Research on Semiconductor Quantum Structures, Bioelectronics and Physics of New Information Technologies

Quantum dots (QD) and quantum wires (QWR) are compound semiconductor or metallic structures that confine electrons or holes or both in a potential box having a dimension of few tens of nanometers. These semiconductor structures have exceptional optical and transport properties, which makes them ideal for fundamental research as well as highly potential platforms for nanotechnological, bioelectronics and biotechnology applications. In the enclosed project descriptions we describe few topics that have been in focus during 2003.

All projects include extensive domestic and international collaboration with the following laboratories: Optoelectronics Laboratory,HUT, VTT Electronics, Instituto Nazionale di Fisica della Materia, University of Lecce, MegaGauss Laboratory, University. of Tokyo, Inst. of Industrial Science, University of Tokyo. Center for Teraherz Science, USCB. On natinal level there is a close collaboration with VTT Microelectronics, University of Jyväskylä and University of Oulu.

As a new field we have started research of low power biomorphic neural circuits based on floating gate MOS and SET transistors. In this project neuro-MOS and neuro-SET based neural networks are developed and studied, especially for fast and power efficient signal processing. Neuro-MOS structures, including MOS capacitor based, and neuro-SET structures, are studied and optimized in deep sub-micron line width processes. Power optimisation will be studied, based on physical and architectual ideas from extremely power-efficient biological neurons. New efficient algorithms utilizing the benefits of neuro-structures are developed. Models for simulation of neuro-SET structures are developed. The applicability of floating gate structures - either MOS or SET - to higher level neural architectures, e.g. recurrent or CNN, will be studied.

The ever-decreasing size of the basic components of information processing will give quantum effects an important role in future technologies. Recent developments such as quantum cryptography and the idea of a quantum computer have shown that, rather than being only harmful, these effects can probably be utilized to a great extent. In communications technologies, optical transmission is setting the trend in the development of the networks. The full harnessing of the huge bandwidth provided by light still requires for replacing the switching, routing and processing electronics by all-optical components. Research on nonlinear optical materials and light-induced quantum effects will be crusial in the development of future all-optical processing technologies. Related to optical communications technology, we are studying all-optical swithing and processing using nonlinear materials, combined with novel material structures such as photonic crystals.

We have investigated cold atomic Fermi-gases which can be used for studing important quantum many-body effects such as superconductivity. Cold atomic gases may also serve as a source of atoms in quantum information processing applications. Recent focus has been in the study of supefluidity and its detection by optical means. This research is done in collaboration with the University of Innsbruck, Austria, with Nordita, Denmark, and with the Loomis Laboratory, University of Illinois, USA. We have also participated in the University of Jyväskylä -based experimental research on the idea of using superconducting Josephson junctions as the basic processing element of a quantum computer.

Modelling of nanoscale semiconductor devices

Researchers: Fredrik Boxberg, Roman Terechonkov and Jukka Tulkki

The commercial integrated electronics is mainly based on silicon (Si), while compound semiconductors (CS), e.g. gallium arsenide (GaAs), are used in very special applications. For example, light detectors and emitters and very-high frequency devices are usually fabricated using CS due to the better optical and electron mobility properties of these materials. Quantum mechanical devices are a potential application for CS. For quantum mechanical operation phase coherence is crucial and it has been found difficult to obtain long phase coherence lengths in Si structures. This advantages of CS is due to both a different fabrication technique and different electronic properties. We are studying quantum effect electronics both in Si and CS. The motivations are lower power consumption, faster operation and smaller device size.

We have studied a CMOS fabrication technique for Si quantum wires (QWRs). Figure 34A shows schematically a Si QWR (turquoise color) embedded in silicon dioxide (SiO2). These kind of QWRs could easily be integrated in commercial integrated circuits. We have modelled the growth process of SiO2 on the QWRs and its effect on the fabricated devices. Great straininduced effects on the electronic properties were found. These effects alone could in principle ruin the operation of the devices. So far it has not been possible to fabricate high quality Si QWRs and we argue that one reason is the oxidation-induced strain field.

Currently the main topic of research are the electronic and optical properties of CS devices. We are studying the possibility to fabricate both laser devices and single photon devices (e.g. needed for future quantum information technology). We have been working on quantum wells (QW), QWRs and quantum dots (QDs). Figure 34B shows an 8 nm thick corrugatedQW (the hatched layer) with a corrugation period of 13 nm. The corrugation confines the carrier density into wires. The wire-like carrier densities are shown with turquoise in Fig. 34B. Moreover, we have been studying indium arsenide (InAs) QDs embedded in a GaAs matrix both with and without a covering layer. A typical QD structure is shown in Fig. 34C. The turquoise region corresponds to the InAs QD.

