Synthetic control over the binding configuration of luminescent sp3-defects in single-walled carbon nanotubes

The controlled functionalization of single-walled carbon nanotubes with luminescent sp3-defects has created the potential to employ them as quantum-light sources in the near-infrared. For that, it is crucial to control their spectral diversity. The emission wavelength is determined by the binding configuration of the defects rather than the molecular structure of the attached groups. However, current functionalization methods produce a variety of binding configurations and thus emission wavelengths. We introduce a simple reaction protocol for the creation of only one type of luminescent defect in polymer-sorted (6,5) nanotubes, which is more red-shifted and exhibits longer photoluminescence lifetimes than the commonly obtained binding configurations. We demonstrate single-photon emission at room temperature and expand this functionalization to other polymer-wrapped nanotubes with emission further in the near-infrared. As the selectivity of the reaction with various aniline derivatives depends on the presence of an organic base we propose nucleophilic addition as the reaction mechanism.

Emitted photons were collected with the same objective. The photoluminescence was filtered by a spectrograph (Acton SpectraPro SP2358, grating blaze 1200 nm, 150 lines mm −1 ) and focused onto a gated InGaAs/InP avalanche photodiode (Micro Photon Devices). Arrival times of the detected photons were recorded with a time-correlated single-photon counting module (Picoharp 300, Picoquant GmbH). The instrument response function (IRF) was determined by the fast, instrument-limited photoluminescence decay at the E11 transition (e.g. ~1000 nm for (6,5) SWNTs). All fluorescence decay histograms were fitted with a biexponential model in a reconvolution procedure.

Normalization of Photoluminescence Spectra
To display changes in defect density (see Figure 1b-d and Figure 4a,b of the main manuscript) normalized PL spectra are presented instead of absolute emission intensities for the following experimental reasons: (1) To stop the reaction and remove side products, the functionalization protocol includes a filtration, washing and redispersion sequence (see detailed sp 3 functionalization protocol). It is not possible to simply record the PL spectra during the reaction. As with pristine SWNTs, the yield of the redispersion step varies and depends on parameters such as sonication power, environmental humidity and temperature. Hence, the resulting dispersions have different concentrations of the dispersed SWNTs and consequently different absolute PL intensities.
(2) While the absorption spectrum of the SWNTs is unaffected at lower levels of functionalization, samples with higher defect densities display a significant reduction of the main absorption band. Hence, the effective absorption cross-section for the functionalized SWNTs also depends on defect density preventing an accurate correction for the yield of the redispersion step.
(3) PL measurements were performed by focusing the excitation laser into a cuvette through an objective. This configuration has the advantage that the near-infrared absorption of toluene does not affect the PL spectra due to the extremely short path length within the liquid. However, even slight differences in the quality of the focus have an impact on the absolute PL intensities, while the spectrum is usually unaffected by minor changes in focus.
As a result, comparison of the absolute PL intensities across a sample series is unreliable and only normalized PL spectra are shown. However, direct and reliable values of the emission efficiencies are provided by photoluminescence quantum yield measurements (see below and

Photoluminescence Quantum Yield Measurements
The absolute photoluminescence quantum yield (PLQY) of pristine and functionalized nanotubes in dispersion was determined using an integrating sphere. 1,2 The SWNT dispersions were adjusted to an optical density of < 0.2 cm -1 at the E11 transition and placed in the center of an integrating sphere (LabSphere, Spectralon coating). A laser beam tuned to the E22 transition was directed onto the sample and the signal (scattered laser light and photoluminescence) was transmitted to the spectrometer via an optical fiber. To account for absorption of the solvent at the excitation wavelength, the same measurement was repeated with the pure solvent (toluene). PQLY was calculated as the ratio of emitted to absorbed photons. The wavelength-dependent detection efficiency and losses were corrected by recording a reference spectrum of a stabilized tungsten halogen light source with known spectral power distribution (Thorlabs SLS201/M,300-2600 nm) that was placed in front of the integration sphere.

