Selective Graphene Hydrogenation
- Selective hydrogenation of graphene is a process that attaches hydrogen to specific carbon sites, converting sp² to sp³ bonds and modulating its electronic structure.
- It enables spatial and chemical precision through one-sided, patterned, and curvature-driven strategies, yielding phased outcomes from semiconducting to spin-polarized states.
- Experimental and computational studies demonstrate reversible and tunable hydrogenation using techniques like plasma, electrochemistry, and double ionic gating.
Selective hydrogenation of graphene denotes the controlled attachment of hydrogen to only a chosen subset of carbon sites, lattice regions, faces, layers, or topological motifs in a graphene sheet or a graphene-derived two-dimensional carbon network. Rather than the fully saturated, symmetry-restoring limit of ideal graphane, selective hydrogenation encompasses one-sided full coverage, partial one-sided coverage, curvature-selected adsorption, moiré- or grain-boundary-localized adsorption, stripe- or ring-selective functionalization, and electrically driven layer-selective hydrogenation. Across these regimes, hydrogen converts local bonding to -like coordination, breaks -conjugation, reshapes the density of states near the Fermi level, and can drive outcomes ranging from indirect-gap semiconductors and wide-gap graphane-like phases to spin-polarized edge states, conductor-insulator transitions, anisotropic thermoelectricity, and proton-controlled logic and memory behavior (Pujari et al., 2011, Song et al., 2021, Tong et al., 2024, Tong et al., 31 Mar 2026).
1. Forms of selectivity and hydrogenation motifs
Selective hydrogenation appears in several structurally distinct forms. In the simplest spatial sense, hydrogen can be confined to one side of graphene. “Single-side-hydrogenated graphene” is a fully one-sided C:H phase in which all carbon atoms remain in one plane while hydrogens form a planar layer $1.08$ Å above the carbon plane; it is a non-magnetic semiconductor with an indirect bandgap of $1.89$ eV, a lattice parameter of $2.83$ Å, and C–C bonds elongated to $1.64$ Å, about larger than in graphene (Pujari et al., 2011). By contrast, graphane is the double-sided fully hydrogenated limit, whereas graphone is one-sided partial coverage, nominally about , and is associated in the cited literature with ferromagnetism and a small indirect gap of 0 eV in the unrelaxed geometry discussed there (Pujari et al., 2011).
Selectivity can also be defined in-plane. On Ru(0001), epitaxial graphene can be converted into a millimeter-scale honeycomb-patterned hydrogenated graphene in which graphane forms only over a self-assembled hydrogen template, leaving unhydrogenated graphene holes at the hcp regions of the moiré. The result is effectively a patterned graphane/graphene superlattice with a period of about 1 nm and graphene holes of about 2 nm (Song et al., 2021). In twisted bilayers, selectivity can target one of two atomically adjacent sheets: a strong transverse field produces a charge imbalance between electronically decoupled layers, and only the layer whose density crosses the monolayer hydrogenation threshold undergoes a conductor-insulator transition (Tong et al., 31 Mar 2026).
A further form is chemically or topologically selective hydrogenation of non-hexagonal graphene-derived allotropes. In tetra-penta-deca-hexagonal graphene, hydrogenation occurs mainly at tetragonal rings, creating transverse hydrogen stripes while leaving continuous hexagonal chains along the 3-direction. This ring-selective pattern generates quasi-one-dimensional conductive channels and a Dirac-like cone along 4 after hydrogenation (Oliveira et al., 2024). In Me-graphene, the most favorable initial adsorption site is a specific 5 site; once hydrogenated, it breaks the 6-7 bond and triggers island growth around neighboring 8 and 9 sites (Jr et al., 2020).
2. Microscopic determinants of site preference
The most general microscopic picture advanced in the literature is that graphene’s reactivity is not spatially uniform, even before hydrogen arrives. In an unrestricted broken-symmetry description, the 0-electron system of graphene carries a nonzero population of effectively unpaired electrons. The total number 1 is decomposed atomically as 2, where 3 serves as an atomic chemical susceptibility. Stepwise hydrogenation can therefore be formulated as an algorithmic process in which the carbon atom with the highest 4 is hydrogenated next; after each step the structure is reoptimized and the 5 map is recalculated (Sheka et al., 2011, Sheka et al., 2012). In finite graphene flakes this makes edge carbons the most reactive initially, and after edge saturation reactivity migrates inward.
Curvature is a second decisive selector. On corrugated monolayer graphene grown on SiC(0001), STM shows a quasi-6 landscape with hills and valleys, with a peak-to-peak corrugation of about 7–8 pm over about 9 nm. After exposure to atomic hydrogen, hydrogen-related protrusions are observed only on convex hilltops and never in concave valleys at room temperature (Goler et al., 2013). DFT and the associated reaction-coordinate analysis attribute this to local pyramidalization: convex curvature lowers the chemisorption barrier and increases binding energy, and after the first H adsorbate the barrier for the second H in an ortho configuration becomes effectively barrierless (Goler et al., 2013). This suggests a general rule that local convexity acts as a geometric selector for hydrogen uptake.
