The experimental scheme for laser assisted chlorination is illustrated in Fig. 1a. We utilized ultraviolet (UV) nanosecond laser beam (λ = 213 nm (5.8 eV)) which is aligned parallel to the sample surface under flowing of Cl2 gas. The Cl2 molecules can be photochemically dissociated by the focused UV pulsed laser and generated Cl radicals diffuse to graphene. First, we employed graphene field effect transistor device to explicitly determine the carrier density and mobility induced by laser assisted chlorination. The device was prepared based on mechanically exfoliated monolayer graphene on 20 nm thick high-k (εr ~ 20) HfO2 layer grown on Si wafer, to access ultra-highly doped state, whereas conventional SiO2 dielectric layer has limited access to doping concentration (~ 5\(\times\)1012 cm− 2) set by dielectric breakdown (Figure S1). We performed four-terminal measurement at high vacuum (< 10− 6 Torr) at room temperature. As doping time increased, its charge neutral point (CNP), VCNP, monotonically shifts to + 5.6 V, indicating ultra-high p-type doping concentration (p > 3\(\times\)1013 cm− 2). Meanwhile, the hole mobility (µH) decreased moderately from 4,698 cm2/V·s (pristine) to 2,551 cm2/V·s (doped state), attributed to the increased impurity scattering (Figs. 1b and c). We note that measurement of higher doping concentration beyond this point was limited by dielectric breakdown of HfO2 layer and therefore, the presented doping density is a lower bound limit. Despite the constraint, our result showed much higher doping concentration and higher mobility than liquid ionic gating26, Xe-lamp induced Cl doping24 and Cl plasma treatment20, while at high doping density regime, it presented charge mobility closer to the theoretical limit imposed by phonon scattering (Supplementary Note 2)27 compared to other state-of-the-art doping methods including Li-ion intercalation12 and electrolyte gating14 (Fig. 1d).
The presented high carrier density with high mobility implies that the laser assisted chlorination process fully takes virtue of non-invasive bonding characteristics of chlorination of graphene. To understand the chemical trend in doping, we carried out Raman and X-ray photoelectron spectroscopy (XPS) analysis based on mechanically exfoliated graphene monolayer on SiO2(300 nm)/Si wafer (Fig. 2). Raman spectra near G and 2D peaks (Fig. 2a) showed that both peaks significantly blue shifted from 1584 and 2678 cm− 1 to 1600 and 2689 cm− 1, respectively, while I2D/IG intensity ratio reduced from 2.8 to 1.0, indicating strong p-type doping effect28, 29 (Fig. 2a). Also, no sign of defects (e.g., D peak (~ 1350 cm− 1)) was observed as shown in full spectra in Figure S2. As described in Fig. 2b, the changes in Raman spectra showed monotonic shift and saturation after 5 min. We note that this trend contrasts previous reports on chlorination by plasma treatment, wherein the excessive processing time decreased the doping density20. We attribute this difference to the significantly low momentum energy of Cl radicals generated by laser, owing to the short mean free path of Cl atoms (~ 100 nm in the limit of ideal gas approximation at 400 Torr30), as well as the absence of external driving force such as electric field. UV nanosecond laser in parallel direction can excite Cl2 electronic energy state (ground 1Σg → excited 1Πu) and break Cl2 bonding31, 32 without affecting surface. The generated Cl radicals diffused to the graphene surface while experiencing significant momentum reduction.
XPS analysis also revealed clear evidence of high coverage of chlorine without structural damage (Fig. 2c). As presented in the C 1s XPS spectra of pristine graphene, the narrow sp2 C-C bonding peak (at 284.4 eV) was clearly observed.33 After chlorination process, C-Cl and C-Ox bonding states were evident at 286.5 and 288.3 eV, respectively33, without sign of sp3 C-C bonding formation at 285.4 eV. In addition, the Cl 2p peak was observed after doping. The C:Cl elemental ratio was estimated to be 43.5 % based on the intensity of C and Cl peaks (Figure S4), which is close to the highest reported experimental values20, 21. However, we note that the observed high Cl coverage ratio in graphene deviates from the theoretical prediction based on free standing graphene18: a maximum of 12.5 % Cl coverage ratio was predicted as stable stoichiometry in chlorinated graphene, when it forms ionic bonding by charge transfer complex. Higher coverage ratio exceeding this value would result in a weaker non-invasive bonding state, termed as ‘non-bonding’ where Cl element likely escape from graphene by forming Cl2 molecule, due to the weak interaction between Cl and graphene. Meanwhile, recent experimental study on chlorination of graphene by plasma doping suggested potential substrate effect by showing changes in maximum Cl coverage ratio depending on the type of substrate (dielectric or metal)21. This result implies that the substrate may affect the Cl-graphene interaction, which induced the observed unexpectedly high Cl coverage ratio.
