This thesis is aimed at a very detailed process of a transparent plexiglass used for manufacturing an upstand of a gramophone. The diploma work is mainly focused on various types of the plexiglass, its attributes and the gramophone’s upstand production. Several manufacturing technologies of a plexiglass such as division of a board-type semi product, precise milling, grinding, polishing and thermal treatment are provided in this diploma work through a very detailed description of a production process of a gramophone’s deck. The outcome of this thesis is a description of individual technologies and the economic evaluation within the real business environment of AZ Plastik company.

Design of flat pallet for automatic parking garage. Parking and description of automatic parking systems. Welding and weldability of materials. Electric arc welding and welding by melting electrode in protective gas by MIG/MAG method, welding instruments. Design of parking palette calculation and calculation this palette by MKP method. Process of drawing documentation for parking palette. Design of welding process and WPS. Description of annealing and zinc coating. Economic-technical analysis of palette for automatic parking system production.

The content of this thesis is to understand the principle of ultrasound imaging and mathematical models used to estimate perfusion parameters of concentration curves. Thesis deals with global optimization algorithms for finding parameters, an approximation of the actual data model curves. It also includes a comparison of different methods and used functions.

The PhD thesis deals with the analysis of digital systems described on RT level. The methodology of  data paths analysis is decribed, the data path controller analysis is not solved in the thesis. The methodology is built on the concept of Testable Block (TB) which allows to divide digital component to such segments which can be tested through their inputs/outputs, border registers and primary inputs/outputs are used for this purpose. As a result, lower number of registers is needed to be included into scan  chain - border registers are the only ones which are scanned.  The segmentation allows also to reduce the volume of test vectors, tests are generated for segments, not for the complete component. To identify TBs, two evolutionary algorithms are used, they operate on TB formal model which is also defined in the thesis.

The public transport system in South Africa is in a precarious state, capturing no more than 50% of the passenger market. The three public transport modes that are currently utilized—train, bus, and minibus-taxi—are competing for market share instead of complementing one another. Furthermore, most public transport networks have not been properly redesigned over the past three decades. Improvements were initiated reactively in the past: transit stops and routes were added or removed from the network when demand fluctuated. This reactive process has diminished the confidence of commuters in the public transport networks, forcing commuters to use private transport. A proactive redesign method is needed—one that includes all the modes of public transport, and anticipates an increase in demand and rapid development in geographic areas, while ensuring good accessibility to the network. Current network design models do not include multiple modes of public transport, and are based on the geographical layout of developed cities and their particularities, which makes them unsuitable for the South African environment with its unique land use disparities. This dissertation proposes a multimodal network design model that is capable of designing real world and large scale networks for the South African metropolitan areas. The City of Tshwane Metropolitan Municipality (CTMM) transport network area was used to develop and test the model, which consists of four components. The Geographic Information System (GIS) component has a central role in storing, manipulating, and exchanging the geographic data within the model. For the GIS the appropriate input data is identified, and a design for the geo-database is proposed. The Population Generation Algorithm (PGA) component translates the demographic data into point data representing the transit demand in the study area. The Bus Stop Placement Algorithm (BSPA) component is a metaheuristic that searches for near-optimal solutions for the placement of bus stops in the study area. A novel solution approach proposed in this dissertation uses geographic data of commuters to evaluate the bus stop placement in the study area. The Multimodal Network Design Algorithm (MNDA) component also employs a metaheuristic, enabling the design of near-optimal multimodal networks. The addition of multiple modes to the Transit Network Design Problem (TNDP) is also a novel and significant contribution. The two metaheuristic components are first tested on a test network, and subjected to a comprehensive sensitivity analysis. After identifying suitable parameter values and algorithm settings, the components are applied to the entire CTMM. / Dissertation (MSc)--University of Pretoria, 2009. / Industrial and Systems Engineering / unrestricted

