## Introduction

Hexagonal boron nitride (abbreviated as h-BN in the following) is a remarkable two-dimensional mesh consisting of alternating sp2-bonded boron and nitrogen atoms1. As a wide bandgap insulator2, h-BN is regarded as an ideal substrate and tunneling barrier for graphene devices because of its atomically smooth surface free of dangling bonds and charge traps3. In addition, the exceptional thermal and chemical stabilities of h-BN enable it to function as an antioxidation layer, even at temperatures as high as 1100 °C4. Monolayer h-BN sheets remains stable up to 800 °C in air, whereas graphene oxidizes at temperature above 500 °C5. This makes h-BN a top choice for ultrathin antioxidization coatings. In addition to the potential for application in electronics and corrosion-proof coatings, h-BN holds promise for selective membrane applications under extreme conditions due to its ultimate thinness, flexibility, and mechanical strength4,6,7,8.

Similar to graphene9, h-BN is elastic and impermeable to all gases. High-quality h-BN sheets have a Young’s modulus of ~0.85 TPa and an excellent fracture strength of ~70 GPa8. The sheets can withstand stretching of up to 20%8. This feature makes h-BN a superb material for application in micro-electromechanical systems (MEMSs). A perfect h-BN crystal, even when just a monolayer thick, is typically impermeable to most atoms and molecules under ambient conditions. Only accelerated hydrogen atoms possess sufficient kinetic energy to penetrate through h-BN planes without damaging the lattices10,11. These properties allow bubbles with various shapes to be produced on h-BN12. Recently, several reports have described the trapping of gases by graphene12,13,14,15,16. By forming bubble structures, the adhesion energy between graphene sheets and substrate can be precisely measured17,18. A voltage-controlled graphene micro-lens with variable focuses19 and a laser-driven nano-engine20 have been realized, as the curvature and shape of the bubbles can be controlled to a great extent by variation in the voltage. Although h-BN has a similar honeycomb lattice to graphene, little investigation of h-BN bubbles for gas isolation has been reported. Compared to graphene, h-BN is insulating and more chemically stable, which may greatly extend its application in extreme conditions.

There have been several reports about plasma treatment on 2D materials. Most of them are about plasma applications in surface etching21,22, low-temperature growth23,24 and surface hydrogenation25. In this work, we demonstrate the isolation of hydrogen in h-BN bubbles on the microscale via plasma treatment. The dimension of the bubbles can be regulated by altering the treatment conditions. The diameter of the bubbles ranges from tens of nanometers to several micrometers. The masses of H2 molecules inside the bubbles are believed to have been converted to atomic hydrogen by the plasma, which permeate through the h-BN plane and is then captured inside the h-BN interlayers. Re-formation of H2 molecules deforms the h-BN structure to form bubbles, as the H2 molecules cannot escape from the h-BN net. Our demonstration provides an effective method for fabricating bubbles on h-BN and may also be an approach for extracting/storing hydrogen in h-BN.

## Results

### Bubble formation on h-BN

Figure 1a shows a schematic illustration of the experimental design for the bombardment of the h-BN surface by plasma in different gaseous environments. As shown in Fig. 1, three different gases (Ar, O2, and H2) were fed into the chamber to examine the ability of different elements to penetrate h-BN. The flow rate was kept at 3 sccm, while the pressure was kept at ~3 Pa. Plasmas were then generated by a RF generator with a power of 100 W. All h-BN flakes were treated with the plasma at 350 °C for 150 min. Atomic force microscopy (AFM) images of the h-BN flakes were taken after the plasma treatment and are given in Fig. 1b–d (the corresponding optical images are given in Supplementary Fig. 1). As shown in Fig. 1b–d, the h-BN flakes treated with argon or oxygen plasma appear to be intact, while a large amount of bubbles appeared on the h-BN surface that was treated with the hydrogen plasma (H-plasma). A series of experiments including structural characterization by transmission electron microscopy (TEM) (see Supplementary Figs. 2 and 3), chemical characterization by X-ray photoelectron spectroscopy (XPS) (see Supplementary Figs. 4 and 5 and Supplementary Table 1), Energy-dispersive X-ray spectroscopy (EDX) (see Supplementary Figs. 6 and 7), Fourier Transform infrared spectroscopy (FTIR) (see Supplementary Fig. 8) and mass spectrometry (see Supplementary Figs. 9 and 10) have been conducted to verify the h-BN bubble structure intact and the presence of hydrogen gases inside the bubbles.

