Imaging of Transient Structures Using Nanosecond in Situ TEM

See allHide authors and affiliations

Science  12 Sep 2008:
Vol. 321, Issue 5895, pp. 1472-1475
DOI: 10.1126/science.1161517


The microstructure and properties of a material depend on dynamic processes such as defect motion, nucleation and growth, and phase transitions. Transmission electron microscopy (TEM) can spatially resolve these nanoscale phenomena but lacks the time resolution for direct observation. We used a photoemitted electron pulse to probe dynamic events with “snapshot” diffraction and imaging at 15-nanosecond resolution inside of a dynamic TEM. With the use of this capability, the moving reaction front of reactive nanolaminates is observed in situ. Time-resolved images and diffraction show a transient cellular morphology in a dynamically mixing, self-propagating reaction front, revealing brief phase separation during cooling, and thus provide insights into the mechanisms driving the self-propagating high-temperature synthesis.

Transmission electron microscopy (TEM) has evolved dramatically in recent years with the development of monochromation and spherical aberration correction (13) facilitating sub-angstrom spatial resolution. In situ TEM studies have also progressed from simple heating and cooling experiments to include capabilities such as nanoindentation (4) and environmental cells (5, 6). The TEM is already a powerful tool for material characterization in diverse scientific fields, but there is a need to incorporate fast time-resolution capabilities into EM.

In the past, direct electron imaging has been essential to uncover phenomena that are difficult to distinguish with diffraction (79), for instance, dislocation dynamics. Electron interrogation methods also have the potential to provide higher sensitivity and resolution compared with laser or x-ray methods. This is possible because of bright electron sources, the ability to control and focus electrons for different contrast mechanisms, and the stronger interaction of electrons with matter (10), resulting in a broad class of observable samples and length scales.

Traditional in situ TEM spatially resolves microstructural details of phase, structure, and morphology; however, it cannot collect data with acquisition times less than 1 ms. This is often orders of magnitude too slow to capture the essential material details of interface motion, crystal formation, twinning, and many other fundamental material processes. Many such processes are nonrecurring, necessitating single-shot techniques that capture images or diffraction patterns in a single brief exposure. At nanosecond time scales, such transient data are only attainable with the use of a single, short electron bunch with very high peak current by TEM standards (microamperes to milliamperes). Nanosecond-scale in situ TEM has been achieved using a conventional TEM modified to introduce a laser for stimulation of a photoemission electron source (11, 12). Figure S1 illustrates how a 15-ns yttrium-lithiumfluoride–Nd laser frequency quintupled to 211 nm is directed toward the electron source, generating in excess of 2 × 109 photoemitted electrons in a single 15-ns bunch. Up to 5 × 107 of these electrons can be collimated into a small (micrometer-scale) area to probe the sample. This very high current density enables nanosecond single-shot imaging and electron diffraction. Imaging and diffraction provide complementary information, where diffraction reveals the evolution of crystallographic structure (13) (or amorphous nature) and imaging reveals details of the real-space evolution of fundamental nonrecurring processes in nanoscale materials dynamics.

The single-shot capability provides a pump-probe platform that can acquire a “snapshot” image or diffraction pattern of a transient process. The dynamic processes are stimulated in a TEM specimen by an additional pulsed laser (yttrium-aluminum-garnet–Nd), the pump laser. This laser heats or ablates a small area of the sample, initiating the process that will later be captured by the electron pulse. The time delay between the two pulses can be precisely controlled within a scale ranging from nanoseconds to hundreds of microseconds, with a timing jitter of ±1 ns. Longer multi-microsecond time delays used in this study cannot be achieved in a single laser–plus–beam splitter system without using optical paths of thousands of meters. By accumulating single-shot pump-probe observations with varying time delays, one can develop an understanding of each stage in the evolution of a material process.

Reactive multilayer foils (RMLFs), also termed nanostructured metastable intermolecular composites, are layers of reactant materials that undergo exothermic, self-propagating reactions when layer mixing is induced by an external energy source (Fig. 1). These nano foils exhibit mobile, high-temperature reaction zones where atoms of adjoining layers diffuse across the interfaces, wherein velocity and temperature can be manipulated by composition and geometry of the component materials (1416). They are used as customized heat sources for rapid fuses, biological neutralization, and joining of materials by means of localized heating rather than global device heating (17, 18).

Fig. 1.

Schematic of RMLF reaction-front propagation where mixing is initiated by a laser and continues to travel along the foil until the layers are consumed and converted into reacted intermetallic.

