Solid-State Electronic Devices: Integrated Circuits [part 1]

Home | Articles | Forum | Glossary | Books

Learning Goals:

1. Understand IC scaling issues

2. Describe CMOS process integration

3. Understand the basics of logic gates and CCDs

4. Study the operation of different memory cells--SRAMs, DRAMs, flash

5. Understand IC packaging

Just as the transistor revolutionized electronics by offering more flexibility, convenience, and reliability than the vacuum tube, the integrated circuit enables new applications for electronics that were not possible with discrete devices. Integration allows complex circuits consisting of millions of transistors, diodes, resistors, and capacitors to be included in a chip of semiconductor. This means that sophisticated circuitry can be miniaturized for use in space vehicles, in large-scale computers, and in other applications where a large collection of discrete components would be impractical. In addition to offering the advantages of miniaturization, the simultaneous fabrication of many ICs on a single Si wafer greatly reduces the cost and increases the reliability of each of the finished circuits. Certainly discrete components have played an important role in the development of electronic circuits; however, most circuits are now fabricated on the Si chip rather than with a collection of individual components. Therefore, the traditional distinctions between the roles of circuit and system designers do not apply to IC development.

In this section we shall discuss various types of ICs and the fabrication steps used in their production. We shall investigate techniques for building large numbers of transistors, capacitors, and resistors on a single chip of Si, as well as the interconnection, contacting, and packaging of these circuits in usable form. All the processing techniques discussed here are very basic and general. There would be no purpose in attempting a comprehensive review of all the subtleties of device fabrication in a guide of this type. In fact, the only way to keep up with such an expanding field is to study the current literature. Many good reviews are suggested in the reading list at the end of this section; more important, current issues of those periodicals cited can be consulted for up-to-date information regarding IC technology. Having the background of this section, one should be able to read the current literature and thereby keep abreast of the present trends in this very important field of electronics.

1. Background

In this section we provide an overview of the nature of integrated circuits and the motivation for using them. It is important to realize the reasons, both technical and economic, for the dramatic rise of ICs to their present role in electronics. We shall discuss several main types of ICs and point out some of the applications of each. More specific fabrication techniques will be presented in later sections.

1.1 Advantages of Integration

It might appear that building complicated circuits, involving many interconnected components on a single Si substrate, would be risky both technically and economically. In fact, however, modern techniques allow this to be done reliably and relatively inexpensively; in most cases an entire circuit on a Si chip can be produced much more inexpensively and with greater reliability than a similar circuit built up from individual components. The basic reason is that many identical circuits can be built simultaneously on a single Si wafer (FIG. 1); this process is called batch fabrication. Although the processing steps for the wafer are complex and expensive, the large number of resulting integrated circuits makes the ultimate cost of each fairly low. Furthermore, the processing steps are essentially the same for a circuit containing millions of transistors as for a simpler circuit. This drives the IC industry to build increasingly complex circuits and systems on each chip, and use larger Si wafers (e.g., 12-inch diameter). As a result, the number of components in each circuit increases without a proportional increase in the ultimate cost of the system. The implications of this principle are tremendous for circuit designers; it greatly increases the flexibility of design criteria. Unlike circuits with individual transistors and other components wired together or placed on a circuit board, ICs allow many "extra" components to be included with out greatly raising the cost of the final product. Reliability is also improved since all devices and interconnections are made on a single rigid substrate, greatly minimizing failures due to the soldered interconnections of discrete component circuits.


FIG. 1 A 300-mm-diameter (about 12-inch) wafer of integrated circuits. The circuits are tested on the wafer and then sawed apart into individual chips for mounting into packages. ( Texas Instruments.)

The advantages of ICs in terms of miniaturization are obvious. Since many circuit functions can be packed into a small space, complex electronic equipment can be employed in many applications where weight and space are critical, such as in aircraft or space vehicles. In large-scale computers it is now possible not only to reduce the size of the overall unit but also to facilitate maintenance by allowing for the replacement of entire circuits quickly and easily. Applications of ICs are pervasive in such consumer products as watches, calculators, automobiles, telephones, television, and appliances. Miniaturization and the cost reduction provided by ICs mean that we all have increasingly more sophisticated electronics at our disposal.

