Preface
Chapter 1. Silicon and Silicon Carbide Oxidation
1.1. Introduction
1.2. Overview of the various oxidation techniques
1.3. Some physical properties of silica
1.4. Equations of atomic transport during oxidation
1.5. Is it possible to identify the transport mechanisms taking place during oxidation?
1.6. Transport equations in the case of thermal oxidation
1.7. Deal and Grove theory of thermal oxidation
1.8. Theory of thermal oxidation under water vapor of silicon
1.9. Kinetics of growth in O2 for oxide films < 30 nm
1.10. Fluctuations of the oxidation constants under experimental conditions
1.11. Conclusion
1.12. Bibliography
Chapter 2. Ion Implantation
2.1. Introduction
2.2. Ion implanters
2.3. Ion range
2.4. Creation and healing of the defects
2.5. Applications in traditional technologies and new tendencies
2.6. Conclusion
2.7. Bibliography
Chapter 3. Dopant Diffusion: Modeling and Technological Challenges
3.1. Introduction
3.2. Diffusion in solids
3.3. Dopant diffusion in single-crystal silicon
3.4. Examples of associated engineering problems
3.5. Dopant diffusion in germanium
3.6. Conclusion
3.7. Bibliography
Chapter 4. Epitaxy of Strained Si/Si1-x Gex Heterostructures
4.1. Introduction
4.2. Engineering of the pMOSFET transistor channel using pseudomorphic SiGe layers
4.3. Engineering of the nMOSFET transistor channel using pseudomorphic Si1-yCy layers; SiGeC diffusion barriers
4.4. Epitaxy of Si raised sources and drains on ultra-thin SOI substrates
4.5. Epitaxy of recessed and raised SiGe:B sources and drains on ultra-thin SOI and SON substrates
4.6. Virtual SiGe substrates: fabrication of sSOI substrates and of dual c-Ge / t-Si channels
4.7. Thin or thick layers of pure Ge on Si for nano and opto-electronics
4.8. Devices based on sacrificial layers of SiGe
4.9. Conclusions and prospects
4.10. Bibliography
List of Authors
Index
First published 2011 in Great Britain and the United States by ISTE Ltd and John Wiley & Sons, Inc.
Apart from any fair dealing for the purposes of research or private study, or criticism or review, aspermitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced,stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers,or in the case of reprographic reproduction in accordance with the terms and licenses issued by theCLA. Enquiries concerning reproduction outside these terms should be sent to the publishers at theundermentioned address:
ISTE Ltd | John Wiley & Sons, Inc. |
27-37 St George's Road | 111 River Street |
London SW19 4EU | Hoboken, NJ 07030 |
UK | USA |
www.iste.co.uk | www.wiley.com |
© ISTE Ltd 2011
The rights of Annie Baudrant to be identified as the author of this work have been asserted by her inaccordance with the Copyright, Designs and Patents Act 1988.
Library of Congress Cataloging-in-Publication Data
Silicon technologies : ion implantation and thermal treatment / edited by Annie Baudrant.
p. cm.
Includes bibliographical references and index.
ISBN 978-1-84821-231-2
1. Semiconductor doping. 2. Ion implantation. 3. Semiconductors--Heat treatment. I. Baudrant, Annie.
TK7871.85.S5485 2011
621.3815'2--dc22
2011008131
British Library Cataloguing-in-Publication Data
A CIP record for this book is available from the British Library
ISBN 978-1-84821-231-2
Text not available in this digital edition.
Image not available in this digital edition.
Electronic components based on semiconductors are at the core of all electrical and electronic equipment. They are stakeholders of an increasing share of the objects surrounding and accompanying us. They give to these objects various functions: capture, storage, transmission or information restoration (cameras, phones, TV, etc.), control-command, aid to decision-making, safety, etc.
Nowadays, the components aggregate more than 20% of the value of electrical or electronic equipment, against 7% in 1985. Therefore, there is a global market of the “semiconductor” sector of 270 billion dollars.
These components are indeed at the heart of all applications, for the “digital society”, of which they ensure both the engine and the memory, or for the improvements expected in the fields of transport, aeronautics, health, safety, and electrical energy management close to the user or within distribution systems.
Broadband available to all, the intelligent and efficient management of energy in buildings, telehealth, the intelligent road and zero emission cars or even the replacement of the electrical infrastructure, go inevitably through an increasing use of electronic components with more complex and powerful semiconductors.
The number of delivered components amounts to billions of parts per week (3.25 billion of units/week, at the end of December 2009).
This increasing and continuous penetration of electronic components in equipment was made possible by two other key features of the sector, which have been permanent over the last four decades:
— the monolithic and collective fulfillment or the increasing integration, of complex electronic functions (combination of elementary components) on hardware support, which constitutes an integrated silicon circuit, with a fast renewal of products and technologies;
— the continuous reduction of manufacturing costs and unit prices.
