductor applications.5,6 These advances require separation of the desirable semiconducting (S-CNT) from the as-synthesized electronically heterogeneous metallic and semiconducting mixtures to practically use CNTs in electronic devices. This has been achieved through many different sorting methods that use either DNA,7 surfactants or conjugated polymers8−11 to selectively wrap or adsorb on to S-CNTs in aqueous or organic medium, which are then used in density gradient ultracentrifugation,12 column chromatography,13,14 or gel-based separation15 processes. Among them, the use of conjugated © 2017 American Chemical Society

polymers based on polyfluorenes has resulted in high purity (>99%) sorting of S-CNTs in organic solvents such as toluene in milligram scales.16,17 Though conjugated polymers are essential to sort S-CNTs and further process them into arrays, it is eventually desirable to remove them post processing. Any polymeric or organic residues on the S-CNTs may increase contact resistance of FETs thereby limiting conductance at sub100 nm channel lengths, pose as a barrier to exciton diffusion in photovoltaic cells, or hinder further functionalization for biosensing applications. Hence, in the past few years, a number of reports have emerged on the removal of these sorting Received: May 15, 2017 Accepted: July 31, 2017 Published: July 31, 2017 28859

DOI: 10.1021/acsami.7b06850 ACS Appl. Mater. Interfaces 2017, 9, 28859−28867

Research Article

ACS Applied Materials & Interfaces

of the polymer backbone (Figure 1). The complexation of Re(CO)5Cl to BPy in PFO-BPy (PFO-BPy:Re) at 65 °C in solution triggered up to 85% removal of the PFO-BPy from arcdischarge S-CNTs (diameter = 1.3−1.7 nm) and up to 72% from CoMoCAT S-CNTs (diameter = 0.7−0.8 nm).24 Here, we present a process to quantitatively remove the PFO-BPy from S-CNTs in solution using McAPR in a high boiling solvent. We explore the applicability of this method on aligned CNT arrays that are meaningful for devices. The aligned CNT arrays were created by a recently discovered method called dose-controlled floating evaporative self-assembly (FESA).25 The polymer is removed from arrays of S-CNTs by introducing a thermal annealing step to selectively degrade the polymer prior to implementing the optimized McAPR. These studies highlight the challenges in polymer removal in S-CNT monolayer-film compared to solution. Resulting S-CNT array FET devices show that the electronic properties of pristine CNT are preserved through this process. Our method of stripping PFO-BPy from S-CNT postsorting is mild enough to preserve the electronic structure of the CNTs, avoids the use of any acidic reagents, and we believe is generalizable to any polymer with BPy as a comonomer. Morphology of the aligned S-CNT arrays were characterized by SEM. Optical microscopy and X-ray photoelectron spectroscopy (XPS) were used to characterize the quantitative polymer removal. We further quantitatively describe the effect of polymer removal on FET device performance using short channel devices to analyze the characteristic conductance density (Gon) and contact resistance (2Rc) by comparing before and after polymer removal. The ability to create aligned array of bare S-CNTs is exciting not only for potentially improving the FET device performance but also to create sensors where efficient chemical or biological functionalization requires a pristine surface.

polymers based on either changing the torsional angle of the polymer backbone in response to an acid,18 degradation of the polymer backbone, by acids,19,20 reversible depolymerization of metal coordination polymer,21 and depolymerization of hydrogen-bond supramolecular polymer.22,23 Most of these appoaches are based on designing new conjugated polymers to incorporate both the sorting characteristics as well as degradability aspects, which are synthetically demanding. Recently, we introduced an alternate approach, in which a well-known commercially available polyflourene copolymer (PFO-BPy) that has shown outstanding sorting selectivity to S-CNTs can be removed post sorting by a simple solution treatment method. PFO-BPy sorted S-CNTs have already resulted in devices that outperform even the state-ofthe-art silicon FETs1 and therefore form a good model system for further advances. Our method involved subjecting the sorted PFO-BPy wrapped CNTs to metal complexation (Figure 1) in solution, leading to the unwrapping of PFO-BPy from the