In this thesis, the structural, optical, and electrical properties of wurtzite MgxZn1-xO:Al and MgxZn1-xO:Ga thin films have been investigated in dependence on Mg and dopant concentration. Among the transparent conductive oxides (TCOs), ZnO based compounds have gained renewed interest as a transparent electrode for large scale applications such as defroster windows, at panel displays, touch screens, and thin film solar cells due to low material and processing cost, non-toxicity, and suitable physical properties. In general, these applications require transparent electrodes with lowest possible resistivity of rho < 10^-3 Ohmcm and lower [1]. Recently, it has been reported that Ga and Al doped ZnO thin films can be deposited with respective resistivity of 5x10^-5 Ohmcm [2] and 3 x10^-5 Ohmcm [3] which are similar to the data obtained for other practical TCOs, i.e. the resistivity of about 4x 10^-5 Ohmcm for Sn doped In2O3 (ITO) thin films [4]. Moreover, the bandgap of ZnO can be increased by alloying with Mg offering band alignment between transparent electrode and active (or buffer) layer of the device, e.g. Cu(In,Ga)Se2 solar cells [5]. The tunable bandgap of these transparent electrodes can further increase the efficiency of the devices by avoiding energy losses in the interface region of the layers. From this point of view, this work has been aimed to investigate the doping efficiency and limits in transparent conductive (Mg,Zn)O alloys. For this purpose, the samples investigated in this work have been grown by pulsed-laser deposition (PLD) using a novel, continuous composition spread method (CCS). In general, this method allows to grow thin films with lateral composition gradient(s) [6, 7]. All MgxZn1-xO:Al and MgxZn1-xO:Ga thin films have been deposited on 2-inch in diameter glass, c- or r-plane sapphire substrates using threefold segmented PLD targets in order to grow thin films with two perpendicular, lateral composition gradients, i.e. the Mg composition is varied in one direction whereas the Al/Ga concentration is varied in a perpendicular direction [7, 8]. In order to investigate the influence of the temperature, samples grown at different substrate temperatures in the range of 25 to 600 C were investigated. The optical and electrical measurements have been carried out on (5x 5)mm^2 samples that were cut from the CCS wafers along the respective composition gradients, i.e. Mg and Al/Ga contents. Subsequently, physical properties of thin films have been analyzed for a large range of Al/Ga content between 0.5 and 7 at.%, which corresponds to doping concentrations between 2x 10^20 and 3x 10^21 cm^-3, for different Mg contents x(Mg) ranging from 0.01 to 0.1. It has been found that practically the limiting the dopant concentrations is about 2 x10^21 cm^-3. Further, the electrical data suggests, that the compensating intrinsic defect is doubly chargeable hinting to the zinc vacancy (V_Zn) as microscopic origin. Increasing the dopant concentration above 2 x10^21 cm^-3 leads to a degradation of electrical and structural properties [8]. Further, the influence of growth and annealing temperatures on structural, electrical and optical properties of the films has been studied. For that purpose, Al and Ga doped (2.5 at.% = 1x10^21 cm^-3) Mg0.05Zn0.95O thin films have been chosen from CCS samples grown at T_g = (25 - 600) C . For both doping series, the samples grown at higher temperatures exhibit better crystalline quality compared to the samples grown at lower growth temperatures. As a result, samples grown at higher temperatures reveal higher Hall mobility. For the Al-doping series, the highest free charge carrier density of n = 8.2x 10^20 cm^-3 was obtained for an Mg0.05Zn0.95O:Al thin film grown at 200 C, with corresponding Hall mobility of mu = 13.3 cm^2/Vs, a resistivity of rho = 5.7x10^-4 Ohmcm, and optical bandgap of E_g = 3.8 eV. Interestingly, the free charge carrier density of n = (5 - 8) x 10^20 cm^-3 for samples grown with T_g > 300 C is clearly higher than the value of n = 1.25 x 10^20 cm^-3 that was obtained for the high temperature grown sample, i.e. at T_g = 600 C. Furthermore, for all T_g, Al-doped films have a higher doping efficiency than the Ga-doped counterparts. In order to look deeper into the microscopic origin of this behavior, the samples were post-annealed in vacuum at 400 C. Experimental results showed that the free charge carrier density of Al-doped samples first decreased and saturated afterward with increasing annealing time. On the other hand, the free charge carrier density of the Ga-doped samples first slightly increased and saturated with increasing annealing time. For both doping series, the saturation value of n ~ 1 x 10^20 cm^-3 was very close to the data that has been observed for (i) high temperature grown samples and (ii) the solubility limit of Al in ZnO of 0.3 at.% = 1.2x 10^20 cm^-3, that has been determined by Shirouzu et al. for high temperature grown (T_g > 600 C) Al-doped ZnO [9]. Correspondingly, the optical bandgap also changed, i.e. increased (decreased) for Al- (Ga-) doping series, and approached a constant value of 3.5 0 +- 0.1 eV which is explained by generation of acceptor-like compensating defects, and the solubility limit of the dopants. From XRD data, no secondary phases were found for as-grown and post-annealed films. However, the slight improvement of crystalline quality has been observed on post-annealed samples. Further, it has been shown that the growth and annealing temperatures are important as they strongly affect the metastable state of the solid solution that samples grown at low temperature represent. The low solubility limit of the dopants, i.e. 0.3 at.% for Al in ZnO under equilibrium condition, can be increased by preparing samples by non-equilibrium growth techniques [10]. This is also consistent with experimental results of this work that Al- as well as Ga-doped metastable ZnO and (Mg,Zn)O thin films can be prepared with highest possible doping efficiency for the dopant concentration up to 2.5 at.% when growth or annealing temperatures below 400 C are used.

