An Environment of Challenges 1.4 DEFINING ELECTRONIC MATERIALS Electronic materials span such a wide range of properties that it is hard to define them. Figure 1.3 illustrates the elements used in electronic applications. Clearly, this involves most of the periodic table. Why is it that electronic devices make use of such a wide variety of elements while most applications such as automobiles make se of only a much more modest subset? Consider the possible types of materials on might select. Electronic devices and device processing methods use solids, liquids, gases, and even gas plasmas frequently. This variety is needed to achieve the level of control required to manufacture the current-generation technology. Of the solids, elemental, alloy, and compound materials are used with no class of solids more common than any other. Among compounds, it is common to require a very specifi compound. For example, for advanced dielectrics TiO2 has been considered However, this specific compound must be produced rather than any of the other oxides of titanium. Thus, the particular material required is often very well defined and only that material will do. Apparently, the general class of material does not define an electronic material We can see in Figure 1. 4 that the position of an element on the periodic table does not matter greatly. The properties of elements in the various columns of the periodic table tend to be closely related and hence these elements have similar applications he inert gases are typically unreactive, but compounds such as xenon fluorides do occur and are used occasionally in microelectronic production processes. Inert gases re largely used for etching materials by physical impacts(sputtering, see Chapter 11). The group VIla elements(halogens)are highly reactive, tend to form volatile compounds with many elements, and are used in etching. In compounds involving halogens such as CuBr are applied directly as device materials Group Vla elements produce strongly bound compounds In the case of oxides, these are generally used as dielectrics, while the elements below oxygen form the chalcogenide compounds, most of which are semiconductors The most common elements used in microelectronics are in groups Illa to Va. Group Va elements are largely used in compound semiconductor form, and as dopants in the group IV semiconductors. Group IVa elements include the common semiconductor Si and Ge as well as C. The latter is used in the forms of diamond, graphite, organic compounds or fullerene molecules. The Group Illa elements include the excellent electrical conductor Al, elements found in semiconducting compounds with group Vaelements, and as dopants in group I Va semiconductors Group llb elements form compound semiconductors with the chalcogenides and have a variety of other uses. The most common conductors are the group lb elements Cu and Au with Ag also occasionally used. Transition metals are commonly used in ompound form either as silicides or nitrides, primarily as stable contact materials bridging between Si and a highly conductive metal, or as diffusion barriers. The rare
An Environment of Challenges 11 1.4 DEFINING ELECTRONIC MATERIALS Electronic materials span such a wide range of properties that it is hard to define them. Figure 1.3 illustrates the elements used in electronic applications. Clearly, this involves most of the periodic table. Why is it that electronic devices make use of such a wide variety of elements while most applications such as automobiles make use of only a much more modest subset? Consider the possible types of materials one might select. Electronic devices and device processing methods use solids, liquids, gases, and even gas plasmas frequently. This variety is needed to achieve the level of control required to manufacture the current-generation technology. Of the solids, elemental, alloy, and compound materials are used with no class of solids more common than any other. Among compounds, it is common to require a very specific compound. For example, for advanced dielectrics TiO2 has been considered. However, this specific compound must be produced rather than any of the other oxides of titanium. Thus, the particular material required is often very well defined and only that material will do. Apparently, the general class of material does not define an electronic material. We can see in Figure 1.4 that the position of an element on the periodic table does not matter greatly. The properties of elements in the various columns of the periodic table tend to be closely related and hence these elements have similar applications. The inert gases are typically unreactive, but compounds such as xenon fluorides do occur and are used occasionally in microelectronic production processes. Inert gases are largely used for etching materials by physical impacts (sputtering, see Chapter 11). The group VIIa elements (halogens) are highly reactive, tend to form volatile compounds with many elements, and are used in etching. In rare applications, ionic compounds involving halogens such as CuBr are applied directly as device materials. Group VIa elements produce strongly bound compounds. In the case of