Moiré reconstruction and twist supply a third selector. In rotated graphene bilayers, the upper layer corrugates over the moiré, and carbons near almost-AA stacking are displaced out of plane more strongly than AB-like sites. Ab initio calculations find that these AA-like carbons are the preferred chemisorption sites for atomic H, with adsorption-energy differences as large as 0 meV across inequivalent sites. The preference strengthens as the twist angle decreases, and either electron or hole doping substantially increases the adsorption energy because the van Hove singularities move close to the Dirac point and enhance the density of states near the Fermi level (Brihuega et al., 2017).
Topological defects and grain boundaries modify the mechanism again. In polycrystalline graphene, hydrogen confined to interior grain-boundary sites behaves differently from hydrogen in grain interiors. The reason is not merely stronger binding at defect sites; the local non-hexagonal topology breaks the simple bipartite symmetry that otherwise produces the familiar hydrogen-induced resonant impurity state near the Dirac point. When adsorbates are confined to interior grain-boundary sites, the resonant peak is absent and transport is only weakly perturbed, whereas uniformly distributed hydrogen strongly degrades the mobility; at fixed 1 H coverage, the mobility differs roughly by a factor of 2 between homogeneous and inhomogeneous distributions (Vargas et al., 2018).
3. Routes to realization: atomic hydrogen, plasma, electrochemistry, and double gating
A wide experimental and computational toolbox has been brought to bear on selective hydrogenation. In ultra-high vacuum, atomic hydrogen can be delivered by a thermal cracker. A recent spectroscopic study used a FOCUS EFM-H source with 3 partial pressure 4 mbar and a capillary temperature of about 5 K, reaching doses up to 6 kL and observing either 7 or 8 9 saturation depending on the initial distortion of the suspended monolayer (Apponi et al., 14 Apr 2025). On corrugated graphene/SiC, a thermal hydrogen cracker delivered a flux of $1.08$0 H atoms/(s cm$1.08$1); exposures of $1.08$2, $1.08$3, and $1.08$4 s corresponded to coverages of about $1.08$5, $1.08$6, and $1.08$7, respectively (Goler et al., 2013).
Low-damage Ar/H$1.08$8 plasma hydrogenation provides a technologically convenient route. In a standard RIE system, graphene can be hydrogenated with $1.08$9 H$1.89$0/$1.89$1 Ar at $1.89$2 mbar, $1.89$3 sccm, $1.89$4 MHz, and $1.89$5 W, while the self-bias is tuned to $1.89$6. Under these conditions the estimated ion energy at the graphene surface is $1.89$7–$1.89$8 eV, dominated by $1.89$9, and the process is largely reversible by annealing while avoiding strong sputter damage in contacted devices (Wojtaszek et al., 2011). The same general plasma route, probed later by STM/STS, produces semiconducting patches with an average band gap of about $2.83$0 eV on graphite and CVD graphene/Ni, again reversible by moderate annealing (Castellanos-Gomez et al., 2012).
Electrochemistry supplies a different kind of selectivity because potential and time directly control the hydrogen chemical potential at the interface. In $2.83$1 $2.83$2, epitaxial graphene on SiC exhibits a hydrogen adsorption peak at about $2.83$3 mV vs NHE and a desorption peak at about $2.83$4 mV, implying a conversion potential about $2.83$5 V below hydrogen evolution. Integrating the current in a conversion run gives about one monolayer of hydrogen incorporated into five monolayers of graphene, and the process can be reversed by annealing at $2.83$6C (Daniels et al., 2010).
Double ionic gating extends electrochemical control into a genuine phase-space mapping of reactivity. In monolayer graphene, the gate sum controls carrier density while the gate difference controls the transverse field,
$2.83$7
with experimentally relevant values $2.83$8 V nm$2.83$9 and $1.64$0 cm$1.64$1 (Tong et al., 2024). In this geometry, proton transport can be accelerated toward the limiting electrolyte current while lattice hydrogenation is suppressed, or hydrogenation can be induced at $1.64$2 by pushing $1.64$3 above threshold (Tong et al., 2024). In twisted bilayers, the same logic becomes layer-selective because the field polarizes the two electronically decoupled layers differently, allowing only one to reach the monolayer hydrogenation threshold at fixed total $1.64$4 (Tong et al., 31 Mar 2026).