Figure 34

Figure 34: Schematic drawings of (A) a Si QWR embedded in SiO2, (B) a corrugated InGaAs quantum well and (C) a covered InAs QD island. The turquoise areas correspond to the active region of the device, i.e., high charge carrier density. The red bars correspond to 50 nm in (A) and (C) and 25 nm in (B).

We are developing general tools for strain analysis, band structure calculations and optical models. The final aim is to model photonic devices starting from the structural properties and ending up with properties like the light amplification in the device. For this purpose we need to simulate the strain field, the electronic structure and the recombination of carriers.

The strain in the modelled structures is due to different lattice constants of the epitaxially combined materials. The strain is of the order . Hence, the devices can be built with hardly any dislocations and we know that the strain is completely elastic. However, the strain affects the electronic structure remarkably and it cannot be omitted from the calculation of electronic and photonic properties of quantum structures. The assumption of a dislocationfree structure is one of the corner-stones of our model. The strain and the piezoelectric field in these structures affects the position and the inter band coupling of the energy bands. The strain calculations are based on the elastic continuum theory. The full treatment of the strain in compound semiconductors leads to an electro-elastic coupled problem where the strain is coupled to the piezoelectric field. This coupling is due to the ionic atomic structure of III-V compound materials. The effect is absent in purely one material semiconductors like Si or Ge.

The optical properties rely completely on the underlying electronic bands, which are influenced by the device geometry and crystal orientation. From the electronic structure one can model properties like photon recombination rates, polarization of the emitted light and light amplification. One could in principle extend this model to describe also some basic dynamic photonic processes.

Electronic structure calculations

The electronic and optical properties of semiconductor quantum structures are governed by the band structure. The band structure can be used to calculate the transition matrix elements, for example, which then can be used to predict the intensity of photoemission as a function of the wave length and excitation ( injection ) intensity.

There are only two methods to obtain the band structure: either experimental or theoretical approaches. Development of the computers has made possible to compute semiquantative band structure for compound semiconductors.

Band structure calculations are computationally very intensive. Therefore we have devoloped parallelized computer algorithms to solve the finite difference eigenvalue problem. Our computer code gives us 8-band eigenvalues and eigenfunctions for quantum well, quantum wire and quantum dot structures.

The 8-band model used in our calculations is based on combining a Shrödinger equation based series expansion of the wave function with an experimentally determined parameters. In addition to the experimental data, one also uses information from band structure calculations of the related bulk semiconductors Fig. 35.

Figure 35

Figure 35: On the left: The calculation procedure diagram. On the right:The electron density distribution for conduction ground state band ( step-like quantum wire case ).

In our project we are interested in the electronic states near the band edge. For this case, the -method gives reasonable results. The size of the matrix to be diagonalized can be up to million by million. This enforces us to use many processor computers and partial spectrum eigenvlaue solver library. The PARPACK library allows us calculating the eigenvalues and eigenvectors Fig. 35 under the minimum RAM requirements.

Modelling optical components for access networks

Researchers: Jani Oksanen and Jukka Tulkki

With the long haul network backbone transformed into an optical information highway, the transfer capacity is mostly limited by the electronic bottlenecks in the access and metropolitan area networks. The undisputed success of the optical fibres in the network backbone encourages to resolve the bottlenecks by passing to optical solutions in the other parts of the networks as well. This solution, however, requires all-optical components which do not exist commercially (or at all) at present.

Metropolitan area networks include a large number of separate connections, which makes fast switching devices and inexpensive laser transceivers essential. The goal of this project is to create models and new ideas for the devices needed in expanding the optical network.

In the project we have this far investigated 1) the differences of the quantum well and dot lasers with respect to their chirp under direct current modulation, 2) the operation of an optical amplifier linearized using gain clamping in vertical direction (also known as linear optical amplifier, or LOA) and 3) the use of quantum cascade lasers in free space optical communications. At the moment we are focusing our attention on an all-optical inverter, that could enable fast optical logic with reasonable input powers. The inverter shows promising characteristics for low input signal bitrates (up to ~1Ghz). At higher bitrates it experiences some stability problems we’re hoping to overcome.