The reactive foils contain stored chemical energy in the form of layered structures with <1 nm of interdiffusion at the interfaces (fig. S2). The reaction-front velocity (reaching ∼10 m/s) is related to bilayer thickness, with an exception in thin bilayers (∼10 nm) where intermixing during deposition retards the reaction propagation (15). Maximum temperatures attained during mixing vary greatly based on composition and geometry, reaching upwards of 1750 K (19). Little is known about the dynamic and transient events that transpire in the proximity of the reaction front, although these will govern the structure, mechanical properties, and performance of the reacted materials. The combination of nanoscale geometry and rapid velocity often make traditional characterization methods incapable of direct observation of the self-propagating high-temperature synthesis.

We use dynamic TEM (DTEM) to observe the RMLF reaction front for rapid phase-transition times and metastable morphologies. RMLF experimental samples are composed of five bilayers of Al/Ni0.91V0.09 (at a 2:3 atomic ratio, totaling 125 nm in thickness) and mounted in Cu meshclamping grids. Through fast optical imaging at 2 μs per frame (fig. S3 and movie S1), we found that when initiated by a 1064-nm laser pulse in an external vacuum chamber, the reaction front travels across the sample at a velocity of 13 m/s. Because the spatial resolution was 35 μm per pixel, the optical videos did not show additional definitive information. We then conducted DTEM experiments at the same pump-laser conditions [80-μm1/e2 radius, 12-μJ energy, 3-ns full width at half maximum (FWHM) pulse duration]. The orientation of the experimental setup illustrated in the schematic (fig. S1, inset, and fig. S3) shows how the reaction front radiates out in all directions from the reaction initiation site in the freestanding RMLF membrane. By initiating the mixing reaction several hundred micrometers away from the observation area, the reaction front of the intermetallic formation zone is observed while it is in a steady state of propagation. Delays between the reaction-front arrival and the probe electron pulse were varied from –0.7 to 25 μs, before the arrival and long after the front has passed.

At low magnification with a field of view of ∼500 μm2, the reaction front is visible after 1 μs as it travels away from the drive-laser initiation site (fig. S4). At increased magnification, the bright field images (Fig. 2) demonstrate the transient morphology of the reaction front. The reaction front is closely followed by the elongated fingerlike structure, defining the interface between the reacted and unreacted material. These cellular features of Fig. 2 have a periodicity of ∼600 nm wide and are greater than 40 μm long (equivalent to 3 μs of propagation time), showing a distinct formation and gradual termination, which proves their transient character.

Fig. 2.

RMLF mixing front is defined in (A), and after longer times (2 μs), the structure moves further (B) and is no longer present (5 μs) after the Al/Ni layers have completed mixing (C). The images shown here are plan-view bright field images. Dynamic snapshot images of the mixing reaction-front zone reveal a transient cellular structure.

The images in Fig. 2 are in plan view (with the viewing direction perpendicular to the metallic layers), and the cellular features are normal to the bilayer structure in the direction of the yellow arrow in the schematic Fig. 1. The size scale of the transient structures (600-nm periodicity) greatly exceeds the bilayer periodicity of the RMLF (25 nm); thus, the observed cellular structures are not directly related to the metalliclayer structure but instead suggest a transverse patterning mechanism inherent in the reaction process. This is reminiscent of binary alloy solidification in the midst of slight undercooling and also of solid solution eutectoid transformation (20, 21). These models have shown that similar structures are known to arise from the interplay of thermal gradients and atomic interdiffusion, the same physical principles that govern the RMLF front propagation.

Analysis of 15-ns-resolution diffraction data (Fig. 3) reveals the initial formation of the intermetallic ordered B2 phase, NiAl, from the separate face-centered cubic (fcc) Al and Ni0.91V0.09 layers. The NiAl formed in under 300 ns after the arrival of the front. This implies that much of the interdiffusion is confined to the close proximity of the reaction front, as indicated from the time-resolved images showing the sharp onset of the cellular microstructure. The electron pulse has a 15-ns FWHM time resolution and passes through a condenser aperture so that it interacts with a round sample area that is 11.2 μm in diameter (making the obtained phase information accurate to this dimension), which is far smaller than the width of the metastable cellular formation region imaged in Fig. 2. Diffraction patterns from the unreacted bilayers appear as uniform nanocrystalline ring patterns, consistent with the ∼10-nm grain diameters in the as-grown layers. Diffraction data taken only 2.3 μm ahead of the reaction front shows little indication of a phase transition. As the hot reaction front passes the sample area being probed by electrons, background intensity increases and the diffraction ring patterns become diffuse because of thermal effects of the exothermic mixing reaction. These rings are rotationally averaged to reduce noise. The film remains primarily crystalline throughout the dynamic mixing event, and this is confirmed by the uninterrupted presence of distinct rings. The timing of the hot reactionfront velocity is measured by DTEM imaging of the front of the cellular structure. When this information is combined with the diffraction data, it is evident that the solid NiAl phase begins to form immediately after the arrival of the hot reaction front.