For the driving markets of the memories, of the specific processors and digital circuits, the production is primarily based on CMOS technologies. They are characterized by the most critical dimension of the transistors constituting them. Thus, the “40 nm node” corresponds to the technology, where this lower limit measures 40 nm.
Nowadays, the most aggressive generation of production is the 40 nm node, but the 28 nm node will be proposed in production from the next half-year.
The dimensions' reduction, for which the transistors (elementary components of the integrated circuits) are produced, is at the heart of two important product evolutions:
— the rise in performance (transistor speed, related to the minimal size of these elementary components);
— the integration of a growing number of transistors, today exceeding a billion per circuit.
The evolution of technologies leading to this increasing integration is fast. It follows an empirical prediction made by Gordon Moore (cofounder of Intel), of an integration doubled every 18 months. Unequalled fact in any industrial sector, this technical evolution of the products has lasted for four decades (see Figure 1).
The manufacturing technologies of electronic components have evolved a lot since the invention of the first transistor: we went from an element the size of a 1€ coin to submicron dimensions (45 nm previewed at IBM in spring 2006, while in December 2007, the TSMC group sampled SRAMs in 32 nm technology. Nowadays, we are heading for 22 nm and below).
Nevertheless, some principles have been preserved, and of course the technology is subjected to constraints, resulting from the fundamental properties of the semiconductors, such as they appear in the chapters on the components' physics. Let us note however that 32 nm noticeably represents the thickness of a hundred atomic lines, and that we imperceptibly come near a limit below which we will no longer be able to go, except by completely changing the principles.
Surprisingly, there are no works written in the French language describing these 30 years of history and evolution of technological methods. We can remember:
— that the evolution of CMOS technologies is made up of incremental and radical evolutions;
— that the control of these radical evolutions gives place to the highest costs and are the object of strategic decisions.
We often find a good number of precise publications on each topic, but little or no scientific summaries outlining the “why and how” of the various technical choices.
Among these key points, we have often quoted, over the last ten years:
— around the year 2000, for 130 to 90 nm nodes:
— around the year 2005, for 65 to 45 nm nodes:
- dipping lithography: with this, we can go down below 65 nm. Its unit piece of equipment exceeds 50 million dollars;
— around 2010, for 32 to 22 nm nodes:
— around 2014, for the 18-12 nm node:
All these stages use new techniques and thus new investments. We forget to name the advances made in the course of time by elementary methods, with the constant objective to decrease all dimensions (width, length, thickness) and to increase all the electrical performances (current, speed, low consumption, etc.).
This book thus proposes to the reader a timeline, of the development of active zones, modifying the surface structure of the massive silicon substrate, the basic principles, the implementations according to the technological fields, the limits and constraints, with a focus on the recent advances.
Its objective is not to go through the issue exhaustively for all the elementary processes of a technological assembly, but to point out the essential and fundamental pieces of data, only for the methods aiming to use and to improve the properties of the material or of the silicon semiconductor, for the electronic components, as well as for the microsystems (integrated sensors for example).
Below, a few paragraphs introduce each of the four matters evoked, but leave to the authors (J.J. Ganem, I. Trimaille, J.J. Grob, D. Mathiot and J.M. Hartmann, the best French-speaking specialists on the topic), the pleasure of elaborating on the manufacturing methods and scientific materials (physical and chemical laws),which consolidate or predict the outcomes.
Oxidation is a very important stage in the achievement of integrated silicon circuits, because it is thanks to this specific property that the silicon, which is not a priori a very good semiconductor, has become the most frequently used material in microelectronics. This operation is necessary throughout the modern manufacturing methods of integrated circuits. It is thus essential to know how to carry out an oxide of good quality.
The oxide can be used as:
— an implantation and diffusion mask of dopants;
— a passivating layer on the silicon surface;
— insulation zones between various components of an integrated structure;
— an active zone in MOS (gate oxide) transistors;
— electrical insulation between the adjacent layers to improve the integration and the reduction of dimensions (“spacer” for example, see further);
— electrical insulation between the various levels of metallization or of conducting layers in strongly doped polycrystalline silicon;
— sacrificial layers that can improve the circuit performances and integration. These sacrificial layers can also be used to manufacture microstructures containing polycrystalline silicon and to intervene in integrated microsystems (MEMS: micro electro-mechanical systems).
Ion implantation is a low-temperature process. The implantation process takes the ions of the chosen species, accelerates them with an electrical field, and then makes them scan the surface of the slice, to carry out a uniform pre-deposition.
This method of introducing doping atoms into silicon was developed in the 1960s: it is ion implantation (an ion is a loaded atom). Very briefly, the ion implantation consists of projecting the ions of the adequate doping species towards the slice, through the openings of an oxide mask or of hardened resin.
