Breaking the icosahedra in boron carbide

Significance

The extraordinary hardness of boron compounds is related to their internal structure, which is comprised of 12-atom icosahedra arranged in crystalline lattices. In these hierarchical materials, the icosahedra are easy to image with EM, but individual atoms are not. Here, we show that laser-assisted atom probe tomography can be used to deduce the atomic structure and relative interatomic bond strengths of atoms in boron carbide. To our surprise, the icosahedra disintegrated during the field evaporation process. Statistical analyses of event multiplicity and stoichiometry in the atom probe dataset substantiate that the icosahedra are less tightly bound than their interconnecting chains. Comparisons with quantum mechanics simulations further suggest that this instability plays a role in the amorphization of boron carbide.

Abstract

Findings of laser-assisted atom probe tomography experiments on boron carbide elucidate an approach for characterizing the atomic structure and interatomic bonding of molecules associated with extraordinary structural stability. The discovery of crystallographic planes in these boron carbide datasets substantiates that crystallinity is maintained to the point of field evaporation, and characterization of individual ionization events gives unexpected evidence of the destruction of individual icosahedra. Statistical analyses of the ions created during the field evaporation process have been used to deduce relative atomic bond strengths and show that the icosahedra in boron carbide are not as stable as anticipated. Combined with quantum mechanics simulations, this result provides insight into the structural instability and amorphization of boron carbide. The temporal, spatial, and compositional information provided by atom probe tomography makes it a unique platform for elucidating the relative stability and interactions of primary building blocks in hierarchically crystalline materials.

Icosahedra are commonly observed polyhedra in nature and can be found in a wide variety of molecules, viruses, minerals, and ceramics (1–4). The architecturally efficient icosahedral geometry and tight covalent bonding result in an unusually stable atomic configuration and in many cases, extraordinary properties. The prodigious structural stability, rigidity, and strength of C₆₀ nearly spherical fullerene molecules (Bucky balls) are well-documented and have received considerable attention (5). Elemental boron and boron-based ceramics also form nearly spherical icosahedra (3, 4, 6), and the extreme hardness of these borides has been attributed to the presence of these icosahedra. In the case of boron carbide, one of the hardest structural ceramics, B₁₂ or B₁₁C icosahedra are stacked with rhombohedral symmetry and connected by three atom chains (Fig. 1A) (3, 6).

Fig. 1.

(A) Crystal structure of boron carbide. (B) Bright-field TEM image of a boron carbide atom probe tip. (C) 3D atomic boron, carbon, and gallium maps of the atom probe tip.

For boron carbide, the relative bond strength between the atoms in the icosahedra and chains is still being debated, but the icosahedra are generally thought to be very strong and stable because of their near-spherical shape and the highly delocalized fullerene-like intraicosahedral sp² bonds (3, 7–10). As a result of its high hardness, boron carbide is an excellent candidate for use in personal body armor; however, it undergoes a dramatic loss of ballistic performance in high-energy impacts. This loss is attributed to the formation of nanoscale amorphous shear bands (11). The shear amorphization of boron carbide has been confirmed by indentation experiments (12–16), but a fundamental understanding of the mechanism underlying it has not been fully established. It has been suggested that local shear first breaks the weaker chains and then, displaces and ruptures the stronger icosahedra (16), but recent quantum mechanics (QM) simulations attribute the instability to disintegration of the icosahedra (17, 18).

In this study, we carried out field evaporation experiments on boron carbide using a state of the art UV laser-assisted local electrode atom probe. Atom probe tomography (APT) is conventionally used for both 3D imaging and chemical composition measurements at the atomic scale (19, 20). Here, we embraced another dimension of atom probe data mining by statistically investigating the evaporation events, counting the types and numbers of atoms per laser pulse and deducing the relative stability of boron carbide icosahedra and chains. Our atom probe results show that, in boron carbide, the icosahedra are actually less stable than the chains and that chain–icosahedron bonds are unexpectedly strong, findings in direct contrast to conventional conjecture.