Theoretical models and simulations allow profound insight in the device operation, within the framework of the accuracy of the model and its assumptions. The above devices are studied using analytical and numerical models ranging from band structure calculations to stochastic rate equations. The obtained information can be used in improving the existing devices and possibly in creating new ones as well.

Figure 36

Figure 36: Schematic representation of the linear optical amplifier including a waveguide and a vertical microcavity with highly reflecting distributed Bragg reflectors (DBR).

Quantum Transport in the Ballistic and Coulomb Blockade Regimes

Researchers: Teppo Häyrynen and Jukka Tulkki

If electrons move through a conducting device without scattering the transport of electrons is ballistic. Ballistic transport is observed when the length of the channel is small compared to the mean free path of an electron. In the ballistic transport regime a device can be modeled in terms of transmission probabilities which are calculated for different combinations of source and drain eigenmodes

We have used the mode matching (MM) and Green’s function methods within the Landauer- Büttiker formalism to calculate the conductance in selected two- and three-dimensional channels. To compare these two methods we calculated the conductance of a 2D T-stub (see the inset of Fig. 37(a)).Furthermore we calculated the conductance of a silicon on insulator (SOI) quantum point contact (QPC) by using the MM method at small temperatures where the step like behavior of conductance was observed. The calculated result agrees qualitatively with the measured conductance [M. Prunnila et al., Silicon quantum point contact with aluminum gate, Mat. Sci. Eng. B, 74(1-3):193-196,2000].

Figure 37

Figure 37: (a) The conductance of the a-stub. The shape of the T-stub is shown in the inset. The waveguide is wide and the size of the stub is . The zero of the energy axis correspond to the ground state energy of the lead: , . (b) Calculated conductance of the QPC at T=0K and T=1.5K.

In a single electron transistor (SET) a small conducting island is connected to a source and drain by tunnel junctions. In a tunnel junction a thin layer of insulator separates two conductors (e.g. thin silicon dioxide layer between doped silicon areas). If the junction area is small and the temperature is low the movement of an electron can be blocked by charging energy of the island (i.e. the thermal energy of an electron is smaller than the coulomb energy required to transfer the electron into the island). This is known as the Coulomb blockade phenomenon. If small voltage is applied between drain and source an electron can tunnel into the island raising the Fermi energy of the island and preventing other electrons tunneling into it until either the applied voltage is increased or an electron is tunneled out of the island. Furthermore the number of excess electrons in the island can be controlled with a gate electrode which shifts the energy levels of the island.

Figure 38

Figure 38: Energyband diagram of a double junction system. In the left figure all applied voltages are set to zero and no current flows through the SET. In the right figure electrons can tunnel into the island and out of it because the excited state of the island due to the Coulomb charging lies between the Fermi energies of the left and right leads.

We have simulated the functionality of an SET based exclusive-OR (XOR) logic gate with SIMON (SIMulation Of Nano-structures), a software which makes use of Monte Carlo and Master equation methods for calculations. The two gates of the single electron transistor are the input nodes of the XOR gate and the source of the SET represents the output node.

Figure 39

Figure 39: The switching characteristics of XOR gate at T=40 K. The drain-source voltage is VDS=10mv, the tunneling junctions have parameters and and the gate capacitances are .

Low Power Biomorphic Neural Circuits on Floating Gate MOS and SET Transistors

Researchers: Teppo Häyrynen and Jukka Tulkki

Due to limitations set by the power consumption and algorithm design new ideas and techniques has to be applied in the integration of ever-growing number of electronic components on a single chip. Parallel processing at the system level is one way to increase computational power. Single electron transistors (SET) may offer another way. SET circuits are a challenge for system design because of their sensitivity to fluctuating background charge and the requirement for low temperature. However, SETs might be used in fast and power efficient signal processing circuits. Especially neural approach is a promising methodology to bypass the setbacks of SETs.

Our goal is to implement the energy efficient processes of signaling in biological neurons to electronic circuits. Our purpose is to develop equivalent circuit models for a single neuron as well as for a synapse interacting between neurons. Furthermore our goal is to model a group of interacting neurons.

Figure 40

Figure 40: In the parallel conductance model the plasma-membrane of a neuron consists of crossmembrane proteins which operate as ionchannels and a double layer of lipids. Ionchannels can be modeled with (voltage dependent) conductances and the double lipid layer of lipids with a capacitance.