Bottom-up templating offers another route entirely. In graphene/Ru(0001), multiple cycles of atomic-hydrogen exposure at about $1.64$5C for $1.64$6 min followed by annealing at about $1.64$7C for $1.64$8 h produce first a monolayer-thick intercalated hydrogen network and then hydrogenation of graphene only above that template. Diffusion barriers below $1.64$9 eV across atop/bridge/fcc regions and about 0 eV toward hcp regions explain why the pattern self-organizes into a honeycomb superstructure (Song et al., 2021).
4. Structural and spectroscopic fingerprints
Selective hydrogenation is typically diagnosed through a convergent set of structural and spectroscopic markers. In core-level photoemission, the clearest signature is the growth of an 1 C 1s component. In highly hydrogenated suspended monolayer graphene, the 2 component remains centered near 3 eV, while the hydrogen-induced 4 component appears 5 eV higher in binding energy. The 6 fraction rises from 7 to 8 in the flatter sample and from 9 to 0 in the more distorted sample (Apponi et al., 14 Apr 2025). The broad 1 line shape suggests a superposition of local chemical environments, and the corresponding valence-band spectra indicate coexistence of one-side and two-side hydrogenation in the partially saturated case (Apponi et al., 14 Apr 2025).
Electron energy-loss spectroscopy provides complementary evidence. In the same monolayer system, the 2-plasmon is merely reduced when the sheet saturates at 3 4, but it is completely quenched in the fully hydrogenated sample, indicating the loss of extended 5 domains at the plasmon length scale. Analysis of the low-energy electronic onset through the linear region of 6 yields optical band gaps of 7 and 8 eV for the 9 and 00 01 samples, respectively (Apponi et al., 14 Apr 2025). High-resolution EELS also reveals the C–H stretching mode at about 02 meV, a direct vibrational fingerprint of covalent hydrogenation (Apponi et al., 14 Apr 2025).
On graphite and CVD graphene, scanning tunneling spectroscopy resolves the gap locally rather than in an area average. After Ar/H03 plasma exposure, the histogram of 04 values yields average band gaps of 05 eV for HOPG and 06 eV for CVD graphene on the first hydrogenation cycle; after annealing and rehydrogenation the corresponding values are 07 and 08 eV (Castellanos-Gomez et al., 2012). The same study shows that plasma-treated surfaces become topographically rougher in a partially irreversible way even when the electronic gap itself can be closed again by annealing (Castellanos-Gomez et al., 2012).
Raman spectroscopy remains the standard mesoscale probe. In electrochemically hydrogenated epitaxial graphene, the D peak rises sharply, the D/G ratio increases from about 09 to about 10, a new peak appears near 11 cm12 from C–H stretching, the 2D peak shifts from 13 to 14 cm15, and the resistance rises from about 16 k17 to more than 18 M19 (Daniels et al., 2010). In double-gated graphene devices, hydrogenation likewise produces a strong D band with 20 and a broadened 2D band, while LiTFSI control devices show no D band at any bias (Tong et al., 2024). In twisted bilayers, the difference between single-layer and double-layer hydrogenation is visible in Raman because the 2D band is reduced but still recognizable in the former case and becomes fully smeared in the latter (Tong et al., 31 Mar 2026).
Real-space microscopy exposes the geometry of selectivity directly. STM on graphene/SiC shows para dimers, ortho dimers, and tetramers located only on the convex tops of the quasi-21 hills (Goler et al., 2013). STM on graphene/Ru reveals the conversion from a triangular moiré into a bright honeycomb network surrounding dark hexagonal graphene holes elevated by about 22 Å relative to the pristine regions (Song et al., 2021).
5. Electronic, magnetic, transport, and thermoelectric consequences
The best-known electronic consequence of selective hydrogenation is band-gap engineering. One-sided full coverage produces an indirect gap of 23 eV in SSHGraphene, intermediate between graphene and graphane (Pujari et al., 2011). Moderate plasma hydrogenation of graphite and CVD graphene yields an average gap of about 24–25 eV (Castellanos-Gomez et al., 2012). Highly hydrogenated suspended monolayer graphene reaches optical gaps of 26–27 eV (Apponi et al., 14 Apr 2025). In Me-graphene, hydrogenation is explicitly non-monotonic: the gap evolves from 28 eV in pristine Me-graphene to metallic states at 29 and 30 coverage, a narrow indirect gap of 31 eV at 32, and 33 eV in fully hydrogenated Me-graphane (Jr et al., 2020). This suggests that selective hydrogenation is not merely a method for opening a gap, but a method for choosing among multiple electronic phases.