oxides, these are generally used as dielectrics, while the elements below oxygen form the chalcogenide compounds, most of which are semiconductors. The most common elements used in microelectronics are in groups IIIa to Va. Group Va elements are largely used in compound semiconductor form, and as dopants in the group IV semiconductors. Group IVa elements include the common semiconductors Si and Ge as well as C. The latter is used in the forms of diamond, graphite, organic compounds or fullerene molecules. The Group IIIa elements include the excellent electrical conductor Al, elements found in semiconducting compounds with group Va elements, and as dopants in group IVa semiconductors. Group IIb elements form compound semiconductors with the chalcogenides and have a variety of other uses. The most common conductors are the group Ib elements Cu and Au with Ag also occasionally used. Transition metals are commonly used in compound form either as silicides or nitrides, primarily as stable contact materials bridging between Si and a highly conductive metal, or as diffusion barriers. The rare
12 The materials science of semiconductors earths are not used extensively. This is gradually changing and elements such as halfnium, erbium and gadolinium are making increasingly common appearances zE 口口口6/区 2N> UE a 2 vs the primary application of the various elements in the periodic table in microelectronics. Elements left blank are used very rare applications or are not used at all. Many elements such as Al have a number ations. In the case of al. these include as a contact/metallization, a semiconductor component, a dopant, and in insulators. Element symbols shown in gray are very rarely used
12 earths are not used extensively. This is gradually changing and elements such as halfnium, erbium and gadolinium are making increasingly common appearances. Figure 1.4: Shows the primary application of the various elements in the periodic table in microelectronics. Elements left blank are used only in very rare applications or are not used at all. Many elements such as Al have a number of applications. In the case of Al, these include as a contact/metallization, a semiconductor component, a dopant, and in insulators. Element symbols shown in gray are very rarely used. The Materials Science of Semiconductors
An Environment of Challenges In some cases the lla elements such as calcium are used as conductors or contacts although their reactivity makes them too unstable ly cases. Finally, the group la alkali metals are rarely used because of their reactivity and rapid diffusion rates in many materials, although these too are becoming more common. Both group la and lla elements are increasingly used, for example, in organic electronic devices The choice of elements is thus made based on application and properties and is driven by the resulting performance of devices. Microelectronic applications generally require relatively small amounts of material, making their availability in large quantities less important. The competitive performance-driven market provides an impetus for spending more for starting materials in order to achieve superior results. The cost of he few milligrams of paladium or platinum that might be used is a small part of the final cost of the completed device. This suggests on its face that price per unit volume of an element is not critical. If price and availability are not generally important, one might then ask, what is critical? 1.5 PURITY The fundamental property that most clearly links electronic materials is purity. All of the various classes of compounds that are used are required to contain as few other elements as possible. In many cases extreme measures are taken to prevent contami- nation. Consider the following examples of the scale of contamination problems that must be dealt with in microelectronic applications Aluminum oxide has often been used as the packaging material for military pecification computer chips because of its outstanding resistance to penetration by ontaminants. The original packages of this type were made from standard ceramic- process-grade aluminum oxide powder produced directly from bauxite ore. While this material is relatively pure, there is typically a very low level of uranium oxide contamination. This tiny amount uranium was found to be sufficient that radioactive decay in the oxide led to false data in the information stored in the chips. To avoi this, the aluminum oxide must be decomposed electrochemically to aluminum and oxygen. The aluminum is then converted to a vapor compound and purified by fractional distillation. Finally, the compound is reacted with purified oxygen to produce electronic-grade aluminum oxide. It is a lot of trouble to go through to get rid of a few uranium atoms, but it turns out to be necessary Electronic grade Si is a second example, and represents the greatest challenge materials purity in any application. Virtually all contaminants must be eliminated art per million levels with some impurities, such as transition metals, detectable effects on devices to the part per trillion level. To visualize the purity requirement for Si, imagine that maple trees represent Si atoms and that pine trees represent Fe atoms. The area of the contiguous United States is about 10 million uare kilometers or 10 trillion square meters. If one could plant an extremely dense forest of maple trees, one per 10 m- area, then you could have no more than one pine