The ions used for the doping, such as boron, phosphorus or arsenic, are generally produced from a gas source, guaranteeing a great purity of the source. These gases have a tendency to be very dangerous. When they are implanted in a semiconductor, each doping atom creates a charge carrier (hole or electron according to whether it is a type p or n dopant), thus locally altering the conductivity of the semiconductor.
The ion implantation is also a method used for the preparation of SOI (silicon-on-insulator) substrates from conventional silicon substrates.
Diffusion is the term used to describe the movement of atoms, molecules or particles from a high concentration zone towards a lower concentration zone.
Diffusion is a phenomenon, depending on time and temperature. The diffusion velocity of an atom, molecule or compound from an area of high concentration towards a low concentration zone is a function of time and temperature. The parameter connecting the diffusion velocity to the temperature at a given time, is known as the diffusion or diffusivity coefficient.
The atoms of the dopant must move the silicon atoms of the crystalline structure and take their place to become electrically active. The diffusion process is used in the manufacture of integrated circuits, in order to introduce a controlled quantity of a specific dopant into a specific area of the semiconductor crystal. The diffusion process used to achieve this substitution is divided into distinct stages.
The term epitaxy is of Greek origin and means “to build above”. Deposition by epitaxy is in general the construction of a single-crystal silicon layer on a slice (also a single-crystal). The layer deposited is a crystallographic extension of the substrate, from the point of view of the atomic arrangement (i.e. identical crystalline structure). The substrate can thus be regarded as the “seed” essential to generate a single-crystal growth.
Deposition by epitaxy is a process of CVD. The first use of CVD was the single-crystal silicon deposition at the end of the 1950s. This technique then played a crucial role in this industry, but this chapter will not develop the aspect “material -growth and crystallography of silicon substrates”, instead focusing on the use of the epitaxy technique for the manufacture of active SiGe zones in nano-CMOS new technologies.
We hope that by introducing a comprehensive overview of these techniques to readers, this book will answer the expectations of those students, professors, technicians, engineers or researchers, closely interested in the manufacture of silicon micro nanostructures.
This preface would not be complete without extending to the authors, J.J. Ganem, I. Trimaille, J.J. Grob, D. Mathiot and J.M. Hartmann, my warmest thanks for their rich contributions to this book.
Annie BAUDRANT
May 2011
This chapter is devoted to the physics of silicon and silicon carbide oxidation. We will find in this chapter, an examination of the main techniques of deposition and growth of thin films. The reader will then discover how the laws governing oxidation are established, in particular those concerning silicon oxidation. This remains nowadays the most widespread method in the manufacture of integrated circuits and of MEMS in the broad sense. This chapter has a double purpose. First, it is written to expose in detail the theoretical principles that are particularly interesting for researchers, and secondly, to review a certain number of experimental results, useful in the practice of any process engineer.
The substantial improvement of the electrical and physical characteristics of the SiO2/Si interface leads to an impressive development of integrated circuits. That was made possible by a better understanding, over time, of the way in which silica is manufactured by deposition or growth on silicon.
Although thermal growth is nowadays one of the most frequently used methods, other techniques have also been developed. In section 1.2, we review the main passivation techniques employed today in industry as well as in research laboratories.
In the semiconductor industry, silica can be manufactured by thermal growth from silicon substrates placed at atmospheric pressure in a flux of water vapor, oxygen or a mix of oxidizing gases. Chemical vapor deposition is also usually used to obtain thick passivation films.
In laboratories, more often aiming at the study of physical phenomena governing the production of layers, the techniques used will be based on the physics of the methods (anodic oxidation in an electrolytic environment or by oxygen plasma) or on specific conditions of thermal oxidation (low or high pressure, etc.).
In section 1.3, we detail the silica properties useful for understanding the growth phenomena. We pay particular attention to showing the properties of atomic transport and the solubility of gases in silica. For a more complete review of the physical characteristics of silica, we refer the reader to the works of Bruckner [BRU 70]. For self-scattering and chemical diffusion in glasses, we recommend the works of Frischat [FRI 75].
In section 1.4, we develop the general equations of transport taking place during oxide growth, and the borderline cases arising from various approximate assumptions that can be made on the flux of mobile species in the case of oxidation. Thus, we also could deduct the expressions of the flux of the transported ion species, in the case of anodic oxidation.
In section 1.5, we show how it is possible to give a satisfactory answer to the two following questions: Which are the mobile species in silicon and in silicon carbide oxidation? How do they move? For this purpose, isotopic labeling techniques are also presented, as well as the experimental results.
In section 1.6, we state the transport equations after the simplifying hypotheses. Two cases are considered, where the transported species are either neutral or charged. When they are charged, two cases still emerge. For the films with a thickness lower than 3 nm, where the tunneling currents are established, the Cabrera and Mott theory predicts the growth laws. For the larger thicknesses, equations become very complex.
