Local plasmon engineering in doped graphene

Single atom B or N substitutional doping in single-layer suspended graphene, realised by low energy ion implantation, is shown to induce a dampening or enhancement of the characteristic interband π plasmon of graphene through a high-resolution electron energy loss spectroscopy study in the scanning transmission electron microscope. A relative 16% decrease or 20% increase in the π plasmon quality factor is attributed to the presence of a single substitutional B or N atom dopant respectively. This modification is in both cases shown to be relatively localised, with data suggesting the plasmonic response tailoring can no longer be detected within experimental uncertainties beyond a distance of approximately 1 nm from the dopant. Ab initio calculations confirm the trends observed experimentally. Our results directly confirm the possibility of tailoring the plasmonic properties of graphene in the ultraviolet waveband, at the atomic scale, a crucial step in the quest for utilising graphene’s properties towards the development of plasmonic and optoelectronic devices operating at ultraviolet frequencies. Due to its fascinating properties, graphene is emerging as a highly promising plasmonic material for implementation in devices aimed at applications such as chemical and molecular sensing, ultrafast optical modulation, non-linear optics, photo detection, light sources and quantum optics.1-6 In the terahertz (THz) to mid-infrared (mid-IR) spectral range graphene plasmons are associated with the collective excitation of free charge carriers and exhibit a higher degree of tunability and mode confinement, as well as longer propagation distances than noble metals.1-5 The graphene “charge carrier plasmon” frequency scales as ⁄ . , where EF is the Fermi energy and D is the size of the graphene sheet.1, 4 This means that the plasmon frequency can be tuned by either varying the Fermi level (e.g. through electrostatic gating4, 5, 7, 8 or chemical doping2, 4, 5), modifying the size of the graphene sheet4 (e.g. by making micro to nanoscale graphene ribbons4, 5 or discs3, 5, 9) or a combination of both. Using one or both these “methods” to push the charge carrier plasmon frequency into the near-IR to visible spectrum is of significant current interest in the community.1, 3, 4, 10 Existing graphene-based IR plasmonic devices already exhibit promising properties, such as gate tuneable switching and control of the plasmon wavelength as well as a 40-60 times reduction in plasmon wavelength (as compared to the incident IR illumination).7, 8 While these plasmons can propagate a distance on the order of a few times their own wavelength,7, 8 on par with measurements of plasmons in Au,8, 11 this falls short of that expected for high purity graphene.7, 8 This has been attributed to disorder.12 Indeed, the graphene charge carrier plasmon mobility is expected to decrease to various degrees depending on type and concentration of dopants12, 13 and other defects,14 as well as the specific edge structures of nanoscale ribbons and similar nanoscale geometries.4 Possible strategies for realisation of graphene based plasmonics in the near-IR to visible spectrum, while taking into consideration the above effects (among others), are discussed in Ref. 4 At higher spectral frequencies, in the ultraviolet (UV) range, graphene exhibits interband plasmons resonances attributed to the collective oscillation of π and σ valence electrons.15- 17 These interband plasmons show a remarkable degree of sensitivity to various nano- to atomic-scale structures and defects in graphene: interband plasmon localisation has been attributed to confinement induced by edge states of a ~1.3 nm graphene quantum disc18 and single substitutional Si atoms have been associated with a highly localised enhancement of the interband plasmon response.19 In periodically rippled graphene (on a Ru(0001) surface) the interband π plasmon is confined to ripple “hills” while being significantly dampened in ripple “valleys”.20 Admittedly showing a more limited tunability compared to the charge carrier plasmon,4 the interband π plasmon frequency is nonetheless predicted to progressively red-shift with increasing graphene nanodisc diameter, being the most sensitive to disc diameters below 20 nm.21 However, with the exception of the above studies, reports on other aspects of the interband plasmon response of graphene are lacking in the literature. Such knowledge might open up avenues for future implementation of graphene based plasmonic and optoelectronic devices operating in the UV waveband. With this goal in mind, the present work investigates the modification of the interband plasmon response of graphene associated with two key substitutional dopants, namely boron and nitrogen atoms. The inclusion of B or N atoms in the graphene lattice is the focus of extensive study in the scientific community, with the aim to modify the electronic structure of graphene.22-29 Substitutional B and N atoms have been predicted to induce a shift of the Fermi level,26, 29, 30 resulting in p or n doping akin to that routinely exploited in current semi-conductor technology. Indeed a p and n character has recently been verified in suspended graphene containing single substitutional B and N atom dopants.29 Under certain circumstances, boron and nitrogen doping is also expected to induce a band gap in graphene.24, 25. Key to these proof-of-principle studies, electron energy loss spectroscopy (EELS) in combination with high angle annular dark field (HAADF) imaging in the aberration corrected scanning transmission electron microscope (STEM) are considered particularly useful techniques for identifying individual nano to atomic scale defects in a material and the associated effects on its electronic structure and dielectric response.31 Due to the ideal ‘gentle STEM’32 combination of ultra-high vacuum conditions and low acceleration voltage (which minimises any beam-induced damage to the samples), individual B23, 29 and N23, 29, 33, 34 atom dopants in graphene can be identified directly in an ADF image. So-called “core” EEL spectra (EEL >≈ 50 eV) contain information about the local electronic structure and bonding in graphene,28, 29, 33-35 while “valence” EEL spectra (EEL <≈ 50 eV) contain information about the graphene dielectric response18, 19, 36-41. In combination with simultaneous (STEM) ADF imaging, EEL spectra allow for a direct correlation of defect-induced modifications of the graphene electronic structure28, 29, 33-35 and dielectric response18, 19 with atomic scale structure. These capabilities mean STEM-EELS is an excellent technique for investigating the interband plasmon response induced by individual B and N atom dopants in graphene, as this information can be correlated directly with the atomic structure, all within the same experiment. In the valence loss spectrum of graphene the so-called “π peak” (~ 5 eV) is attributed primarily to the excitation of the π interband plasmon,37, 39-43 superimposed on a sum over π-π* interband transitions.37, 39-42 A recent controversy concerning the interpretation of the π peak41, 44 was addressed in subsequent theoretical studies which show that (for STEMEELS) the π loss peak is most appropriately interpreted as predominantly due to the excitation of the interband π plasmon of both doped17 and dopant free15, 16 graphene. In following with this, we use STEM-EELS to show that a single B or N substitutional atom dopant induces dampening or enhancement of the graphene interband π plasmon, respectively, with an estimated relative 16% decrease (B) or 20% increase (N) in quality factor, and, a ~1 nm localisation in both cases. This effect was observed independently using two separate STEM-EELS systems (with different yet complementary experimental parameters) and is significantly more pronounced than that previously reported for single Si atoms.19 Ab initio calculations were carried out on the largest supercells possible whilst keeping the computational costs tractable to validate the experimental results. While our theoretical spectra broadly reproduce the trends observed experimentally, it is suggested that the supercell sizes used in the present work (in practice limited by associated computational costs) are simply too small to accurately predict the relevant properties of the experimentally probed systems and faithfully reproduce all the details of the loss function.Nevertheless, our combination of state-of-the-art experimental and theoretical results demonstrates that the plasmonic properties of graphene can be tailored at the atomic scale, using an implantation technique already extensively used in semi-conductor industry.23, 29 Atomic scale plasmon engineering of graphene might prove valuable in the quest for utilising graphene’s properties towards the development of plasmonic and optoelectronic devices operating in the UV waveband. Indeed, recent reports propose utilising the π interband plasmon response associated with graphene nanopores as a sensing mechanism for DNA nucleotides.45, 46