To understand the mechanism behind the interesting bubble formation is important. It is well known that gaseous molecules subjected to a strong electromagnetic field decompose into a plasma state, producing masses of charged ions whose particle radii are considerably decreased relative to the original atoms. The hexagonal honeycomb lattice of h-BN possesses a lattice constant (ah-BN) of ~250.4 pm26. A hydrogen atom has a diameter of ~240 pm (with a van der Waals radius of ~120 pm27,28), which is less than the value of ah-BN and could pass through the valence electrons concentrated around the BN atoms. Moreover, large amounts of protons, which have a diameter of ~1.68 fm29,30, are generated in the plasma state. These extremely small particles can easily penetrate the BN. The accelerated energetic protons and electrons could then recombine to form hydrogen molecules. These hydrogen molecules are not easily able to pass through the h-BN honeycomb due to their large kinetic diameter of ~289 pm31, which is substantially larger than ah-BN. Thus, the pressure from the trapped gas causes bubbles to form on the surface of the h-BN flakes. In contrast, the atomic diameters of both argon (~376 pm) and oxygen (~304 pm) (with the van der Waals radii of ~188 pm and ~152 pm, respectively27,28) are considerably larger than ah-BN, and as a result, they cannot penetrate the h-BN honeycomb barrier.

The mechanism of unimpeded penetration of atomic hydrogen into the h-BN interlayers can also be ascribed to two additional reasons. The first one is the polarization of the electron density distribution in this kind of 2D material. Recent experiments have demonstrated that h-BN possesses the best porosity among 2D materials11. The concentration of valence electrons around the N atoms results in strong polarization of the BN bonds, making h-BN much more porous than graphene and MoS2. The second reason is the stacking order. It has previously been confirmed in the former work that bi- and tri-layered graphene are also nonconductive to protons due to their staggered lattices (AB stacked), in which the electron cloud of the first layer clogs the pores of the second11. In contrast, multilayered h-BN exhibit excellent proton conductivity probably due to its favoring the AA′ stacking structure, which has been found to be the most stable one among the five possible stacking structures (AA′/AA/AB/AB′/A′B) in recent calculations32. The aligned pores in the out-of-plane direction also allow protons to move unimpededly throughout the space. This characteristic structure of multilayer h-BN could facilitate its use in hydrogen storage applications.

To study the stacking order of the h-BN multilayer, a scanning transmission electron microscopy (STEM) investigation was carried out. The h-BN flakes are firstly exfoliated onto a quartz substrate, and then annealed to remove the residue of resist in an oxygen flow at 800 °C. After cooling down to room temperature, a group of wrinkles are found on the surfaces. The wrinkles, which are found along the armchair direction of h-BN crystals33,34, help us determine the cutting direction of the focused ion beam. The specimen obtained for the STEM investigation is similar to that shown in Supplementary Fig. 3a.