■

RESULTS AND DISCUSSION In this study, we primarily focused on sorted S-CNTs using PFO-BPy from raw arc-discharge CNTs by duplicating protocols established in our prior work.1 S-CNTs in this diameter range (1.3−1.7 nm) provide a balance of high conductance and high on/off ratio necessary for highperformance FETs. The absorption spectrum showed complete absence of peaks in the 600−850 nm range where metallic CNTs appear and show a dominant S33 peak from the semiconducting tubes (Arc@PFO-BPy) (Figure 2a and Figure S1).24 To obtain bare S-CNTs, McAPR was conducted at two

Figure 1. Schematic representation of the envisioned mechanism for stripping of PFO-BPy from the S-CNT surface by complexation with pentacarbonylchlororhenium(I) (Re(CO)5Cl).

S-CNT surface.24 Through a combination of simulations and experiments the metal-chelation-assisted polymer removal (McAPR) was attributed to likely changes in the extent of noncovalent as well as electrostatic interactions between the polymer and CNT, and contributions from increased stiffness

Figure 2. (a) Absorption spectra of PFO-BPy (green), Arc@PFO-BPy (black), Re-treated (65 °C)+washed (red), and Re-treated (110 °C)+washed (blue). Inset shows that the S33 peaks from the CNTs remain unchanged by the treatment. (b) Schematic illustration of partial removal and complete removal of PFO-BPy with the picture of the isolated Arc@PFO-BPy CNT before and after Re-treatment in toluene. Upon metal complexation, the CNTs crash out of solution into aggregates, which is accompanied by a color change as the metal complexed polymer is left in solution. 28860

DOI: 10.1021/acsami.7b06850 ACS Appl. Mater. Interfaces 2017, 9, 28859−28867

Research Article

ACS Applied Materials & Interfaces

Figure 3. XPS spectra of (a) N(1s) peak from Arc@PFO-BPy, Re-treated (65 °C), and Re-treated (110 °C) in toluene (Bottom to top), (b) corresponding N(1s) peak after additional washing steps. (c) Comparison of N(1s)/C(1s) peak area ratio’s of Arc@PFO-BPy and Re-treated +washed tubes as a function of Re-treatment temperature. N(1s) peaks were normalized to the C(1s) peak at 285.5 eV in the inset of (c).

Figure 4. Raman spectra from the solution-processed CNTs drop casted on to silicon substrate, showing (a) G-band, and D-band of Arc@PFO-BPy (black), Re-treated (110 °C) (brown), Re-treated (65 °C)+washed (red), and Re-treated (110 °C)+washed (blue), and the corresponding (b) ID/IG ratio, and (c) 2D-band shifts with the schematic illustration of Re-treated Arc@PFO-BPy (d).

We further analyzed the extent of metal complexation based on XPS measurement and analysis of N(1s) from PFO-BPy.1 In the XPS N(1s) spectra of Re-treated Arc@PFO-BPy, three peaks at 398.8, 400.3, and 402.2 eV were observed. These peaks correspond to the N(1s) binding energy from PFO-BPy when uncomplexed (red), complexed with Re(CO)5Cl but bound to the CNT(blue), and complexed but not bound to the CNT (green), respectively (Figure 3a). After the postwash steps, the complexed N(1s) peaks (blue at 400.3 eV, and green at 402.2 eV) disappeared (Figure 3b), and accordingly the N(1s)/C(1s) ratio decreased to near zero within the XPS detection limits for the Arc@PFO-BPy treated under reflux conditions (Figure 3c). As the degree of complexation increases, we anticipate a larger change in the stiffness of the polymer chain that can result in