An Environment of Challenges 13 In some cases the IIa elements such as calcium are used as conductors or contacts, although their reactivity makes them too unstable in many cases. Finally, the group Ia alkali metals are rarely used because of their reactivity and rapid diffusion rates in many materials, although these too are becoming more common. Both group Ia and IIa elements are increasingly used, for example, in organic electronic devices. The choice of elements is thus made based on application and properties and is driven by the resulting performance of devices. Microelectronic applications generally require relatively small amounts of material, making their availability in large quantities less important. The competitive performance-driven market provides an impetus for spending more for starting materials in order to achieve superior results. The cost of the few milligrams of paladium or platinum that might be used is a small part of the final cost of the completed device. This suggests on its face that price per unit volume of an element is not critical. If price and availability are not generally important, one might then ask, what is critical? 1.5 PURITY The fundamental property that most clearly links electronic materials is purity. All of the various classes of compounds that are used are required to contain as few other elements as possible. In many cases extreme measures are taken to prevent contamination. Consider the following examples of the scale of contamination problems that must be dealt with in microelectronic applications. Aluminum oxide has often been used as the packaging material for militaryspecification computer chips because of its outstanding resistance to penetration by contaminants. The original packages of this type were made from standard ceramicprocess-grade aluminum oxide powder produced directly from bauxite ore. While this material is relatively pure, there is typically a very low level of uranium oxide contamination. This tiny amount uranium was found to be sufficient that radioactive decay in the oxide led to false data in the information stored in the chips. To avoid this, the aluminum oxide must be decomposed electrochemically to aluminum and oxygen. The aluminum is then converted to a vapor compound and purified by fractional distillation. Finally, the compound is reacted with purified oxygen to produce electronic-grade aluminum oxide. It is a lot of trouble to go through to get rid of a few uranium atoms, but it turns out to be necessary. Electronic grade Si is a second example, and represents the greatest challenge in materials purity in any application. Virtually all contaminants must be eliminated to part per million levels with some impurities, such as transition metals, having detectable effects on devices to the part per trillion level. To visualize the purity requirement for Si, imagine that maple trees represent Si atoms and that pine trees represent Fe atoms. The area of the contiguous United States is about 10 million square kilometers or 10 trillion square meters. If one could plant an extremely dense forest of maple trees, one per 10 m2 area, then you could have no more than one pine
The materials science of semiconductors tree in all of the United States Now thats a weed problem! One trillion Si atoms fill cube -3 um on a side. No more than one fe atom can be allowed in such a volume of typical Si. This is more than 1000 times the purity requirement of other The actual situation in submicron devices is even worse than this already extreme value. It was estimated above that a o 1 um-scale device control volume contains 25 million atoms. Within this volume, a single atom has a bulk atom fraction of 40 parts per billion. This indicates that a single Fe atom could ruin the device in whose control volume it occurs. If a single impurity atom can ruin a device, and if one can only tolerate one bad device in 1000 to achieve a working circuit, then only one in x10 atoms could be a particularly destructive impurity such as Fe. The fact is that current circuits are designed with redundancy such that single failed transistors generally do not ruin the entire chip. Even so, it is necessary to account for the possibility of impurities such as Fe in the design of the devices and to achieve and aintain a low level of contamination in every step of the circuit fabrication We can now see why raw material price is not an issue with most elements used in semiconductor devices. The cost of purification is generally the bulk of the cost of the material. This rule of thumb does not apply to all materials and devices. Because price is directly related to supply and demand for elements, certain rare elements can be prohibitively expensive for large-scale processes. In general, however, a highe price stimulates greater production, keeping the price roughly constant. On the other and, some devices such as solar cells and