Fuelling the Digital Chemistry Revolution with Language Models.

The RXN for Chemistry project, initiated by IBM Research Europe - Zurich in 2017, aimed to develop a series of digital assets using machine learning techniques to promote the use of data-driven methodologies in synthetic organic chemistry. This research adopts an innovative concept by treating chemical reaction data as language records, treating the prediction of a synthetic organic chemistry reaction as a translation task between precursor and product languages. Over the years, the IBM Research team has successfully developed language models for various applications including forward reaction prediction, retrosynthesis, reaction classification, atom-mapping, procedure extraction from text, inference of experimental protocols and its use in programming commercial automation hardware to implement an autonomous chemical laboratory. Furthermore, the project has recently incorporated biochemical data in training models for greener and more sustainable chemical reactions. The remarkable ease of constructing prediction models and continually enhancing them through data augmentation with minimal human intervention has led to the widespread adoption of language model technologies, facilitating the digitalization of chemistry in diverse industrial sectors such as pharmaceuticals and chemical manufacturing. This manuscript provides a concise overview of the scientific components that contributed to the prestigious Sandmeyer Award in 2022.

Conversion of a Patterned Organic Resist into a High Performance Inorganic Hard Mask for High Resolution Pattern Transfer.

Polyphthalaldehyde is a self-developing resist material for electron beam and thermal scanning probe lithography (t-SPL). Removing the resist in situ (during the lithography process itself) simplifies processing and enables direct pattern inspection, however, at the price of a low etch resistance of the resist. To convert the material into a etch resistant hard mask, we study the selective cyclic infiltration of trimethyl-aluminum (TMA)/water into polyphthalaldehyde. It is found that TMA diffuses homogeneously through the resist, leading to material expansion and formation of aluminum oxide concurrent to the exposure to water and the degradation of the polyphthalaldehyde polymer. The plasma etch resistance of the infiltrated resist is significantly improved, as well as its stability. Using a silicon substrate coated with 13 nm silicon nitride and 7 nm cross-linked polystyrene, high resolution polyphthalaldehyde patterning is performed using t-SPL. After TMA/H2O infiltration, it is demonstrated that pattern transfer into silicon can be achieved with good fidelity for structures as small as 10 nm, enabling >10× amplification and low surface roughness. The presented results demonstrate a simplified use of polyphthalaldehyde resist, targeting feature scales at nanometer range, and suggest that trimethyl-aluminum infiltration can be applied to other resist-based lithography techniques.

Capillary assembly as a tool for the heterogeneous integration of micro- and nanoscale objects.

During the past decade, capillary assembly in topographical templates has evolved into an efficient method for the heterogeneous integration of micro- and nano-scale objects on a variety of surfaces. This assembly route has been applied to a large spectrum of materials of micrometer to nanometer dimensions, supplied in the form of aqueous colloidal suspensions. Using systems produced via bulk synthesis affords a huge flexibility in the choice of materials, holding promise for the realization of novel superior devices in the fields of optics, electronics and health, if they can be integrated into surface structures in a fast, simple, and reliable way. In this review, the working principles of capillary assembly and its fundamental process parameters are first presented and discussed. We then examine the latest developments in template design and tool optimization to perform capillary assembly in more robust and efficient ways. This is followed by a focus on the broad range of functional materials that have been realized using capillary assembly, from single components to large-scale heterogeneous multi-component assemblies. We then review current applications of capillary assembly, especially in optics, electronics, and in biomaterials. We conclude with a short summary and an outlook for future developments.

Programmable Assembly of Hybrid Nanoclusters.

