1 Introduction
Knowledge of the mechanical properties of volume-confined materials (e.g., thin films, nanopillars, simple nanoparticles (NPs), and composite nanoparticles (CNPs), etc.) is an essential prerequisite in designing micro- and nanoelectro-mechanical systems (MEMS and NEMS) for deploying nanotechnology safely and reliably. NEMS have become an exciting frontier for the next generation devices for sensing and computing, since they were shown to achieve high operating frequencies (up to 10
9 Hz) with extreme sensitivities when size shrinks below 100 nm.
(1, 2) When dealing with mechanical properties at the nanoscale, parameters like the surface stress are expected to be involved in most size effects (e.g., extremely large fractions at grain boundaries with a special atomic structure may occur in these systems) responsible for the very high strength, toughness, fatigue life, and wear resistance that have indeed been observed.
Nanoindentation is currently the principal platform to investigate mechanical properties and has become a unique tool permitting the
in situ mechanical probing necessary to assess the reliability and durability of structural nanocomponents. It has now long been used to study the elastic, plastic, and fracture properties of surfaces of bulk samples and thin films,
(3-5) and in fact, the importance of the atomistic-level understanding from the nanoindentation processes is now widely recognized.
(6, 7) In the past decade, it has become possible to perform controlled compression and bending tests on nanostructures smaller than a micrometer, such as NPs,
(8-13) nanowires,
(14-16) and nanopillars.
(17-22) Limiting our scope to NPs, earlier compression tests on Si nanospheres were carried out by Gerberich and co-workers,
(8) finding the estimated hardness (i.e.,
50 GPa) in dislocation-free NPs (40−100 nm diameter) to range up to 4 times more than in bulk Si (i.e.,
12 GPa). Additionally, peculiar phenomena such as a ductile behavior in a normally brittle material and reverse plasticity were also observed,
(9) and then qualitatively reproduced and confirmed in different subsequent experiments.
(12, 13) To date, though, those experiments have focused exclusively on homogeneous NPs, while the mechanical properties of the many and more complex multiphase nanostructures that can be produced nowadays seem to have neither been explored nor reported yet. This situation can be partially attributed to the fact that the degree of complexity of modeling and analysis suddenly increased when moving away from simpler NPs. Nevertheless, this lapse is in strident contrast with the wealth of synthetic procedures that have been devised, in which the final nanostructures are built from very different materials and in very different morphologies (core−shell, onion-like, dimers, aggregates, etc.).
(23, 24) Furthermore, the general foundations of nanoplasticity and the identification of meaningful length scales underlying the plastic phenomenon also represent active research subjects themselves by a theoretical standpoint. Several hypotheses have been suggested or speculated about dislocation dynamics in metals, reverse plasticity, and hardening mechanisms,
(4, 9, 10, 22, 25) consequently revamping this debate after more recent
in situ compression tests monitored via transmission electron microscopy (TEM). So far, though, the discussion about nanoscale plasticity has focused on single crystal metals (mostly FCC) but is now broadening to other material systems.
(26) Herein, we report on quantitative microcompression tests performed on novel CNPs consisting of cobalt boride nanoparticles coated with silica. The mechanical properties of such core−shell nanocomposites are deduced from experimental data by means of contact theory analysis. This material exhibits a complex three-regimes response that entails both elastic and plastic deformation and may result from several concurring deformation processes (i.e., mechanical, chemical, and magnetic). These results are interpreted in consideration of the pre-existent distribution of radial defects found in the cobalt boride core of these CNPs. Such defects represent a signature feature which sets them apart from any defect-free, single phase nanoparticle reported previously and are of great relevance to the novel plastic phenomena.
3 Results and Discussion
Within the mentioned context and while most literature deals with single phase NPs, herein we report a nanoindentation study on the mechanical properties of silica-coated cobalt boride nanoparticles (CNPs).
(27, 30) In the past, Co−B materials drew considerable attention mainly because of a number of functional properties relevant to their soft ferromagnetism and additional applications such as surface coating (e.g., anticorrosion), catalysis for hydrogen storage, or fuel cells. Crystalline metal borides are also known for their mechanical properties and high hardness,
(31-33) but much less is known of these alloys when amorphous.
