Ultrafast low-energy electron diffraction in transmission resolves polymer/graphene superstructure dynamics

See allHide authors and affiliations

Science  11 Jul 2014:
Vol. 345, Issue 6193, pp. 200-204
DOI: 10.1126/science.1250658

Probing interfaces with electrons

When molecules move on surfaces, they behave differently from when inside a solid. But surface layers give off limited signals, so to probe these systems, scientists need to act fast. Gulde et al. developed an ultrafast low-energy electron diffraction technique and used it to study how a polymer moved and melted on a graphene substrate (see the Perspective by Nibbering). After hitting the sample with a laser pulse, energy transferred across the graphene-polymer interface, the polymer film became less orderly, and an amorphous phase appeared.

Science, this issue p. 200; see also p. 137


Two-dimensional systems such as surfaces and molecular monolayers exhibit a multitude of intriguing phases and complex transitions. Ultrafast structural probing of such systems offers direct time-domain information on internal interactions and couplings to a substrate or bulk support. We have developed ultrafast low-energy electron diffraction and investigate in transmission the structural relaxation in a polymer/graphene bilayer system excited out of equilibrium. The laser-pump/electron-probe scheme resolves the ultrafast melting of a polymer superstructure consisting of folded-chain crystals registered to a free-standing graphene substrate. We extract the time scales of energy transfer across the bilayer interface, the loss of superstructure order, and the appearance of an amorphous phase with short-range correlations. The high surface sensitivity makes this experimental approach suitable for numerous problems in ultrafast surface science.

The investigation of atomic-scale dynamics with high spatiotemporal resolution yields insights into ultrafast structural reorganizations associated with energy transfer or phase transitions. Substantial progress was made in establishing methods for the time-resolved structural analysis of bulk media, including ultrafast implementations of x-ray crystallography (13), high-energy electron diffraction (46), and microscopy (79), as well as time-resolved x-ray and electron spectroscopy (10, 11). In contrast, structural dynamics at surfaces, interfaces, and ultrathin films remain largely elusive, as the surface signal in both x-ray and high-energy electron diffraction is typically masked by large bulk contributions. This limits our ability to study quasi–two-dimensional (2D) systems exhibiting characteristic phase transitions and topologically controlled ordering (12, 13), as well as the dynamics of surface reconstructions and complex adsorbate superstructures (14, 15). Ultrafast electron scattering in grazing incidence enhances the surface signal (14, 16) but faces particular challenges in quantitative diffraction analysis.

Optimal surface sensitivity would be attained with an ultrafast implementation of low-energy electron diffraction (LEED). At electron energies of tens to a few hundreds of electron volts, scattering cross sections are strongly increased, which allows for probing depths of only a few monolayers and has made LEED a widely used tool for surface structure determination. However, at such low energies, it has proven exceedingly difficult to implement pulsed electron sources that fulfill the requirements of an ultrafast diffraction experiment (1719), that is, short pulse duration and low beam emittance. Laser-triggered electron emission from nanoscale photocathodes (20, 21) is expected to resolve some of these issues (2224), providing well-collimated low-energy electron pulses and a temporal resolution that is comparable to state-of-the-art ultrafast x-ray or high-energy electron diffraction. Motivated by these prospects, we have undertaken the development of a new diffraction apparatus.

We have developed transmission ultrafast LEED (T-ULEED) based on a nanometric needle photocathode and demonstrate its capability to resolve atomic-scale structural dynamics of surfaces and monolayer films with a temporal resolution of a few picoseconds. Specifically, we studied the ultrafast laser-driven dynamics of a polymer superstructure on freestanding monolayer graphene.

