Preserving π-conjugation in covalently functionalized carbon nanotubes for optoelectronic applications

Covalent functionalization tailors carbon nanotubes for a wide range of applications in varying environments. Its strength and stability of attachment come at the price of degrading the carbon nanotubes sp2 network and destroying the tubes electronic and optoelectronic features. Here we present a non-destructive, covalent, gram-scale functionalization of single-walled carbon nanotubes by a new [2+1] cycloaddition. The reaction rebuilds the extended π-network, thereby retaining the outstanding quantum optoelectronic properties of carbon nanotubes, including bright light emission at high degree of functionalization (1 group per 25 carbon atoms). The conjugation method described here opens the way for advanced tailoring nanotubes as demonstrated for light-triggered reversible doping through photochromic molecular switches and nanoplasmonic gold-nanotube hybrids with enhanced infrared light emission.


Supplementary Note 1: Synthesis and characterization of the functionalized tubes
In this supplementary note we characterize our functionalization products. The intermediate azide and the final product are examined by thermogravimetric analysis (TGA), nuclear magnetic resonance (NMR), elemental analysis (EA), and infrared spectroscopy (IR). Raman spectroscopy identifies the nanotube species. High-resolution scanning transmission electron spectroscopy, coupled with local electron energy loss spectroscopy, allows to locally detect the presence of the functionality onto the nanotube sidewalls. 2-D excitation-emission spectroscopy proofs that the functionalized nanotubes emit light.

Schematic summary of the reaction steps
The synthetic steps towards the functionalized SWNTs are described in the methods section of the main manuscript text and schematically summarized in Supplementary Figure 1.

Comparison between covalently and non-covalently triazine-functionalized SWNTs
To prove the effective covalent attachment of the triazine onto the SWNTs, we performed a blank reaction designed to prevent the covalent attachment of the triazine onto the SWNTs. We followed the synthetic steps of the original reaction but without adding the sodium azide and thus without converting the 2,4,6-1,3,5-trichloro-triazine into 2-azido-4,6-dichloro-1,3,5-triazine. In this way the 2,4,6-1,3,5-trichloro-triazine can only non-covalently attach to the SWNTs.
For performing the blank reaction, SWNTs (1 g) were added to N-methyl-2-pyrrolidone (150 ml) and sonicated for 1 h. The mixture was then stirred at room temperature for one additional hour.
Sodium sulfate (1 g) was added to the organic phase and stirred for 1 h. After filtration the solvent was evaporated. The product was used for the next reaction without further purification.
In the IR spectra of (4,6-diphenoxy-1,3,5-triazin-2-yl)-L-cysteine (Supplementary Figure 12) the absorbance band at 2375 cm -1 is assigned to the thiol group. This result proves that cysteine is attached to the triazine ring through amino functional group and that the thiol group remains unreacted.

Raman Scattering
Supplementary Figure 13 depicts the Raman spectra of the radial breathing modes (RBM) of pristine (SWNT) and triazine functionalized (SWNT-low and SWNT-high) nanotubes excited at -39/65-633 nm. The lines of the (7,6), (8,4), (7,5), and (8,3) nanotube species are observed at these excitation energies. [7] No change in the RBM is observed for the different degrees of functionalization. The D and G bands of the Raman spectra are reported in Fig. 2c of the main text.

HR(S)TEM and EELS studies
Transmission electron microscopy (TEM) and spatially-resolved electron energy loss spectroscopy (SR-EELS) studies were performed on the functionalized nanotubes.
Supplementary Figure 14a is a high resolution (HRTEM) micrograph where a bundle of singlewalled nanotubes is visible. Some moieties are observed at the surface of these SWNTs. In order to locally analyse these objects, having access to chemical information, we performed spatiallyresolved EELS (SR-EELS) measurements. This is a powerful tool to investigate such nanomaterials at (sub-) nanometer scale [8][9][10][11][12][13][14][15].
We indicates that this nitrogen corresponds to a pyrrolic-like atomic configuration [8][9][10][11][12][13]. Thus, the presence of this nitrogen, its atomic configuration, and the atomic configuration of the associated carbon, reveal that this nitrogen is associated to the presence of triazine.
Furthermore, it is also confirmed by the fact that these molecules are the only source of nitrogen in the samples. From the local TEM analyses, we conclude that the observed moieties correspond to the triazine functionalized molecules localized at the surface of the SWNTs.