Low and Variable Temperature Spectroscopy of SWNTs
Low-temperature photoluminescence spectra of individual, PFO-BPy-wrapped, functionalized (6,5) SWNTs embedded in polystyrene and dense films of nanotubes were recorded using a closed-cycle liquid helium optical cryostat (Montana Instruments Cryostation s50) with an adjustable temperature between 3.8 K and 300 K. The nanotube samples were excited with a continuous wave laser diode (OBIS, Coherent Inc., 640 nm) through an infrared 50x long working distance objective (Mitutoyo, N.A. = 0.42) mounted outside the cryostat. The laser power was typically ~100 µW and the polarization was adjusted with a λ/2 plate to match the orientation of individual nanotubes. PL spectra were acquired with a thermoelectrically cooled InGaAs camera (NIRvana 640ST, Princeton Instruments) mounted on a corresponding spectrograph (IsoPlane SCT-320, Princeton Instruments).

Measurements
Room-temperature photoluminescence and autocorrelation measurements were performed on individual functionalized (6,5) SWNTs embedded in polystyrene with a home-built confocal microscope with slip-stick positioners (ANPxy101 and ANPz102, attocube systems). A wavelength-tunable Ti:sapphire laser (Mira, Coherent) in continuous wave mode served as the excitation source and was tuned to 995 nm to be in resonance with the E11 transition of (6,5) SWNTs. The excitation laser was focused onto the sample with an apochromatic objective (LT-APO/IR/0.81, attocube systems) and PL was collected by the same objective and spectrally filtered using a tunable long-pass filter (TLP01-1116, Semrock). A spectrometer (Acton SP2500, Roper Scientific) coupled with a liquid-nitrogen cooled InGaAs camera (OMA V: 1024-1.7, Roper Scientific) were used to record photoluminescence spectra. For time-resolved photoluminescence and pulsed photon correlation in a standard Hanbury-Brown and Twiss setup, the sample was excited using a supercontinuum laser (SuperK EXTREME, NKT Photonics) with a 6 ps pulse width and a 78 MHz repetition rate, tuned to 995 nm by a set of spectral filters. Emission was directed onto a superconducting single photon detector (TCOPRS-CCR-SW-85, Scontel) and photon detection events were recorded using a timecorrelated single-photon counting module (PicoHarp300, PicoQuant).
While the concentrations of the reactive reagent (e.g., aniline derivative) and DMSO were kept constant, the degree of functionalization could be controlled by the amount of base (KO t Bu), reaction time or temperature. SWNT dispersions (after removal of excess polymer) were used as the starting material and the nanotube concentration was always kept at 0.54 mg L -1 (corresponding to an E11 absorbance of 0.3 cm -1 for (6,5) SWNTs). 3 Filtration of the nanotube dispersion and redispersion in pure solvent is used to remove excess wrapping polymer and was performed for all reactions presented in this study for higher reproducibility. However, this step is mainly needed for reactions with UV irradiation to avoid light attenuation at 365 nm due to strong absorption by the polymer. The filtration step is not necessary when the functionalization is performed in the dark and the concentration of the wrapping polymer is below ~0.3 g L -1 .

Reagents
All chemicals used in the described functionalization reactions were purchased from Sigma Aldrich and used without further purification: dimethylsulfoxide (anhydrous, ≥99.9%), tetrahydrofuran (anhydrous ≥99.9% inhibitor free).
The quality of DMSO was found to have a significant impact on the reactivity but not on the selectivity. In order to achieve comparable reactivities as presented in this work, freshly dried DMSO is recommended. Stock solutions of KO t Bu and DMSO should be stored under inert gas atmosphere, but the functionalization itself can be performed in an open flask at room temperature.
Step by Step Reaction Protocol 1. Filter polymer-wrapped SWNTs dispersion over a PTFE membrane (Merck Millipore JVWP, 0.1 µm pore size) and wash the resulting filter cake three times (each 5 minutes) with hot toluene (80 °C) to remove excess wrapping polymer.
2. Redisperse the washed filter cake by ultrasonication and adjust the SWNT concentration to an optical density of >1.0 cm -1 at the E11 absorption peak. It is important that the reagent (e.g., aniline derivative) and DMSO are present in the reaction mixture before the addition of KO t Bu/THF to prevent undesired side-reactions.