In the laser-pump/electron-probe scheme (Fig. 1A), the sample is excited out of equilibrium by amplified femtosecond laser pulses (800 nm wavelength, 80-fs pulse duration, repetition rate 10 kHz, focal diameter about 100 μm). To minimize hot electron emission from graphene (25), the pump pulse is temporally stretched to 3 ps by dispersion, which, however, is still sufficiently short to resolve the processes described below. The pump-induced structural dynamics are probed by ultrashort electron pulses emitted from a sharp tungsten tip (50-nm radius of curvature), triggered by the second harmonic of the laser. These electron pulses (up to 100 electrons per pulse) are collimated and focused onto the sample at variable electron energies using an electrostatic lens assembly in a geometry that we have recently studied numerically (22). Scattered electrons are subsequently recorded in a transmission geometry by a phosphor screen microchannel plate detector (MCP, Hamamatsu F2226-24P). With our laser-triggered low-energy electron source, diffraction patterns can also be recorded in back-reflection and for a range of electron energies, as demonstrated in the supplementary materials (26) (figs. S6 to S8).

Fig. 1 Ultrafast low-energy electron diffraction in transmission and polymer/graphene bilayer system.

(A) Schematic of the laser-pump/electron-probe setup. (B) Diffraction pattern of polymer/graphene bilayer system (at increased electron energy of 1 keV; integration time 1 s). (C) Projection image of one of the samples (450 eV). Dark areas indicate bilayer coverage. (D) Idealized sketch of stripe-like polymer superstructure domains of two different orientations. (E) Zoom into area denoted in (D). Solid (dashed) arrows indicate lattice vectors for real (reciprocal) space unit cell of graphene. Chain spacing denoted by a. (F) Central part of diffraction pattern indicating Miller indices and reciprocal distance a* = 2π/a = 1.47 Å−1.

The electron pulse duration and the spatial and temporal overlap (delay time, Δt = 0) of the laser-pump and electron-probe pulses are determined via a transient-electric-field effect near a bare transmission electron microscopy (TEM) copper grid. Upon excitation of a single copper grid bar with high peak intensity (fluence up to 30 mJ/cm2, unstretched pump pulses), a dense electron cloud is emitted (25, 27), which may lead to a spatial deflection of the passing electron pulse (Fig. 2A). Projection images of the grid before and after Δt = 0 are shown in Fig. 2B, acquired by defocusing the pulsed electron beam. The central distortion in the lower image indicates the extension of the pump-induced electron cloud. Using a collimated electron beam passing a single mesh cell, the effective temporal resolution of the diffraction experiment can be evaluated. Figure 2C shows the delay-dependent electron signal within a detector segment corresponding to a few-micron-sized mesh region. The observed transient displays a 2-ps width for 450 eV electron energy. This represents an upper bound for the electron pulse duration when using a partial beam because both plasma-cloud formation and beam traversal times limit the probing speed of the transient-electric-field method at low energies to the order of a picosecond. Averaging over the deflection of the entire collimated beam yields an upper limit of about 6 ps.

Fig. 2 Temporal characterization of low-energy electron pulses.

(A) Schematic of the transient-electric-field effect used for determination of the electron pulse duration. (B) Projection images of bare TEM grid recorded before and after time zero. (C) Delay-dependent intensity change for collimated beam at an electron energy of 450 eV (open circles). Red line, error function fit to transient intensity change. Assuming a Gaussian pulse profile and a step-shaped transient-electric-field response yields an electron pulse duration (full width at half maximum) of about 2 ps.

Using these experimental capabilities, we have studied the dynamics of a freestanding polymer/graphene bilayer. The relevance of this model system entails at least two aspects. From the perspective of polymer physics, considerable interest exists in thin polymer films on solids and the role of spatial confinement on the dynamics or key characteristics of phase and glass transitions (28, 29). Ultrathin films of monolayer dimensions, on the other hand, will exhibit structural and dynamical features governed by short-range polymer-substrate interactions involving in-plane atomic corrugation (30, 31). Furthermore, increasing efforts are devoted to developing graphene-based heterostructures, with polymers serving both as important compounds in graphene transfer and as interface candidates (32, 33).