Supplementary Note 2: Computational studies
The structure of triazine-functionalized carbon nanotubes were calculated for armchair and zigzag nanotubes. For this paper we concentrate on the exemplary (

TGA analysis
Thermogravimetric analysis (TGA) showed that the amount of SWNT was reduced by 12% and 36% due to the triazine functional groups of the SWNT-high and SWNT-low samples, respectively (Supplementary Figure 19a). Based on these mass ratios, the DFG of SWNT-high was 1/25 and of SWNT-low was 1/100. These results are in excellent agreement with the elemental analysis results.
-44/65-Each reaction was repeated at least ten times for evaluating the reproducibility of the functionalization. The nitrogen contents slightly changes from batch to batch, with a maximum deviation from the mean value of 1% (SWNT-low) and 1.3% (of SWNT-high), respectively (Supplementary Figure 19b). The nitrogen content is an indicator for DFG, proving that our approach produced SWNT with a well-defined amount of functionalities.

IR spectroscopy
In the IR spectra, characteristic bands of aziridine rings usually appear in the 1000-1250 cm -1 region and are assigned to symmetric aziridine ring breathing modes and to the C-N bonds in which carbon is sp 3 hybridized (sp 3  The sp 3 and carbon-oxygen species are due to defect sites in the SWNTs and their inherent oxidation. These constitute 30.9% of the total carbon content, leaving 69.1% of sp 2 -hybridized carbon.
The inset graph of Supplementary Figure 22 shows the C 1s spectra of the functionalized samples. They closely resemble the spectra of the pristine SWNT but are energetically shifted.
We fitted the spectrum of SWNT-high by combining the same peak-area ratios and peaks we

Quantification of the Covalent Triazine Functionalization
The functionalization ratios of triazine groups (FR XPS (tri)) on the SWNTs were determined by relating the respective total triazine N 1s spectral area of each sample to its sp 2 C1s area (i.e., the sum of sp 2 and attachment C 2 -unit peak areas). To this end measurement statistics, relative elemental sensitivity factors [22], and the stoichiometry were taken into account. We define  -48/65-background [23]. SWNT-high and SWNT-low were fitted with two peaks having an area ratio of 1:3 that we assigned to the single bridging (peak B) and the three triazine nitrogen atoms (peak A). The total triazine N 1s peak areas were used to determine the FR XPS (tri) and DFG XPS (tri) values given in Supplementary Table 5.
In the N 1s spectrum of the spiropyran-functionalized SP-SWNT an additional signal at E B = 405.49 eV (peak C) is clearly visible. We assigned it to the nitro group of the spiropyran molecule assembled at the triazine anchor group [24], thereby proving its successful synthesis.
The N 1s signal between E B = 397 and 402 eV for SP-SWNT is similar to that of SWNT-low.
The additional intensity is due to the second nitrogen atom of the molecular switch (peak D).
We determined the FR XPS (spiro) of the spiropyran groups with respect to the available triazine the last part of this note we analyse the optically-triggered effects due to the spiropyran-tomerocyanine isomerization when SP-SWNT are exposed to UV radiation. -51/65-