The classic Deal and Grove theory, describing silicon oxidation is detailed in section 1.7. It is based on the fact that oxidizing species (O2, H2O) are dissolved in silica in interstitial positions and migrate towards the SiO2/Si interface, to form there the new oxide by a chemical reaction assumed to be of the first order. This theory leads to the linear-parabolic growth kinetics, which we largely resort to in industry. They give a good account of the growth kinetics in the case of oxidation under water vapor, and in the case of oxidation under O2, for thicknesses higher than 20–30 nm.
Whereas the Deal and Grove theory assumes that species diffuse through oxide without interacting with it, Breed and Doremus have proposed a theory developed in section 1.8, assuming the interaction of the diffusing species with the silica network. The concentration profiles of the various species that we can expect from it are compared to the experimental results.
In section 1.9, we examine the growth kinetics of film slower than 30 nm and those of very thin films (< 3 nm) obtained by thermal growth, for which the Deal and Grove theory is no longer satisfactory. Several alternative models are stated to explain the initial regime of fast growth.
Finally, in section 1.10, we show how the important oxidation parameters, kP and kL, are affected by the experimental oxidation conditions such as the pressure, temperature and crystalline direction of the substrate and doping.
Several English journals, already published, explain the classical theories linked to the subjacent oxidation mechanisms. Those show the crucial interest for the industry to control and understand silicon oxidation, which must produce increasingly thin functional oxide layers. This chapter relies on the work of Rigo [RIG 86], while updating them by taking into account all the latest breakthroughs of the field.
There are many techniques used to form silica films on silicon. They are included in two categories, according to whether they rely on an oxidation or deposition process.
In oxidation processes, the film growth depends on the injection phenomena taking place at the two interfaces: oxide/gas and oxide/substrate. Those are a priori rather well defined. In all cases, the oxide film itself always plays a fundamental role. The atomic transport, from one interface to another, of at least one component through oxide is necessary to the growth. The oxide composition is determined by the external environment and by the laws of thermodynamics. Oxidation can be carried out in two ways:
— Thermal oxidation: heating a silicon sample at high temperatures under an oxidizing atmosphere, increases the hopping probability of the atoms. Indeed, the hopping probability being proportional to exp(−W/kBT), raising the temperature thus increases the transport of most mobile species. Moreover, fluxes at the interfaces are then increased. At this stage, it is necessary to stress that mobile species can be either electrically charged or neutral.
— Anodic oxidation: the injection and transport of charged species (anions and cations) are caused by an important electrical field (10 MV.cm − 1), which decreases the potential barrier that the ions must overcome. This type of growth can be carried out at room temperature.
In deposition processes, the films’ growth depends on phenomena taking place outside the silica film or its surface. The main part of the film already formed does not play any role in the following growth stages (if the film is maintained at a sufficiently low temperature). The composition of the formed film should vary considerably according to the techniques used: evaporation, sputtering of a material or chemical vapor deposition. We will find a comparison of the physical and chemical properties of these films in relation to their formation process in the summary article by Pliskin [PLI 77].
Among all the methods employed in the silicon industry, thermal oxidation is the most widely used. It can be done either with dry oxygen (dry oxidation), or with water vapor (wet oxidation).
The intermediate case, i.e. oxidation in an O2+H2O mixture, will not be discussed in this chapter. The two types of thermal oxidation will be developed in distinct sections. However, they are sometimes used sequentially, during the manufacture of integrated circuits (for example, wet oxidation followed by a stage of dry oxidation), to grow thick field oxides.
Typically, dry oxidation is carried out in a device schematized in Figure 1.1b. The furnace is a quartz tube heated by the Joule effect. After being cleaned, the silicon wafers are placed in a socket. The latter is slowly introduced in the center of the furnace, in an oxygen flux, so as to avoid thermal shocks. When we use a pure O2 flux, the oxidation is carried out with a PO2 oxygen pressure equal to the atmospheric pressure.
By using a mixture of O2 and of inert gases, we can oxidize with lower partial oxygen pressures.
To form very thin oxides (a few nm), we will use the rapid processing systems (RTF — rapid thermal furnace) principle. They rely on a rapid heating system (halogen lamps), in order to increase the temperature of a silicon wafer from the room temperature to 1,000°C in a fraction of a second. These furnaces allow very short processing times at high temperature, to obtain thin oxides with good electrical properties. A rapid thermal processing system under controlled atmosphere is schematized in Figure 1.1a.
Dry oxidation, although very simple, has some disadvantages:
— the process can become complicated if we need to grow several films of different thicknesses;
— to obtain a Si-SiO2 interface of good quality, the oxide must be annealed (for example, under a N2 flux, after stopping the O2 flux).
A pre-oxidation stage can be necessary (for example, under a flux of chlorinated compounds).