In dieser Arbeit wird eine Beschreibung von Monte-Carlo-Verfahren zur Lösung komplexer Optimierungsaufgaben mit Hilfe von Markov-Ketten durchgeführt. Nach einer kurzen Einführung werden Lösungsmenge solcher Aufgaben und der physikalische Zustandsraum komplexer Systeme identifiziert. Zunächst wird die Dynamik von Zufallswanderern im Zustandsraum mit Hilfe von Master-Gleichungen modelliert. Durch Einführung von Performanzkriterien können verschiedene Optimierungsstrategien quantitativ miteinander verglichen werden. Insbesondere wird das Verfahren Extremal Optimization vorgestellt, dass ebenfalls als Markov-Prozess verstanden werden kann. Es wird bewiesen, dass eine im Sinne der genannten Kriterien beste Implementierung existiert. Da diese von einem sogenannten Fitness Schedule abhängt, wird dieser für kleine Beispielsysteme explizit berechnet. Daran anschließend wird die Zustandsdichte komplexer Systeme betrachtet. Nach einem kurzen Überblick über vorhandene Methoden folgt eine detaillierte Untersuchung des Verfahrens von Wang und Landau. Numerische und analytische Hinweise werden gegeben, nach denen dieser Algorithmus innerhalb seiner Klasse wahrscheinlich der Optimale ist. Eine neue Methode zur Approximation der Zustandsdichte wird vorgestellt, die insbesondere für die Untersuchung komplexer Systeme geeignet ist. Abschließend wird ein Ausblick auf zukünftige Arbeiten gegeben.

Doping allows us to modify semiconductor materials for desired properties such as conductivity, bandgap, and / or lattice parameter. A small portion replacement of the highly mismatched isoelectronic dopants with the host atoms of a semiconductor can result in drastic variation of its structural, optical, and / or electronic properties. Here, the term "mismatch" describes the properties of atom size, ionicity, and / or electronegativity. This thesis presents the fabrication of two kinds of highly mismatched semiconductor alloys, i.e., Ge(1-x)Sn(x) and GaAs(1-x)N(x). The structural and optical properties of the prepared Ge(1-x)Sn(x) and GaAs(1-x)N(x) have been investigated. The results suggest an efficient above-solubility doping induced by non-equilibrium methods of ion implantation and ultrashort annealing. Pulsed laser melting promotes the regrowth of monocrystalline Ge(1-x)Sn(x), whereas flash lamp annealing brings about the formation of high quality GaAs(1-x)N(x) with room temperature photoluminescence. The bandgap modification of Ge(1-x)Sn(x) and GaAs(1-x)N(x) has been verified by optical measurements of spectroscopic ellipsometry and photoluminescence, respectively. In addition, effective defect engineering in GaAs has been achieved by flash lamp annealing, by which a quasi-temperature-stable photoluminescence at 1.3 µm has been obtained. / Dotierung ermöglicht es, die Eigenschaften von Halbleitermaterialien, wie Leitfähigkeit, aber auch Bandabstand und / oder Gitterkonstanten gezielt zu verändern. Wenn ein Halbleiter mit einer kleinen Menge unterschiedliche Fremdatome dotiert wird, kann dies in einer drastischen Modifikation der strukturellen, optischen und / oder elektronischen Eigenschaften resultieren. Der Begriff "unterschiedlich" bedeutet hier die Eigenschaften von Atomgröße, Ioniztät und / oder Elektronegativität. Diese Doktorarbeit beschreibt die Herstellung von zwei Arten von stark fehlangepassten Halbleiterlegierungen: Ge(1-x)Sn(x) und GaAs(1-x)N(x). Die strukturellen und optischen Eigenschaften von Ge(1-x)Sn(x) und GaAs(1-x)N(x) wurden untersucht. Die Ergebnisse deuten auf eine effiziente Dotierung oberhalb der Löslichkeit, induziert durch die Nicht-Gleichgewichtsverfahren Ionenimplantation und Ultrakurzzeit-Ausheilung. Gepulstes Laserschmelzen ermöglicht das Nachwachsen von monokristallinem Ge(1-x)Sn(x), während die Blitzlampenausheilung in der Bildung von GaAs(1-x)N(x) hoher Qualität mit Photolumineszenz bei Raumtemperatur resultiert. Die Änderung der Bandlücke von Ge(1-x)Sn(x) und GaAs(1-x)N(x) wurde durch die optischen Methoden der spektroskopischen Ellipsometrie und Photolumineszenz verifiziert. Darüber hinaus konnte in ausgeheiltem GaAs eine quasi-temperaturstabile Photolumineszenz bei 1,3 µm beobachtet werden.