To carry out the field evaporation experiments, we shaped consolidated boron carbide samples with close to B₄C stoichiometry into very sharp needles with ∼20-nm radius using a focused ion beam (FIB). An example of an APT needle is provided in the bright-field transmission EM (TEM) micrograph in Fig. 1B. One common concern of using FIB-prepared specimens for APT data analysis is the structural and chemical changes induced by the high-energy Ga⁺ beam (21). Boron and carbon were found to be uniformly distributed throughout the specimens, whereas gallium was only observed in the top few nanometers of each specimen (Fig. 1C). This result establishes that analyzed volumes taken away from the damaged layer are free of gallium implantation and that the detected field evaporation behavior is from the pristine boron carbide.

The atom probe experiments are conducted under the principles of field evaporation. Application of a very intense electric field polarizes surface atoms (or molecules), and superposition of rapid voltage or laser pulses ionizes and liberates individual atoms or molecules from the surface. The liberated ion is then accelerated by the intense electric field and captured on a position-sensitive detector that records its spatial and chemical information. Successive field evaporation events are used to create a 3D dataset of the sample (22). Details about the principles of APT can be found in SI Materials and Methods.

SI Materials and Methods

Materials.

The dense boron carbide was consolidated using spark plasma sintering at 1,900 °C for 5 min. The stoichiometry was measured to be B4C by X-ray diffraction, Raman spectroscopy, and electron energy loss spectroscopy.

Boron carbide rods with 1 × 1-mm² cross-sections were obtained by slicing the dense sample using a diamond saw. Needle-shaped tips were produced by electropolishing the rods in 5% (mass/mass) KOH–water solution. The final shaping was carried out in an FIB (FIB Zeiss; Auriga), milled at 30 keV, and cleaned at 5 keV.

Atom Probe Experiments.

Atom probe experiments use the principles of field evaporation. Under a very intense electric field, surface atoms or molecules at the tip of a specimen are polarized, and high-energy voltage pulses can be used to field evaporate individual atoms from conducting metals and alloy specimens. These atoms are ionized during the evaporation process, accelerated by the intense electric field, and recorded by a detector that determines their spatial and chemical identities (Fig. S1). The traditional pulsed voltage approach has also been used to study some semiconductors, but only those with very small band gaps (e.g., highly doped Si) (30). Semiconductors, such as pure Si (room temperature band gap 1.12 eV), are known to be very difficult to field evaporate with voltage pulses. Being a semiconductor, we first attempted to field evaporate our boron carbide specimens using the pulsed voltage technique, but no signal was detected up to 8,000-V DC voltage and 160-V pulsed voltage. Our inability to field evaporate boron carbide with voltage pulses can be attributed to its large band gap (the room temperature band gap of boron carbide is estimated to be 1.6–2.0 eV) (7).

Fig. S1.

Schematic drawings showing the field evaporation process in one laser pulse. (A) Ground state where laser does not illuminate the specimen. (B) Laser pulse illuminating the specimen surface, increasing the specimen temperature and causing thermal perturbation (28). The atom subjected to the highest field ionizes and is repelled by the electric field. The liberated atom is captured by the detector. Note that the drawing is not to scale.

The recent development of laser-assisted APT has made it possible to characterize a broader array of semiconductors and insulators. Like the traditional pulsed voltage approach, laser-assisted APT achieves 3D atomic-scale analysis by using a high electric field to remove individual atoms (or molecules) from material surfaces and a position-sensitive detector to determine their location and chemical identities. Thermal vibrations associated with rapid laser pulsing can displace a surface atom (or molecule) slightly away from the tip and cause its electrons to channel into the tip, resulting in a charged ion. After created, this ion is pulled off the surface and accelerated toward a detector by the intense electric field. The field evaporation experiments for this study were performed using a 355-nm UV laser-assisted atom probe (Cameca 4,000× Si). The laser illumination frequency was 250 kHz, and the laser dwell time was 8 ps per pulse. High DC voltages (4–8 kV depending on the specimen tip radii) were applied to generate an intense electric field. Altogether, six separate experiments were carried out—two with a 30-pJ laser pulse at 50 K, two with a 50-pJ laser pulse at 50 K, and two with a 30-pJ laser pulse at 20 K.