We prepared an ultrathin atactic poly(methyl methacrylate) (PMMA) layer adsorbed to freestanding monolayer graphene on a TEM grid (26). A projection image of a partially covered grid is shown in Fig. 1C. Local diffraction images from selected sample regions, as shown in Fig. 1B, can be recorded by passing the electron beam through a single mesh cell. The diffraction pattern displays peaks corresponding to the sixfold symmetry of single crystalline graphene and additional spots closer to the central beam stop. The latter are associated with the polymer adsorbate, which forms a superstructure that is orientationally linked to the graphene substrate.

We use cryo-TEM in local diffraction mode to identify the polymer order on the scale of few tens of nanometers, finding a reduced symmetry of the diffraction pattern with only two superstructure peaks (fig. S4). This implies that the polymer chains locally exhibit a stripe-like order, registered to the graphene substrate, and with a well-defined periodicity a = 4.26 Å, doubled relative to that of the graphene lattice (Fig. 1, D and E). The angular correlation with the substrate allows for three equivalent superstructure domains rotated by 60° with respect to each other (Fig. 1F) (denoted by half-integer Miller indices). The stripe-like structure follows from the tendency of the polymer strands to adapt to the periodically corrugated adsorption potential by forming 2D folded-chain crystals on the substrate. Similar behavior has been reported for PMMA adsorption on graphite and mica surfaces (34, 35), even for atactic polymers and with the microscopic structure depending on tacticity (35). In contrast to the ordering perpendicular to the chains (36), the diffraction images show no clear sign for pronounced order of the atactic polymer along the chains.

We have found that the polymer superstructure can be reversibly melted by laser excitation. Figure 3 summarizes the changes in the diffraction maps induced by the optical pump. The graphene and superstructure diffraction spots of the bilayer system before and after excitation are shown in Fig. 3A. A normalized difference image is displayed in Fig. 3B. Whereas no prominent changes are observed for the graphene peaks, the intensity of the superstructure spots substantially decreases. The structural evolution is illustrated in Fig. 3C, showing difference images at three pump-probe delays. These difference maps display two main features, namely a reduction of the superstructure intensity (blue spots) and an increase of diffraction at smaller scattering angles corresponding to an in-plane momentum transfer k|| of less than 1.25 Å−1 (red disc).

Fig. 3 Time-resolved diffraction maps of superstructure melting.

(A) Diffraction images for negative delay (left) and Δt = 600 ps (right). Pump fluence: 6 mJ/cm2. Electron energy: 450 eV. (B) Difference of images shown in (A) with intensity In normalized to the superstructure peak intensity I0 = I(Δt < 0) at negative delay times: Int)=[I(Δt) − I0]/I0. (C) Normalized difference images for central region [dashed rectangle in (B)] for several delay times. (D and E) Relative change of radial diffraction intensity (averaged over azimuthal angles) for isotropic (D) and sixfold symmetric (E) components (26). (The darker region around the bottom superstructure spot arises from reduced MCP quantum efficiency.)

To differentiate the evolution of the inner disc from that of the superstructure spots, we decompose the diffraction changes into an isotropic and a sixfold symmetric component in Fig. 3, D and E, respectively. Specifically, we compute angular averages of the images in azimuthal segments around and in between the spots, and plot the diffraction changes along the radial direction (fig. S2). The increase of small-angle scattering (Fig. 3D) is peaked around k|| = 1.1 Å−1, which is also discernible as a ring shape in Fig. 3C (central panel). The analysis in Fig. 3, D and E, shows a strong dependence on scattering angle in the time scales of positive and negative signal changes. For example, while the spot intensity decrease is nearly complete after 160 ps (Fig. 3E, blue line), substantial relaxation still occurs in the nearly isotropic small-angle region at later times (Fig. 3D).

Figure 4 presents a more quantitative picture of the diffraction changes based on time curves and fluence dependencies. The delay-dependent reduction of the superstructure diffraction intensity, representing the loss of crystalline order, occurs on a time scale of ~105 ps (Fig. 4A, blue triangles). This temporal evolution is accompanied by an inward shift of the remaining spot intensity (Fig. 4B), which amounts to an increase of the polymer interchain distance of about 0.23 Å or 5%. The concurrent appearance of the small-angle diffraction disc is evaluated in Fig. 4A (red circles, averaged over the entire disc region). We find that the time scale of the diffraction increase is in fact a strong function of the scattering angle, with the slowest dynamics occurring at the smallest momenta, as quantified in the inset of Fig. 4A.