Such CNPs bear therefore great expectations in terms of mechanical behavior in view of the characteristic morphology displayed and of a distinctive set of nanoscale properties. Preliminarily, the characterization of the internal structure and the chemical composition via (HR)TEM and ICP (inductively coupled plasma) techniques revealed the complex composite nature of these CNPs. These nanocomposites are formed from smaller subunits of amorphous cobalt boride that precipitate and agglomerate during the synthesis process, driven by their high surface area (i.e., surface energy) and magnetic nature. These phenomena lead to spherical assemblies further coated with silica in which the internal bonds established between subunits and with the outer silica shell influence the final mechanical properties. Figure
1 includes (HR)TEM images of the CNPs yielding a size distribution of the magnetic core centered around 110 ± 20 nm and CNPs fulfilling the range between 70 and 150 nm. The (HR)TEM of a 150 nm CNP in Figure
1b (as synthesized) demonstrated the distribution of a large number of pre-existing radial “disruptions” in the magnetic cobalt boride core, that arise from its aggregated nature. Presently, we cannot rule out the possibility of these defects being nanovoids induced by incomplete agglomeration, although that may not necessarily explain their radial distribution. Those defects are closely spaced and extend from the surface deep into the core for several tens of nanometers. Structurally, they appear as material discontinuities and represent weak spots in the nanostructure, supposedly sensitive to shear stress in the defect direction. This scenario bears evident resemblance with the shear transformation zones (STZs) usually underlying the plastic deformation in glasses and amorphous alloys, where STZs are localized defects subject to stress induced transformations which release strain energy irreversibly, as reviewed by Schuh et al.
(34) However, while the underlying plastic deformation mechanism is indeed very similar, the STZs in the original theory have an orientation determined by the stress that is most likely to induce transformation of the STZ, whereas the orientation of CNP defects is here predetermined. The relation between the STZ framework and the CNPs should be further investigated in future work.
(35) Since TEM micrographs render the in-plane projected traces of the defects, it is currently impossible to ascertain whether they are spread over more than one plane in the spherical core. A closer observation of these radial discontinuities by means of dark field scanning transmission electron microscopy (STEM) analysis (Figure
1c,d) reveals average distances between the subunits forming the CNPs to be around 1−2 nm. The presence of these defects is an element of novelty compared to the defect-free homogeneous NPs tested to date.
(8-13) Additionally, selected area electron diffraction (SAED) analysis on a CNP confirms its amorphous nature, both on the silica shell and on the Co
2B magnetic core. This can be explained when considering the rate of crystallization to form nanoparticles by means of the coprecipitation method. Indeed, the rate of crystallization is usually neglected because constrained by reaction kinetics as minimizing thermal diffusion to deliver an amorphous solid. Boron is also a factor inducing amorphization of crystalline cobalt. Notably, the amorphous nature likely bears relevant implications on the CNP hardness, since the development of amorphous metallic materials (commonly termed as metallic glasses) shows an outstanding increase in strength over conventional crystalline ones, as underlined by Burgess and Ferry.
(36) Further evidence on the conformal core−shell composition were shown through a STEM (scanning transmission electron microscopy) analysis. STEM-XEDS elemental mapping of the composite (Figure
1e) shows the relative elemental distribution, with red areas corresponding to Co (Kα line) (as Co
2B) and green areas corresponding to Si (Kα line) (as SiO
2). Concerning the chemical composition, ICP analysis reflected boron content in the magnetic core of the CNPs to be around 10 wt %, which corresponds to the primary product (Co
2B) of the coprecipitation reaction followed in the synthetic procedure.
(37, 38) By a mechanical standpoint, the structural complexity of CNPs translates into a variety of possible deformation mechanisms that can be understood in terms of the interactions between the cobalt boride subunits themselves and with the outer silica shell. These interactions, generally electrostatic or due to van der Waals forces, constitute the basic bonding within the CNPs, potentially producing a more stable nanocomposite. Indeed, the interfacial interactions between the subunits inside the composite play a crucial role when governing their quality and physical characteristics.