Field evaporation under laser illumination is thought to be a thermally assisted process with rapid temperature changes (31, 32). Previous investigations of laser-induced heating of atom probe tips have focused primarily on metals. The calculated temperature rises for tungsten and other metal tips were estimated to be ∼300 K in normal experimental conditions (29) and up to ∼800 K in extreme conditions (29, 33, 34). It is difficult to directly assess the tip temperature of laser-illuminated semiconductors and insulators, because the temperature in the specimen depends on its thermal diffusivity, thermal conductivity, specific heat,, and density, all of which are temperature-dependent (35). There is, however, strong experimental evidence to show that semiconducting GaAs nanowires remain crystalline during laser-assisted field evaporation experiments (36, 37). Du et al. (36, 37) performed atom probe experiments with laser energies (10–100 pJ) similar to those used in this study (30–50 pJ), and they obtained field desorption maps that revealed clear crystallographic poles, which indicate that the entire specimen, including the very surface atoms before evaporation, remained crystalline. The absence of melting indicates that the tip temperature stayed below the melting point of GaAs (1,511 K), which is about one-half of that for boron carbide (3,036 K). Inspired by these experiments on GaAs, we interrogated our boron carbide datasets and found the presence of lattice planes in the reconstructed volume (Fig. 3), confirming that no melting occurred on the atom probe tip surface in this study.

Fig. 3.

(A) A typical mass spectrum of boron carbide produced with UV laser pulses. The purple swatches at m/q = 130, 65, and 44 indicate where signals from B₁₁C icosahedra would have been but are absent. Note that the dataset is free from gallium damage. (B) Zoomed in view of A showing that most boron and carbon atoms were detected as individual ions. (C) Atomic composition of single and multiple events. Single events are boron-rich and result from icosahedra, whereas multiple events are carbon-rich and associated with chains. Dashed lines indicate the measured bulk composition. (D) Chemical correlation statistics of multiple events highlight the role of icosahedral–chain interactions.

Moreover, the mass resolution is high in all experiments, and thermal tails were not severe, as shown in Fig. S2. The absence of severe thermal tails suggests that the thermal energy dissipates quickly so that there is no excessive heat building up between pulses.

Fig. S2.

Mass spectra of boron carbide field evaporated at (A) 50 pJ and 50 K, (B) 50 pJ and 30 K, and (C) 30 pJ and 30 K. Note that the y axis is in log scale and that none of the datasets have severe thermal tails that would be an indication of excessive heating.

Data Interpretation.

It has been known to the atom probe community that APT tends to underestimate boron content in pure boron and boron-doped materials. Such underestimation is because boron tends to form multiple events, and because of the dead time of the detector, some B ions in the multiple events failed to be detected. Up to 36% loss of total boron signal in pure boron has been reported (38). In this work, we observed that, as the event multiplicity increases, the boron content decreases. Such decrease in boron detection is not associated with the multiple event-related detector dead time. If substantial boron signals were lost in the multiple events, it would have led to an overall underestimation of boron in B4C. However, the measured stoichiometry by APT agrees well with the nominal composition. This consistency provides strong experimental evidence that the detected signals accurately reflect the chemistry of the field-evaporated species.

Moreover, the field-evaporated atoms and molecules captured by the detector provide spatial and chemical analyses of the liberated species. The x–y position of the ion on the detector provides spatial information, and the TOF of the ion points to its chemistry. The z position is determined by the order in which the atoms are detected. As atomic layers are field evaporated by successive laser pulses, the specimen becomes slightly blunter. Small increments in DC voltage are required to continue field evaporation.