In biological neurons a signal is carried by ions, mainly by -ions, producing a current trough the plasma-membrane and along the neuron. Thus the neuron can be dealt as an electrical components. The starting point of the study is a transmission line formalism for modeling the signal propagation in a neuron. A well known model for the plasma-membrane of neuron is the so called parallel conductance model. In this model the cross-membrane proteins that operate as ionchannels are modeled with parallel conductances and the double layer of lipids is modeled with a capacitance. Furthermore the cytoplasm of a neuron can be modeled with a resistance. Currently we are developing a dynamical model for the plasma-membrane of a neuron. Our objective is to take into account the essential properties of biological neurons.

Photonic Crystals

Researchers: Researchers: Anu Huttunen and Päivi Törmä*
*Department of Physics, University of Jyväskylä.

Photonic crystals are periodic dielectric structures. The periodicity creates bandgaps for light, i.e., light in a certain wavelength region cannot propagate in the photonic crystal. The periodicity, and thus the bandgap, can be in either one, two or three dimensions. Twodimensional photonic crystals embedded with defects could be used e.g. as waveguides for integrated optics and a defect inside a three-dimensional photonic crystal can act as a microcavity. Photonic crystals are a very attractive solution to various problems in telecommunications and may become the key material for integrated optics.

We study thin slabs of one- and two-dimensional photonic crystals. We show that varying the boundary material results in changes in the band gap and that this effect can be utilized for reflecting light traveling along a one-dimensional photonic crystal slab by changing the boundary material abruptly (see Fig. 41). We have suggested that in two-dimensional photonic crystals, the same effect could probably be used for guiding of light by patterning the boundary material. We also study photonic crystal structures made of Kerr-nonlinear material.

Figure 41

Figure 41: (a) Energy density profiles of a Gaussian pulse in a photonic crystal slab at different times. The frequency of the pulse can propagate in the photonic crystal when boundary material is air, but falls into a band gap when boundary material is GaAs. This can be seen from the intensity profile as the pulse is reflected. (b) The fraction of the energy density of a Gaussian pulse that is reflected inside the slab from the point where the boundary material above and below the photonic crystal slab changes from air to a material with dielectric constant . Solid line indicates dielectric boundary materials and dots indicate metals: gold, silver, and aluminum.

We also study photonic band gap fibers regarding ultra-fast optical systems. The cladding layer of photonic band gap fibers is periodic in the plane of the fiber cross-section. The goal is to design and demonstrate efficient amplification of optical pulses in photonic crystal fibers with amplifying medium. Tight optical confinement can be achieved in photonic crystal fibers by proper design of the geometry leading to enhanced amplification. Also dispersive and nonlinear effects are very different from conventional optical fiber.

Cold Degenerate Atomic Fermi Gases

Researchers: Mirta Rodriguez, J. Kinnunen1 , P. Pedri2, L. Santos2 and Päivi Törmä1
1Department of Physics, University of Jyväskylä
2Quantum Optics Group, University of Hannover (Germany)

The remarkable achievement of Bose-Einstein condensation (BEC) in alkali gases has stimulated the trapping and cooling of also the Fermionic isotopes.

Atomic gases can be efficiently and accurately manipulated. They are dilute and weakly interacting thus offering the ideal tool for studying fundamental quantum statistical and manybody physics.

The most prominent phenomena for the fermionic samples is the experimental realization of a superfluid Fermi gas. Novel forms of fermionic superfluidity in alkali gases have been reported during the present year. Fermionic superfluidity happens at low enough temperatures due to the attractive interaction between the atoms trapped in different hyperfine states.

We are studying different manisfestations of superfluidity in these novel systems:

  1. We have proposed the use of on-resonant light to probe the order parameter in order to detect the superfluid transition by analyzing the atomic current.
  2. A purely quantum transport phenomena such as Bloch oscillations can also be observed in these Fermi gases when they are loaded in a periodic optical potential. We have analyzed the bloch oscillations both in the normal and superfluid regime and proposed the different value of the amplitude of the oscillation as a signature of superfluidity.

Fermi-Bose and Bose-Bose mixtures are one of the central topics in the field of atomic gases. We have considered the scissors mode in these multicomponent systems.

Figure 42

Figure 42: Fermi sea of neutral atoms at T=0 loaded into a magnetic or optical one-dimensional harmonic trap.