X-ray photoelectron spectroscopy
Our XPS study provided strong evidence for an electronic interaction between the triazine unit and the functionalized SWNT. The C 1s XP signal of a C 2 unit under the bridge to the triazine unit shifted by 3.5 eV to higher binding energy. A corresponding shift to lower binding energy was observed for the N 1s signal of the bridging nitrogen atom with respect to the nitrogen in the The XPS peak shifts were independent of the actual amount of SWNT material covering the substrate. This implies the induction of a negligible photovoltage and/or a low-resistance junction between the Au substrate and the drop-coated SWNT film. Therefore the Fermi levels E F of the electron analyzer and the sample were aligned in the XPS measurements and the observed shift of the C 1s peak position with increasing functionalization ratio directly reflected the shift of the SWNT Fermi level. E F shifts upwards towards the conduction band of the semiconducting SWNT. When referenced to the analyzer Fermi level the C1s XPS signal must shift downwards towards higher binding energy E B (see Supplementary Figure 28).
In Supplementary Figure 29 we plot FR XPS (tri) as a function of the observed C 1s peak shift E for the SWNTs. We also investigated two additional triazine-functionalized samples, prepared with the same procedures, one with less (SWNT (*) -low) and one with more (SWNT (*) -high) -52/65-functional groups based on SWNTs coming from different production methods and batches (CoMoCAT ® ) and referenced against their respective pristine SWNTs.

Raman characterization of the metallic tubes.
The Raman spectra confirm the functionalization of metallic nanotubes through changes in peak width. The G band of metallic carbon nanotubes consists of two bands, labelled as G + at ca. 1580 cm -1 due to the transverse optical phonon (TO), and Gat ca. 1540 cm -1 due to the longitudinal optical phonon (LO). The LO mode responsible for the Gband is very sensitive to shifts of the Fermi energy level and its broadening can be used to quantify the doping in metallic SWNTs [26]. We monitored the G band of metallic nanotubes with various functionalization levels under 532 nm laser excitation, where metallic tubes were resonantly excited. For each sample we acquired at the same spot the G and the radial-breathing mode (RBM) spectra to control the CNT chirality distribution at the analysed position ( Supplementary Figure 30a,b). We identified the metallic (9,3), (8,5), and (7,7) species.
To calculate the Fermi energy shift, we used the model of the LO peak broadening discussed in Ref. [26]. A shift in the Fermi level away from its intrinsic position (crossing of valence and conduction band) continuously increases the LO phonon lifetime and decreases the full width at half maximum (FWHM) of the LO peak. As a rule of thumb one half of the maximum FWHM indicates the E F at half the LO energy. The maximum FWHM value for the analysed species (d=0.85 nm) is 100 cm -1 and the phonon energy is 187 meV. [27] We considered a Lorentzian dispersion model for fitting our data (Supplementary Figure 30d), obtaining values of the doping levels of -70 meV for the pristine SWNT, -40 meV for SWNT-low and SP-SWNT, and +20 meV for SWNT-high. The positive sign of the E F shift was obtained from the 6 cm -1 decrease in frequency of the 2D Raman mode between the pristine SWNT and SWNT-high. [26,28] It agrees -53/65-with the increase in Fermi energy obtained from the XPS data. Overall, E F from Raman is smaller than observed in XPS. This is explained by the limited number of small-diameter tubes accessed in Raman scattering, whereas XPS obtains a mean doping level across all nanotubes.

Effect of the SP-to-MC isomerization
Under UV irradiation, SP converts into MC. As a consequence of the SP -to-MC isomerization, the emission of the tubes gets quenched. In Supplementary Figure 31a  nanotubes' G band consists of two sub-bands, labelled G + and Gas well but arising from different modes than in metallic tubes. [26,27] For semiconducting SWNTs, the LO phonon contributes to the G + band (at 1590 cm -1 ) while the TO contributes to the Gband (at ca. 1565 cm -1 ). Doping thus affects differently semiconducting SWNT than the metallic SWNT, shifting the position of the G + band. This can be used to quantify the shift of the Fermi energy [28]. The Raman spectra of SP-SWNT and MC-SWNT (obtained by UV light irradiation of the SP-SWNT to promote SP-to-MC conversion) are shown in Supplementary Figure 32. The only difference -54/65-between the two spectra is given by a shift of 2 cm -1 of the G + band position, suggesting an upshift of 0.2 eV of the Fermi energy from comparing with the result of Ref. [28].