Based on the suggested Cl binding mechanisms, we obtained the theoretically estimated doping concentration in graphene. The charge transfer rate between Cl and graphene is small in case of non-bonding state (0.03e per single Cl) due to the weak interaction18. Considering carbon density of graphene, (3.82 × 1015 cm− 2), in case of C2Cl ratio, the charge density, p can be as high as 5.73 × 1013 cm− 2, despite the small charge transfer rate. In case of ionic bonding (charge transfer complex), which can induce 12.5% of maximum Cl coverage ratio (C8Cl), the charge transfer rate is 0.27e per single Cl18, which can result in p = 1.27 × 1014 cm− 2. Thus, we expect that the combination of non-bonding and ionic bonding states of Cl can readily induce the experimentally observed high doping concentration.
The Cl dopant can be reversibly removed by photothermal process. We introduced CW green laser (λ = 532 nm (2.3 eV)) at normal direction with 2 µm (1/e2) of focal size (Fig. 3a). After Cl removal by laser (25 mW and 1 min), the G and 2D Raman peaks downshifted to 1588 and 2677 cm− 1, respectively, and the I2D/IG intensity ratio restored to 2.5 without D peak generation, corresponding to the pristine graphene state. By subsequent re-chlorination process, hole doped state was uniformly restored over the entire graphene sheet (Fig. 3b). Based on this procedure, arbitrary doped patterns can be reversibly formed without defect as demonstrated in Fig. 3f. To elucidate the effect of CW green laser in the Cl removal process, we carried out heat transfer simulation and Kelvin probe force microscopy (KPFM) mapping. KPFM mapping was performed to resolve the spatial distribution of changes in chemical potential of graphene due to the removal of Cl beyond the optical diffraction limit of Raman probing. From the simulated temperature profile (Fig. 3c), the full width half maximum (FWHM) of developed steady state temperature profile on graphene was estimated as ~ 1.2 µm (Supplementary Note 10). Meanwhile, KPFM images obtained from Cl-removed spot (Fig. 3d) showed a wider distribution (~ 1.7 µm, at 25 mW for 1 min) as depicted by plotting the line profile overlaid on the temperature distribution induced by the laser (Fig. 3e). Such trend suggests that the laser induced Cl desorption was dominated by thermal process. Photochemical routes typically tend to generate features smaller than the Gaussian profile laser beam focal spot size32, 34.
Such rewritable and highly doped patterns in graphene can demonstrate writing and erasure of photoactive junction in graphene-based photodetector as schematically depicted in Fig. 4a. We mapped the photocurrent response from the two-terminal graphene device fabricated on hBN/SiO2(300 nm)/Si substrate with Pd electrodes. A CW laser beam (532 nm) with 1 µm (1/e2) spot size at 100 µW power was raster scanned while collecting source drain current with zero bias voltage. The positive current indicates the excited hole moving toward source. As shown in Fig. 4b-i, pristine graphene showed significant photocurrent at metal/graphene junctions in opposite polarities due to the built-in potential developed by Fermi level pinning35. Furthermore, irregular photocurrent appears at the graphene channel area, attributed to the local electron-hole puddle or unexpected local doping during the electron beam lithography process. After chlorination process (Fig. 4b-ii), photocurrent at the channel area was eliminated as the high doping concentration overwhelmed any non-uniform doping fluctuation in graphene. Also, the distance between the photocurrent peaks appeared at two opposite metal/graphene junctions of the chlorinated device increased to 7.1 µm, whereas its pristine state showed 5.8 µm, as depicted in Fig. 4c. The high doping concentration rendered sharp band bending near electrodes and the peak position of the photocurrent shifted toward the metal35, 36. After local Cl-removal at the center of the channel area, the photocurrent clearly appeared in a symmetric distribution of opposite polarities (Fig. 4b-iii). The formation of p-p−-p junctions in the graphene channel resulted in opposite band bending as schematically described in Fig. 4d-iii.
We attribute the observed photocurrent in channel area to the photothermoelectric (PTE) effect, which originates from the local non-uniformity in Seebeck coefficient modulated by the gradient of density of states, represented by Mott formula9, 37. It has been reported that PTE effect at graphene unipolar junctions (i.e., p-p− or n-n− junctions) imposes photocurrent directions opposite to photovoltaic (PV) effect38, where the excited charges are drifted by built-in chemical potential. Considering the chemical potentials developed in p-p−-p junctions as illustrated in the energy band diagram in Fig. 4d, if the PV effect dominates the photocurrent, the photocurrent should flow in opposite directions to our experimental results, implying that the observed photocurrent at doped junctions is driven by PTE effect. On the other hand, at metal/graphene junction, the photocurrent mechanism cannot be differentiated solely by its flow direction as the photocurrents by both PV and PTE effects have the same direction39, 40. (Supplementary Note 13)
Next, the re-chlorination process can reversibly erase the photocurrent junction formed at channel area and the photodetector was ready for generation of a new photoactive pixel (Fig. 4b-iv). The line profiles of the photocurrent maps show the identical shapes in the chlorinated and re-chlorinated states of the photodetector (Fig. 4c). Thus, this result shows that the developed laser assisted reversible chlorination mechanism enables rewritable photoactive pixels in graphene photodetectors.