Falco C.M.J.M. van Delft

Parallel computation with molecular-motor-propelled agents in nanofabricated networks

Significance Electronic computers are extremely powerful at performing a high number of operations at very high speeds, sequentially. However, they struggle with combinatorial tasks that can be solved faster if many operations are performed in parallel. Here, we present proof-of-concept of a parallel computer by solving the specific instance {2, 5, 9} of a classical nondeterministic-polynomial-time complete (“NP-complete”) problem, the subset sum problem. The computer consists of a specifically designed, nanostructured network explored by a large number of molecular-motor-driven, protein filaments. This system is highly energy efficient, thus avoiding the heating issues limiting electronic computers. We discuss the technical advances necessary to solve larger combinatorial problems than existing computation devices, potentially leading to a new way to tackle difficult mathematical problems. The combinatorial nature of many important mathematical problems, including nondeterministic-polynomial-time (NP)-complete problems, places a severe limitation on the problem size that can be solved with conventional, sequentially operating electronic computers. There have been significant efforts in conceiving parallel-computation approaches in the past, for example: DNA computation, quantum computation, and microfluidics-based computation. However, these approaches have not proven, so far, to be scalable and practical from a fabrication and operational perspective. Here, we report the foundations of an alternative parallel-computation system in which a given combinatorial problem is encoded into a graphical, modular network that is embedded in a nanofabricated planar device. Exploring the network in a parallel fashion using a large number of independent, molecular-motor-propelled agents then solves the mathematical problem. This approach uses orders of magnitude less energy than conventional computers, thus addressing issues related to power consumption and heat dissipation. We provide a proof-of-concept demonstration of such a device by solving, in a parallel fashion, the small instance {2, 5, 9} of the subset sum problem, which is a benchmark NP-complete problem. Finally, we discuss the technical advances necessary to make our system scalable with presently available technology.

Template masters for substrate conformal imprint lithography generated by charged particle nanopatterning techniques

Substrate Conformal Imprint Lithography (SCIL™), developed within Philips Research, is a large area replication technology, which allows flexible nano-imprinting, even around defects. It uses templates (stamps) with a high modulus poly(dimethyl)siloxane (PDMS) pattern layer bonded onto a glass sheet with a low modulus PDMS intermediary layer. This template sheet is attached to a grooved vacuum plate. By sequentially pressurizing and evacuating the grooves, controlled contact with the resist layer and smooth release after resist curing can be established. The PDMS stamps are cast from a nanostructured silicon wafer, which serves as the template master. Charged Particle Nanopatterning (CHARPAN) techniques based on ion multi-beam projection techniques, establish a promising route for generating such nanometer resolution template masters. 2D structures have been written in the CHARPAN tool using Hydrogen (H3+) ions in a high resolution negative tone e-beam resist, Hydrogen Silsesquioxane (HSQ). The CHARPAN tool can also be operated with heavier sputter ions (Ar+, Xe+ etc.) enabling maskless and resistless 3D direct nanopatterning of a silicon template master. CHARPAN generated 2D and 3D template masters, the PDMS stamps cast from these masters and the resulting SCIL imprinted structures, show that at least a 20 nm resolution is feasible for this particular combination of technologies. The combination of CHARPAN and SCIL opens up new possibilities for low cost, fast and flexible 2D and 3D manufacturing of nano-devices in several application fields, e.g. in life sciences related test structures and devices.

Classification of impact-assisted etch mechanisms

Abstract Previously, a mechanistic framework has been constructed for impact-assisted etch reactions in e.g. Reactive Ion Etching. The consecutive reaction steps (adsorption, surface reaction and product desorption) are assumed to be activated thermally and in parallel mechanically by fast particle impacts. The rate determining step in the etch mechanism determines the etch profile. The framework discerns four basic types of etching. In this work, the etch mechanisms of Fe/FeCrB multilayers, quartz, alumina, zirconia, magnesia and silicon have been studied in an HCl plasma (Alcatel GIR300) as a function of pressure, RF input power and surface temperature. With the help of our framework the results can be rationalized and a proper choice of etch conditions is possible.

Mechanistic framework for dry etching, beam assisted etching and tribochemical etching

Abstract A mechanistic framework is presented for impact assisted etch reactions. The consecutive reaction steps are assumed to be activated thermally and in parallel mechanically by fast particle impacts. The model explains the complicated temperature dependencies observed in dry etching and beam assisted etching, and it correlates the side wall profiles to the rate determining steps in the etch mechanisms. The framework is used to describe Reactive Ion Beam Etching (RIBE) experiments on InP with Ar + ions and chlorine, in comparison with our recent Reactive Ion Etching (RIE) experiments of magnetic alloys in HCl plasmas. The framework is also applicable to other non-thermally activated etch reactions, as encountered in tribo-chemical etching and laser chemical etching.