Control experiment on the effect of UV irradiation
To confirm that the observed effects are due to the SP-to-MC conversion and rule out UV exposure damages, we exposed to UV irradiation two samples with similar characteristics but one with-(SP-SWNT) and the other without (SWNT-low) spiropyran. In Supplementary   Figure 33 the temporal change of the absorbance at 300 nm is monitored. While the onset of the merocyanine absorption band can be observed in SP-SWNT, no changes occur in the SWNT-low sample, ruling out any damage by the UV light exposure. -55/65-

Supplementary Note 5: Nanoplasmonic hybrids
In this Supplementary Note we report on the characteristics of the nanoplasmonic hybrids Au@SWNTs. Additionally, we discuss a set of control experiments for plasmonic enhancement in SWNTs.

Optical and morphological properties of the AuNPs
For the synthesis of the AuNPs, please refer to the Methods section of the main manuscript text.
TEM micrographs of the AuNPs show a distribution of the particle size between 10 and 20 nm (Supplementary Figure 34). For such an ensemble absorption spectroscopy reveals the presence of the plasmon absorption band of the AuNPs peaked at 520 nm.

2D emission spectroscopy of the Au@SWNTs hybrids: Emission enhancement
Once coated with the AuNPs, we observe enhancement of the emission of the nanotubes. The relative change of the emission intensity depends on the specific species and ranges between 20% and 100%. In Supplementary Figure 36 we plot the relative change of the emission intensity vs. excitation wavelength of the considered species (green dots, the curve is a guide to the eye) and compare it with the absorption spectrum of the Au@SWNT hybrids (blue curve).
-56/65-Interestingly, there is a systematic red-shift between the plasmonic bands of the gold particles and the excitation resonance profile of the carbon nanotubes. The dependence of the enhancement on excitation wavelength with respect to the plasmonic response will be studied in a future paper.

Nanoplasmonic hybrids -control experiments: Role of the covalently-bound thiols and contribution of the triazine-based covalent attachment
To highlight the importance of the thiol-based attachment in our hybridization process, we performed two sets of control experiments: One control shows the relevance of the thiol-triazine based immobilization of the Au nanoparticles on the SWNTs and the other compares standard covalent functionalization with our cycloaddition-based method.
The first set of control samples were prepared following the very same procedure as for the Au@SWNTs hybrids but starting each time from different sets of tubes, one for each step of the synthesis leading to the SH-SWNTs. In Suppplementary Figures 37a,b we compare the emission of the pristine tubes (SWNT) with the emission of the hybrids obtained by mixing gold with pristine tubes (Au+SWNT) and tubes functionalized with triazine groups but without the thioltermination (Au+SWNT-high). The emission from the hybrids is always weaker than the emission from the pristine tubes, highlighting the key role of the thiol-induced immobilization of the AuNPs at the nanotubes sidewalls.
In the second set of control experiment, we show that the other key element towards successful emission enhancement is the preservation of the pristine optoelectronic properties of the tubes.
We generate a set of COOH-covalently functionalized carbon nanotubes (COOH-SWNTs) by following the standard acid treatment. [29] To produce the carboxylic groups onto the tubes sidewalls, we dissolved 1 g of SWNTs in a mixture of 150 ml of H 2 SO 4 and 50 ml of HNO 3 , -57/65-which was heated up to 60°C and stirred for 45 min. Afterwards, the solution was filtered and washed repeatedly with distilled water to reach a pH value of 7. Subsequently, the sample was dried overnight at 60°C. Figure 38a compares the Raman spectra of pristine (SWNT) and carboxylated (COOH-SWNT) tubes. The increase in the intensity of the D band after the carboxylation of the tubes is a clear indication of the increased defectivity introduced by the covalent approach. The optical properties of the tubes are destroyed by the functionalization. We ensured a mild functionalization to preserve some emission from the functionalized tubes.
To attach the thiol-groups onto the tubes, 100 mg of the COOH-SWNTs were dispersed in thionyl chloride (10 ml) by 30 min sonication. Then they were refluxed for 6 h. The excess of thionyl chloride was evaporated and dry DMF was added to the residual compound. Cystein (100 mg) and triethylamine (0.11 ml) were added to the mixture, which was then stirred for 2 h at 25°C and for 12 h at 70°C, respectively. The mixture was cooled and centrifuged (1100 rpm,  -59/65-

