Hexagonal boron nitride (h-BN) is a famous non oxide ceramic known and appreciated for its numerous interesting properties, such as a high thermal conductivity, an excellent thermal chock resistance, a high resistance against oxidation (850 °C under air), a good chemical inertness [1
Based on this non exhaustive list of important properties, it is easy to understand the key role of h-BN for real or potential applications. Actually, h-BN can be found many fields ranging from cosmetics to metallurgy passing by aeronautics or battery [1
Another very interesting specificity of h-BN is linked to its structural similarity to graphite doing of themselves twin materials [3
]. Actually, the h-BN progression is always directly linked to graphite counterparts. Examples are numerous, such as the preparation of boron nitride fibers in replacement of the carbon ones to be used at high temperature under air [4
] the synthesis of boron nitride nanotubes after the discovery of the CNTs in the early 1990s [9
], the preparation of analogue boron nitride fullerene/nanocage somewhat latter [12
] and more recently the myriad of studies to get boron nitride monolayers within the explosion of graphene [3
]. All these examples show that the shape given to the ceramic is at least as important as the material itself, meaning that the shaping is essential to ensure potential applications.
The production of commercial h-BN powders is now completely mastered from an oxygen-containing boron compound reacting with a nitrogen-containing source [1
]. The use of such powders is clearly convenient to provide sintered pieces but really restrains the possibility to reach others specific shapes (fibers, foams, coatings, etc.)
In the latter case, a versatile alternative has to be found. The Polymer Derived ceramics (PDCs), developed in the late 1960s by Chantrell and Popper [19
] for a non-oxide system is undoubtedly a good choice [20
]. This route consists in the synthesis of very pure preceramic polymer from tailored molecular precursor. The preceramic polymer (liquid, soluble or fusible) can be then easily shaped into the convenient shape by means of dip- spin-coating, casting, melt-spinning [21
. Followed by an appropriate thermal treatment. The preceramic polymer body is then converted into hexagonal boron nitride while maintaining the desired shape. Obviously, the selection of the starting precursor is essential and should follow guidelines. Among all available precursors [22
], borazine (B3
) or polyborazilene (PBN) (–(B3
], the latter being directly derived from the polycondensation of the former, are very attractive because they are prepared through a reproducible synthesis, whitout any contaminant within compounds and a perfect B/N stoichiometry... PBN, which presents the advantage of a lower volatility and a better ceramic yield, is always preferred to borazine [30
For half a decade, our group focused attention on thick h-BN coating deposited on metallic substrate especially on titanium to get self-lubricating devices available in aeronautics industry [31
]. Actually, thanks to its lamellar structure combined to its important oxidation resistance even at high temperature under air, h-BN is one of the best candidates for protecting metal components in engineering. However, to give a high protectiveness, the ceramic may reach a high level of crystallization, which becomes a key parameter to manage. Traditionally, thin h-BN coatings are prepared through conventional sputtering [34
], PLD [35
], CVD and derivatives [37
], these latter being the most studied [43
]. Moreover, on top of the heavy equipment needed [44
], coatings often display a low crystallinity, oxygen contamination or a non stoichiometric B/N ratio [43
]. In this context, the PDCs route appears as a relevant alternative to get advanced coatings. The use of a liquid phase allows a simple shaping by means of dip- or spin-coating able to control the thickness [45
]. A review from 2011 [46
] reports studies on the synthesis of h-BN coatings by preceramic polymers. Among them, we can cite the dibromoborane-dimethyl sulfide [47
], the tris(alkylamino)borane [48
], the alkylaminoborazine and derivatives [30
], the borazine [30
] and the PBN [30
]. Hence, there are many precursor sources, but, the best results, in terms of homogeneity and purity, are systematically obtained with PBN. Actually, by controlling the polycondensation step, PBN can be prepared as a liquid polymer easy to handle and to shape. Besides, an appropriate thermal treatment allows easily converting the green preceramic coating into a contaminant-free crystallized h-BN one, the ceramic crystallization remaining a key parameter to get efficient coated devices.