Some of the most important advantages of miniaturization pertain to response time and the speed of signal transfer between circuits. For example, in high-frequency circuits it is necessary to keep the separation of various components small to reduce time delay of signals. Similarly, in very high-speed computers it is important that the various logic and information storage circuits be placed close together. Since electrical signals are ultimately limited by the speed of light (about 1 ft/ns), physical separation of the circuits can be an important limitation. As we shall see in Section 5, large-scale integration (LSI) of many circuits on a Si chip has led to major reductions in computer size, thereby tremendously increasing speed and function density. In addition to decreasing the signal transfer time, integration can reduce parasitic capacitance and inductance between circuits. Reduction of these parasitics can provide significant improvement in the operating speed of the system.

We have discussed several advantages of reducing the size of each unit in the batch fabrication process, such as miniaturization, high-frequency and switching speed improvements, and cost reduction due to the large number of circuits fabricated on a single wafer. Another important advantage has to do with the percentage of usable devices (often called the yield) which results from batch fabrication. Faulty devices usually occur because of some defect in the Si wafer or in the fabrication steps. Defects in the Si can occur because of lattice imperfections and strains introduced in the crystal growth, cutting, and handling of the wafers. Usually such defects are extremely small, but their presence can ruin devices built on or around them. Reducing the size of each device greatly increases the chance for a given device to be free of such defects. The same is true for fabrication defects, such as the presence of a dust particle on a photolithographic mask. For example, a lattice defect or dust particle 1/2 um in diameter can easily ruin a circuit which includes the damaged area. If a fairly large circuit is built around the defect it will be faulty; however, if the device size is reduced so that four circuits occupy the same area on the wafer, chances are good that only the one containing the defect will be faulty and the other three will be good. Therefore, the percentage yield of usable circuits increases over a certain range of decreasing chip area. There is an optimum area for each circuit, above which defects are needlessly included and below which the elements are spaced too closely for reliable fabrication.

1.2 Types of Integrated Circuits

There are several ways of categorizing ICs as to their use and method of fabrication. The most common categories are linear or digital, according to application, and monolithic or hybrid, according to fabrication.

A linear IC is an IC that performs amplification or other essentially linear operations on signals. Examples of linear circuits are simple amplifiers, operational amplifiers, and analog communications circuits. Digital circuits involve logic and memory, for applications in computers, calculators, microprocessors, and the like. By far the greatest volume of ICs has been in the digital field, since large numbers of such circuits are required. Because digital circuits generally require only the " on-off" operation of transistors, the design requirements for integrated digital circuits are often less stringent than for linear circuits. Although transistors can be fabricated as easily in an integrated form as in a discrete form, passive elements (resistors and capacitors) are usually more difficult to produce to close tolerances in ICs.

Integrated circuits that are included entirely on a single chip of semi conductor (usually Si) are called monolithic circuits (FIG. 1). The word monolithic literally means "one stone" and implies that the entire circuit is contained in a single piece of semiconductor. Any additions to the semi conductor sample, such as insulating layers and metallization patterns, are intimately bonded to the surface of the chip. A hybrid circuit may contain one or more monolithic circuits or individual transistors bonded to an insulating substrate with resistors, capacitors, or other circuit elements, together with appropriate interconnections. Monolithic circuits have the advantage that all of their components are contained in a single rigid structure that can be batch fabricated; that is, hundreds of identical circuits can be built simultaneously on a Si wafer. On the other hand, hybrid circuits offer excellent isolation between components and allow the use of more precise resistors and capacitors. Furthermore, hybrid circuits are often less expensive to build in small numbers.

2. Evolution of Integrated Circuits

The IC was invented in February 1959 by Jack Kilby of Texas Instruments. The planar version of the IC was developed independently by Robert Noyce at Fairchild in July 1959. Since then, the evolution of this technology has been extremely fast paced. One way to gauge the progress of the field is to look at the complexity of ICs as a function of time. FIG. 2 shows the number of transistors used in MOS microprocessor IC chips as a function of time. It is amazing that on this semilog plot, where we have plotted the log of the component count as a function of time, we get a straight line over four decades, indicating that there has been an exponential growth in the complexity of chips. The component count has roughly doubled every 18 months, as was noted early on by Gordon Moore of Intel corporation. This regular doubling has become known as Moore's law.

The history of ICs can be described in terms of different eras, depending on the component count. Small-scale integration refers to the integration of 1-10^2 devices, medium-scale integration to the integration of 10^2-10^3 devices, LSI to the integration of 10^3-10^5 devices, very large-scale integration to the integration of 10^5-10^6 devices, and now ultra large-scale integration (ULSI) to the integration of 10^6-10^9 devices.