To oxidize under an atmosphere of pure oxygen, we must reduce the contamination sources to a minimum. They were listed by Revesz [REV 79] and are primarily CO2 and H2O. The gas flux always contains traces of contaminants, which are mainly water vapor and hydrocarbons. The latter will react with oxygen in the hot area of the furnace, to form CO2 and H2O. It is thus necessary to burn these hydrocarbons (for example by injecting the gas into a small quartz tube containing quartz wool heated to 900°C) [IRE 78a]. H2O and CO2 are then eliminated by sending gas through zeolites and/or by the means of a low temperature trap.
A third possible contaminant is H2. Above 1,000°C, the hydrogen diffuses through the quartz tube from the outside. Various techniques have already been conceived (in the laboratory), in order to avoid this phenomenon. We then use a tube with double or triple inner surfaces, in between which an inert gas circulates [REV 69, VAN 72].
Lastly, the native oxide and the adsorbed impurities are also contamination sources. The silicon substrate is covered at room temperature with a native oxide layer (thickness ~1 nm) and with adsorbed impurities. This layer can be removed before the oxidation, through 10 minute processing under H2 at 1,200°C [REV 69], followed by a neutral gas annealing.
Another means reported by several authors [KAM 77, HOR 78] would consist of carrying out oxidations with low partial oxygen pressures. This is made possible by mixing O2 with inert gases (N2 or Ar). However, knowing that neutral gases always contain traces of impurity, it is almost impossible to simultaneously obtain dry oxygen and low values of PO2.
The vapor flux with atmospheric pressure is obtained by heating a balloon containing very high purity water. The entry channel of the vapor into the furnace is itself heated to avoid any condensation. By mixing the vapor with O2, we can also carry out the oxidation under wet oxygen. In order to improve the purity of the vapor, we resort to H2O produced by pyrogenic synthesis [DEA 78]. It is obtained by the reaction of O2 with H2, simultaneously introduced and heated, before they reach the oxidation furnace.
At the same time as oxidation, there are other methods widely spread in the semiconductor industry, including CVD (chemical vapor deposition), evaporation and the sputtering of targets. We will limit ourselves to the case of the CVD technique.
Chemical vapor deposition can be carried out by pyrolysis of silane SiH4 in the presence of oxygen. The basic equations are then as follows:
Derivatives of silane such as Si(OR)4 and the oxidizing compounds such as N2O and CO2 can also be used.
Whereas the high partial pressures of silane PSiH4 and the high temperatures (600–1,000°C) support the formation of H2O, low pressures and low temperatures lead to the formation of H2. Silica films are formed by a heterogeneous reaction on the surface. However, in the gas phase, due to nucleation process, a homogeneous cloud of colloidal particles also occurs. This is a significant contamination source. Indeed, some of these particles condense in a white powdery coating on the surfaces at low temperatures, such as the walls of the reactor.
There are several types of CVD techniques at room temperature:
— low temperature with normal pressure (low temperature CVD –LTCVD);
— low pressure (low pressure CVD – LPCVD);
— enhanced plasma (plasma enhanced CVD – PECVD).
LTCVD reactors function under normal pressure conditions and at temperatures lower than 500°C. In most commercial facilities, the substrate is heated by the Joule effect, whereas the walls and entry points of gas are cooled to remove heterogeneous nucleation in the gas phase. The complete elimination of the contaminating particles constitutes one of the most serious difficulties of this technique.
The pressure can vary between 0.5 and 1 Torr and the temperature between 700 and 800°C. This way, we manage to increase the mass transfer of the gas phase towards the substrate of an order of magnitude. It is the surface reaction that determines the oxide growth rate. Thus, we decrease the effects related to mass transfer the variations. Since the walls of the low pressure reactor are hot, the films deposit there in the same way as on the substrate. The schematic diagram of a LPCVD reactor is presented in Figure 1.2.
A glow discharge plasma is generally created with pressures ranging between 0.01 and 1 Torr. This plasma splits up the gas molecules to form very reactive species, in order to carry out a deposition in chemical phase of low temperature films. J.R. Hollahan mentions the silica film deposition for temperatures ranging between 250 and 350°C [HOL 79].
An article by W. Kern and R.S. Rosler [KER 77] reviews the basic principles, the deposition parameters, the advantages and the limitationss of various CVD techniques depositing dielectric passivation films.
We mention here other methods used in the laboratory to grow silica films on silicon.
An oxide is grown in an electrolytic cell by implementing a potential difference between the silicon wafer, acting as an anode, and a cathode plate immersed in an electrolyte or a gas plasma [DEL 71].
The aqueous solutions containing salt traces are not desirable for silicon oxidation, because they cause its corrosion. The oxide thus formed is porous.
The following non-aqueous solutions have been used to form compact SiO2 films [CRO 71a, DRE 66, SCH 57]:
A schematic representation of an anodic oxidation cell is given in Figure 1.3.
Oxidations are performed at a constant current (typically 1 to 10 mA.cm-2). As the oxide thickness grows, the cell voltage increases up to a pre-established value. This value is then maintained constant until the density of the current decreases, to reach a value of about 10 µA.cm-2.