different reaction conditions to remove the PFO-BPy from Arc@PFO-BPy. Arc@PFO-BPy was dispersed in toluene with excess Re(CO)5Cl salt and reacted at 65 and 110 °C (reflux condition) to obtain the “Re-treated Arc@PFO-BPy” samples. In the absorption spectrum, the peak corresponding to the PFO-BPy at 355 nm disappears from the “Re-treated” CNTs when the reaction was conducted under reflux conditions. Quantification of the absorption peak shows that ∼85% of the PFO-BPy was removed at 65 °C, while complete (100%) removal was achieved at 110 °C (Figure 2a,b). This treatment results in precipitation of S-CNT, which are collected and then washed sequentially with toluene, chloroform, THF, and methanol to remove adsorbed metal complexed polymer and excess salt to obtain “post-wash” samples. 28861

DOI: 10.1021/acsami.7b06850 ACS Appl. Mater. Interfaces 2017, 9, 28859−28867

Research Article

ACS Applied Materials & Interfaces the polymer unwrapping from the CNTs. The changes in rigidity of the polymer chain upon complexation were inferred from MD simulations in our earlier publication.24 However, the low-temperatures at which the reaction was carried out in the earlier studies, still left ∼20% of polymers on the CNTs. At higher temperatures, the solubility of the salt and the complexed polymer increase in toluene, most likely leading to higher reaction efficiency. To evaluate if the elevated temperature and the Re-treatment leads to any irreversible defect formation in the Arc tubes, we used Raman spectroscopy (Figure 4) to characterize the Dband (1348 cm−1), G-band (1589 cm−1), and 2D-band (2672 cm−1) modes. The integrated intensity ratio of the D-band to the G-band, ID/IG, is related to the defect density and doping levels.26 The ID/IG ratio of Arc@PFO-BPy was 0.12(±0.03), which increased marginally to 0.15(±0.04) after Re-treatment, and decreased to 0.09(±0.03) in postwash samples deposited on silicon substrate (Figure 4b).24 While these observed differences in the ID/IG ratio are relatively small to elaborate, these results show that the Re-treatment at high-temperature does not irreversibly create additional defects in the Arc tubes. In general, the ID/IG ratio of the bare Arc27 tubes can be different from the polymer wrapped tubes28 due to noncovalent interactions, and therefore, removal of these macromolecules can account for a modest decrease in this value. The 2D band clearly shifts showing increased doping due to Re-treatment, which is in fact reversed in the postwash samples (Figure 4c). Though we achieved 100% polymer removal to release bare S-CNTs in solution, we could not further process these CNTs into aligned arrays due to bundling of the tubes.29,30 Next, we employed the McAPR on the as-fabricated FESA thin-film SCNT arrays (Figure 5a), with a packing density of 47 tubes μm−1 (Figure 5b).31 PFO-BPy plays a critical role to stabilize the S-CNTs in the ink, hence enabling the well-spaced and individualized S-CNTs in the FESA deposited arrays. Thus, we fabricated the S-CNT arrays first (schematic in Figure 5ai),16,25,31 and we subjected the arrays to McAPR (Retreatment at 110 °C and postwash). While these conditions were sufficient to completely remove the polymer from the CNTs in toluene solution, on the FESA films this method was ineffective (Figure 5c) if the arrays have not been thermally annealed before the treatment steps. The as-fabricated S-CNT array FETs were subjected to a 400 °C thermal annealing1 (Figure 5aii) in vacuum, to improve the efficiency of polymer removal. The removal of PFO-BPy was characterized by quantifying the N (1s) peak in the XPS. The complete disappearance of the N(1s) peak (Figure 5d and Table S1) and the Re (4f) peak (Figure S2) in the thermally annealed+Retreated+washed sample confirmed the effectiveness of this process. The complete survey scans of all the samples are provided in Figure S3. Though the nature of the C(1s) peak does not change significantly between these samples, the intensity of this peak (when normalized to Si(2p) peak) reduces as the polymer is removed. Earlier FTIR studies have shown that the thermal treatment eliminates the alkyl side chains of PFO-BPy in solid state. However, the exact chemical nature of the polymer backbone following this thermal treatment is unclear. To study the effect of thermal treatment on the polymer structure, a thin film of the neat polymer PFO-BPy was subjected to 400 °C annealing in vacuum. The absorption maximum (λmax) of thermal annealed polymer film at 351 nm was blue-shifted by ∼18 nm (Figure 6 b) compared to the parent polymer PFO-BPy