When functionalization is performed in the dark:
9. Protect the glass vial from light and stir the reaction mixture for the desired duration at room temperature. Reaction times can range between 15 min and 180 min.

When functionalization is performed under UV light irradiation:
10. Irradiate the glass vial with UV-light (365 nm, here SOLIS-365C, Thorlabs, 1.9 mW/mm 2 ) under continuous stirring. Reaction times usually vary between 10 min and 45 min.

Work up:
11. After the desired functionalization time has elapsed, pass the reaction mixture through a PTFE membrane (e.g. Merck Millipore JVWP, 0.1 mm pore size) and wash the filter cake with approximately 5 mL MeOH and 5 mL toluene on the filtration setup.
12. Redisperse the filter cake in the desired amount of toluene with a low concentration of fresh wrapping polymer (e.g. 0.1 g L -1 ) by bath sonication for 20 min. Addition of wrapping polymer is not strictly necessary, but increases the colloidal stability of the dispersion for characterization. Note that for high defect densities (D/G + ratio greater than 0.2) the yield of the redispersion process starts to decline due to increasing aggregation of the functionalized SWNTs.

Supplementary Note 1: Concentration Effects
Following the general procedure described above, (6,5) SWNTs were also functionalized using different concentrations of 2-iodoaniline and KO t Bu. The ratio of 2-iodoaniline to base was kept at 1:2. As the the functionalization was mostly conducted with a large excess of reagent

Photoluminescence
To investigate the impact of temperature on defect emission we recorded PL spectra of thin films of (6,5) SWNTs functionalized with 2-iodoaniline (in the dark and under UV illumination) from 4 K to 330 K ( Supplementary Figures 8 and 9) in an optical cryostat as well as of dispersions between 278 K and 308 K ( Supplementary Figures 10 and 11) in a Peltierbased temperature-controlled cuvette holder (Fluorolog). Note that the relative intensity of defect emission varies strongly between PL measurements of dispersions and thin films (see Supplementary Figures 9 and 11). This effect can be assigned to the power-dependence of the defect state emission compared to E11, which is more pronounced for E11* − defects (see Supplementary Figure 6).
Starting at 4 K, the nanotube thin films showed a relative increase in defect emission with increasing temperature (both for E11* and E11* − ) reaching a maximum between 180 K and 220 K. This increase in defect emission correlates well with the model of a potential barrier around the defect site as proposed by Kim et al. 4 In this context, the increase in defect emission is attributed to a higher exciton trapping efficiency. After reaching this maximum, the defect emission steadily decreased again. This decrease in defect emission was investigated in more detail for dispersions and closer to room temperature.
The temperature-dependent distribution of localized and mobile excitons is commonly associated with a certain detrapping energy of the localized excitons, which is however quite different from the optical trap depth. As described by Kim where k is the Boltzmann constant and is a correction factor. The van't Hoff plots for the E11* and E11* − emission bands display good linear fits to the data and extraction of ΔEthermal.
According to this analysis the E11* defect exhibits a detrapping energy of approximately 16 79 meV, while the apparent detrapping energy for E11* − is surprisingly low with ~25 meV (see Supplementary Figures 10 and 11). Note that for the determination of the detrapping energy of E11* − defects, samples with different defect densities were used. Kim et al. previously reported that the detrapping energy increases at higher defect density. 5 While a similar trend can be observed here, the detrapping energy of E11* − defects increases only slightly from 23 to 27 meV. Such a low detrapping energy is in clear contrast to the increased defect state PL lifetime of E11* − . Previously, the strong correlation between defect state lifetime and optical trap depth was suggested to originate from phonon-assisted thermal detrapping. 6 Thus, we expect that a different temperature-dependent non-radiative decay mechanism, such as multiphonon decay, is dominant for E11* − defects within the high temperature range. This finding further highlights the complex relationship between optical trap depths and thermal detrapping. SWNTs at room temperature (r.t., red line). The higher signal at 1060 nm at low temperature is assumed to originate from defects unintentionally introduced during initial processing of SWNTs (dispersion etc.). Their number and emission intensity vary even for untreated ("pristine") SWNTs. 7
Characteristic nanotube transitions are indicated as E11, E22 and PSB (phonon sideband). Additional SWNT species and absorption bands of the wrapping polymers are indicated.