Hybrid nanoparticle clusters (often metallic) are interesting plasmonic materials with tunable resonances and a near-field electromagnetic enhancement at interparticle junctions. Therefore, in recent years, we have witnessed a surge in both the interest in these materials and the efforts to obtain them. However, a versatile fabrication of hybrid nanoclusters, that is, combining more than one material, still remains an open challenge. Current lithographical or self-assembly methods are limited to the preparation of hybrid clusters with up to two different materials and typically to the fabrication of hybrid dimers. Here, we provide a novel strategy to deposit and align not only hybrid dimers but also hybrid nanoclusters possessing more complex shapes and compositions. Our strategy is based on the downscaling of sequential capillarity-assisted particle assembly over topographical templates. As a proof of concept, we demonstrate dimers, linear trimers, and 2D nanoclusters with programmable compositions from a range of metallic nanoparticles. Our process does not rely on any specific chemistry and can be extended to a large variety of particles and shapes. The template also simultaneously aligns the hybrid (often anisotropic) nanoclusters, which could facilitate device integration, for example, for optical readout after transfer to other substrates by a printing step. We envisage that this new fabrication route will enable the assembly and positioning of complex hybrid nanoclusters of different functional nanoparticles to study coupling effects not only locally but also at larger scales for new nanoscale optical devices.

Explaining the Transition from Diffusion Limited to Reaction Limited Surface Assembly of Molecular Species through Spatial Variations.

Surface assembly is often decomposed into two classes: diffusion and reaction limited processes. The transition between the two cases is complex because the dynamics are so different. In this article, we simulate, explain, and experimentally discuss the evolution of the spatial distribution for surface assemblies with diffusion limited and reaction limited processes. Explicitly, we demonstrate that diffusion limited and reaction limited processes show some temporal differences, but more importantly, we show that the spatial arrangements are different enough to discriminate between the two cases. Using fundamental properties, such as the diffusion constant, we calculate the evolution of the spatial profile and derive from physical, heuristic models the assembly rate for reaction and diffusion limited processes based on the individual particle's interactions with the surface. Finally, we confirm the spatial profile differences between diffusion and reaction limited cases by experimentally measuring the surface assembly between two molecules of similar size, but having different assembly routes. Unique to our description is that we have derived and simulated everything through the particle picture in place of ensemble descriptions such as the diffusion equation, and we show the equivalence between our heuristic formulas and those derived from the diffusion equation.

Testing the Equivalence between Spatial Averaging and Temporal Averaging in Highly Dilute Solutions.

Diffusion relates the flux of particles to the local gradient of the particle density in a deterministic way. The question arises as to what happens when the particle density is so low that the local gradient becomes an ill-defined concept. The dilemma was resolved early last century by analyzing the average motion of particles subject to random forces whose magnitude is such that the particles are always in thermal equilibrium with their environment. The diffusion dynamics is now described in terms of the probability density of finding a particle at some position and time and the probabilistic flux density, which is proportional to the gradient of the probability density. In a time average sense, the system thus behaves exactly like the ensemble average. Here, we report on an experimental method and test this fundamental equivalence principle in statistical physics. In the experiment, we study the flux distribution of 20 nm radius polystyrene particles impinging on a circular sink of micrometer dimensions. The particle concentration in the water suspension is approximately 1 particle in a volume element of the dimension of the sink. We demonstrate that the measured flux density is exactly described by the solution of the diffusion equation of an infinite system, and the flux statistics obeys a Poissonian distribution as expected for a Markov process governing the random walk of noninteracting particles. We also rigorously show that a finite system behaves like an infinite system for very long times despite the fact that a finite system converges to a zero flux empty state.

Sub-10 Nanometer Feature Size in Silicon Using Thermal Scanning Probe Lithography.

High-resolution lithography often involves thin resist layers which pose a challenge for pattern characterization. Direct evidence that the pattern was well-defined and can be used for device fabrication is provided if a successful pattern transfer is demonstrated. In the case of thermal scanning probe lithography (t-SPL), highest resolutions are achieved for shallow patterns. In this work, we study the transfer reliability and the achievable resolution as a function of applied temperature and force. Pattern transfer was reliable if a pattern depth of more than 3 nm was reached and the walls between the patterned lines were slightly elevated. Using this geometry as a benchmark, we studied the formation of 10-20 nm half-pitch dense lines as a function of the applied force and temperature. We found that the best pattern geometry is obtained at a heater temperature of ∼600 °C, which is below or close to the transition from mechanical indentation to thermal evaporation. At this temperature, there still is considerable plastic deformation of the resist, which leads to a reduction of the pattern depth at tight pitch and therefore limits the achievable resolution. By optimizing patterning conditions, we achieved 11 nm half-pitch dense lines in the HM8006 transfer layer and 14 nm half-pitch dense lines and L-lines in silicon. For the 14 nm half-pitch lines in silicon, we measured a line edge roughness of 2.6 nm (3σ) and a feature size of the patterned walls of 7 nm.