The oxidation process can last from 1 to 10 hours, depending on the desired oxide thickness. With these experimental parameters, we obtain SiO2 films of 0.6 nm per volt. In practice, it is rather easy to carry out oxidations up to 300 V (180 nm of silica). For values of higher voltages, it is necessary to be freed from the corrosion phenomena of impurities, such as fluorine ions [CRO 71b].
Voltages going up to 500 V (i.e. X = 300 nm) can be reached using a bath of dihydroxydiethyl ether containing 6.7 × 10-3 mole/liter of Al(NO3)3, 9H2O.
With this type of oxidation method, the film growth is due to the ionization current, representing unfortunately only 1% of the total current. Indeed, the main part of the current (99%) is an electronic current circulating through the oxide during growth.
We can also dope the oxides during the anodic growth thanks to suitable salts, such as diethylphosphate [SCH 64]. The doped oxides can be used as diffusion sources of controlled composition [SCH 65].
This technique has some advantages:
— oxidations are carried out at room temperature;
— we can produce homogeneous films and repeatable thicknesses even in the case of low thicknesses, of a few nm.
It however involves disadvantages:
— the method is performed in a wet environment;
— the insulating properties of the obtained materials are not as good as those reached by thermal oxidation. They are deteriorated by the presence of water, coming from the external environment [DRE 66, NAN 70], as well as by the possible contamination of the electrolyte by alkaline ions.
In 1963, Miles and Smith [MIL 63] suggested using plasmas to make oxides grow. Since then, this technique was applied to various metals (such as Ta, Nb, Al, Mg, etc.) and to semiconductors (Ge, Si, GaAs, etc.). A review discusses the physical mechanisms implied for this type of oxidation [GOU 81].
In this method, the electrolyte is substituted by a gaseous plasma of oxygen. The oxide surface is bombarded by a large number of charged and neutral species (molecules, atoms, ions, free radical and electrons), whose states of charge and concentrations depend on the various phenomena taking place in the plasma. These will vary with the type of discharge and with the experimental parameters, such as geometry, pressure, etc.
This method is called “plasma oxidation”, when the substrate is electrically floating, with respect to the plasma, or “plasma anodization”, when the substrate is positively biased. During plasma oxidation, no current circulates through oxide, following the example of what occurs during thermal oxidation. On the other hand, in the case of plasma anodization, an electrical current goes through the oxide, as in the case of the electrolytic anodic oxidation. The latter is the only way to obtain important thicknesses. Several types of discharge and geometry can be used.
This is obtained by applying a continuous voltage (around 1 kV) between the anode and the cathode immersed in a gas at low pressure (10-2 to 1 Torr). The electronic density is of the order of 1010cm-3. A common device is presented in Figure 1.4 [OHA 70].
The highly energetic ions due to the high voltage can produce sputtering of the electrodes. This can then involve a strong contamination of the oxide film by the sputtered cathode material, even if the sample is not facing the electrode [LES 78]. This contamination affects the oxide characteristics, as well as the growth kinetics. The ion bombardment on the substrate can also involve an ablation of the oxide film itself. In order to reduce this effect, the sample must be located as far as possible from the electrodes. Moreover, the electrode must be made of a material with low sputtering rates (Al, Si, etc.).
Copeland and Pappu [COP 71] reported a low growth rate of oxide (100 nm in 4–5 hours) at low temperatures (32–55°C).
The hot cathode emits an electron flux producing a plasma with a higher electron density (1010−1011 cm-3), when the plasma is magnetically confined [GOU 80]. This discharge can be maintained with a low DC voltage (lower than 50 V) and the plasma is thus free from very energetic particles. The main problem is the sputtering of the hot cathode in the oxygen atmosphere. High quality SiO2 films have been produced by this method, with fast growth rates (90 nm in 10 minutes), by setting the substrate at a temperature of 225°C [LIG 70].
This discharge is obtained by applying a RF voltage (a few MHz) [PUL 73, PUL 74] or a microwave (a few GHz) [KRA 67, LIG 65]. The oxygen pressure can vary between 0.01 and 1 Torr. The advantage of this technique is that the electron density can reach 1013cm-3 [LIG 65]. The major disadvantage lies in the fact that the energetic particles can damage the oxide.
By using the microwave discharge, we can obtain strong oxidation rates: 200 nm in 5 minutes [LIG 65] and 600 nm in 2 hours [KRA 67], at a substrate temperature of 300–400°C.
In the case of the RF discharge, we reach 3 nm/min [PUL 73] or 5.2 nm/min [PUL 74], with a substrate temperature around 200°C. Chang et al. [CHA 80] have shown that, by mixing a small quantity of CF4 with O2, we considerably increased growth rates with a contamination of fluorine not exceeding 0.5 atom for 100.