(39) Poor bonding in between and with the silica shell may introduce, on the other hand, artificial defects such as voids that can consequently have a deleterious effect on the mechanical properties. Furthermore, their magnetic nature can also be related.
Individual
in situ compression tests of such CNPs were carried out to explore both their elastic and plastic behaviors, placing greater emphasis on plastic phenomena. Remarkably, both reverse and irreversible plasticity are unexpectedly observed, just like in single-phase defect-free NPs. Tests were performed on 20 randomly selected CNPs at a variety of loads between 30 and 300 μN, without crashing them. Such tests are subdivided into seven groups based on the applied maximum load (see Table S1 in the
Supporting Information). The
P−δ curves of the first six groups reported in Figure
2b−g demonstrate good reproducibility of the CNP response, especially at low strains. Overall, the CNP force response can typically appear to be divided into three regimes during the loading part, i.e., a quasi-linear elastic response (zone I), a plastic response with low hardening rate (zone II), and a final hardening with high hardening rate (zone III), as depicted in Figure
2a for a test run up to
P = 60 μN. The black curve corresponds to the force−displacement relationship during loading (compression), while the blue curve corresponds to the same relationship during the unloading (indenter tip withdrawal) after holding at the maximum load for 2 s (see the
Supporting Information, S2). During the unloading process, the tip was lifted and had to overcome the adhesion forces between itself and the CNPs before detouching (these adhesion forces may cause the load to be negative at the end of the unloading process).
An analogous behavior corresponding to this type of three-zones response was already described by Raichmann and co-workers for indented agglomerates held together by only a few strong bonds.
(40) In their study, the intermediate slope (zone II) was associated with interparticle forces followed by a higher slope, which they found characteristic of either agglomerate densification or deformation of individual particles. Analogously, when dealing with CNPs with cores consisting of smaller subunits as in our case, the mechanical behavior should primarily be related to the agglomerated nature revealed by the HRTEM images. Particularly, the behavior of the different CNPs within the same statistical sample may be associated with the number and orientation of the different internal interactions between subunits (holding them together to form the CNP) and the size of the CNPs themselves.
The six groups of compression curves in Figure
2 allowed characterizing the response of the CNPs during loading in each of the three zones. In general, since zone I ends roughly at 20 μN, we almost always tested the CNP up to zone II (around 40 μN) or III (>40 μN). The displayed responses, especially in group 1, demonstrate that data is highly repeatable in zone I but scatters in the plastic regime, as expected. However, tests in other groups confirmed the good repeatability of the response up to 60 μN.
The moderate slope in zone II can then be ascribed to a characteristic deformation mechanism, possibly proceeding by sliding and reordering of the intersubunit interactions (mainly van der Waals forces, electrostatic and/or magnetic exchange, and dipole−dipole interactions). To study these forces locally, one should operate a probe with spatial resolution of less than a nanometer, and with appropriate force resolution, of the order of tens of nN or less, presently out of reach. One more aspect to ponder is the degree of decohesion of the surface atoms in every subunit forming the CNPs, which in the case of this type of composites may be substantial. Such decohesion of the inner-surface atoms comes primarily from their coordination, and the extent of surface softening depends on strain-induced surface rearrangements.
(41) At a certain penetration depth (up to 15 nm), the slope becomes much steeper (zone III), a fact that can be explained by weighing several contributions, including permanent plastic deformation, subunits hardening, or densification inside the CNP upon loading and the magnetic nature of the composite. On the one hand, for example, these deformation mechanisms may involve the permanent breaking of the interactions established between the subunits forming the CNPs. However, on the other hand, the characteristic dynamic relaxation behavior associated with the magnetic NPs should also be considered as an active hardening mechanism. In fact, by taking into account the threshold size of the subunits forming the CNPs, sensible magnetic interactions reportedly arise so that their magnetic dipoles can be reoriented during deformation, thus resulting in absorption of energy and enhanced stiffness.