Characterization of (7,5) SWNTs Functionalized with 2-Iodoaniline
Supplementary Figure 14. a, Averaged Raman spectra of (7,5) SWNTs functionalized with 2-iodoaniline and after different reaction times. Inset: zoom-in on the D-mode region. b, Integrated E11* − /E11 emission area ratios vs. integrated Raman D/G + ratios as a metric for defect density. c, Absorption spectra of (7,5) SWNTs functionalized with 2-iodoaniline after different reaction times. The defect density is still too low to observe significant defect absorption.

Supplementary Note 3: Mechanistic Considerations
Reference and control experiments revealed that defect emission bands located at 1130-1180 nm (E11*) and ~1250 nm (E11* − ) appear for various reaction conditions (see Supplementary Tables 3 and 4). As the functionalization process involves multiple components, solvents and reagents, the role and impact of each shall be summarized and discussed briefly here. It should also be noted that preference for specific defect configurations by introducing sp 3 -defects in close proximity to already existing defect site may also play a minor role. 8

Toluene:
Toluene was chosen as the main solvent because polymer-sorting of SWNTs is typically performed in this medium and the employed aniline derivatives show good solubility. This solvent choice facilitates the direct functionalization after the SWNT sorting process without further processing steps and ensures stable dispersions.

THF:
THF can be used to drastically speed up the rate of the functionalization reaction. While polymer-wrapped SWNTs remain stable in this solvent, the solubility of KO t Bu is greatly increased. An enhanced reactivity can be observed by comparing the reaction times of functionalization performed in toluene vs. those in THF. Hence, small fractions (8.3 vol%) of THF were added to the toluene reaction mixture in the standard protocol.

Potassium tert-butoxide (KO t Bu):
While KO t Bu represents a key reagent in many organic transition-metal-free coupling reactions 9-11 it has been applied for carbon nanotube chemistry only once to the best of our knowledge. 12 Upon combining aniline derivatives with the base KO t Bu, fast deprotonation of the amine group (or other acidic protons) is expected. The resulting anion may then proceed to react with the nanotube. Increasing the amount of base shifts the equilibrium of this deprotonation step and consequently increases the rate of functionalization.
It is important to note that while KO t Bu is most commonly used as a strong organic base, it can also form radicals in the presence of appropriate electron acceptors. 11 As shown by the control reactions performed in this study, KO t Bu itself is able to lead to emissive defect states in the E11* and E11* − region without the presence of an aniline derivative. Thus, while the exact interactions with carbon nanotubes are unknown, it has the potential to follow nucleophilic as well as radical reaction paths with the nanotubes, possibly leading to E11* and E11* − emission bands.

Dimethylsulfoxide (DMSO) / Dimsyl anion:
DMSO is a common co-solvent for KO t Bu and is known to dramatically increase its basicity. 13 Hence, when functionalization is performed in the dark, the addition of DMSO may shift the equilibrium of deprotonated aniline derivatives resulting in increased functinalization rates and higher selectivity towards E11* − defects. Furthermore, in the presence of small amounts of a base, a dimsyl anion can be formed that behaves like a typical carbanion and represents a strong organic base. 14 Thus, in the absence of aniline derivatives it can potentially initiate nucleophilic reaction paths leading to E11* − defects with high selectivity.
Upon UV light excitation (~350 nm) the dimsyl anion can act as an electron donor and radical source. 15 Under UV illumination in the presence of DMSO we observed E11* emission bands, which we associate with such a radical reaction path.
Additionally, weak E11* bands were observed when the functionalization was performed without the addition of KO t Bu under UV illumination. The absence of E11* − defects is consistent with the proposed functionalization mechanism as no nucleophilic species can be formed.