Anodic plasma has the following characteristics (for silicon, as well as for other materials):
— very low efficiency of the ionic current (less than 1%);
— if the total current is constant, the oxidation rate generally decreases as growth proceeds;
— the electron bombardment on the substrate surface plays a crucial role in oxidation. The growth rate is enhanced with the electron temperature, but strongly decreases, if we prevent the electrons from reaching the sample [OHA 73].
This technique offers the following advantages:
— this is a process in a dry environment;
— growth rates are significant at low temperatures.
It also has disadvantages, the main ones being:
— the oxides thus produced have in general high densities of electron traps and thus require annealing, which can be at high temperature (approximately 1,000°C) [LIG 65]. This disadvantage can be mitigated by reducing the bombardment of the energetic ions on the surface of the substrate. That was done [CHA 77, TSU 78] by removing the plasma source of the process chamber, where the high frequency discharge is taking place.
Ultra-thin layers (1–10 nm) have a very important advantage, because they illustrate the first stages of silicon oxidation. They are also very important practically in the manufacture of very high integration silicon components. These films are often grown under a low oxygen pressure by a mixture of O2 with an inert gas. However, as mentioned previously, this oxidation mode is very sensitive to water vapor contamination. With a set up under ultra-high vacuum, we can use very low partial oxygen pressures, while minimizing contaminations. Moreover, a very clean silicon surface, free from natural oxide, is easily obtained by heating under an ultra-high vacuum.
This type of technique was described and used by Hopper et al. [HOP 75] and Smith et al. [SMI 82]. The wafers were electrically heated, thanks to a current crossing them from a sample-holder made of platinum or molybdenum. The basic vacuum of the device was tiny until approximately 10-10 Torr. The reaction chamber was then drained three times with pure oxygen containing at maximum 3 ppm of water vapor. The pressure there was then reduced to a value lower than 10-7 Torr. Later, the sample was heated up to 1,200°C – 1,350°C for 5–10 minutes, to remove the superficial oxide and the carbonaceous contaminants. Immediately after this cleaning process, the substrate temperature is reduced to the desired value and the oxygen is introduced from bakeable UHV leakage valves, until we reach the desired pressure.
The adjustable experimental parameters have been examined between the following values: 700 – 950°C, 50 – 1,200 Torr in the case of Hopper et al. [HOP 75] and 890 – 1,130°C, 5.10-5 – 5.10-2 Torr for Smith et al. [SMI 82]. The latter have determined the critical values of oxygen pressure and substrate temperature, for which the oxide growth is made possible. These critical conditions are illustrated in Figure 1.5 [SMI 82].
Derrien et al. [DER 82] measured the oxidation kinetics for pressures ranging between 10-4 – 10-2 Torr and 700 – 1,050°C. The silicon surface is thermally cleaned under an ultra-high vacuum at 1,100°C, sometimes leaving very little carbon traces. Figure 1.6 [ROC 84] shows a furnace containing a quartz tube heated by the Joule effect, connected to an ultra-high vacuum room. It enables oxidations in a very wide range of pressures (10-6 - 60 Torr), under an ultra-dry oxygen atmosphere (less than 3 ppm H2O).This oxygen can be strongly enriched in 18O (beyond 99%), to study the oxidation mechanisms. The purity and the labeling of the gas are controlled by a mass spectrometer. However, in this device, the water vapor contamination is always possible, if the gas oxygen reacts with hydrogenated species diffusing through the inner surface of the quartz tube.
Katz and Howells [KAT 79] described a thermal scanning processing system under flowing water vapor at high pressure (see Figure 1.7).The metal walls of the furnace are covered with quartz. Due to their coefficient of thermal expansion, they limit the processing temperature to a maximal value of 750°C. High purity deionized steam is injected into the furnace with a pressure of 20 atm. The pressure regulation is ensured by the opening of a load shedding valve, thanks to a solenoid. The steam flow rate at 20 atm is maintained at 4 liters/min. The oxidation kinetics carried out under such conditions at 725°C were published in [KAT 79]. The latter shows results identical to those obtained at 1,050°C under water vapor at atmospheric pressure.
Thus, 1 µm of oxide is obtained after a 5–hour processing. The oxides carried out at high pressure present higher refractive indexes (1.474 against 1.469), higher densities (2.31 g.cm-3 against 2.21 g.cm-3) and lower etching speed rates in buffered HF solutions (855 Å/min against 922 Å/min), than oxides obtained at 900°C under atmospheric pressure of water vapor. Nevertheless, this type of oxide presents similar electrical characteristics, after being carefully submitted to suitable annealing processing (the measured interface defects are about 3 × 1010 cm-2).
In this section, we present some properties of silica, useful for a better understanding of silicon oxidation and the growth models developed in this chapter.
During silicon oxidation, the oxide growth proceeds because a phenomenon of atomic transport takes place through the film (as shown in the following sections).
Thus, we must consider the transport properties of silica, which should depend in theory on its atomic structure, and thus on its physical characteristics.