In the simplest case, one pulse liberates a single ion from the specimen surface, and it is termed as a single event. In many ceramic systems, one laser pulse can release two or more ions from the specimen surface, and such an event is termed as a multiple event. It has generally been assumed that atoms in the multiple events were from random locations on the specimen surface. Recent studies by Müller et al. (39) and Yao et al. (40) showed the opposite—most of the atoms in each multiple event were spatially correlated. Analysis of the data obtained here also reveals a spatial correlation of the atoms in multiple events (an example is shown in Fig. S3). If the atoms from the same multiple event were randomly distributed, no peaks would be present in the histogram. The majority of atoms from the same pulse were less than 2 nm apart. The strong spatial correlation of the atoms in the multiple events suggests that they were released from the same site rather than from random locations and originally bonded together. The spacing is slightly larger than an icosahedron, because the positively charged atoms repel each other during the early stage of the flight and their detected position deviated slightly from the original position.

Fig. S3.

(A) Spatial correlation histogram and (B) cumulative distribution plot of double events, triple events, and quadruple events in a boron carbide specimen field evaporated at 50 pJ and 30 K.

Multiple events during field evaporation can take place in two scenarios. (i) The chemical bonds of a few atoms on the tip surface break in a correlated event. The liberated ions are accelerated and captured by the detector at the same time and recorded as a multiple event. (ii) A molecule is first field evaporated and then, accelerated toward the detector but decomposes into individual ions during the flight. On arriving at the detector, the decomposed ions are captured simultaneously and recorded as a multiple event. In this study, the multiple events associated with the field evaporation of boron carbide are most closely associated with the first scenario. That is to say that the detection of B and C ions for a single laser pulse results from individual B and C atoms being released from the specimen surface at the same time. If the second scenario was true (that is to say if molecules were evaporated and then decomposed during the flight), there would be a slight delay in the detection of the corresponding ions, because the molecule is heavier than the ions. Saxey (41) proposed that this delay would create distinctly curved tracks in the multiple event correlation histogram, and good examples of this can be found in figure 6 in ref. 41 and figure 6 in ref. 42. No such traces were observed in our Saxey plots of multiple event correlation histograms (Fig. S4). The absence of the traces indicates that in-flight decomposition of molecular ions did not occur in our boron carbide experiments and that the chemical bonds between individual atoms were broken before they left the specimen surface.

Fig. S4.

Correlation histogram of double events in boron carbide. No curved decomposition tracks were observed.

Ab Initio Simulation Methods.

The quantum mechanics calculations were performed with the Vienna ab initio simulation package using the PBE functional and the projector augmented wave method to account for the core–valence interactions. A kinetic energy cutoff of 500 eV for the plane wave expansions was used. The convergence criteria were set to 1 × 10⁻⁴ eV energy difference for solving the electronic wave function and 1 × 10⁻² eV/Å force for geometry optimization. Γ-Point sampling was used to sample Brillouin zone. Ab initio dynamical simulations were performed in the canonical ensembles (fixed number of atoms, fixed volume, and fixed temperature of the system) using Nosé thermostats (damping constant 100 fs) to control the temperatures. Newton’s equations of motion were integrated using the Verlet algorithm with a time step of 2 fs.

The free surface was created from a perfect crystal. The crystal was constructed with a along [21–30], b along [01–10], and c along [0001] and 90 atoms per cell. We created the (0001) surface [(111) surface when viewed in rhombohedral symmetry] by a sequence of tensile deformation along the [0001] direction using a step of 0.02 strain. The atomic positions were optimized at each step. Dipole corrections and surface spin polarization were allowed in the slab simulations. We found that surface is created at a 0.28 tensile strain. We annealed the surface from 800 to 300 K within 8 ps. We found that the icosahedra remain intact during these procedures. We used the same procedure to create the (100) surface. However, the icosahedra were already broken during the application of the tensile step. Because the APT experimental results do not depend on the crystal orientations, we chose a (111) surface for the field evaporation simulations, which is expected to be the most stable configuration for surface icosahedra.