Figure 2a illustrates that the incident electron beam (e-beam) for the STEM investigation is parallel with the $$[11\bar 20]$$ crystallographic axis of h-BN. The cross-sectional schematic of the armchair edge-on view of h-BN multilayer is presented in Fig. 2b. Figure 2c is a Wiener filtered STEM medium angel annular-dark-field (MAADF) image showing the edge-on view configuration of h-BN multilayer, which reveals the layer stacking of the h-BN. The intensity profiles along two selected lines (red/blue) in Fig. 2c are showed in Fig. 2d, exhibiting apparent synchronism of the peak positions. Besides, the average peak-to-peak distances measured in red and blue profiles are both ~2.26 Å, which are close to 2.166 Å expected in h-BN. The minor deviation of the spacing in STEM measurement might be due to the limitation of the scale calibration of the instrument. This result shows that the stacking order of our h-BN is either AA′ or AA in our h-BN multilayer, which is in line with earlier predictions. Beside the cross-sectional STEM image, we also carried out a plan-view STEM investigation (shown in Supplementary Fig. 11) of another h-BN multilayer. The image confirms that the h-BN multilayer we fabricated is in AA′ stacking. These images provide direct evidence that the multilayered planes in h-BN are not staggered so that the electron cloud of each layer does not block the pores of the successive layer. As such, atomic hydrogen could tunnel through multilayered h-BN.

### Control on size of h-BN bubbles

Both the duration and the environmental temperature of plasma treatment have prominent effects on the density and sizes of bubbles, which provide a feasible way to fabricate h-BN bubbles with controllable dimensions. To understand the dependence of the bubble dimensions on the duration of the H-plasma treatment, we varied the treatment duration while holding the other conditions constant (100 W RF, 350 °C, 3 sccm H2, ~3 Pa) and examined the variations in the dimensions of the generated bubbles. The results of this experiment are given in Fig. 3. As shown in Fig. 3a, the h-BN samples obtained in different duration of H-plasma treatment possessed bubbles with obviously different densities and size distributions. The bubbles treated for 90 min were generally smaller than 100 nm in diameter and had a higher density in the distribution, while plasma treatment for 150 min led to a lower bubble density and a larger bubble size with diameters greater than 3 μm. The dependence of diameter distribution on treatment duration is plotted in Fig. 3b. Extended plasma treatment favors the formation of larger bubbles. An H-plasma treatment for 500 min could fabricate larger bubbles. As shown in Supplementary Fig. 12a–c, bubbles with diameters of ~20 μm were obtained on h-BN.

In addition to the treatment duration, the environmental/sample temperature is another key factor that influences the dimensions of the bubbles. Thus, the sample temperature was also adjusted while keeping the other conditions constant (100 W RF, 120 min, 3 sccm H2, ~3 Pa) to investigate the variations in the bubble dimensions. Figure 3c presents the density and diameter fractions of the h-BN bubbles for different sample temperatures. Almost all of the bubbles produced at room temperature (30 °C) were <100 nm in diameter, and the relative fraction of large bubbles gradually increased when the sample temperature was elevated from 30 to 350 °C. Meanwhile, the density of the bubbles dramatically decreased with increasing sample temperature. Furthermore, the bubble shown at the bottom-left of Supplementary Fig. 13d formed by merging three individual bubbles. The observation reveals that the bubbles can migrate and merge at higher sample temperature. Similar phenomena were also observed in Fig. 3a and Supplementary Fig. 14. The hydrogen molecules inside smaller bubbles, normally possessing higher inner pressure than relatively large bubbles, move more easily at higher sample temperatures in-between h-BN interlayers15. Hydrogen gas in small bubbles is more likely to pass through the inter-planar channels and combine to form a larger bubble with a decreased inner pressure. Such combination leads to a dramatic decrease in bubble density at high sample temperatures. Detailed investigation in the stability of bubbles on h-BN is given in Supplementary Figs. 1520.

We then further explored the relationship between the height/aspect ratio of the bubbles and their diameter. The maximum height, hmax, and the quotient of hmax to the radius R (called the aspect ratio) of each bubble are taken from all observed h-BN bubbles by AFM. The results are presented in Fig. 3d. It is found that hmax approximately increases linearly with bubble diameter, while the aspect ratio remains almost unchanged.