Figure 5. (a) Schematic illustration of polymer removal from the SCNT arrays for (i) as-fabricated (“assembled”), (ii) assembled +annealed (P1), (iii) assembled+annealed+Re-treated (110 °C)+washed (P2), and (iv) assembled+annealed+Re-treated+washed +annealed (P3). (b) SEM image of S-CNT arrays fabricated by FESA after full treatment (P3). XPS spectra of FESA film without thermal annealing (c), and with thermal annealing (d) showing the N(1s) peak from (blue) control Arc@PFO-BPy arrays, (black) Re-treated, (orange) washed CNT in film.

(λmax = 369 nm). Additionally, when the polymer thin film was exposed to 365 nm light, the fluorescence emission color changes from blue (PFO-BPy) to green-yellow (annealed) (Figure 6b inset picture) upon thermal treatment. Polyfluorene polymers are well-known for their blue light emitting properties32,33 however, undesired green-yellow band emissions are typically observed under photo-oxidative conditions and thermal annealing as a result of formation of excimers and aggregates.34 This yellow emission is attributed to the presence of either carbonyl (CO)35 or H-bonding36 groups in the polymer backbone. These films also become insoluble in solvents due to potential cross-linking during thermal annealing. The annealed PFO-BPy thin-film was reacted with Re(CO)5Cl, to further understand the polymer backbone modification. The absorption spectrum of Re-treated thin film exhibits a new peak at 402 nm (Figure 6b) with a red shift of ∼50 nm. This red-shifted absorption peak is attributed to the formation of Re complex with the BPy groups present along the polymer backbone, as stated in our previous publication (PFOBPy:Re).24 These results indicate that thermal annealing leaves the BPy groups intact, hence allowing the metal complexation. Though the exact structure of the polymer backbone subjected to 400 °C annealing is unclear, changes in the absorption spectrum and emission color upon thermal annealing suggests two possibilities: namely, generating a carbonyl (-CO) or hydroxyl groups (−OH) upon scission 28862

DOI: 10.1021/acsami.7b06850 ACS Appl. Mater. Interfaces 2017, 9, 28859−28867

Research Article

ACS Applied Materials & Interfaces

Figure 6. (a) Chemical process depicting the thermal annealing of PFO-BPy spin coated film on Si/SiO2 substrate and subsequent Re treatment. Thermal annealing removes the side groups on the fluorene unit; however, the nature of the new substitution is unknown, which is represented by X. (b) UV−vis absorption spectrum of PFO-BPy (black), annealed (red), and annealed+Re treated (blue) film. (c) FTIR spectrum of PFO-BPy (black) and annealed (red) film, (d) Comparison of UV−vis absorption spectrum of PFO-BPy thin film (black) and thermal annealed spin coated thin film of PFO-BPy wrapped CoMoCAT (6,5) CNT (CNT@PFO-BPy+annealed red curve) on quartz substrate showing a blue shift in the λmax, and (e) GPC data comparing the molecular weight and dispersity (Đ) of the parent PFO-BPy polymer from American Dye Source (black) with the oligomers that was extracted from the FESA films of Arc@ PFO-BPy, which clearly shows degradation of the polymer backbone.