Nanoscale etching of resists in view of a mechanistic framework

Abstract In our previously developed mechanistic framework for dry etching, the three consecutive reaction steps (reactant chemisorption, surface reaction and product desorption) are assumed to be activated thermally and in parallel mechanically by fast particle impacts. According to this model, profiles with perpendicular side walls should be obtained if (a) the rate determining step in the etch mechanism is not the etch product desorption, and if also (b) the rate determining step is activated mechanically only. Both criteria are shown to be fulfilled in the case of oxygen Reactive Ion Etching (RIE) of (photo)resists, which is in accordance with the observed etch profiles. The rate determining step is shown to be most probably the (dissociative) oxygen chemisorption on the resist.

Something has to give: scaling combinatorial computing by biological agents exploring physical networks encoding NP-complete problems

On-chip network-based computation, using biological agents, is a new hardware-embedded approach which attempts to find solutions to combinatorial problems, in principle, in a shorter time than the fast, but sequential electronic computers. This analytical review starts by describing the underlying mathematical principles, presents several types of combinatorial (including NP-complete) problems and shows current implementations of proof of principle developments. Taking the subset sum problem as example for in-depth analysis, the review presents various options of computing agents, and compares several possible operation ‘run modes’ of network-based computer systems. Given the brute force approach of network-based systems for solving a problem of input size C, 2C solutions must be visited. As this exponentially increasing workload needs to be distributed in space, time, and per computing agent, this review identifies the scaling-related key technological challenges in terms of chip fabrication, readout reliability and energy efficiency. The estimated computing time of massively parallel or combinatorially operating biological agents is then compared to that of electronic computers. Among future developments which could considerably improve network-based computing, labelling agents ‘on the fly’ and the readout of their travel history at network exits could offer promising avenues for finding hardware-embedded solutions to combinatorial problems.

Reply to Einarsson: The computational power of parallel network exploration with many bioagents

REPLY TO EINARSSON : The computational power of parallel network exploration with many bioagents

Manufacturing sub-50-nm gratings using E-beam lithography and electroplating

In this paper we present a new method for manufacturing steeply profiled grating structures composed of narrow lines and spaces embedded in transition metal layers. We focus on making EUV transmittive grating structures typically consisting of rectangular lines that are down to 40 nm wide, around 100 nm tall, up to 100 nm long, spaced at a 1-2 micrometers pitch, and embedded in 100 nm thick nickel-iron alloy absorber layers. The method comprises the use of a molybdenum plating base deposited on a silicon nitride coated silicon wafer, electron beam writing of down to 40 nm wide and 100 nm tall lines in negative tone hydrogen silsesquioxane (HSQ) resist, and electroplating of the desired metal absorber layer in between the resist lines, using the molybdenum layer as plating base. After processing both the HSQ resist and the thin plating base can remain respectively between and below the grating structure because these materials are relatively transparent to EUV radiation. The presence of HSQ in the plated metal spaces results in a flat top surface, preventing the adhesion of contaminants. Our measurements reveal the presence of rectangular HSQ lines, characterized by near vertical side walls, a high line-width uniformity and a low line-edge roughness. These structures are subsequently embedded in a homogeneously grown absorber, characterized by a low small-scale surface roughness and high-quality overall flatness. The process window, in terms of exposure dose and pre-exposure resist treatment, has been well established for various line width and line pitch settings.

DUV resists in negative tone high resolution electron beam lithography

Shipleys chemically amplified DUV resists UVN-2 (negative tone) and UV-5 (positive tone) have been studied for their high resolution capabilities in electron beam lithography. UV-5 is also capable of negative tone behaviour in case of e-beam overexposure. This effect is shown to be due to direct electron beam cross linking superimposed on the normal catalytically induced positive tone behaviour. The ultimate resolution for 100 nm thick UVN-2 is below 50 nm for single lines, when using the shortest possible developing time in order to prevent swelling. The best obtained positive tone resolution of UV-5 is 50 nm for single trenches and 120 nm for 1:1 lines and spaces. The best obtained negative tone resolution of overexposed UV-5 is below 90 nm. This negative tone behaviour is accompanied by a 60% swelling of a 100 nm thick layer, which swelling is most probably related to the relatively long developing time. This swelling is not observed for a 485 nm thick UV-5 layer.