More recently, motivated by the veritable craze for graphene (G), the current 2D superstar material, we have been witnessing to an increasing demand in BN nanolayers to be used as graphene substrate to keep advantages of its remarkable transport properties [54
]. Actually, the common use of silica as insulating substrate to get a handled device remains insufficient to prepare efficient systems, its surface imperfections significantly affects the supported graphene carrier mobility. In that context, many studies claim the potential of h-BN as suitable dielectric substrate for graphene [55
]. Actually, the perfect lattice matching of h-BN with graphene [58
] coupled with the fact that h-BN has an atomically flat and dangling bond-free surface, will ensure no charge traps at the G/h-BN interface and consequently will increase of graphene’s carrier mobility [59
]. These constraints explained efforts to target uniform thickness, large domain size and high crystallinity. From these guidelines, two options can be distinguished, either a growth of h-BN by CVD or an exfoliation of large single crystals. Today, many studies are devoted to the first way, exploring a variety of transition metals as substrate, such as Ni and Fe films, Pt foil, Cu-Ni alloys, [61
], the most common remaining Cu foils [73
]. Even if results are more and more relevant, with the increasing size of domains, this route needs heavy equipment and reproducibility is strongly dependent on the used apparatus, which really limits its development. Besides, the production is limited. Keeping these issues in mind, the second way, involving exfoliation of single crystals, is relevant if an accessible source of pure and highly crystallized crystals becomes available [76
]. It is well-known that the group of Watanabe at the National Institute for Materials Science (NIMS) [77
] can prepare this kind of high quality single crystals using the High Pressure High Temperature (HPHT) process, but the whole procedure runs through severe production conditions and long treatment time [76
] making its widespread restricted. Commercial sources [17
] are more reachable but the quality remains characterized by a high level of defects and relatively small crystalline areas [76
]. In that latter case, synthetic material can suffer of a lack of crystallization, often linked to the synthetic process used.
In this review, we focused our attention on three main strategies developed in the group in order to increase the h-BN crystallization for supported thick coatings and self-standing nanosheets prepared through the PDCs route. Our approach is based either on the process, or on the chemistry of reactants (sometimes on both). If the easiest way to enhance crystallization consists in a simple increase of the final treatment temperature, this is not systematically allowed, for apparatus limitations reason, due to a cost restrictions or due to substrate specificities (thermal sensitivity for example). Alternatives have, then, to be found. Our first studies are focused on the ceramization process. Hence, we explored the possibility of preparing h-BN coatings on metallic substrates using an innovative IR irradiation annealing process. Since the metallic substrate reflects IR radiation, only the outer preceramic coating is irradiated, preserving the subjacent substrate integrity. The second strategy is directly linked to the chemistry, with the additivation of the polymeric precursor with a crystallization promoter, acting as a catalyst to favor the crystallization occurring during the ceramization step. Finally, to go further, we implemented a third strategy combining the additivation and the use of a particular sintering process, which is the Spark Plasma Sintering (SPS).
3. Second Strategy: Li3N as Crystallization Promoter
Besides the process that can be adjusted to enhance the ceramic crystallization conditions, another tactic may consist in playing with the chemistry of reactants involved during the synthesis. Actually, there are intrinsic relationships existing between properties of both precursors (or solution) and the resulting final material. Hence, either modifying the chemical nature of the used precursor or additivating it with a crystallization promoter would play a key role on the final crystallinity of the ceramic. In that way, as already mentioned in the introduction, the used of PBN as preceramic precursor is particularly relevant [30
] since it is prepared through pure borazine that initially displays the expected targeted hexagonal arrangement [85
]. Moreover, since the polymer is obtained without any solvent addition, only hydrogen is expected to be lost [25
] and no contaminant is present, whose presence may slow down the crystallization process. Some studies report the addition of calcium carbonate [1
] or lithium nitride to increase the final h-BN crystallization. However, Li3
N is more classically used for conversion of h-BN to c-BN at relatively low temperature and pressure [90
]. Based on these relevant results, we decided to add lithium nitride (Li3
N) in pure PBN in order to lower the onset crystallization temperature [95
3.1. Evidence of Li3N Effect
First to highlight the key role of this additive in crystallization enhancement of the final ceramic, we performed Thermo Gravimetric Analyses (TGA) on two reference samples; pure PBN and PBN mixed with Li3
N micro powders (5wt. %), respectively named PBN and PBN-Li3
N (Figure 6
Comparison of both curves shows three main differences. First, a lower weight loss for PBN-Li3
N sample at 600 °C: 13% vs.