Figure 7. (a) Representative high-resolution SEM of aligned S-CNT array in short channel device (Lch = 140 nm) and corresponding schematic illustration, where the CNTs directly span through the source and drain electrode (b). (c) Transfer characteristic of a short channel S-CNT array FET (Lch = 140 nm) at VDS = −0.1 V. (d) Resistance extracted at VDS = −0.1 V and VGS − VT = −5 V at several channel lengths for assembled +annealed (P1) (Blue), assembled+annealed+Re-treated+washed (P2) (Orange), and (iii) assembled+annealed+Re-treated+washed+annealed (P3) (Green). Each data point represents the mean and deviation of 5 device measurements. A linear fit of resistance versus channel length was performed to extrapolate contact resistance.

leaving the conjugated backbone intact as ν (CC) 1430 cm−1 is preserved (Figure 6c); (2) there is no evidence of carbonyl group in the 1670−1750 cm−1 range; (3) the emergence of additional strong peak at 1093 cm−1 (Figure S4) as compared

of the alkyl side groups on the fluorene units. By careful examination of the FTIR spectrum (Figure 6c) of PFO-BPy thin film before and after thermal annealing, the following can be concluded: (1) the alkyl side groups are removed, while 28863

DOI: 10.1021/acsami.7b06850 ACS Appl. Mater. Interfaces 2017, 9, 28859−28867

Research Article

ACS Applied Materials & Interfaces

300−350 nm (Figure S6) are P1 devices: 108 ± 22 cm2/(V*s), P2: 90 ± 35 cm2/(V*s), P3: 99 ± 11 cm2/(V*s). The mobility is statistically invariant which is consistent with the other performance behaviors. Next, we utilize the transmission line method (TLM) to extrapolate contact resistance for devices prepared with each surface treatment condition. Here, the 2Rc for S-CNT array FETs with P1, P2, and P3 treatments are in the range of 0.37− 0.73, 0.39−0.47, and 0.36−0.62 kΩ μm, respectively (Figure 7d). The 2Rc is also calculated on a per-tube basis (extrapolated using the average packing density of 50 tubes μm−1) to be 18.4−36.3, 19.7−23.5, and 18.2−30.8 kΩ/tube for the P1, P2, and P3 treatments, respectively. The similar performance observed for devices fabricated with each surface suggests the additional processing to remove the residual PFO-BPy does not adversely affect the charge transport at the contacts (or along the channel) or the 2Rc. Similarity in 2Rc before and after the removal of the remaining PFO-BPy polymer wrapper suggests that the polymer wrapper does not significantly contribute to contact resistance. Previous works have shown that the vacuum annealing step used for the P1 assembled+annealed treatment completely removes the insulating side chains from the PFOBPy polymer, while the backbone remains partially intact.1 Although a 2Rc of 7 kΩ/tube has been achieved for largediameter CNTs (>2.0 nm) with ohmic contacts, we speculate that the somewhat larger 2Rc observed here is either caused by an intrinsic Schottky barrier at the Pd-CNT interface (which has been observed previously for S-CNTs with similar diameter range as was used here of 1.3−1.7 nm),39,40 or additional contact resistance exacerbated by inter-CNT interactions in the thin film.41,42 While, recent advances in unwrapping polymers and supramolecular dispersants from CNTs in film have been coupled with demonstration of solution-processed short channel FETs,19 thin-film transistors,29 and sensors,43 a direct quantitative measurement of remaining residual dispersant and its effect on device performance has been elusive. The value in this work is the coupling of the evidence of polymer removal from the CNTs in-film and its direct effect on CNT channel and contact resistance.