lighting products must be as inexpensive as possible. In these devices, even small amounts of expensive material can be a problem 1.6 PERFORMANCE e usefulness of an electronic material most of the rest is performance. Performance can have many aspects including electronic properties of the material such as conductivity, fre carrier mobility, etc, and physical and chemical properties such as mechanical strength, stability against diffusional mixing or reaction with adjacent materials, and many more. Electronic and optical properties are related to the way in which electrons interact with the atomic structure of the material. Chemical properties depend upon the atomic bonding and the possible reactions that can take place between one material and others that it touches The performance of electronic materials can affect the lifetime, speed, efficiency,o other aspects of the device behavior. Part lifetimes are generally limited by chemical reactions or motions of atoms over time. Current technology devices are becoming
14 tree in all of the United States. Now that’s a weed problem! One trillion Si atoms fill a cube ~ 3 µm on a side. No more than one Fe atom can be allowed in such a volume of typical Si. This is more than 1000 times the purity requirement of other applications. The actual situation in submicron devices is even worse than this already extreme value. It was estimated above that a 0.1 µm-scale device control volume contains 25 million atoms. Within this volume, a single atom has a bulk atom fraction of 40 parts per billion. This indicates that a single Fe atom could ruin the device in whose control volume it occurs. If a single impurity atom can ruin a device, and if one can only tolerate one bad device in 1000 to achieve a working circuit, then only one in 2x1018 atoms could be a particularly destructive impurity such as Fe. The fact is that current circuits are designed with redundancy such that single failed transistors generally do not ruin the entire chip. Even so, it is necessary to account for the possibility of impurities such as Fe in the design of the devices and to achieve and maintain a low level of contamination in every step of the circuit fabrication. We can now see why raw material price is not an issue with most elements used in semiconductor devices. The cost of purification is generally the bulk of the cost of the material. This rule of thumb does not apply to all materials and devices. Because price is directly related to supply and demand for elements, certain rare elements can be prohibitively expensive for large-scale processes. In general, however, a higher price stimulates greater production, keeping the price roughly constant. On the other hand, some devices such as solar cells and lighting products must be as inexpensive as possible. In these devices, even small amounts of expensive material can be a problem. 1.6 PERFORMANCE Purity is typically only part of the equation determining the usefulness of an electronic material – most of the rest is performance. Performance can have many aspects including electronic properties of the material such as conductivity, free carrier mobility, etc., and physical and chemical properties such as mechanical strength, stability against diffusional mixing or reaction with adjacent materials, and many more. Electronic and optical properties are related to the way in which electrons interact with the atomic structure of the material. Chemical properties depend upon the atomic bonding and the possible reactions that can take place between one material and others that it touches. The performance of electronic materials can affect the lifetime, speed, efficiency, or other aspects of the device behavior. Part lifetimes are generally limited by chemical reactions or motions of atoms over time. Current technology devices are becoming The Materials Science of Semiconductors
An Environment of Challenges so small that motion of atoms over even a few atomic distances could cause the device to become inoperative. This tolerance is nowhere more evident than in the gate capacitor of a metal-oxide field-effect transistor(MOSFET). Current generation devices with tenth-micron minimum gate lengths have SiO2+Si3 N4 gate insulator (dielectric)thicknesses of 2.5 nm(7 layers of molecules) or less. Atoms diffusing into this oxide can cause defect states that can disrupt the performance of the insulator or cause the transistors in the device to latch in the on or off state. part of he solution to this situation is to purify the materials to such an extent that impurities that might move are eliminated. This is the reason why buildings that house integrated circuit fabrication lines(fabs")cost hundreds of millions of dollars Purification does not work when a material one wishes to use intentionally as part of the device is intrinsically inclined to move and cause trouble. Such is the reason why it has taken years to switch from Al to Cu as the metal connecting devices in integrated circuits Copper diffuses rapidly and causes very large problems if it gets into the active device regions. The solution has been to design exceptional diffusion barrier materials with which to surround