17% for PBN sample is observed, the difference (Δm1
) indicating the additive preservation. The second difference corresponds to a higher weight loss for the additivated sample between 600 and 1000 °C interpreted as the removal of the additive or a product derived from the latter stage. Based on previous articles, we can assume that at this temperature range, a reaction occurs between Li3
N and the former BN to get Li3
]. This specie can exist through two different structures α-Li3
melting at 860 °C and 920 °C, respectively, so likely to be eliminated under the N2
flow. Finally, the last divergence comes from the behavior of both products up to 1200 °C. Actually, the TGA curve recorded on the BN derived from the PBN-Li3
N sample stays stable up to 1400 °C, meaning that the sample is not changing. Besides, the TGA curve from the BN produced from the pure PBN gains mass (Δm2
) and this non-expected behavior can be attributed to oxidation (with residual air in the chamber) [53
] meaning that the sample is not stable and sensitive to oxidation probably because of a worse crystallization [98
Further structural characterization by means of Raman spectroscopy and X-ray diffraction are performed on both TGA residues. First, both Raman spectra (Figure 7
) exhibit a characteristic signature of hexagonal boron nitride, with a broad signal around 1367 cm−1
typical of the E2g
mode peak of h-BN [83
]. The first difference comes from the FWHM measured to be 52 cm−1vs.
and the La
] to be 8.6 nm vs.
38 nm respectively for the pure and additivated samples, meaning a better crystallization of the latter sample. The second difference comes from an additional signal within the PBN-Li3
N derivative sample spectrum at 810 cm−1
and assigned to boron oxide [99
] confirming the compound oxidation as previously suggested by TGA.
XRD patterns, proposed on Figure 8
, also present the signature of hexagonal boron nitride with a main signal at around 26.8° corresponding to the (002) diffraction peak, according to the JCPDS file 009-0012. Moreover, examination of values reported on the Table 2
and corresponding to the main (observed or calculated) [33
] XRD parameters clearly shows that the sample prepared with the pure PBN displays a lower crystallization compared to this obtained from PBN-Li3
The fact that this latter XRD pattern displays a shoulder towards the low 2θ values (~25°), meaning a dispersion in the crystallites size, can be related to its very short stay at 1400 °C. By tailoring the ceramization parameters one can expect to enhance even more the crystallization state.
3.2. Influence of the Annealing Temperature on h-BN Crystallization
In order to understand the chemical and structural mechanism occurring with the ceramic derived from the additivated precursor, a series of BN samples incorporating Li3
N (5wt.%) and annealed at 600, 800, 1000, 1200 and 1400 °C (denoted, respectively LiBN01 to LiBN05) is prepared. Diffractogramms recorded on each sample are gathered in Figure 9
. LiBN01 sample is completely amorphous, in good accordance with the TGA curve that shows that the catalytic effect is not expected at this lowest temperature. Regarding LiBN02, we evidenced a signature of turbostratic BN coupled with other signals attributed here to the Li3
. Our assomption is again confirmed with the formation, between 800 and 1000 °C of Li3
from a reaction between Li3
N and the former BN. From 1000 to 1400 °C, differences within the XRD patterns remain marginal, with refining signals, meaning a decrease in the measured FWHM and calculated d002
values obtained on the (002) diffraction peak as attested in Figure 10
. In conclusion, the beneficial effect of the addition of Li3
N within the PBN is clearly visible from 1000 °C and even more accentuated at 1200 and 1400 °C.