Fig. 4 Structural dynamics: Time constants and fluence threshold.

(A) Delay-dependent superstructure diffraction spot (blue triangles) and disc (red circles) intensities, normalized to negative delay. Solid lines are exponential fits to experimental data. (Inset) Momentum-dependent time constants of small-angle scattering increase. (B) Expansion of crystalline PMMA component. Pump fluence for A and B: 6 mJ/cm2. (C) Fluence-dependent superstructure diffraction intensity (green squares) and expansion (blue triangles), evaluated at Δt = 600 ps. Solid lines, guides to the eye. Dashed line, threshold fluence.

We have recorded similar dynamics for multiple samples and observe a distribution of characteristic superstructure melting-time constants of 95 ± 25 ps, which is likely caused by slight variations in sample preparation and storage times. Recrystallization after laser-induced melting of the superstructure occurs on time scales of few tens of microseconds, rendering the repetition rate of 10 kHz suitable for reversible studies.

To corroborate the notion of superstructure melting, we have carried out measurements as a function of pump fluence. Figure 4C displays the fluence-dependent change in superstructure peak intensity (green diamonds) relative to the unperturbed sample, together with the associated structural expansion of the crystalline phase (blue triangles), each evaluated at a delay of 600 ps. Both observables exhibit a threshold behavior at a fluence of 3 mJ/cm2 (dashed line). Beyond this threshold, a strong linear decrease in diffraction intensity is found, and it is accompanied by a kink in the otherwise linear fluence-dependent lattice expansion. These marked changes in the physical properties of the superstructure upon differential energy deposition evidence the qualitative characteristics of a phase transition.

We interpret the structural dynamics as follows: The initial state of the system is characterized by strongly physisorbed polymer chains within a periodic adsorption potential. The presence of both polymer-substrate and polymer interchain interactions determines the equilibrium state as displaying 2D folded-chain crystallites with three equivalent domains orientationally correlated to the graphene substrate. The widths of superstructure diffraction peaks from various samples indicate typical correlation lengths spanning at least 5 to 10 chain spacings of 4.26 Å each. The strength of the PMMA peaks relative to those of graphene suggests both considerable overall crystallinity and substrate coverage. Because the polymer is optically transparent, the laser excitation primarily deposits energy in the graphene, giving rise to a lattice temperature increase of about 540 K at an incident fluence of 3 mJ/cm2 (26). Whereas the graphene degrees of freedom equilibrate on times comparable to the pulse duration used, coupling to the overlayer is considerably slower. The structural dynamics at fluences below the threshold for loss of crystalline order are valuable to estimate the threshold temperature for superstructure melting and the energy coupling times. In the absence of a phase transition, the energy transfer to the adsorbate causes thermal expansion of up to 0.05 Å. Using the thermal expansion coefficient of bulk PMMA for a rough estimate yields a threshold temperature of about 165°C (26), close to the melting temperature of the bulk polymer (values ranging from 130° to 160°C). This expansion occurs on a comparatively short time scale of about 43 ps (fig. S1). As the thermal expansion itself should be more rapid given the PMMA sound velocity and the estimated domain sizes, this time constant likely represents the actual energy transfer time across the graphene/polymer interface. Assuming a monolayer polymer film (about 5 Å thickness), this time constant implies thermal Kapitza resistances around 8 × 10−8 m2K/W, which is well in the range of literature values for a polymer/carbon nanotube interface (26).