Fig. 3.

Dynamic single-shot diffraction with 15-ns time resolution of regions before, during, and after the exothermic mixing reaction front has passed. The times indicated at right are in relation to the reaction front, set at t = 0 s. The crystal structure clearly changes from separate fcc Al/Ni and Ni/V layers to an intermetallic B2 structure NiAl phase within 300 ns after the arrival of the hot reaction front. a.u., arbitrary units.

In the reaction-front region, the intermetallic NiAl nucleates and continues to grow near the hot reaction front as the material enters a two phase field of NiAl + liquid, consistent with known thermodynamics of Ni/Al binary mixtures (22). The Al atoms gain mobility because pure Al melts at the relatively low temperature of 933 K, far below the expected temperature of the reaction front. Basic calculations of heat emitted by the mixing of Al and Ni indicate that the temperature reaches ∼1700 K [ΔH ≈ 77 kJ mol–1, cp Ni/Al = 24 J mol–1 K–1 (23, 24), where ΔH is the change in enthalpy and cp is heat capacity at constant volume], consistent with the measured ∼2% thermal expansion from the NiAl diffraction, using α = 15 × 10–6 K–1 (where α is the coefficient of linear thermal expansion) (25). The as-grown layers are stoichiometrically close to NiAl at a 2:3 Ni-rich ratio, and the final reacted phase is NiAl (the B2 phase revealed by diffraction as in Fig. 3). Therefore, knowing that crystalline solid is present at all detectable times, it is logical that the propagating zone is moving along a thermodynamic trajectory through the NiAl + liquid phase field. This result would imply that a fraction of material at the reaction front is liquid and near equilibrium, enhancing the material transport.

The observed contrast of the cellular features may be enhanced by the liquid phase because of increased thermal diffuse scattering. On occasion, near a defect that acts as a heat sink, a small vicinity of the fully reacted foil retains cellular-type structures. The observation of such remnant features with mass-thickness contrast in high-angle annular dark field scanning TEM (fig. S5) reveals that after cooling to room temperature, the regions of narrow, dark intensity in Fig. 2 correspond to thickness modulations in the projected z direction. We confirmed the material thickness modulation using electron energy-loss spectroscopy to measure the ratio of inelastically scattered–to–elastically scattered electrons at varied positions along the specimen. The rows of increased thickness further support the idea that liquid, which may coalesce to reduce surface energy and cause this thickness modulation, is one of the phases present. This hypothesis is consistent with the thermodynamics and explains how the thickness modulations ∼600-nm apart could form on the observed time and length scales (∼100 ns judging from the micrometer-scale sharpness of the reaction front in Fig. 2A and the 13-m/s propagation speed). These time and length scales are not consistent with an alternative hypothesis that the structures arise from purely solid-state diffusive processes, which are orders of magnitude too slow, even at 1700 K (26).

The dark intensity between the cells fades away in the last micrograph of the series in Fig. 2, at a point in time where the reaction is long since complete. The solid solubility range of the NiAl B2 phase increases as the temperature drops (22), so that at room temperature nearly all of the excess Ni could be reabsorbed into a stable homogeneous B2 structure with a Ni:Al ratio close to 3:2. Post-mortem TEM examination indicates that this happens in nearly all cases, with frozen-in structures (such as those in fig. S5) being rare exceptions associated with defects. The cellular morphology is not normally present at completion.

We have obtained single-pulse nanosecondscale TEM data in both diffraction and imaging modes, which are necessary to study the propagation and behavior of energetic nanolaminates in situ. With the use of 15-ns imaging, we have observed transient structures produced by the self-propagating high-temperature synthesis, revealing lines of mass-thickness contrast due to cellular phase formation of an ordered B2 NiAl phase and liquid. At such high formation temperatures (∼1700 K), these materials are now known to exhibit transverse self-ordering reminiscent of cellular binary solidification mechanisms. We have established that the DTEM is proficient for nanosecond science in a TEM for direct nanoscale characterization of irreversible, dynamic phenomena. It is notable and exciting to find spontaneous, rapid formation of ordered structures in materials far from equilibrium, which is also an important step for essential comprehension of RMLF performance in applications.

Supporting Online Material

Figs. S1 to S5

Movie S1

References and Notes

View Abstract

Navigate This Article