The progress in device performance of modern microelectronic technology is mainly driven by down-scaling. In the near future, this road will probably reach a point where physical limits make even more down-scaling impossible. The substitution of single components materialwise over the last decades, like high-k dielectrics or metal gates, has been a suitable approach to foster performance improvements. In this scheme, the integration of high-mobility III-V compound semiconductors as channel materials into Si technology is a promising route to follow for the next one or two device generations. III-V integration, today, is conventionally performed by using techniques like molecular beam epitaxy or wafer bonding which utilize solid phase crystallization but suffer to strain due to the lattice mismatch between III-V compounds and Si. An alternative approach using sequential ion beam implantation in combination with a subsequent flash lamp annealing is presented in this work. Using this technique, nanocrystals from various III-V compounds have been successfully integrated into bulk Si and Ge as well as into thin Si layers which used either SOI substrates or were grown by plasma-enhanced chemical vapour deposition. The III-V compounds which have been fabricated are GaP, GaAs, GaSb, InP, InAs, GaSb and InxGa1-xAs with variable composition. The structural properties of these nanocrystals have been investigated by Rutherford backscattering, scanning electron microscopy and transmission electron microscopy, including bright-field, dark-field, high-resolution, high-angle annular dark-field and scanning mode imaging, electron-dispersive x-ray spectroscopy and energy-filtered element mapping. Furthermore, Raman spectroscopy and X-ray diffraction have been performed to characterise the nanocrystals optically. In Raman spectroscopy, the characteristic transversal and longitudinal optical phonon modes of the different III-V compounds have been observed. These signals proof that the nanocrystals have formed by the combination of ion implantation and flash lamp annealing. Additionally, the appearance of the typical phonon modes of the respective substrate materials verifies recrystallization of the substrate by the flash lamp after amorphisation during implantation. In the bulk Si samples, the nanocrystals have a circular or rectangular lateral shape and they are randomly distributed at the surface. Their cross-section has either a hemispherical or triangular shape. In bulk Ge, there are two types of precipitates: one at the surface with arbitrary shape and another one buried with circular shape. For the thin film samples, the lateral shape of the nanocrystals is more or less arbitrary and they feature a block-like cross-section which is limited in height by the Si layer thickness. Regarding crystalline quality, the nanocrystals in all samples are mainly single-crystalline with only a few number of stacking faults. However, the crystalline quality in the bulk samples is slightly better than in the thin films. The X-ray diffraction measurements display the (111), (220) and (311) Bragg peaks for InAs and GaAs as well as for the InxGa1-xAs where the peaks shift with increasing In content from GaAs towards InAs. The underlying formation mechanism is identified as liquid phase epitaxy. Hereby, the ion implantation leads to an amorphisation of the substrate material which is then molten by the subsequent flash lamp annealing. This yields a homogeneous distribution of the implanted elements within the melt due to their strongly increased diffusivity in the liquid phase. Afterwards, the substrate material starts to recrystallize at first and an enrichment of the melt with group-III and group-V elements