Spin physics depends strongly on pattern. In honeycomb-patterned hydrogenated graphene on Ru(0001), the unhydrogenated graphene holes bounded by graphane host spin-polarized zigzag-edge states. Spin-polarized DFT on the freestanding patterned lattice identifies an antiferromagnetic semiconductor ground state, with energy differences between trial spin configurations ranging from 34 to 35 meV per supercell and spin density localized along the zigzag edges (Song et al., 2021). By contrast, in polycrystalline graphene a common misconception does not hold: hydrogen does not necessarily create the same resonant impurity physics everywhere. When confined to interior grain-boundary sites, hydrogen-induced resonant states are inhibited and charge transport is weakly sensitive to functionalization, whereas uniform hydrogenation strongly degrades mobility (Vargas et al., 2018).
Selective hydrogenation can also control proton electrochemistry. In double-gated monolayer graphene, proton transport and lattice hydrogenation are decoupled in the 36 plane: large 37 at moderate 38 drives proton currents toward the limiting electrolyte current while graphene remains conductive, whereas high electron density 39 cm40 at 41 produces a four-order-of-magnitude conductor-insulator transition through hydrogenation (Tong et al., 2024). In twisted bilayers, a strong 42 polarizes the two electronically decoupled layers so that only one reaches the hydrogenation threshold; the resulting layer-selective insulating state is accompanied by proton transport through the bilayer and enables parallel NOT, NOR, NAND, and XOR logic functions in the same device (Tong et al., 31 Mar 2026).
Thermoelectric consequences are especially clear in ring-selectively hydrogenated tetra-penta-deca-hexagonal graphene. After hydrogenation, the Seebeck coefficient reaches about 43V/K near the Fermi level at 44 K in both directions, 45 rises to about 46 at 47 K and about 48 at 49 K, while 50 is reduced to about 51, giving 52–53. The power factor peaks at about 54 near 55 K along 56, whereas 57 is almost suppressed (Oliveira et al., 2024). This is noteworthy because, in graphene itself, hydrogenation is usually discussed as a route to reduced conductivity and enhanced scattering, while in this patterned non-hexagonal allotrope it enhances transport selectively along one crystallographic axis (Oliveira et al., 2024).
6. Reversibility, constraints, and unresolved issues
A recurring result across the literature is that the electronic effects of selective hydrogenation are often reversible even when the accompanying topography is not. In Ar/H58-plasma-treated graphene, moderate annealing between 59C and 60C progressively reduces 61, and for exposures shorter than 62 h the ratio falls below 63 after 64C (Wojtaszek et al., 2011). STM/STS measurements show that hydrogenation-induced gaps of about 65 eV disappear after annealing at 66C for 67 min, although the roughened topography is only partially restored (Castellanos-Gomez et al., 2012). In electrochemically hydrogenated epitaxial graphene, the C–H Raman mode disappears after annealing at 68C, and the resistance drops from above 69 M70 to about 71 k72 after 73 h, confirming dehydrogenation while leaving some residual disorder (Daniels et al., 2010). Double-gated devices go further: hydrogenation and dehydrogenation are electrically reversible and can be cycled more than 74 times with less than 75 variation in ON/OFF ratio (Tong et al., 2024).
Another point that the literature makes explicit is that idealized graphane or graphone are not generic end states. Graphane with regular chairlike packing forms only when the graphene membrane is fixed over its perimeter and accessible to atomic hydrogen from both sides; if either the membrane fixation or two-sided accessibility is absent, other morphologies such as tablelike, mixed chair/boat, rolled, corrugated, or amorphous graphane-like products are obtained (Sheka et al., 2011, Sheka et al., 2012). Likewise, one-side hydrogenation on different substrates does not generically produce a large ordered graphone domain: reactive molecular-dynamics simulations find that hydrogenation rates depend strongly on substrate and temperature, and that large hydrogenated domains are unlikely to form; instead, one obtains uncorrelated cluster domains (Woellner et al., 2016). This directly counters the common simplifying picture of a uniformly half-hydrogenated sheet.
Structural damage sets an upper limit on selectivity at high temperature or flux. In biphenylene carbon, hydrogenation at 76 K drives H/C incorporation to about 77–78 before the membrane collapses through large defective areas and structural holes, a behavior stated to be similar to what was previously observed for graphene (Splugues et al., 2017). Plasma studies also show that long Ar/H79 exposures eventually produce vacancy-dominated defects that do not anneal away at 80C (Wojtaszek et al., 2011). Even when the primary goal is band-gap engineering rather than etching, the distinction between reversible C–H formation and irreversible carbon removal must therefore be maintained experimentally.
The cumulative picture is that selective hydrogenation is best understood not as a single process but as a family of coupled chemical, electronic, and mechanical boundary-value problems. The decisive variables are the side of access, local curvature, defect topology, twist angle, moiré registry, substrate interaction, proton chemical potential, electric field, and charge density. This suggests that future control will come less from increasing hydrogen flux and more from engineering the lattice and electrostatic environment so that hydrogen binds only where it is wanted, only on the side or layer intended, and only long enough to deliver the desired phase.