Ejecting charged atoms in atom probe in the physical experiments consists of two steps: field ionization and field evaporation. Because the electron transfer dynamics is extremely challenging to simulate using density functional theory (DFT), we removed four electrons from the model to mimic the postionized state. Then, we applied a very strong electric field along with thermal perturbations to emulate the thermally assisted field evaporation process. The value of 30 V/Å was used as the electric field strength, comparable with the experimental estimation (28). Thermal perturbation was introduced by heating the charged model from 300 to 2,050 K within 8 ps to interrogate which bonds are broken and which possible species can be released from the system. Note that, because of the extremely small size of simulation cell and the extreme high heating rate of 2.2 × 10¹⁴ K/s, we need to go to much higher temperature in simulations to observe the key processes of bond breaking and icosahedra deconstruction. Thus, the temperature to kindle field evaporation in the simulation is much higher than the physical experiments. It is also worth mentioning that the melting point of boron carbide is 3,036 K. Even at a temperature as high as 2,000 K, no melting should have occurred. The destruction of the icosahedra in the simulation is purely attributed to the interplay of strong electric field and thermal perturbation rather than melting—same as the experimental observation.

To gain quantitative insight on the relative stabilities of the surface chain vs. icosahedra, we computed the minimum energy (E) to extract the B-C from the chain and the icosahedron in the slab model (Fig. S5). The minimum energy is described as

$$E=E_{slab-BC}+E_{BC}-E_{slab},$$

where E is the minimum extraction energy, E_slab-BC is the slab energy after extracting B-C from chain or icosahedron, E_BC is energy of the B-C diatomic cluster, and E_slab is slab energy before extracting the B-C cluster.

Fig. S5.

QM simulation model showing the locations of B-C diatomic clusters from the chain and icosahedron. The removal of each increases the overall system energy, and such energy was used to quantify the stability of chains and icosahedra in boron carbide.

The minimum energy is 9.18 eV to extract the B-C from the icosahedron, whereas it requires 12.15 eV to extract the B-C from the chain structure. The results indicate that the B-C in the icosahedra is less stable than that in the chains.

Results and Discussion

During the laser-assisted field evaporation process, laser illumination provides thermal energy to the atom probe tip to assist in the field evaporation process. A pertinent question is whether the tip is melted by the laser pulse, but the dataset that we collected provides strong evidence to show that boron carbide tips do not melt during field evaporation. A detailed discussion of the thermal effects that occur during laser illumination can be found in SI Materials and Methods and Fig. S1, but the definitive evidence against melting is in the fact that we observed no loss of molecular-level crystallinity. The rhombohedral symmetry was retained (for example, Fig. 2), and (001) planes with a lattice spacing of 0.42 nm can be observed. This crystallinity provides clear evidence that the observed breaking of icosahedra in APT is not associated with high-temperature melting and opens an avenue for atomic-level characterization of the icosahedra.

Fig. 2.

Crystallographic information highlighted within the boron carbide atom probe dataset. (A) Boron carbide reconstruction with the region of interest shown in the green cylinder. (B) Clear evidence of (001) lattice planes in boron carbide. (C) 1D spatial map taken normal to the detected lattice planes showing peak to peak distances that correspond to the planar spacing of rhombohedral boron carbide.

The field evaporation characteristics of boron carbide in the atom probe can first be reasoned in a thought experiment. The icosahedra were expected to be very stable, and the intense electric field, assisted by laser-pulsed illumination, was expected to break the intericosahedral and chain–icosahedra bonds, with whole icosahedra evaporated as molecular ions. QM calculations (17, 18) predict that the most stable atomic configuration for crystalline boron carbide is (B₁₁C)CBC, with 12 atoms (B₁₁C) in each icosahedron and three atoms (CBC) in the chain. This configuration is supported but not yet confirmed by aberration-corrected scanning transmission electron microscopy (STEM) images (16). This APT study was originally undertaken to experimentally differentiate between the existence of B₁₁C and B₁₂ icosahedra. In the atom probe data, the peaks corresponding to B₁₁C icosahedra on the mass spectrum would appear at about 130, 65, and 44 Da for charge states of 1+, 2+, and 3+, respectively.