takes place due to segregation. When the temperature is low enough, the III-V compound semiconductor starts to crystallize using the recrystallized substrate material as a template for epitaxial growth. In order to gain control over the lateral nanocrystal distribution, an implantation mask of either aluminium or nickel is introduced. Using this mask, only small areas of the samples are implanted. After flash lamp treatment, nanocrystals form only in these small areas, which allows precise positioning of them. An optimal implantation window size with an edge length of around 300nm has been determined to obtain one nanocrystal per implanted area. During an additional experiment, the preparation of Si nanowires using electron beam lithography and reactive ion etching has been conducted. Hereby, two different processes have been investigated; one using a ZEP resist, a lift-off step and a Ni hard mask and another one using a hydrogen silsesquioxane resist which is used directly as a mask for etching. The HSQ-based process turned out to yield Si nanowires of better quality. Combining both, the masked implantation and the Si nanowire fabrication, it might be possible to integrate a single III-V nanocrystal into a Si nanowire to produce a III-V-in-Si-nanowire structure for electrical testing. / Der Fortschritt in der Leistungsfähigkeit der Bauelemente moderner Mikroelektroniktechnologie wird hauptsächlich durch das Skalieren vorangetrieben. In naher Zukunft wird dieser Weg wahrscheinlich einen Punkt erreichen, an dem physikalische Grenzen weiteres Herunterskalieren unmöglich machen. Der Austausch einzelner Teile auf Materialebene, wie Hoch-Epsilon-Dielektrika oder Metall-Gate-Elektroden, war während der letzten Jahrzehnte ein geeigneter Ansatz, um die Leistungsverbesserung voranzubringen. Nach diesem Schema ist die Integration von III-V-Verbindungshalbleiter mit hoher Mobilität ein vielversprechender Weg, dem man für die nächsten ein oder zwei Bauelementgenerationen folgen kann. Heutzutage erfolgt die III-V-Integration konventionell mit Verfahren wie der Molekularstrahlepitaxie oder dem Waferbonden, welche die Festphasenkristallisation nutzen, die aber aufgrund der Gitterfehlanpassung zwischen III-V-Verbindungen und Silizium an Verspannungen leiden. In dieser Arbeit wird ein alternativer Ansatz präsentiert, welcher die sequenzielle Ionenstrahlimplantation in Verbindung mit einer darauffolgenden Blitzlampentemperung ausnutzt. Mit Hilfe dieses Verfahrens wurden Nanokristalle verschiedener III-V-Verbindungshalbleiter erfolgreich in Bulksilizium- und -germaniumsubstrate sowie in dünne Siliziumschichten integriert. Für die dünnen Schichten wurden hierbei entweder SOI-Substrate verwendet oder sie wurden mittels plasmagestützer chemischer Gasphasenabscheidung gewachsen. Die hergestellten III-V-Verbindungen umfassen GaP, GaAs, GaSb, InP, InAs, InSb und InxGa1-xAs mit veränderbarer Zusammensetzung. Die strukturellen Eigenschaften dieser Nanokristalle wurden mit Rutherford-Rückstreu-Spektroskopie, Rasterelektronenmikroskopie und Transmissionselektronenmikroskopie untersucht. Bei der Transmissionelektronenmikroskopie wurden die Hellfeld-, Dunkelfeld-, hochauflösenden, “high-angle annular dark-field” und Rasteraufnahmemodi sowie die energiedispersive Röntgenspektroskopie und die energiegefilterte Elementabbildung eingesetzt. Darüber hinaus wurden Ramanspektroskopie- und Röntgenbeugungsmessungen durchgeführt, um die Nanokristalle optisch zu charakterisieren. Mittels Ramanspektroskopie wurden die charakteristischen transversal- und longitudinal-optischen Phononenmoden der verschiedenen III-V-Verbindungen beobachtet. Diese Signale beweisen, dass sich unter Verwendung der Kombination von Ionenstrahlimplantation und Blitzlampentemperung Nanokristalle bilden. Weiterhin zeigt das Vorhandensein der typischen Phononenmoden der jeweiligen Substratmaterialien, dass die Substrate aufgrund der Blitzlampentemperung rekristallisiert