(42) These superparamagnetic nanostructures (e.g., nanoparticles embedded in polymer
(42) or Co
2B subunits in the case of the silica-coated cobalt boride nanospheres) are characterized by magnetic moments aligned in a preferred direction given by their crystallographic or morphological nature and their position with respect to the other magnetic nanostructures in the surroundings (magnetic interactions). In both cases, because the compression process changes the conformation of the surroundings, in principle, it also modifies the magnetic interactions and therefore the magnetic moment directions, which can be translated into magnetic energy changes associated with the strain-induced variations in magnetization direction.
3.1
Elastic Modulus and Hardness Extracting the Young’s modulus
ECNP and hardness
HCNP of CNPs is nontrivial. First, the mechanical properties of bulk cobalt boride do not seem to be well-known in the literature. In addition, the macroscopic properties of pure cobalt and (amorphous) silica cannot be taken as a reference, since they are known to independently vary considerably. An adapted refined Oliver−Pharr method
(43) (see a brief description in the
Supporting Information) was deployed to obtain approximate values of
ECNP and
HCNP from 30 μN tests summarized in Table
1. Calculations over just the first six tests yielded
ECNP = 159 ± 31 GPa and
HCNP = 4.2 ± 1.1 GPa. The modulus estimate agrees with results from finite-element (FEM) analysis using a reference for interpreting the experimental data (see Figure S3 in the
Supporting Information). It is also interesting to notice the relatively high hardness, consistent with the hardening effect reported for single phase NPs and with the cited observation by Burgess and Ferry.
(8, 36) Table 1. Summary of the Main Information and Results of Tests at 30 μN on Seven Randomly Selected CNPs (Group #1), i.e., the Particle Diameter (
D), the Residual (δ
C) and Maximum Displacements (δ
MAX), Young’s Modulus (
E), and Hardness (
H)
a | CNP # | D (nm) | δC (nm) | δMAX (nm) | ECNP (GPa) | HCNP (GPa) |
---|
group 1, PMAX = 30 μN | 1 | 119 | 6 | 12 | 141 | 3.7 |
| 2 | 114 | 8 | 13 | 170 | 4.0 |
| 3 | 111 | 6 | 11 | 163 | 4.3 |
| 4 | 117 | 9 | 16 | 162 | 3.6 |
| 5 | 91 | 5 | 10 | 205 | 6.3 |
| 6 | 123 | 9 | 19 | 110 | 3.2 |
outlier | (7) | (75) | (4) | (9) | (310) | (9.4) |
| average (1−6) | 159 | 4.2 |
| mean square error (1−6) | 31 | 1.1 |
The seventh observation (the CNP 75 nm in diameter) was discarded because it would inflate the
ECNP due to the much larger pressures associated with the smaller CNP radius. This is evident from the plot in Figure
3a, showing the size effect of the CNP diameter on the elastic properties. Related to this aspect, Varghese and co-workers reported the Co
3O
4 nanowire Young’s modulus to be size-dependent, exhibiting lower elastic modulus with increasing diameter (for diameters above 100 nm).
(16) Also, Chen et al. and Liu et al. reported the same behavior, either explained on the basis of a contribution from the surface stiffness
(7) or attributed to the reduced defect density as decreasing size, respectively. The latter reduced defect density explanation can also be invoked in the case of our CNPs, since the number of subunits of the magnetic core—and therefore the number of their interfaces—is decreased as size shrinks, similarly to the reduction of defect density in the cited crystalline nanowires. Such a pressure influence was already reported for homogeneous NPs by Mook and co-workers, who assumed a linear model,
E(
p) =
E0(
p) +
k·
p to describe their
E*−
p data (gathered with loads up and above 80 μN).
(11) Accordingly, Figure
3b displays the stiffening trend of modulus with force (pressure) in our data
ECNP (discarding data above 90 μN that seemed unrealistic). Interestingly, despite the fact that also the
HCNP data in Figure
3c exhibit a hardening tendency, no
p dependence appears to have ever been set forward for hardness, and there does not seem to be an evident physical rationale to do so here.
