Photon Antibunching in the Photoluminescence Spectra of a Single Carbon Nanotube

We report the first observation of photon antibunching in the photoluminescence from single carbon nanotubes. The emergence of a fast luminescence decay component under strong optical excitation indicates that Auger processes are partially responsible for inhibiting two-photon generation. Additionally, the presence of exciton localization at low temperatures ensures that nanotubes emit photons predominantly one by one. The fact that multiphoton emission probability can be smaller than 5% suggests that carbon nanotubes could be used as a source of single photons for applications in quantum cryptography.

Figure 1

Characterization of CoMoCat carbon nanotube samples by AFM (a), (b) and optical confocal microscopy (c). The topography image shows an isolated nanotube with a height profile of 0.8 nm (measured in the region indicated by the rectangle). The length of the nanotube is 800 nm. The length distribution (b) of nanotubes was obtained from the analysis of different samples with AFM. The mean nanotube length in our samples is $⟨L⟩=520\mathrm{nm}$. Local photoluminescence map (c) of an area of $3\times 5\mu {\mathrm{m}}^2$ recorded for two orthogonal linear polarizations (indicated by symbols) of the excitation laser. The false color image was acquired in a spectral bandwidth of 15 nm at 77 K by scanning the sample in discrete steps of 100 nm relative to the optical spot of 450 nm diameter. The two bright emission spots in the upper and lower graph originate from individual nanotubes with an emission wavelength of 885 and 872 nm, respectively.

Figure 2

Photoluminescence and photon correlation spectra of a single carbon nanotube at three different temperatures. The nanotube was excited with a Ti:sapphire laser through a phonon sideband 70 meV above the emission energy in cw mode and fs-pulsed mode for PL and photon correlation measurements, respectively. The left panel (a) shows the photoluminescence intensity as a function of wavelength for a given temperature. The right panel (b) depicts for each temperature the corresponding unnormalized correlation function $G^{(2)}(\tau )$ measured in a Hanbury-Brown–Twiss setup. The strongly inhibited correlation signal of the central peak at $\tau =0$ indicates the suppressed two (or more) photon emission probability [given as numbers of the normalized correlation function $g^{(2)}(0)$ in the right panel] and is a hallmark of nonclassical light.

Figure 3

Low-temperature time-resolved PL data recorded for pulsed excitation of a single CNT with PL peak emission at 860 nm. The PL intensity saturates with increasing excitation power, as plotted in (b). At low excitation powers, two decay processes with characteristic decay times of 35 and 250 ps are present. The average ratio of the amplitudes A1 and A2 associated with the strength of the 35 and 250 ps decay processes, respectively, is 74 and independent of pump power [open circles in (c)]. A third, fast decay component of 15 ps emerges at high pump powers; its amplitude A3 is shown as open diamonds in (c) normalized to the slowest decay amplitude A2 (the solid line is a guide to the eye). The corresponding bi- and tri-exponential decay fits to the time-resolved PL intensity (data offset for clarity) are shown as solid lines in a logarithmic plot (a); dashed lines represent best bi-exponential trial fits to the high power spectrum omitting the fast 15 ps decay component.

Figure 4

Time evolution of single carbon nanotube photoluminescence at 4.2 K for excitation powers (continuous-wave) of $30\mu \mathrm{W}$ (a, upper panel) and $3\mu \mathrm{W}$ (a, lower panel) are shown in gray-scale representation (1 sec integration time per spectrum). High excitation power causes spectral jitter of the luminescence between two states separated by $\sim 10\mathrm{meV}$ on a time scale of minutes. The jitter was absent for low excitation powers. Photoluminescence spectra (b, offset for clarity) and photon cross-correlation spectrum (c) of another single carbon nanotube. Starting with a single resonance (upper panel) the nanotube emission spectrum exhibited a splitting of 7 meV after one day of observation (central panel) and a further spectral shift of 6 and 1 meV for the higher and lower energy peak, respectively, after the second day (lower panel). The two emission peaks were spectrally selected with band-pass filters (10 nm bandwidth) for cross correlation (c) and autocorrelation measurements (not shown).