sind, nachdem sie durch Ionenimplantation amorphisiert wurden. In den Bulksiliziumproben besitzen die Nanokristalle eine kreisförmige oder rechteckige Kontur und sind in zufälliger Anordnung an der Oberfläche verteilt. Ihr Querschnitt zeigt entweder eine Halbkugel- oder dreieckige Form. Im Bulkgermanium gibt es zwei Arten von Ausscheidungen: eine mit willkürlicher Form an der Oberfläche und eine andere, vergrabene mit sphärischer Form. Betrachtet man die Proben mit den dünnen Schichten, ist die laterale Form der Nanokristalle mehr oder weniger willkürlich und sie zeigen einen blockähnlichen Querschnitt, welcher in der Höhe durch die Siliziumschichtdicke begrenzt ist. Bezüglich der Kristallqualität sind die Nanokristalle in allen Proben mehrheitlich einkristallin und weisen nur eine geringe Anzahl an Stapelfehlern auf. Jedoch ist die Kristallqualität in den Bulkmaterialien ein wenig besser als in den dünnen Schichten. Die Röntgenbeugungsmessungen zeigen die (111), (220) und (311) Bragg-Reflexe des InAs und GaAs sowie des InxGa1-xAs, wobei sich hier die Signalpositionen mit steigendem Gehalt an Indium von GaAs zu InAs verschieben. Als zugrundeliegender Bildungsmechanismus wurde die Flüssigphasenepitaxie identifiziert. Hierbei führt die Ionenstrahlimplantation zu einer Amorphisierung des Substratmaterials, welches dann durch die anschließende Blitzlampentemperung aufgeschmolzen wird. Daraus resultiert eine homogene Verteilung der implantierten Elemente in der Schmelze, da diese eine stark erhöhte Diffusivität in der flüssigen Phase aufweisen. Danach beginnt zuerst das Substratmaterial zu rekristallisieren und es kommt aufgrund von Segregationseffekten zu einer Anreicherung der Schmelze mit den Gruppe-III- und Gruppe-V-Elementen. Wenn die Temperatur niedrig genug ist, beginnt auch der III-V-Verbindungshalbleiter zu kristallisieren, wobei er das rekristallisierte Substratmaterial als Grundlage für ein epitaktisches Wachstum nutzt. In der Absicht Kontrolle über die laterale Verteilung der Nanokristalle zu erhalten, wurde eine Implantationsmaske aus Aluminium beziehungsweise Nickel eingeführt. Durch die Benutzung einer solchen Maske wurden nur kleine Bereiche der Proben implantiert. Nach der Blitzlampentemperung werden nur in diesen kleinen Bereichen Nanokristalle gebildet, was eine genaue Positionierung dieser erlaubt. Es wurde eine optimale Implantationsfenstergröße mit einer Kantenlänge von ungefähr 300 nm ermittelt, damit sich nur ein Nanokristall pro implantierten Bereich bildet. Während eines zusätzlichen Experiments wurde die Präparation von Siliziumnanodrähten mit Hilfe von Elektronenstrahllithografie und reaktivem Ionenätzen durchgeführt. Hierbei wurden zwei verschiedene Prozesse getestet: einer, welcher einen ZEP-Lack, einen Lift-off-Schritt und eine Nickelhartmaske nutzt, und ein anderer, welcher einen HSQ-Lack verwendet, der wiederum direkt als Maske für die Ätzung dient. Es stellte sich heraus, dass der HSQ-basierte Prozess Siliziumnanodrähte von höherer Qualität liefert. Kombiniert man beides, die maskierte Implantation und die Siliziumnanodrahtherstellung, miteinander, sollte es möglich sein, einzelne III-V-Nanokristalle in einen Siliziumnanodraht zu integrieren, um eine III-V-in-Siliziumnanodrahtstruktur zu fertigen, welche für elektrische Messungen geeignet ist.

Using an optimized KOH activation procedure we prepared highly porous graphene scaffoldmaterials with SSA values up to 3400m² g⁻¹ and a pore volume up to 2.2 cm³ gˉ¹, which are among the highest for carbon materials. Hydrogen uptake of activated graphene samples was evaluated in a broad temperature interval (77–296 K). After additional activation by hydrogen annealing the maximal excess H2 uptake of 7.5 wt% was obtained at 77 K. A hydrogen storage value as high as 4 wt% was observed already at 193 K (120 bar H₂), a temperature of solid CO₂, which can be easily maintained using common industrial refrigeration methods.