the Cu conductors to prevent Cu escape. The performance of Cu is poor in terms of chemical stability but its electrical performance is sufficiently hat it outweighs other considerations Furthermore the chemical properties can be mitigated by good materials design. The debate over performance of a material relative to potential problems it may cause goes on in current technology. The high dielectric constant materials based on Ba compounds are attractive as potential replacements for Si-based dielectrics. Ba, like Cu, can cause significant problems in the wrong parts of the circuit. However, it provides potentially substantial improvements in capacitors used for data storage Manufacturers are now gradually beginning to introduce Ba compounds into their processes but only for selected devices. Other materials, more attractive than B: ompounds are also making their appearance, making the motivation to pursue b: Performance is most obvious in semiconductors, where electronic transport pheno- mena and optical properties are critical. For example, transistor switching-speeds can be limited by the time necessary for an electron to transit the control volume. The probability that an electron will fill a hole at lower energy and give off the excess energy in the form of light is likewise essential to the performance of light emitting devices. Optical detection and photovoltaic systems rely on a high probability of the reverse process, producing free electrons by light absorption. Great efforts have been made over the years to explore the periodic table in search of new semiconductors However, it is essential to consider all aspects of performance in a material, as the GaAs integrated circuit community has discovered he years. Theoretically electrons can be more easily accelerated in GaAs than in Si and live a shorter time Both of these contribute to faster device speeds. Why then has gaAs not replaced
An Environment of Challenges 15 so small that motion of atoms over even a few atomic distances could cause the device to become inoperative. This tolerance is nowhere more evident than in the gate capacitor of a metal-oxide field-effect transistor (MOSFET). Current generation devices with tenth-micron minimum gate lengths have SiO2+Si3N4 gate insulator (dielectric) thicknesses of 2.5 nm (~7 layers of molecules) or less. Atoms diffusing into this oxide can cause defect states that can disrupt the performance of the insulator or cause the transistors in the device to latch in the on or off state. Part of the solution to this situation is to purify the materials to such an extent that impurities that might move are eliminated. This is the reason why buildings that house integrated circuit fabrication lines (“fabs”) cost hundreds of millions of dollars. Purification does not work when a material one wishes to use intentionally as part of the device is intrinsically inclined to move and cause trouble. Such is the reason why it has taken years to switch from Al to Cu as the metal connecting devices in integrated circuits. Copper diffuses rapidly and causes very large problems if it gets into the active device regions. The solution has been to design exceptional diffusion barrier materials with which to surround the Cu conductors to prevent Cu escape. The performance of Cu is poor in terms of chemical stability but its electrical performance is sufficiently good that it outweighs other considerations. Furthermore, the chemical properties can be mitigated by good materials design. The debate over performance of a material relative to potential problems it may cause goes on in current technology. The high dielectric constant materials based on Ba compounds are attractive as potential replacements for Si-based dielectrics. Ba, like Cu, can cause significant problems in the wrong parts of the circuit. However, it provides potentially substantial improvements in capacitors used for data storage. Manufacturers are now gradually beginning to introduce Ba compounds into their processes but only for selected devices. Other materials, more attractive than Ba compounds are also making their appearance, making the motivation to pursue Ba compounds smaller. Performance is most obvious in semiconductors, where electronic transport phenomena and optical properties are critical. For example, transistor switching-speeds can be limited by the time necessary for an electron to transit the control volume. The probability that an electron will fill a hole at lower energy and give off the excess energy in the form of light is likewise essential to the performance of light emitting devices. Optical detection and photovoltaic systems rely on a high probability of the reverse process, producing free electrons by light absorption. Great efforts have been made over the years to explore the periodic table in search of new semiconductors. However, it is essential to consider all aspects of performance in a material, as the GaAs integrated circuit community has discovered over the years. Theoretically, electrons can be more easily accelerated in GaAs than in Si and live a shorter time. Both of these contribute to faster device speeds. Why then has GaAs not replaced