Of course, these boundaries are somewhat fuzzy. The next generation has been dubbed gigascale integration. Wags have suggested that after that we will have RLSI, or "ridiculously large-scale integration."

The main factor that has enabled this increase in complexity is the ability to shrink or scale devices. Typical dimensions or feature sizes of state-of-the-art dynamic random-access memories (DRAMs) at different times are also shown as a semilog plot in FIG. 3. Once again, we see a straight line, reflecting an exponential decrease in the typical feature sizes with time over four decades. Clearly, one can pack a larger number of components with greater functionality on an IC if they are smaller. As discussed in Section 6.5.9, scaling also has other advantages in terms of faster ICs which consume less power.


FIG. 2

Moore's law for integrated circuits: Exponential increase in transistor count as a function of time for different generations of microprocessors. The dashed line indicates projections based on the International Technology Roadmap for Semiconductors (ITRS). Notice that the transistor count in the future may not increase at the same rate as in the past, due to practical constraints such as economics and power dissipation.


FIG. 3 Exponential decrease in typical feature size with time for different generations of dynamic random-access memories ( 16-kb to 32-Gb DRAMs). For reference, sizes of blood cells, bacteria, and viruses are shown on the um scale. Dimensions below 100 nm are considered to be in the realm of nanotechnology.

While scaling represents an opportunity, it also presents tremendous technological challenges. The most notable among these challenges lie in lithography and etching. However, since scaling of horizontal dimensions also requires scaling of vertical geometries, there are also tremendous challenges in terms of doping, gate dielectrics, and metallization. In addition, small features and large chips require device fabrication in extremely clean environments. Particles that may not have caused yield problems in a 1-um IC technology can have catastrophic effects for a 22 nm process, which requires purer chemicals, cleaner equipment, and more stringent clean rooms. In fact, the levels of cleanliness required bypassed the best surgical operating rooms early in the evolution shown in FIG. 3. The cleanliness of these facilities is designated by the class of the clean room. For instance, a Class 1 clean room, which was state-of-the-art in 2000, has less than 1 particle of size 0.2 um or larger per cubic foot. There are more of the smaller particles and fewer of the larger ones. Obviously, the lower the class of a clean room, the better it is. A Class 1 clean room is much cleaner than a Class 100 fabrication facility, or "fab." As one might expect, such high levels of cleanliness come with a hefty price tag: A state-of-the-art fab in 2014 comes equipped with a price tag of about 5-10 billion dollars.

In spite of the costs, the economic payoff for ULSI is tremendous. Just for calibration, let us examine some economic statistics at the dawn of the third millennium. The total annual economic output of all the countries in the world, or the so-called gross world product (GWP), is about 85 trillion US dollars. The US gross national product is about 16 trillion dollars, or about a fifth of the GWP. The worldwide IC industry output is about 350 billion dollars, and that of the entire worldwide electronics industry in which these ICs participate is about 2 trillion dollars. As a single industry, electronics is one of the biggest in terms of the dollar amount. It has surpassed, for example, automobiles (worldwide sales of about 50 million cars annually) and petro chemicals. About 1 billion smart phones are sold annually worldwide.

Perhaps even more dramatic than these raw economic numbers is the growth rate of these markets. If one were to plot IC sales as a function of time, one again finds a more-or-less exponential increase in sales with time over three decades. Of great importance to the consumer, the cost per electronic function has dropped dramatically over the same period. For example, the cost per bit of semiconductor memory (DRAM) has dropped from about 1 cent/bit in 1970 to about 10^-5 cent per bit today, an improvement of five orders of magnitude in 45 years. There are no parallels in any other industry for this consistent improvement in functionality with such a lowered cost.

Although ICs started with bipolar processes in the 1960s, they were gradually supplanted by MOS and then by CMOS devices, for reasons discussed in Sections 6 and 7. Currently, about 90, of the IC market is MOS based and about 8, BJT based. Optoelectronic devices based on compound semiconductors are still a relatively small component of the semiconductor market (about 4,), but are expected to grow in the future. Of the MOS ICs, the bulk are digital ICs. Of the entire semiconductor industry, only about 14, are analog ICs. Semiconductor memories such as DRAMs, SRAMs, and non volatile flash memories make up approximately 25% of the market, microprocessors about 25,, and other application-specific ICs (ASICs) about 20.

 

Top of Page PREV.   NEXT   Guide Index HOME