### Mechanisms of the plasma-driven effect

In order to understand the mechanism behind the plasma process, exploring hydrogen plasma itself is essential35. We carried out modeling of hydrogen plasma in a tube using COMSOL Multiphysics. The spatial distribution of different parameters in the tube are plotted in Supplementary Fig. 21. As shown in Supplementary Fig. 21e, f, k, l, the ion density at the sample area decreases while the thermal velocity increases, when the sample was heated from room temperature to 350 °C. The results indicate that protons in H-plasma can be injected much faster into h-BN interlayers at a higher environmental temperature with comparatively lower spatial density. This simulation exhibits the variation tendency of dimension/density when elevating the environmental temperature (heated by the furnace) of the sample (as shown in Fig. 3c). The furnace temperature (Supplementary Fig. 21a, b), electron density (Supplementary Fig. 21c, d), electron temperature (Supplementary Fig. 21g, h), and pressure (Supplementary Fig. 21i, j) are also given to help in giving images shown plasma-driven kinetics.

It is noticed that the electron density (Supplementary Fig. 21c, d) is two orders in magnitude lower than the ion density (Supplementary Fig. 21e, f) at the sample position according to simulation. This phenomenon indicates that the quasi-neutrality of plasma at the sample position is broken. According to the setup showed in Supplementary Fig. 22a–d, h-BN sample was placed 30 cm away from the RF coil (the core region of the plasma). Along the axial direction of tube furnace, there is a steep electron gradient which leads to the unbalance between electron density and ion density, and thus, the quasi-neutrality of plasma is not sustained. To understand this phenomenon further, we then carried out modeling in Debye length when the furnace was set at 350 °C in Supplementary Fig. 21q. Normally, Debye length depends on density of ion number and environmental temperature. Debye length increases when ion number density decreases or when the environmental temperature increases. It is also found that the Debye length, which is much higher than the diameter of the tube (0.042 m), reached its maximum of ~0.5 m at the sample position. It indicates the destruction of quasi-neutrality at the sample position. This also interprets why the electron density is two orders lower than the ion density at the sample position.

Besides the sample temperature and treatment duration, the position where the h-BN sample was placed also brings different size/distribution of bubble on h-BN. The schematic of experimental setup as well as the results of bubble distribution are exhibited in Supplementary Figs. 23 and  24, respectively. It is obvious that bubbles in the plasma core area are small but dense. while bubbles at other area show relatively large but thin. The electric potential difference (Supplementary Fig. 21m, n) and the ion number density distribution (Supplementary Fig. 21o, p) are also simulated by COMSOL Multiphysics to analyze the bubble distributions in the supporting information. Obviously, the formation of bubbles is a plasma-driven effect. The size and distribution of bubbles on h-BN are generally determined by the comprehensive influence of ion density, Debye length and potential gradient at the position where the h-BN samples are placed.

### Low-temperature AFM measurement

To determine the type of gases inside the h-BN bubbles, low-temperature AFM measurements were carried out to explore their temperature evolution. The experiment was inspired by a very recent literature about bubble structure on bulk transition metal dichalcogenides36. Details about measurement are given in the Methods section. Figure 4a shows an optical image of bubbles on an h-BN flake. After transferring it into a vacuum chamber equipped with an AFM, the sample was cooled down gradually. Figure 4b shows topographic AFM images of the same area measured at 34 and 33 K respectively. It is clearly shown that the bubbles deflate at 33 K while they are inflated with gases at 34 K. As shown in Fig. 4c, the height profiles of line-scans across a bubble (indicated by dashed lines in Fig. 4b) clearly show that the bubble remains at ~34 K and disappears at ~33 K. The deflating/swelling processes were reproducible via cooling/heating the sample. The highest temperature at which the bubbles become flat was recognized as the transition temperature (Ttransition). We measured a total of 58 bubbles. Three other typical samples are given in Supplementary Figs. 2527. The transition temperature of each bubble was recorded when they become flat during cooling. The number of bubbles which have the same Ttransition is counted. The histogram of Ttransition as function of temperature at an interval of 1 K is plotted in Fig. 4d. As shown in Fig. 4d, Ttransition of all the bubbles measured can be described as a Gaussian distribution, and exhibits an average value equal to 33.2 ± 3.9 K. This value is very close to Ttransition of H2 (33.18 K) among all possible gases (see Supplementary Table 2). The result strongly indicates that the gases inside the bubbles are indeed hydrogen molecules.