3.2
Strain Hardening Plasticity The reduction in slope (zone II) observed in the CNP responses in Figure
2 is associated with irreversible plastic deformation with strain hardening. The tendency of these load responses from different CNPs to become scattered after an initial overlap indicate that irreversible plasticity is a stochastic process influenced by the nanostructural disorder (i.e., radial disruptions and other defects) of these CNPs. Unlike what was reported by Gerberich et al.
(8) (and by nearly any cited work on NPs), the CNPs are not defect-free single crystals and the deformation is not a dislocation-mediated process but rather one related to the radial defects in the HRTEM image in Figure
1, in a manner very similar to the aforementioned STZs. Therefore, on the basis of the estimated
HCNP (i.e., a measure of the average pressure), the radial defects that are randomly located within the maximum shear region are thought to constitute intrinsic slip sites, activated under an average critical shear stress well above 1 GPa (see shear fields from FEM results in Figure S3 in the
Supporting Information). The fact that the network of defects is not totally random but displays a radial pseudo-order is crucial to this rationale. Eventually, as more native defects become involved under increasing shear stresses, these deformation zones will increase in density and come to interact, thus causing strain hardening. In this view, the passage between zones II and III could just express the transition from a regime of
dilute plastic deformation in a few intersubunit spots to one of high interdefect interactions. By the end of zone II, different plastic systems can in fact begin to interfere and form obstacles (e.g., just like jogs and kinks in metal dislocation) responsible for the increased hardening rate in zone III. Only future TEM monitored in situ tests will allow deeper insight of this plastic phenomenon and refine our formulation to account for other secondary mechanisms (e.g., intersubunit plasticity, densification, magnetic coupling, etc.).
The hypothesis about the plastic nature of the deformation in zones II and III was confirmed here by a series of repeated tests to investigate further the strain hardening behavior of CNPs. In the past, hardness and stiffness of individual Si NPs were found to increase after each run in similar experiments.
(8) Figure
4a displays data from three repeated runs on CNP #5 at increasing load levels (i.e., 30, 50, and 90 μN). Noticeably, the characteristic three-zones response is only observed in the first hit, while the responses from the following hits denounce an increased hardness and seemingly just one mechanical regime, without any singularity or evident slope reduction that hint to the mentioned zones II−III transition. These findings seem consistent with the formation of irreversible nonmovable defects that would pin intersubunits slippage and interfere with eventual plastic deformation.
None of the known nanoscale strain hardening mechanisms reported in the literature seem to apply in the present situation. In particular, recent work
(22) about the TEM analysis of microcompression tests in the context of metal nanoscale pillars demonstrated that the initial compression may induce a so-called mechanical annealing, clearing the nanostructure from the native content of dislocations and other lattice defects, and resulting in a harder material as well. In our case, though, the strengthening is believed to arise from the previous classical strain hardening mechanism because of a number of factors, namely, (i) the pre-existing distribution of radial defects in the CNPs, (ii) the existence of the strain gradient in these CNPs as opposed to cylindrical pillars, (iii) the nonmetallic and noncrystalline nature of the cobalt boride core, and (iv) the presence of the silica outer layer which may constrain motion and deformation of the Co
2B subunits.
To highlight the strengthening effect of this strain hardening mechanism, we have performed other tests proceeding in reverse. Rather than just increasing the maximum load from below, we examined the responses of a CNP subjected to repeated tests at increasing load level after one first hit at high load. Two examples are shown for CNP #12 (b) and CNP #15 (c) in Figure
4. In the former case, after the initial compression at 60 μN, which reached zone III, two more tests at 60 and 80 μN rendered an almost elastic response with residual deformations of about 2 nm that are unusually small for that load level. Such results demonstrate that the first test induces a strengthening transformation on the material, seemingly permanent. This conclusion is supported by the four tests on CNP #15, i.e., an initial compress at 90 μN and three tests at 30, 50, and 90 μN.