to the neat polymer indicates the formation of an ether functional group37,38 C−O (between the polymer chains) or Si−O bonds with the dielectric SiO2; (4) the BPy groups are largely intact as the ν (CN), at 1565 cm−1 (Figure 6c) is preserved, which likely allows metal complexation and removal of the polymer from the CNTs; and (5) in the 1400 to 1600 cm−1 region, upon thermal annealing the vibrational spectrum is less dense, which might indicate chain scission. To follow these changes in the PFO-BPy wrapper on the CNT surface we created a thin film (∼2 nm) of PFO-BPy wrapped CoMoCAT tubes on a transparent quartz substrate so that direct UV−vis measurements can be made on the film. Our earlier studies24 had shown that the McAPR process is equally effective in solution in unwrapping PFO-BPy from the CoMoCAT tubes which are a smaller-diameter nanotube rich in (6,5) chirality, with potential applications in photovoltaic cells. UV−vis measurements on these PFO-BPy wrapped (6,5) CoMoCAT S-CNT films which were thermally annealed (CNT@PFO-BPy +annealed) show a similar blue shift in the absorption spectrum (Figure 6d). On the basis of these studies, the FESA films of Arc@ PFOBPy FESA were subjected to McAPR treatment followed by washing. The collected washing solvent was analyzed by GPC and UV−vis measurements. However, the concentration of the extracted polymer was too low to get NMR information on the polymer structure. Analysis of the molecular weight of this extract by GPC (Figure 6e) shows oligomers with a molecular weight of 1.9 K (Đ ∼ 1.9, with a DP of 4−8) compared to 32K for the parent polymer. The UV−vis analysis of this extract (Figure S5) shows an absorption maxima at 387 nm which corresponds to the Re-complexed PFO-BPy species.24 These results clearly show that the process of thermal annealing removes the alkyl side chains, leading to chain scission to create oligomers, while leaving the BPy units intact (Scheme S1). Hence metal complexation to BPy group is able to lower the noncovalent interactions with the CNT, change the rigidity of the oligomers24 and effectively lift the residues off the surface. Control experiments where the FESA film was subjected to thermal annealing+ solvent washing steps without the Retreatment step did not lead to removal of the strongly adsorbed oligomers from the CNTs. We fabricated and analyzed short channel devices where the CNTs directly span the source-drain electrodes to study the effect of McAPR processing on the contact resistance and doping levels (Figure 7a,b). S-CNT FETs were fabricated using three surface treatments: (i) assembled+annealed (P1) (Figure 5aii), (ii) assembled+annealed+Re-treated+washed (P2), (Figure 5aiv), and (iii) assembled+annealed+Re-treated+washed +annealed (P3). The additional annealing step in the P3 treatment was to ensure complete removal of solvent and ambient adsorbates prior to contact deposition. The current− voltage characteristics are presented in Figure 7c for devices representative of each surface treatment with Lch = 140 nm. All FETs exhibit p-type behavior. The on-state conductance density (Gon = IDS/(Wch*VDS)) is extracted at VGS − VT = −5 V and VDS = −0.1 V and reaches 890, 860, and 670 μS μm−1 for the P1, P2, and P3 treatments, respectively. The average and standard deviation of the on-state conductance for the devices within a range of 150−350 nm (where the average Lch = 250 nm) are 550 ± 209, 500 ± 160, and 459 ± 190 μS μm−1. The on/off conductance modulation is also similar to 106.2±0.8, 106.0±0.9, and 106.0±1.5. Assuming a parallel plate capacitance the calculated lower bound for the mobility of transistors with Lch =

■

CONCLUSION In conclusion, we have developed a process to quantitatively remove the PFO-BPy from S-CNTs in solution using McAPR. In solution, the polymer can be quantitatively removed using McAPR method in toluene solvent above 100 °C, followed by sequential washing with hot toluene, chloroform, and methanol, as characterized by XPS analysis. However, its implementation on a FESA aligned monolayer of S-CNT array was only partially effective. An additional thermal annealing at 400 °C in vacuum of the FESA aligned S-CNT arrays was necessary to quantitatively remove the polymer via McAPR process. The annealing process preferentially removes the side chains from the polymer backbone and also results in the scission of the polymer backbone into oligomers. However, these oligomers are strongly adsorbed on the CNT surface. FTIR studies show that the BPy groups stay largely intact hence allowing metal complexation, which results in lifting of the oligomers off the CNTs. These studies highlight the challenges in polymer removal in S-CNT monolayer-film versus solution. Resulting SCNT array FET devices show that the electronic properties of pristine CNT are preserved through this process. Our method of stripping PFO-BPy from S-CNT postsorting is mild enough to preserve the electronic structure of the CNTs, avoids the use 28864