### Determining the penetration depth of atomic hydrogen in the h-BN multilayer

Increasing the power of the RF source may lead to bursting of the bubbles, which leaves many pits on the h-BN surface. These pits make it possible to study the penetration depth of atomic hydrogen. Figure 5a shows an optical micrograph of an h-BN flake treated for 180 min at 350 °C with an RF power of 400 W. Numerous pits are observed on the h-BN surface. Figure 5b, c depict the AFM image of a deep pit highlighted by the red box in Fig. 5a. The AFM height image in Fig. 5b provides depth information about the pit, whose depth is given in the insert. The amplitude error image provided in Fig. 5c shows a clearer picture of the morphology around the pit. The bottom of the pit is smooth, which indicates that no atomic hydrogen penetrated the bottom layer. Furthermore, folds are observed in the h-BN flaps around the outside of the pit, as illustrated in Fig. 5d. The inset in Fig. 5b shows that the rupture depth of the folded border step is approximately 15.22 nm, which indicates that ~23 layers of h-BN were penetrated by atomic hydrogen in the 400 W treatment. Figure 5e presents the statistics of the penetration depth of atomic hydrogen in the h-BN sample shown in Fig. 5a. The penetration depth of the h-BN flake ranges from ~11 to ~23 layers. The histogram shown in Fig. 5e reveals that the 400 W RF power provided the H-plasma with enough kinetic energy to penetrate at most ~23 atomic layers underneath the h-BN surface.

### Characterizations of Raman spectroscopy

Raman spectroscopy was also performed to characterize the h-BN bubble. As shown in Fig. 6a, a typical bubble with a round shape was measured. The bubble had a diameter of ~2.36 μm and a height of ~122 nm. The Raman spectrum was measured at three spots from the flat area next to the bubble to the center of the bubble. The results are shown in Fig. 6b. The Raman spectra of h-BN clearly show only one peak near 1366 cm–1, which corresponds to the E2g vibration mode of h-BN. As shown in the inset of Fig. 6b, the E2g peak position and full width at half maximum (FWHM) of each spectrum were extracted and plotted as a function of the measurement site. The peak position in the Raman spectrum appeared at the center of the bubble redshifted (~3 cm–1) relative to that of the position outside of the bubble, with almost no change in the FWHM. The observed redshift in the peak position is due to the presence of stretching strain on the bubble caused by expansion37,38. Further investigation in the temperature dependence of Raman spectra of h-BN bubble is given in Supplementary Fig. 28.

### Hydrogen extraction

The demonstrated fabrication of hydrogen bubbles on h-BN via plasma treatment inspired us to consider the possibility of extracting hydrogen from hydrocarbons and mixtures of gases. Three different gases (C2H2, CH4 and Ar + H2, 5% of H2) were fed into the chamber for plasma treatment under the same experimental conditions (350 °C, 100 W, 120 min, ~3 sccm flow rate, ~3 Pa). Excitingly, all three samples formed bubble structures on the h-BN surface after the plasma treatment. The experimental illustration and AFM images of the h-BN flakes are given in Fig. 7a–f. Considering that the diameters of carbon and argon atoms are much larger than the lattice constant of h-BN, it is likely that only atomic hydrogen could penetrate the h-BN layers. The results indicate that hydrogen could be separated from hydrocarbons and from the mixture of argon and hydrogen.