The basic silica structure is represented in Figure 1.8. It is a tetrahedron, with in its center, a silicon atom linked to 4 oxygen atoms located at the corners. Each oxygen atom belongs to 2 adjacent tetrahedrons, while being linked to 2 silicon atoms. The mean value of inter-atomic distances was measured by Mozzi and Warren [MOZ 69] in the fused quartz with a 2.20 g/cm3 density.
They report the following values:
— 1.62 Å for the Si-O bond;
— 2.62 Å for the O-O bond;
— 3.12 Å for the Si-Si bond.
The stability zones of silica and polymorphic transformations are represented in Figure 1.9; and the various structures in Figures 1.10, 1.11 and 1.12.
Revesz [REV 80] allocated the wide distribution of measured angles, Φ, for the Si-O-Si bond to the flexibility of the Si-O bond that has three components (covalent-σ, covalent-π and ion (40%)).The contribution of each of its components would vary with Φ without significantly changing the total bonding energy.
Indeed, several configurations of the same energies could coexist (see Table 1.1).
Moreover, it was shown that in the crystalline structures, the π bonding was enhanced along the directions parallel to axis c.
The properties of vitreous silica depend on its thermal history and on the impurities it can contain [HET 62] (OH, H, Al, etc.). That could explain the wide range of parameters given in Table 1.2.
Refraction index n (at 546 nm) |
Density ρ (g.cm-3) |
Coefficient of thermal expansion (10-6.K-1) |
---|---|---|
1.460 | 2.200–2.206 | 0.5–1.2 |
The porosity and the transparency of the structure are two of the most important physical characteristics of silica films. We can quantify them by using mass density values. However, for thin films this is not easily directly measurable, contrary to the refraction index. Two equations connect the density ρ, with the refraction index n, of an environment.
The Gladston-Dale equation:
[1.1]
The Lorentz-Lorentz equation:
[1.2]
with k1 and k2, constants where k2 can also be expressed according to the polarizability α and to the molar mass M of silica:
These equations (see Figure 1.13) are based on the hypothesis that the basic material remains identical in its atomic structure, but with different porosities and optical transparency.
To understand the transport mechanisms taking place during oxidation, we should not only know the SiO2 structure, but also examine the way in which the species diffuse and/or possibly react in amorphous silica. For example, the rare gases diffuse without reacting, whereas hydrogen reacts with its diffusive environment for temperatures higher than 500°C. The species that will be examined are those that we generally encounter during oxidation and annealing processes.
The diffusion of rare gases helps us to define two important concepts: solubility and diffusivity.
The rare gases (He, Ne, Ar, Xe) diffuse interstitially without reacting with the silica network.
Solubility is the aptitude of a network to dissolve a gas which is immersed there at a given temperature and pressure. It is best described by the C* parameter. The latter is the maximum concentration of gas dissolved in the material, after having reached the thermodynamic equilibrium.
The solubility process can be described by two distinct models: the free volume model, suggested by Doremus [DOR 94] and the complete statistical thermodynamics model, developed by Shackelford and other scientists [SHA 72, SHE 77].
The free volume model relies on the idea that the solid network contains a certain accessible volume Vs, remaining constant with the temperature and the pressure, as long as the structure is not modified. In that case, C* can be expressed as follows:
[1.3]
with Cg, the atomic concentration per unit of gas volume, and Vox, the volume of the solid.
In the model derived from statistical mechanics, the equilibrium between the gaseous and dissolved states of the same species requires us to know that the Gibbs free energies in the gaseous state Gg equals that in the dissolved state Gs. Thus, C* can be expressed as follows:
[1.4]
where p is the gas pressure and C0 is the concentration of the solubility sites (2.3 × 1021 and 1.3 × 1021 cm-3 respectively for He and Ne). K(T) is expressed:
[1.5]
where WSG is the difference between the potential energy that an atom has at rest, on the one hand, in the dissolved state, and on the other hand, in its gaseous environment
(-1.5 kcal/mole and -2.9 kcal/mole respectively for He and Ne). The expression of f(T) depends on the type and on the degree of freedom assumed [STU 70].
Diffusivity is the aptitude of a species to diffuse in an environment. This physical characteristic is the best represented by the diffusion coefficient D, given by the Arrhenius equation:
However, the experimental results show that D0 and W0 would also depend on T, in the range of temperatures considered [PER 71, SHE 77, SWE 61]:
— Sweets [SWE 61] suggests that it could be due to the fact that the considered crystalline forms of silica present strong phase reversal in the range of temperatures [160°C – 280°C].
— Perkins and Begeal [PER 71] expose models derived from the statistical model. They predict that the pre-exponential factor D0 should be independent from the temperature, if the partition function fq corresponding to the displacement in the direction of the diffusion is the same as that of a linear oscillator. If fq is the partition function of a linear translation, then D0 must be proportional to T1/2.
D0