DOI: 10.1021/acsami.7b06850 ACS Appl. Mater. Interfaces 2017, 9, 28859−28867

Research Article

ACS Applied Materials & Interfaces

treatment, the excess Re(CO)5Cl and loose polymer chains were washed off by immersing the sample in pure toluene, THF, and methanol at temperature setting of 110, 60, 60 °C for 12 h, respectively, on a Belly Dancer shaker. The washed S-CNT arrays were dried under high vacuum for 12 h to remove adsorbed solvent. Fabrication of Short-Channel S-CNT FETs. Following FESA deposition of aligned CNTs, the CNT channels were patterned using electron beam lithography. The resulting CNT channels were protected by 4 μm × 10 μm regions of poly(methyl methacrylate) (PMMA) and an oxygen reactive ion etch was used to etch the unwanted regions of CNTs and then the remaining PMMA was lifted off with acetone. The following surface treatments were explored prior to defining the device electrodes: (i) extensive toluene rinsing and vacuum anneal (400 °C) were performed (further details reported in elsewhere)1 (assembled+annealed (P1)), (ii) Re-treated + washed following surface treatments in (i) (assembled+annealed+Re-treated +washed (P2)), and (iii) a combination of (i)+(ii) followed by an additional vacuum annealing step at 400 °C (assembled annealed+Retreated+washed+annealed (P3)). A second electron-beam lithography step was performed to define source-drain electrodes with various channel lengths. Contacts were formed via thermal deposition of Pd (30 nm) and lift-off in acetone. The devices were measured under ambient conditions directly after electrode patterning with no further treatments.

of any acidic reagents, and likely is generalizable to any polymer with BPy as a comonomer. Similarity in 2Rc before and after the removal of the remaining PFO-BPy polymer wrapper backbone suggests that the backbone does not significantly contribute to contact resistance. While FETs provide a model platform to study the effect of polymer wrapper and the removal process on the electronic properties of the S-CNTs, these studies provide the basis for applications beyond FETs. For example, the ability to completely remove the polymer wrapper in aligned CNT arrays without adversely affecting the device properties opens up applications into photovoltaics and biosensing applications. For the latter, it is important to have complete access to the CNT surface for effective functionalization with suitable probes.

■

EXPERIMENTAL SECTION

Characterization. UV−visible measurements were taken using a Shimadzu PC-2401 spectrophotometer. Raman spectra of carbon nanotubes were obtained using an Aramis Horiba Jobin Yvon Confocal Raman microscope at an excitation wavelength of 532 nm and 6 W laser power. The silicon peak at 520 cm−1 was used to calibrate the wavenumber, and the spot size was ∼1 μm2. X-ray photoelectron spectroscopy (XPS, Thermo Scientific K-alpha) with microfocused monochromatic Al Kα X-ray source was used for elemental analysis. The hemispherical analyzer with 125 mm mean radius was utilized in analyzer energy with 200 μm selected area aperture. CoMoCAT (6,5) S-CNTs Dispersion by PFO-BPy in OrthoDichlorobenzene and Formation of Thin Film on Quartz Substrate. Enriched (6,5) nanotubes were separated with a method adapted from Ozawa et al.,44 in which 35 mg of raw nanotube soot (Sigma-Aldrich, SG65i) is dispersed by ultrasonication (Fisher Model 500) with 2 mg/mL PFO-BPy (American Dye Source, Inc.) at 40% amplitude for 15 min. The resulting slurry is centrifuged at 300 000g for 10 min. The supernatant is removed, and rotary evaporated to concentrate, then centrifuged at 150 000g for 24 h. The pellet is redispersed into hot tetrahydrofuran and then centrifuged again. This procedure is repeated 3−5 times until the mass ratio, measured by absorbance, of polymer to nanotube is