As shown in Fig. 7d–f, the bubbles formed after the ethyne plasma treatment had diameters of predominately <300 nm and a high density, while the bubbles produced by the methane plasma possessed a much larger diameter range from 2 to 4 μm with a much lower density. The exact causes for the phenomena need comprehensive investigation in future. At this moment, the main reason, we believe, could be that the dissociation energy of methane is actually lower than that of the ethyne39. When all experiments were carried out under similar conditions, more atomic hydrogen in CH4 plasma can be generated and then more efficiently get into h-BN bubbles than the case in ethyne plasma. The plasma treatment with 5% hydrogen in argon also led to the formation of bubbles on the h-BN surface but with an extremely low density. To characterize the bubbles further, the plots of hmax and the aspect ratio with respect to the bubble diameter are presented in Fig. 7g–i. Similar to the bubbles fabricated by the plasma of pure hydrogen, in all three plots, hmax had a nearly linear increasing relationship with the bubble diameter, while the aspect ratio remained nearly unchanged and thus was independent of the bubble diameter. Interestingly, although the bubbles produced from CH4 had much larger diameters (2–4 μm) than those produced from C2H2 (<300 nm), they had almost the same hmax (<10 nm), resulting in an extremely low aspect ratio (hmax/R = ~0.005) for the bubbles produced from methane.

## Discussion

We demonstrate an approach for isolating hydrogen in h-BN bubbles by plasma treatment. The dimensions of the bubbles range from nanometers to micrometers and depend on the duration and sample temperature of the plasma treatment. Most of the bubbles undergo minimal changes over 40 weeks. Additionally, the penetration depth of atomic hydrogen reaches approximately 23 h-BN layers. The Raman spectra of a bubble reveal that stretching occurs on the bubble surface. Finally, we demonstrate that hydrogen gas can be extracted from hydrocarbons and mixed gases by the plasma treatment and isolated between h-BN layers in bubbles. The demonstrated ability to fabricate hydrogen bubbles on h-BN flakes might be used to extract hydrogen from hydrocarbon/gas mixtures or to fabricate nano/micro-electromechanical systems.

## Methods

### Fabrication and characterization of hydrogen bubbles on h-BN

The experimental setup is given in Supplementary Fig. 22. First, h-BN flakes were produced by micromechanical exfoliation of single-crystal h-BN and then deposited on quartz substrates that were previously cleaned with oxygen plasma. The substrates with h-BN flakes were loaded into a furnace equipped with an RF generator (13.56 MHz, MTI Corporation) with a tunable power from 100 to 400 W. The sample temperature in the sample chamber could be controlled from room temperature to 1000 °C. An advanced pump (GX100N Dry Pumping System, Edwards) was connected to the chamber to control the flow rate. After the samples were heated to the preset temperature, hydrogen (oxygen or argon) gas with a flow rate of 3 sccm (~3 Pa) was introduced into the chamber, and then the plasma was formed. The plasma treatment normally took 90, 120 or 150 min. The h-BN flakes were finally removed from the sample chamber for characterization under an optical microscope (Eclipse LV150, Nikon), an atomic force microscope (Dimension Icon, Bruker), and a Raman spectrometer (532 nm, WITec).

### Low-temperature AFM measurement of the bubbles

After bubbles were produced on h-BN flakes, optical microscope and AFM were used to locate their positions. Subsequently, the substrate with the h-BN flakes was transferred into the vacuum chamber of a variable temperature SPM (Scanning probe microscope) system (attoDRY 1100 by Attocube). The chamber was filled with 4He in a pressure of 5 mbar for heat exchange. The sample temperature can be controlled in the range from 4 to 300 K. All the AFM scanning was conducted in contact mode. In order to minimize the damage to the bubbles, AFM tips with a force constant of 0.02–0.77 N m−1 were used. The sample was cooled at a rate of 0.5 K h−1 while AFM scanning was carried on the h-BN bubbles. The highest temperature at which the bubbles became flat was recorded as the transition temperature. Afterwards, the samples were warmed up for several Kelvins to observe the expansion of bubbles in order to exclude the possible case of gas leakage.

### Scanning transmission electron microscopy

Electron microscopy experiments were conducted using a Nion UltraSTEM100 scanning transmission electron microscope, operated at 60 kV in near-ultrahigh vacuum (2 × 10−7 Pa). The beam current during the experiments varied between 8 and 80 pA, corresponding to dose rates of 5–50 × 107 e Å−2 s−1. The beam convergence semi-angle was 35 mrad and the semi-angular range of the medium-angle annular-dark-field detector was 60–200 mrad.