The proposed crystallization mechanism is based on previous studies reported by Bezrukov [100
] and other groups [97
]. At around 600–700 °C, Li3
formed by reaction between Li3
N and preceramic BN, melts and infiltrates amorphous BN and dissolves it, leading to highly crystallized hexagonal boron nitride.
The main interest in this additivation is to drastically decrease the crystallization temperature from traditionally, at least 1500 °C [25
], to 1000–1200 °C. To our knowledge, we are the first to demonstrate the additive activity of Li3
N when adding in a preceramic polymer and the heat treatment of that mixture.
3.3. Thick h-BN Synthesis
From this important result we decided to combine the use of an additivated polymer to a RTA thermal annealing to get crystallized h-BN coating over titanium substrate while preserving the latter integrity and saving time [31
In that way, pure PBN is mixed with micro sized Li3N powders, in 1, 3 or 5wt. % and stirred for 30 min in a gloves box. Still under argon, each solution is deposited on a titanium substrate, previously cleaned with ethanol. Coated samples are then kept at 200 °C for 1 h to avoid further important polymer volatilization. Finally, samples are placed in the RTA furnace and heated at 1200 °C under N2 underwent 9 flashes following a program described above. The 3 resulting samples are named TiBN1, TiBN3 and TiBN5. The naked substrate undergoing the same annealing is named TiBN0.
Observations made by SEM and proposed on a previous paper [31
] show recovering and almost homogenous coatings presenting, moreover, more and more visible defects due to the original presence of the powder.
In order to study the ceramic crystallization, XRD was performed on the whole sample series. The corresponding patterns are proposed in Figure 11
Only regarding peaks corresponding to h-BN at 26.7°, 42.3° and 44.4° (2θ), they appear finer with the crystallization promoter ratio increasing. Besides, measured FWHM(002)
and calculated d002
values reported in Table 3
demonstrate that most the sample incorporates Li3
N in the pre-mixture, most it becomes crystallized. It has to be noticed that all the other signals display on the XRD patterns are attributed to either titanium or (TiN + TiB2
) respectively, the substrate and the interface (according to the JCPDS files 087-0633 and 089-3923).
In summary, we demonstrate here that it is possible to get crystallized h-BN coating on titanium at only 1200 °C for a total treatment duration of a few minutes. Increasing the IR irradiation time could still increase crystallization but in that case, the thermal conduction occurring from the coating to the substrate will modify the titanium structure (phase transition: 980 °C), which is not allowed. To the best of our knowledge this is the best compromise between the crystallization rate and the substrate preservation. Futher studies [31
] have shown that coatings prepared with an additivated precursor and treated under RTA present very interesting mechanical and adhesion properties for tribological applications.
3.4. BNNS Preparation
If the availability of thick h-BN coatings on metallic substrate is very interesting for mechanical and tribological applications, recent works, essentially based on graphene heads towards 2D nanostructures preparation. In this new context, we decided to take advantage of the good results obtained on the PBN additivation to favor the synthesis of well-crystallized BNNSs [103
Hence, we simply prepare a bulk sample using PBN additivated with Li3
N (5wt. %) annealed, under N2
, at 1400 °C (1 °C/min) for 1 h after a 1-hour pre-stabilization at 200 °C (on a hot plate). Here again, high crystallization of the obtained white powder is clearly proved by Raman spectroscopy and XRD. Raman scattering spectrum, reported on Figure 12
, is typical of h-BN with a unique signal at 1367 cm-1
displaying a weak FWHM value (17 cm−1
Regarding , the XRD pattern (Figure 13
) in comparison with commercial h-BN (HENZE BNP GmbH support) is also characteristic of highly crystallized h-BN with thin and well defined diffraction peaks at 2θ = 26.9°, 41.8°, 44°, 55.2°, 75.9° and 82.1° corresponding to (002), (100), (101), (004), (110) and (112) crystallographic planes, respectively according to the JCPDS file 009-0012. However, we can note the absence of the (102) crystallographic plan expected at 50.3° (2θ). In addition, taking into account the fact that calculated relative intensity of each peak (Table 4
