In conventional semiconductors, it is well known that a built-in electric field is essential for separating photoexcited electrons from holes to generate photocurrent. However, for the zero-bandgap semiconductor graphene, this concept was challenged when scientists found that the Seebeck effect gives rise to a photocurrent in a graphene p–n junction after the local electron temperature is increased by photoexcitation1. From then on, the two above effects, known as the photovoltaic and photothermoelectric effects respectively, have been dictating the design of graphene photodetectors for applications such as high-bandwidth optical communications. Now, writing in Nature Nanotechnology, Qiong Ma et al. have found another mechanism for photocurrent generation in intrinsic graphene2. Unlike the two well-known effects, which usually become prominent in graphene at a high doping level, this new effect only dominates when graphene is completely undoped.

The new mechanism originates in the asymmetry of electron–hole velocities in graphene. For example, electrons and holes have an intrinsic velocity difference of about 104 m s−1 when they are 0.1 eV away from the Dirac point (the point at which conduction and valence bands touch). This asymmetry implies that whenever electrons and holes are excited by photons, they will diffuse at different speeds, separate, and produce a local electric field naturally. Yet, as the diffusion has no directional preference, the local electric fields are circularly symmetric around the excitation point and cancel each other out. As a result, there is no total electric field to drive current in the electrical circuit. However, as shown in Fig. 1a, when light is shone on the edge of the graphene, the diffusion is no longer circularly symmetric, and this gives rise to a residual electric field that produces photocurrent in the circuit. Therefore, graphene with well-designed edges can serve as a photocurrent source.

Fig. 1: Photocurrent generation from velocity asymmetry of electron and hole pairs in intrinsic graphene.
figure 1

a, Shining a light at the edge of graphene excites photoelectrons and holes. The difference in diffusion velocity generates a local electric field, which drives a photocurrent in the electrical circuit. b, In undoped graphene, the initial photoelectron relaxes (indicated by the top arrow) in the first few hundred femtoseconds by exciting another electron from the valence band to the conduction band (known as interband transition and indicated by the bottom arrow). The electron–electron scattering is confined to the same axis in momentum space to satisfy energy and momentum conservation. This collinear process maintains the direction of the initial wave vector, conserving the diffusion direction of electrons and the total photocurrent. The purple and blue balls represent electrons before and after scattering. c, Top view of b. All the wave vectors have to be confined to one straight line. d, In doped graphene, the same relaxation process is dominated by intraband transitions in the conduction band. The initial photoelectron can change the value and direction of its wave vector (top arrow) by exciting another electron from the Fermi sea (bottom arrow). This means that the direction of initial carrier diffusion is altered. The process relaxes the total photocurrent more easily than interband transitions in undoped graphene. e, Top view of d. The straight black arrows show the wave vectors. The wave vector summation of the initial carriers (blue balls) matches the summation of the final carriers as shown by dotted lines, owing to momentum conservation.

Full size image

A key feature of the new mechanism is that it dominates only in undoped graphene. This is due to the electron–electron scattering dynamics: this scattering occurs within the first few hundred femtoseconds after photoexcitation3,4 and is followed by electron–phonon scattering which subsequently thermalizes the whole system. In undoped graphene, electron–electron scattering is dominated by interband excitations. An initially photoexcited electron could relax by exciting another electron (or many electrons) in the Fermi sea, creating an additional electron and hole in the conduction and valence bands, respectively. Owing to energy and momentum conservation, the initial and final wave vectors of these carriers are confined to the same axis, as shown in Fig. 1b and c. Such a collinear process maintains the direction of the initial wave vector, which means that the initial photocarriers remain at the same diffusion speeds and directions in real space. Therefore, because of the linear band structure of graphene, the unique electron–electron scattering in undoped graphene conserves the photocurrent and guarantees an effective photodetection5. In sharp contrast to undoped graphene, in doped graphene the scattering is dominated by intraband transitions, as shown in Fig. 1d and e. In intraband scattering, the wave vector of the initial photoelectron can change its direction and still satisfy energy and momentum conservation. This is because the other electron in the Fermi sea can change its momentum to compensate this direction change, as shown by arrows in Fig. 1d and e. This scattering means that the direction of initial carrier diffusion is altered, and the total current is not conserved.

This new mechanism may open possibilities for the design of graphene-based photodetectors, as it does not require any junction as a photocurrent source. In the photovoltaic and photothermoelectric effects, metal–graphene or graphene p–n junctions are essential for electron–hole separation. Devices based on these effects may sometimes suffer from complicated fabrication processes. Luckily, with the new mechanism only the geometric shape of graphene is relevant to photocurrent generation, which will greatly simplify the fabrication process. Additionally, the detector can be totally turned off by gate doping, which gives another degree of freedom in controlling the photodetection.