To our surprise, APT experiments showed no such signals (Fig. 3A). All measured peaks have a mass to charge ratio of less than 25 Da. An expanded view of these peaks (Fig. 3B) reveals that they are predominantly boron and carbon ions, with a very small number (∼0.3%) of C₂⁺ complex ions. The icosahedra were completely disintegrated during field evaporation in the atom probe. This finding was consistent for all six tested specimens and insensitive to crystal orientation, direct current (DC) voltage (4–8 kV), specimen temperature (20–50 K), and laser energy (30–50 pJ) (Fig. S2).

Furthermore, closer examination of the evaporation statistics revealed that icosahedra and chains in boron carbide were field evaporated differently. In this study, ∼35% of the atoms were detected as single events, and ∼65% of the atoms were detected as multiple events (bar chart in Fig. 3C). Single events corresponded to the release of individual atoms, whereas a multiple event occurred when two or more atoms were liberated per laser pulse (23). Multiple events can be further partitioned into double events, triple events, etc. depending on the exact number of atoms detected per laser pulse. When the atomic fractions of boron and carbon were plotted as a function of event multiplicity (solid lines with squares in Fig. 3C), they were found to be different and deviate from the bulk composition (dashed lines in Fig. 3C). The measured boron content in the single events is ∼89%, very close to the B₁₁C icosahedron stoichiometry of 92%. This correlation suggests that most of the single events can be attributed to atoms from the icosahedra. In the case of double events, the boron and carbon contents match the bulk composition, suggesting that equal proportions of icosahedra and chains were field evaporated. In higher-multiplicity events, the corresponding carbon concentrations are ∼35%, much higher than the icosahedral carbon content of ∼8%, suggesting that the higher-order multiple events involve atoms from the carbon-rich chains. As explained in SI Materials and Methods, this increase is not associated with multiple event-related detector dead time. This ∼35% carbon concentration is significantly lower than that of CBC chains (at 67%), implying that some fragmented icosahedra were also evaporated in this scenario, which will be discussed shortly.

The apparent distinctions in the field evaporation behaviors of icosahedra and chains provide a direct comparison of their stabilities. The APT community has shown that atoms that are comparatively weakly bonded (e.g., metals) are more often evaporated as single events, whereas strongly bonded atoms (e.g., ceramics and polymers) tend to generate substantial numbers of complex ions and multiple events (22–25). We also note that most atoms detected in each multiple event were observed to be spatially correlated on the detector plane (Fig. S3), implying that the atoms that evaporated in a single laser pulse came from the same region of the sample and were originally bonded together (Fig. S4). In this series of experiments, most single events are attributed to the icosahedra, with multiple events associated with chains. This finding suggests that, contrary to common belief, the icosahedra are actually less atomically stable than the chains.

More detailed analysis of the chemical correlation (i.e., the statistics of atom species evaporated from each pulse in multiple events) offers additional insight on the relative instability of icosahedra. Inspection shows that most multiple events involve carbon atoms and that many involve more than one carbon atom (Fig. 3D). Events with multiple carbon atoms cannot come from a single icosahedron. Moreover, events with multiples of both boron and carbon (e.g., 2B2C, 3B2C, and 2B3C) could not have been produced by evaporation of a single chain or a single icosahedron. Coevaporation of neighboring icosahedra could, in principle, produce counts with two C atoms, but the overall stoichiometry of multiple atom events (Fig. 3C) does not support this interpretation. Instead, the observed chemical correlation indicates that portions of a chain and an adjacent icosahedron evaporate simultaneously, which points to the existence of strong chain–icosahedra bonding in boron carbide.