) are consistent with theoretical ones only for the (002) and (004) signals, suggest the synthesis of ceramic with 2D dimension. Two supplementary peaks can also be detected at 2θ = 42.6° and 45.6° and attributed to rhombohedra BN (r-BN) (JCPDS file 045-1171). Structure of r-BN is similar to that of h-BN with a stacking of alternative atomic layers ABC-ABC instead of AB-AB for h-BN [16
In order to confirm the sample microstructure, the raw powder is dispersed in ethanol under ultrasound, for a chemical exfoliation step, then observed by TEM. Figure 14
presents representative micrographs obtained on exfoliated BN. Figure 14
b displays diffraction contrats that are characteristics of the perfect sample crystallization. Misorientations between h-BN layers may explain those patterns as shown on the image at higher resolution (Figure 14
a) where perfectly crystallized sheets are stacked on each other in a disorderly manner. This is also revealed by multiple spots in the corresponding Selected Area Electron Diffraction (SAED) patterns obtained on different sample areas (Figure 14
a inset), and exhibiting severals sets of six hexagonal spots rotated with various angles, demonstrating the achievement of h-BN single crystals stacked respectively in a AA’ structure along the c
-direction or in a more complex turbostratic structure. This relative rotation may be attributed to the stain induced by the rolled structure and the folding of sheets on themselves. The High Resolution Transmission Electron Microscopy (HRTEM) performed on a fine zone of one sheet shows the perfect atomic in-plane arrangement characteristic from the hexagonal lattice (Figure 14
c). Several dark lines with interatomic distance measured at 3.33 Å are experimentally measured on the edges of the sheet, giving its number of layers. The diffraction pattern associated with this sheet clearly exhibits hexagonal bright dots, representative of single crystal.
It is shown that, by adding a crystallization promoter into the preceramic polymer, it is possible to get well-crystallized powders easily exfoliated to get 4- 3- 2- layers and even monolayers [103
]. These sheets are used to get h-BN epitaxial graphene van der Waals heterostructures by deposing a drop of a solution of h-BN layers on ethanol on an epitaxial graphene, demonstrating the interest of these starting materials [105
]. However, in any case, the BNNSs size remains critical to consider applications as graphene substrate or to modify the electronic properties. In that way, alternative ways have to be found in order to increase the domains’ size and preserving the high crystallization level.
4. Third Strategy: Combining PDCs with SPS
Once we have been able to get pure and well-crystallized h-BN nanosheets using a crystallization promoter, the next step is to increase the lateral size of the obtained flakes. This step is of primary importance for the use of h-BN under different configurations: as an insulating substrate for graphene [57
] or for creating new van der Waals heterostructures [61
]. Our strategy is to combine the advantages of the PDCs route to those of the Spark Plasma Sintering (SPS) (Figure 15
) process in order to synthesize important amount of pure, well-crystallized and large-scale (>10 μm) h-BN flakes, which could be further easily exfoliated into BNNSs.
To perform the experiment, liquid PBN, mixed with Li3N micro powder (5 wt. %) is preceramized at 650 °C. The latter amorphous BN powder is then sintered by SPS in order to favor crystallization and increase crystals size. After sintering, we obtained a very dense white pellet of 2 cm in diameter.
4.1. Characterization of the Bulk h-BN Pellet
The white pellet (Figure 16
) is first investigated by X-ray photoelectron spectroscopy before and after an Ar+ sputtering of the surface. Results obtained before abrasion of the surface gives an elemental B/N ratio of 0.97, very close to the one of stoichiometric BN and consistent with reported XPS data for BN [109
]. Carbon and oxygen contaminants are also observed attributed to extrinsic pollution as they disappear on the second analysis performed after 1 μm abrasion (4 keV Ar+ sputtering).
SEM performed on the same sample shows a homogeneous stacking of h-BN flakes (up to 30 μm) preferentially oriented perpendicularly to the pellet surface i.e.
aligned in the compression direction and the external current (Figure 17
a). A detailled flake is also presented on Figure 17
b to prove the large flake size.