These findings evidently bear important implications for MEMS and NEMS by a design and manufacturing perspective. However, caution is required at the moment, as the exact deformation mechanism(s) underlying these plastic events remain controversial. We reiterate that a handful of other possible origins can be invoked for the irreversible plastic deformation in the complex deformation process above. We make no attempt to survey all possibilities but briefly outline two instances. First, alternative meanings of zone II could encompass the possibility of such a regime to be a consequence of a failure of the composite structure. For example, indentation of the CNP, delaminating between core and outer silica skin, or failure of the silica layer due to excessive shear could all induce in principle a similar load response. However, below 80 μN, atomic force microscopy (AFM) imaging highlighted neither indentation marks of significant depth (as compared to the diameter) nor detectable singularities in the particle cross section that could hint to these types of failure modes. Second, besides purely mechanical arguments, other chemo-physical mechanisms may intervene, such as diffusion-assisted subunits boundary deformation, or the magnetic nature of the composites. As for the latter, we ought to acknowledge that these CNPs—and magnetic NPs in general—above a threshold diameter show a nonzero induced dipolar magnetization, the direction of which depends on the N
el relaxation process caused by the orientation of the magnetic vector within the particle itself. This N
el relaxation process may dominate the dynamic response behavior of the magnetic CNPs depending on the relaxation time.
(44, 45) The characteristic N
el relaxation time (τ
N) is related to the stiffness of the nanostructured material and can concur to explain their mechanical properties. As already established for a polymer with magnetic NPs embedded in it,
(42) if τ
N of the nanocomposites indicates relaxation times comparable to or longer than the time scale of the nanoindentation process (this is indeed the case of the CNPs analyzed as being in their blocked state during the indentation process, see Figure S4 in the
Supporting Information), there is an energy penalty associated with the deflection of the magnetization. However, the determination of the exact magnetism-related mechanism is left for future research.
3.3
Reverse Plasticity Several authors have observed a partial recovery of the plastic deformation after a certain elapsed time after test.
(9, 46-48) Zou and Yang
(13) indeed reported a total recovery of the plastic deformation of a silica NP after a 30 μN indentation, as verified via AFM. In dislocation-free metal nanostructures, this plastic deformation is likely caused by the nucleation or “insertion” of a number of dislocations at the contact surface. These defects are associated with large internal stresses which render them unstable upon load removal and make them disappear by egression back out from the same free surface. This problem has actually been investigated with atomistic models.
(49) Such a reverse plasticity has also been observed in our CNP tests. To prove its occurrence, the particle in test #9 was compressed twice sequentially at the same load level of 40 μN and then imaged with the nanoindenter before and after each test. The loading curves corresponding to the two hits are plotted in Figure
4d and appear almost identical due to the reversible deformation. In fact, during the first test, a total residual deformation of 6 nm (see Table S1,
Supporting Information) was recorded, but AFM imaging (Figure S5,
Supporting Information) confirmed the total recovery afterward like what was reported by Zou and Yang.
(13) After the second hit, only a local (unsymmetrical) flattening of the CNP is noticeable at the contact zone with the tip, which may have been likely triggered by either an imperfect tip positioning or by the local failure at a weak spot in the composite structure. The occurrence of reverse plasticity may be somewhat unexpected here, due to the presence of radial defects and the amorphous structure of the CNPs (most literature reports this phenomenon in dislocation free crystalline metal NPs). In the CNPs, this phenomenon may be caused by the confining action of the outer silica layer that blocks the egression of agglomerates sheared off from the Co
2B core toward the outer layer. This hypothesis needs to be investigated further but would provide an interesting strengthening mechanism to inhibit permanent plastic deformation in MEMS and NEMS. In the future, it should be ascertained whether the reverse plastic deformation in CNPs can trace back to some sort of rupturing−reforming process of intersubunit bonds in the core during loading and unloading cycles. Noteworthily, the CNP response in Figure
4d is limited to zone I, which we associate with the occurrence of total reverse plasticity. When the deformation reaches zones II and III, irreversible strain hardening was indeed generally observed, as discussed above. Such results point out that residual deformations from a nanomechanical test should not be trusted alone but always checked against (visco-)elastic reverse deformation for accuracy. As for MEMS and NEMS design and manufacturing, reverse plasticity represents a peculiar nanoscale behavior of volume constrained matter to be either avoided/limited or exploited in material-based operations, depending on the engineering needs.