To further understand the destruction process of icosahedra in APT, we emulated field evaporation with QM simulations [Perdew–Burke–Ernzerhof (PBE) flavor of Density Functional Theory] using the Vienna ab initio simulation package (26, 27), which allowed visualization of the bond-breaking sequence and determination of the relative stability of the icosahedra and chains (Movie S1). To mimic the atom probe experiment in a simplified model, we created the (111) surface of the B₄C crystal, so that intact icosahedra might be most easily achieved. Other orientations [for example, (100)] tend to result in broken icosahedra on the surface, so that the stability of intact icosahedra cannot be directly assessed. To charge the surface atoms with broken bonds, we removed four electrons from the system. To mimic the high DC voltage on the specimen in the experiment, we applied a strong electric field 30 V/Å (28) perpendicular to the surface. Because laser illumination is fundamentally a thermally assisted process (29), we introduced thermal perturbations by heating the tip from 300 to 2,050 K over 8 ps. We note that the simulation requires an overestimation of the temperature expected in experiments because of the extremely small simulation cell and extremely high heating rate. Nevertheless, the QM simulations were used to gain insight into local bonding and its influence on the measured instability of the icosahedra.

The QM simulations cannot reproduce the ejection of ions from the surface, but they do provide an opportunity to monitor the evolution of local bonding and its effect on the arrangement of atoms both within and between neighboring icosahedra and chains. The ground-state structure of the (111) surface at 300 K is shown in Fig. 4A. All icosahedra (labeled 1–3) are intact, and the surface carbon atom (C) in the chain that was cut to create the surface stays connected to boron atoms (B₁ and B₂) in neighboring icosahedra. As the energy was increased, numerous intraicosahedral bonds in icosahedron 2 were broken, but atom C remained attached to atoms B₁ and B₂ (Fig. 4B). The observation that the intraicosahedral bonds were the first to break mimics our experimental finding that icosahedra were destroyed before they could be ionized and ejected from the surface. Both experiments and simulations indicate that, although icosahedra have strong sp² bonds and near-spherical geometry, they are not as stable as previously anticipated. By contrast, the chain–icosahedral interaction was found to be surprisingly stable. It is worth noting that icosahedra 1 and 3 remained intact when icosahedron 2 was broken, which further supports our interpretation that the breaking of icosahedra is caused by local interactions. Increasing the energy further provided additional insight on which chemical species may be liberated from the specimen surface. When icosahedra 1 and 2 were broken (Fig. 4 C and D), species, such as B2 and B2C, were released and are responsible for the multiple hit events that were recorded in the APT dataset (Fig. 3D). The QM simulations further suggest that, even at the highest energies, the icosahedra break sequentially, which can be used to explain the well-defined crystal lattice information in the acquired atom probe data (Fig. 2). To further examine the relative stability of the icosahedron and chain, we constructed a slab model with a surface icosahedron and an exposed C-B chain fragment and computed the minimum energy to extract a B-C diatomic unit from both the surface chain and the icosahedron (Fig. S5). The details of these calculations are described in SI Materials and Methods, and the results indicate that the required extraction energy for the surface chain is greater than for the surface icosahedron: 12.15 vs. 9.18 eV. This finding is consistent with our APT experiments evidencing that the icosahedra fragment easier than the chains.

Fig. 4.

The structural evolution of surface boron carbide icosahedra and chains at (A) the ground state, (B) a higher applied energy state, where one icosahedron starts to disintegrate, and (C and D) even higher applied energy states, where small molecular species can detach from the surface. Solid lines indicate the periodic boundaries, and shaded atoms in the bottom layer serve as the fixed boundary.

Conclusions

In summary, the combination of atom probe experiments and QM simulation provides unique insight into the structural stability of boron carbide. Here, we show that the icosahedra in boron carbide are not as stable as previously anticipated and that the chain–icosahedron bonds are stronger than expected, which corroborates the hypothesis suggested by previous QM simulations that chain–icosahedra interactions trigger shear amorphization (17). This observation provides a key to understanding amorphization and the loss of ballistic performance in boron carbide. Furthermore, investigating the evaporation statistics using APT allowed us to conduct a characterization of the relative stability and the interactions of icosahedra and chains in boron carbide. This approach can be applied to directly compare the structural stability and chemical bond strengths of the building blocks in other hierarchical materials as well.