Above the fluence threshold for superstructure melting, qualitatively different structural features are observed, including (i) a loss of polymer-substrate registration, (ii) the appearance of an amorphous phase with enhanced low spatial frequency components, and (iii) an accelerated expansion of remaining crystallites. Notably, these processes occur on time scales longer than the mere layer heating and reflect the intrinsic polymer reorganization (see Table 1): (i) Crossing the threshold, increased thermal fluctuations release a growing fraction of mobile polymer chain segments out of substrate registration, in particular at domain boundaries and chain folds. Considering the diffraction peak decrease, this process occurs on a 100-ps time scale. (ii) Subsequently, mobile segments form expanded amorphous structures, which are no longer angle-correlated to the graphene substrate, and contribute to the appearance of the inner diffraction disc. Depending on their spatial frequencies, these rearrangements proceed over durations beyond 300 ps for the range of scattering momenta imaged here (Fig. 4A, inset). The peaked feature in the radial diffraction intensity of the amorphous phase implies a preferred spatial correlation period of 5.7 Å, which is substantially expanded by 25% with respect to the crystalline phase. (iii) The continuous expansion of the crystalline phase above the threshold excludes an interpretation of the melting process based on a gradual exchange between two coexisting thermodynamic phases at a pinned temperature.

Table 1 Experimentally observed time scales in the PMMA/graphene bilayer system.

View this table:

The sequential relaxation processes observed here are a direct result of the hierarchy of coupling strengths, from the relatively weak interchain interaction over the substrate surface registration to the structural constraint given by the polymer backbone. Presently, it remains an open question whether the enhanced expansion of remaining crystallites above threshold is caused by a loss of crystallinity—e.g., by domain shrinkage—or if itself serves as a major driver of the melting process. The observed time scales are relatively short in comparison with very recently reported intrinsic fluctuations within polymer-graphene heterostructures at similar temperatures (31). This suggests that the loss of superstructure registration involves a release of considerable stress present in the ordered phase. We expect that a Frenkel-Kontorova–type model and dynamical excitations therein may lead to a detailed microscopic picture (30), provided that the dimensionality of the system and the internal polymer degrees of freedom are considered.

The compact experimental implementation of time-resolved LEED is adaptable to conventional ultrafast optics setups and ultrahigh vacuum apparatuses, rendering its application feasible in the broader surface science community. The present operation conditions (energies at the upper end of the LEED range and transmission geometry) were chosen for an optimal combination of low pulse duration and high scattering efficiency providing monolayer sensitivity. Time-resolved LEED measurements in back-reflection from surfaces, including reconstructions and adsorbate layers, will require lower energies than employed here. To demonstrate the suitability of our approach for more general ultrafast surface studies, we provide back-scattering diffraction images and a pulse characterization at lower energy in the supplementary materials (figs. S8 and S5, respectively). We anticipate that further developments toward miniaturized electron guns will in the future allow for femtosecond LEED studies of bulk surfaces at energies of 100 eV and below, opening up the exploration of a whole arena of unconventional phases and transitions in solid-state and molecular surface science.

Supplementary Materials

Materials and Methods

Figs. S1 to S8

References (3748)

References AND NOTES

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Whereas the (10) diffraction spot of PMMA overlaps with that of graphene, the (3/2 0) PMMA spot is not observed, which is most likely a result of the chain form factor or disorder.
  3. See, e.g., (40). In the measurements presented, the total dose on a given sample position was limited to minimize degradation effects on the observed structural dynamics. This was achieved by restricting the integration time and number of delay times measured.
  4. Acknowledgments: We thank M. Müller for helpful discussions on polymer dynamics. Supporting sample characterizations by H. Schuhmann, M. Seibt, S. Strauch, H. Stark (TEM imaging), S. Dechert, and M. Sivis (Raman spectroscopy), as well as James E. Evans and Nigel D. Browning (high-resolution TEM, Pacific Northwest National Laboratory), are gratefully acknowledged. This work was partially funded by the Deutsche Forschungsgemeinschaft (DFG-ZuK 45/1 and DFG-SFB 1073). M.G. was financially supported by the German National Academic Foundation. A.M.W. and H.K.Y. gratefully acknowledge support from the Alexander von Humboldt Foundation.
View Abstract

Navigate This Article