Enhanced Second-Harmonic Generation from Sequential Capillarity-Assisted Particle Assembly of Hybrid Nanodimers.

We show enhanced second-harmonic generation (SHG) from a hybrid metal-dielectric nanodimer consisting of an inorganic perovskite nanoparticle of barium titanate (BaTiO3) coupled to a metallic gold (Au) nanoparticle. BaTiO3-Au nanodimers of 100 nm/80 nm sizes are fabricated by sequential capillarity-assisted particle assembly. The BaTiO3 nanoparticle has a noncentrosymmetric crystalline structure and generates bulk SHG. We use the localized surface plasmon resonance of the gold nanoparticle to enhance the SHG from the BaTiO3 nanoparticle. We experimentally measure the nonlinear signal from assembled nanodimers and demonstrate an up to 15-fold enhancement compared to a single BaTiO3 nanoparticle. We further perform numerical simulations of the linear and SHG spectra of the BaTiO3-Au nanodimer and show that the gold nanoparticle acts as a nanoantenna at the SHG wavelength.

Hybrid Colloids Produced by Sequential Capillarity-assisted Particle Assembly: A New Path for Complex Microparticles.

Colloidal particles have long been under the spotlight of a very diverse research community, given their ubiquitous presence in a broad class of materials and processes, and their pivotal role as model systems. More recently, intense efforts have been devoted to the development of micro- and nanoparticles combining multiple materials in objects with a controlled architecture, hence introducing multiple functionalities and a prescribed symmetry for interactions. These particles are often called hybrid colloids or colloidal molecules, given the analogy with classical molecules presenting well defined structures and chemical compositions. Here, we review the recent progress made in our group to fabricate a broad library of hybrid colloids exploiting a novel assembly route, which uses capillary forces at the moving edge of an evaporating droplet for the sequential composition of colloidal clusters, whose geometry and chemistry can be independently programmed.

Hybrid colloidal microswimmers through sequential capillary assembly.

Active colloids, also known as artificial microswimmers, are self-propelled micro- and nanoparticles that convert uniform sources of fuel (e.g. chemical) or uniform external driving fields (e.g. magnetic or electric) into directed motion by virtue of asymmetry in their shape or composition. These materials are currently attracting enormous scientific attention as models for out-of-equilibrium systems and with the promise to be used as micro- and nanoscale devices. However, current fabrication of active colloids is limited in the choice of available materials, geometries, and modes of motion. Here, we use sequential capillarity-assisted particle assembly (sCAPA) to link microspheres of different materials into hybrid clusters of prescribed shapes ("colloidal molecules") that can actively translate, circulate and rotate powered by asymmetric electro-hydrodynamic flows. We characterize the active motion of the clusters and highlight the range of parameters (composition and shape) that can be used to tune their trajectories. Further engineering provides active colloids that switch motion under external triggers or perform simple pick-up and transport tasks. By linking their design, realization and characterization, our findings enable and inspire both physicists and engineers to create customized active colloids to explore novel fundamental phenomena in active matter and to investigate materials and propulsion schemes that are compatible with future applications.

Understanding How Charged Nanoparticles Electrostatically Assemble and Distribute in 1-D.

The effects of increasing the driving forces for a 1-D assembly of nanoparticles onto a surface are investigated with experimental results and models. Modifications, which take into account not only the particle-particle interactions but also particle-surface interactions, to previously established extended random sequential adsorption simulations are tested and verified. Both data and model are compared against the heterogeneous random sequential adsorption simulations, and finally, a connection between the two models is suggested. The experiments and models show that increasing the particle-surface interaction leads to narrower particle distribution; this narrowing is attributed to the surface interactions compensating against the particle-particle interactions. The long-term advantage of this work is that the assembly of nanoparticles in solution is now understood as controlled not only by particle-particle interactions but also by particle-surface interactions. Both particle-particle and particle-surface interactions can be used to tune how nanoparticles distribute themselves on a surface.