The preferential orientation of the flakes is confirmed by X-ray diffraction recorded on both pellet surface and pellet cross-section (Figure 18
). Even if both patterns exhibit thin, intense and well-separated peaks assigned to expected crystallographic planes from an h-BN crystal (JCPDS file 009-0012), the comparison of both diffractograms clearly shows a preferential orientation of the h-BN crystallites along the a
-axis (characteristic of the covalent B-N bond), that is parallel to the applied load.
The good crystallinity of the sample deduced from the XRD patterns is also observed on the Raman spectrum (Figure 19
). The well-defined symmetric single peak recorded at 1366 cm−1
, which is characteristic of the E2g
vibration mode of h-BN crystal [33
], presents a FWHM value of 7.7 cm−1
comparable to the best values reported for the h-BN single crystals obtained by HPHT [77
] and strongly improved compared to 17 cm−1
for PDCs BNNSs presented above. This last result evidences a very low defect density combined with a very large crystallite size.
In summary, by coupling the PDCs method described above with SPS process, we can synthesize large micrometers-crystallized domains (up to 30 μm) of pure h-BN growing in the sintering load direction. From this sample, we have then prepared Boron Nitride Nano Sheets (BNNSs) by physical (tape) or chemical (in ethanol, ultrasonication: 1 min at 25 W, Hielscher UP400S and centrifugation) exfoliation.
4.2. Characterization of Single Crystal Flakes
Flakes can be exfoliated by the tape method then deposited onto a Si substrate, covered by 80 nm SiO2
oxide layer. An optical examination reveals 10 to 300 nm (Figure 20
b) thick pieces with side size up to 40 μm (Figure 20
c). Raman spectrum recorded on one flake (Figure 20
a) displays the single peak of h-BN located at 1366 cm−1
with a 7.5 cm−1
FWHM, reflecting once more the excellent crystallinity of the flakes.
TEM observations of 5–30 μm flakes confirmed the high crystalline quality of samples (Figure 21
). Actually, diffraction contrast (dark areas correspond to crystallized zones of the flake in Bragg’s position) (Figure 21
a,b) visible on a reprensentative flake (Figure 21
a) attests the crystallinity of the final ceramic. Figure 21
c represents a TEM image of the edge of the flake showing the perfect staking of BNNSs composing the flake.
4.3. Characterisation of the BNNS
The size distribution of the crystals before chemical exfoliation, and of the BNNSs after chemical exfoliation, has been respectively studied from SEM and optical images [76
]. After exfoliation, BNNSs thickness slows down to several nanometers and the BNNSs size (10 to 15 μm vs.
30 μm for crystals) is lowered. The surface roughness of h-BN deposited onto SiO2
substrate has then been measured by AFM and the RMS (Root Mean Square) value has been determined to 0.7 nm, similar to the one measured for standard silica. This means that our samples would be flat enough to be used as graphene substrates or to produce Van der Waals heterostructures. A droplet of the chemical solution was then simply deposited onto a 300 mesh holey carbon copper grid in order to perform TEM observations. Figure 22
a proposes a typical TEM image obtained on 5–10 μm size BNNSs. Another image (Figure 22
b) still performed on a typical sample shows how the BNNSs delaminate from each to other leading to a large amount of nanolayers. Finer observation by HRTEM allows to get image of a 4- layers BNNS, the thickness being determined counting the dark lines. SAED pattern (Figure 22
c, inset), obtained from the high resolution image (Figure 22
), shows six bright spots hexagonally distributed, demonstrating the achievement of well crystallised h-BN. This SAED pattern can be interpreted as the AA’ atomic stacking along the c-direction.
Finally, EELS spectra are recorded on thin BNNSs and show only boron and nitrogen K-edges without any trace of lithium, carbon or oxygen, corroborating previous XPS results on the excellent purity of samples.
To conclude, by combining PDCs route with SPS method, we associated the synthesis of high purity ceramic to the high crystallinity sintering. We obtained very pure and well crystallized large flakes (up to 30 μm) orientated along the compressed direction. By exfoliating these flakes, we can obtain a large amount of self-standing h-BN few-layers whose quality may open new perspectives for many applications as graphene substrate or to produce van der Waals heterostructures.