Electrically driven single-photon emission from an isolated single molecule

Electrically driven molecular light emitters are considered to be one of the promising candidates as single-photon sources. However, it is yet to be demonstrated that electrically driven single-photon emission can indeed be generated from an isolated single molecule notwithstanding fluorescence quenching and technical challenges. Here, we report such electrically driven single-photon emission from a well-defined single molecule located inside a precisely controlled nanocavity in a scanning tunneling microscope. The effective quenching suppression and nanocavity plasmonic enhancement allow us to achieve intense and stable single-molecule electroluminescence. Second-order photon correlation measurements reveal an evident photon antibunching dip with the single-photon purity down to g (2)(0) = 0.09, unambiguously confirming the single-photon emission nature of the single-molecule electroluminescence. Furthermore, we demonstrate an ultrahigh-density array of identical single-photon emitters.


Supplementary Note 1. Estimation of the total emission rate and quantum yield
The hemisphere photon collection efficiency for one optical detection channel is about 11%. The average transmittance of filters is around 95%. The detection efficiency of SPADs is about 37% around 1.90 eV. In the experiment shown in main-text Fig. 4d, the detected photon counts by two SPADs are 42 kHz and 34 kHz at an excitation current of 100 pA, respectively. Assuming an isotropic emission behaviour over the hemisphere, the total emission rate can be estimated to be 2 MHz.
Such emission rate corresponds to a quantum yield (i.e., an electron-to-photon conversion quantum efficiency) as high as 310 3 photons per electron, which is of the same order of magnitude as those reported in the literatures 1,2 . The high quantum yield greatly facilitates the demonstration of single-molecule single-photon emission.
The quantum yield is improved by the adoption of a series of strategies, such as the selection of decoupling layers and emitting molecules, the adoption of silver as both tip and substrate materials, and the fine tuning of the nanocavity plasmon resonance as well as the selection of the excitation position over the molecule. The decoupling layer is mainly used to prevent fluorescence quenching due to direct electron exchange between the molecular emitter and the metal substrate. The fine tuning of the nanocavity plasmon resonance is used to provide strong plasmonic enhancement to overcome the quenching caused by the dipole-dipole energy transfer between the molecular emitter and the metal electrodes. The excitation position is selected above the lobe of the flat-lying molecule to generate strong molecular emission thanks to the symmetry breaking 3,4 .  Figure 1, the electrical excitation of a molecule in a double-barrier junction requires the creation of a hole (k1) and then the capture of an electron (k2) to generate an exciton 5,6 , which is followed by the exciton decay (k3).

As shown in Supplementary
Following the method used in the literature 6,7 , the measured time constant τ0 from antibunching curves in a three-state model can be expressed as: where A=k1+k2+k3 and B=k1k2+k2k3+k3k1.
Assuming that the electron capture process is much faster than the other two processes (i.e., k2 >> k1, k3) as proposed in Ref. 6,Eq. (1) could be simplified as: The rate constant k1 is assumed to be proportional to the tunnelling current I and can be expressed as: where α is the exciton creation efficiency and e is the elementary charge. We would like to point out that the mechanisms of the excitation and decay process in the molecular electroluminescence here could be more complicated than the model

Supplementary Note 5. The exclusion of multi-electron excitation process
The multi-electron excitation process is unlikely to play a role in our experiment based on the following two considerations: (1) As far as we know, multi-electron excitation model is usually used to explain the energy up-conversion phenomena observed in tunnelling electron induced plasmon emission 15 and molecular fluorescence 16 , when the energy of a tunnelling electron is smaller than the energy of emitted photons. Nevertheless, in our experiment the excitation energy of the tunnelling electron (2.5 eV) is larger than the molecular optical bandgap and the associated photon energy of molecular fluorescence (~1.9 eV), and thus the molecule is most likely to be excited through one-electron excitation process.
(2) As shown in previous references, the photon yield per electron in multi-electron excitation process (10 -8~1 0 -7 photons per electron for plasmon emission 15 , ~1×10 -5 photon per electron for molecular emission 16 ) is reported to be much smaller than that in one-electron excitation process (~10 -4 photons per electron for plasmon emission 15 , ~3×10 -4 photons per electron for molecular emission 16 ). The optimized photon yield per electron of molecular fluorescence reported here (~3×10 -3 photons per electron) is much larger than the yield in the multi-electron process, but relatively close to that in the one-electron process.
Therefore, the excitation mechanism is believed to be dominated by the one-electron excitation process, rather than the multi-electron process.