Supplementary Methods
Infrared (IR) spectra were recorded using a JASCO spectrometer. Ultrasonic bath (Model: SONOREX, RK255 HZ) was used to disperse materials in solvents. TGA measurements were recorded by a STA 409 apparatus (from Netzsch) in temperatures ranging from 25-800ºC with a 10ºC/min heating rate under air. Elemental analysis was performed using ELEMENTAR apparatus with three columns and detector for carbon, nitrogen, hydrogen and sulfur elements. amounts of substance and their spectra needed to be normalized. Considering the comparable shape of the C 1s spectra and the fact that the SWNTs were not exposed to any harsh treatments, we assumed that their carbon backbones remained intact and that the established functionalization ratios (FR) did not significantly reduce the amount of sp 2 -hybridized carbon atoms. Therefore we chose to normalize all SWNT XP spectra with respect to the carbon backbone, i.e., the intensity of the sp 2 -carbon component.
High resolution TEM was performed employing an imaging-side aberration-corrected FEI Titan-Cube microscope working at 80 kV, equipped with a Cs corrector (CETCOR from CEOS -60/65-GmbH). Spatially-resolved electron energy loss spectroscopy (EELS) measurements were performed on probe-corrected scanning TEM (STEM) FEI Titan Low-Base operating at 80keV (fitted with a X-FEG® gun and Cs-probe corrector (CESCOR from CEOS GmbH)).
Furthermore, in order to avoid the effects of electron beam damage, these measurements have been performed using a liquid-nitrogen-cooled cryo-holder at -170º C. EEL spectra were recorded using the spectrum-imaging (SPIM in 2D or spectrum-line (SPLI) in 1D) mode [15,30] in a Gatan GIF Tridiem ESR 865 spectrometer. The convergent semi-angle was of 25 mrad, the collection semi-angle was of 30 mrad and the energy resolution ~ 1.0 eV. To filter the noise in the experimental data, the background corrected EEL spectra showed in Supplementary   Figure 14f were smoothed using a Savitzky-Golay filter (second-order polynomial).
The transmission electron microscopy (TEM) samples were prepared by dispersing the NTs powders in ethanol. The dispersions were ultrasonicated and subsequently deposited on holey carbon 3 mm copper grids.
Raman spectra of metallic nanotubes were acquired using an XploRa spectrometer (Horiba), excitation wavelength at 532 nm, equipped with charge-coupled device, 2400 lines/mm gratings and edge filter to block Rayleigh-scattered light. Frequencies were calibrated using a cyclohexane reference sample.
The electrospra y ionization-time of flight (ESI-TOF) measurements were performed on an Agilent 6210 ESI-TOF from Agilent Technologies, Santa Clara, CA, USA. The solvent flow rate was adjusted to 4 µL/min and the spray voltage set to 4 kV. The drying gas flow rate was set to 15 psi (1 bar). All the other parameters were adjusted for a maximum abundance of the relative [M+H] + . 1 H NMR and 13 C NMR spectra were recorded on a Bruker Avance 400 spectrometer -61/65-(Bruker Corporation, Billerica, MA, USA) (at 295 K). Tetramethylsilane was used for internal calibration at 125 MHz with complete proton decoupling.