Programmable colloidal molecules from sequential capillarity-assisted particle assembly.

The assembly of artificial nanostructured and microstructured materials which display structures and functionalities that mimic nature's complexity requires building blocks with specific and directional interactions, analogous to those displayed at the molecular level. Despite remarkable progress in synthesizing "patchy" particles encoding anisotropic interactions, most current methods are restricted to integrating up to two compositional patches on a single "molecule" and to objects with simple shapes. Currently, decoupling functionality and shape to achieve full compositional and geometrical programmability remains an elusive task. We use sequential capillarity-assisted particle assembly which uniquely fulfills the demands described above. This is a new method based on simple, yet essential, adaptations to the well-known capillary assembly of particles over topographical templates. Tuning the depth of the assembly sites (traps) and the surface tension of moving droplets of colloidal suspensions enables controlled stepwise filling of traps to "synthesize" colloidal molecules. After deposition and mechanical linkage, the colloidal molecules can be dispersed in a solvent. The template's shape solely controls the molecule's geometry, whereas the filling sequence independently determines its composition. No specific surface chemistry is required, and multifunctional molecules with organic and inorganic moieties can be fabricated. We demonstrate the "synthesis" of a library of structures, ranging from dumbbells and triangles to units resembling bar codes, block copolymers, surfactants, and three-dimensional chiral objects. The full programmability of our approach opens up new directions not only for assembling and studying complex materials with single-particle-level control but also for fabricating new microscale devices for sensing, patterning, and delivery applications.

Accurate Location and Manipulation of Nanoscaled Objects Buried under Spin-Coated Films.

Detection and precise localization of nanoscale structures buried beneath spin-coated films are highly valuable additions to nanofabrication technology. In principle, the topography of the final film contains information about the location of the buried features. However, it is generally believed that the relation is masked by flow effects, which lead to an upstream shift of the dry film's topography and render precise localization impossible. Here we demonstrate, theoretically and experimentally, that the flow-shift paradigm does not apply at the submicrometer scale. Specifically, we show that the resist topography is accurately obtained from a convolution operation with a symmetric Gaussian kernel whose parameters solely depend on the resist characteristics. We exploit this finding for a 3 nm precise overlay fabrication of metal contacts to an InAs nanowire with a diameter of 27 nm using thermal scanning probe lithography.

Insights into mechanisms of capillary assembly.

Capillary assembly in a topographical template is a powerful and flexible method for fabricating complex and programmable particle assemblies. To date, very little attention has been paid to the effects that the trap geometry--in particular the trap depth--has on the outcome of the assembly process. In this paper, we provide insights into the mechanisms behind this directed assembly method by systematically studying the impact of the trap depth and the surface tension of the suspension. Using confocal microscopy, we investigate the assembly process at the single-particle level and use these observations to formulate a simple mechanical model that offers guidelines for the successful assembly of single or multiple particles in a trap. In particular, single particles are assembled for shallow traps and moderate surface tensions, opening up the possibility to fabricate multifunctional particle dimers in two consecutive assembly steps.

Cascaded assembly of complex multiparticle patterns.

A method for the cascaded capillary assembly of different particle populations in a single assembly cycle is presented. The method addresses the increasing need for fast and simple fabrication of multicomponent arrays from colloidal micro- and nanoscale building blocks for constructing nanoelectronic, optical, and sensing devices. It is based on the use of a microfluidic device from which two independent capillary bridges extend. The menisci of the capillary bridges are pulled over a template with trapping sites that receive the colloidal particles. We describe the parameters for simultaneous, high-yield assembly from both menisci and demonstrate the applicability of the process by means of the size-selective assembly of particles of different diameters and also by the fabrication of two-component particle clusters with defined shape and composition. This approach allows the fabrication of multifunctional particle clusters having different functionalities at predetermined positions.