Aniline Derivatives:
The addition of 2-iodoaniline or 2-fluoroaniline was found to greatly increase the functionalization rate for E11* and E11* − defects when THF was used as a solvent. When functionalization is performed in toluene, vastly different reactivities and selectivities towards E11* and E11* − defects were found compared to functionalization in the absence of aniline derivatives. Thus, a functionalization path via aniline intermediates is highly likely. It has to be noted that E11* defects were observed for functionalization with 2-fluoroaniline, thus a dehalogenation step is not necessary for the introduction of E11* defects.

UV-Light Illumination:
UV-light illumination was found to increase the reaction rate for the introduction of E11* as well as E11* − defects. The latter is expected to originate partially from heating effects during the illumination process. Importantly, no reaction of the wrapping polymer with SWNTs was observed upon illumination.

General Remarks about the Importance of the Base
Several control and test reactions highlight the importance of the base in this reaction system.
While a base that purely engages in a nucleophilic reaction pathway is desirable, the following aspects should be considered before testing alternative bases: 1. The highest selectivity towards the introduction of sp 3 -defects showing E11* − emission was observed for anilines, which are very poor acids with a pka of 28.7 (2-fluoroaniline in DMSO 16 ), thus a strong base is needed for effective formation of the reactive aniline anion.
2. Strong bases such as alkyl lithium compounds lead to significant side-wall functionalization of SWNTs [17][18][19] and may not follow the desired functionalization path.
3. The employed base should not act as reducing agent as this can lead to a Billups-Birch type functionalization. 20 4. To ensure high reproducibility it would be beneficial to perform the functionalization process under ambient conditions. Thus, bases that are highly sensitive toward oxygen and water should be avoided.
In summary, KO t Bu in DMSO as co-solvent represents an excellent system as it is a strong base (pka 32.2), a poor reducing agent and bench-stable in the dark.

Supplementary Note 4: Non-Aniline Reagents
In contrast to functionalization with aniline derivatives, a reduced reactivity and selectivity towards E11* − defects is observed for non-aniline reagents. This behaviour probably originates from the molecular differences of the carbanion intermediates. The stabilization and position of the reactive carbanion can differ vastly between different substance classes. For example, Li et al. showed that the reaction of C60-fullerenes with indole in the presence of KO t Bu/DMSO occurs via functionalization in C3-position of indole. 21 In contrast to that, phenol attacked in the C4-position. 22 For anilines the lowest relative energy of the carbanion is expected in C2position, 23 however, steric interaction may limit an attack in this position.
Overall the lowest energy path along the potential energy surface of the reaction can vary between different reagents and lead to significant changes in reactivity for the creation of E11* − defects. When the formation of E11* − is reduced, other functionalization processes (e.g., radical functionalization) may be favoured kinetically. This could lead to additional shoulders and sidebands in the PL spectrum and overall lower selectivity for one specific defect emission.
This concept is supported by the very similar PL spectra obtained for reactions with 2iodophenol and thiophenol, as they represent similar substance classes und are expected to follow similar reaction paths. Lastly, for anilines the introduction of E11* − defects can be controlled through the deprotonation equilibrium and thus KO t Bu concentration (see Figure 1b). This equilibrium depends on the pka values of the reagent.

Functionalization with Non-Aniline Reagents
Supplementary Figure 17. a-c, Normalized PL spectra of (6,5) SWNTs functionalized in the dark with indole (a), 2-iodophenol (b) or thiophenol (c) and 2 eq. of KO t Bu for 30 minutes in the dark in toluene/DMSO/THF. The concentration of indole, 2-iodophenol and thiophenol was kept at 29.30 mmol L -1 .

E11* − Optical Trap Depths of Functionalized (6,5) SWNTs
Supplementary Table 5. Summary of optical trap depths for E11* − defects obtained by reaction of (6,5) SWNTs with seven different reagents. Different functional groups did not yield significant changes in optical trap depth.

Supplementary Note 5: Impact of Oxygen/Water on Functionalization
When the reaction is performed under inert conditions a strong increase in the degree of functionalization is observed. This is expected as the functionalization process may be inhibited under atmospheric conditions due to multiple effects: (1) Quenching of the base by moisture can lead to a reduced formation of reactive carbanionic intermediates.
(2) Oxidation of negatively charged SWNT intermediates under regeneration of the carbon double bond.