Invention of Silicon: History of the Semiconductor Foundation

Semiconductor Material Record

How Silicon Became the Chip Platform

Follow the linked advances in junction control, crystal purity, surface chemistry, wafer processing, integration, and MOS scaling.

1 / 7 files opened

All Material Devices Process Scale Next file

Device physics file

The Silicon p-n Junction

Selected file: The Silicon p-n Junction. Choose another file to update this evidence card.

Selected

Observed effect

On February 23, 1940, Russell Ohl tested a silicon sample whose electrical response changed sharply under light. An internal boundary divided regions with different electrical behavior.

Material explanation

Bell Labs work connected the boundary to impurities that produced electron-rich n-type silicon and hole-rich p-type silicon. Their interface rectified current and generated a voltage under illumination.

Device path

The controllable junction supplied the physical structure later used in silicon diodes, bipolar transistors, solar cells, and the isolation of early integrated-circuit components.

A useful semiconductor needed more than partial conductivity; it needed regions whose boundaries could steer charge predictably.

Dates identify different forms of evidence: laboratory observation, working device, patent filing, public demonstration, production, and broad adoption did not occur together.

Silicon became the foundation of semiconductor electronics through a linked set of material and manufacturing advances, not through one invention. Engineers learned to purify it, grow it as a nearly perfect single crystal, alter selected regions with dopants, stabilize its surface with silicon dioxide, and reproduce microscopic devices across a wafer. Those capabilities made reliable silicon transistors possible in 1954 and provided the production platform for planar integrated circuits and MOS technology soon afterward.

Milestone	What was established	Why it mattered
1940 silicon p-n junction work	Russell Ohl identified a boundary between differently conducting regions in silicon	Made rectification and photovoltaic action explainable through controlled material regions
1950-1955 purification	Zone and float-zone methods removed unwanted impurities to very low levels	Allowed deliberate doping to govern electrical behavior
1954 silicon transistor	Bell Labs made the first laboratory device; Texas Instruments produced commercial devices	Brought lower leakage and wider temperature operation than early germanium transistors
1957-1959 surface stabilization	Thermally grown silicon dioxide reduced harmful surface effects	Supported passivation, masking, insulation, and insulated-gate control
1959 planar process	Jean Hoerni retained oxide over junctions and processed devices through patterned openings	Improved reliability and made one-sided batch fabrication practical
1959-1960 silicon integration	Noyce’s interconnection concept and Last’s team produced planar monolithic circuits	Joined several components and their wiring on one manufacturable die
1959-1964 MOS development	Atalla and Kahng made the first working MOSFET; commercial devices followed	Opened the path to dense, low-power logic and memory

Why Silicon Replaced Germanium in Most Electronics

The first transistor demonstrated at Bell Labs in December 1947 was made from germanium. Germanium was easier to purify and process with the methods then available, so it dominated the first years of transistor production. Silicon was more difficult: it melts at a higher temperature, reacts with many container materials when molten, and demanded better control of contamination.

Once those manufacturing problems were reduced, silicon offered a more useful operating range. Its wider energy band gap suppresses thermally generated charge carriers more effectively than germanium does. In practical devices, that means lower reverse leakage and more stable switching as temperature rises. Early silicon transistors could therefore work in environments that caused contemporary germanium units to leak badly or stop functioning as intended.

The decisive advantage was not electrical behavior alone. Silicon can grow a dense, adherent layer of silicon dioxide when heated in an oxidizing atmosphere. That oxide can protect the surface, insulate a gate, block dopants in selected areas, and support metal interconnections. Germanium oxide does not provide the same stable production partner. Silicon thus supplied both the active semiconductor and a compatible surface material for fabricating circuits.

From Detector Crystals to Controlled Junctions

Semiconducting crystals were used before physicists could fully explain them. Point-contact crystal detectors converted radio-frequency signals into current that headphones or later amplifier stages could handle. Greenleaf Whittier Pickard patented a silicon crystal detector in 1906, but such detectors depended on finding a sensitive spot with a fine wire contact. Performance varied from one piece of material to another.

Radar research during the Second World War forced better control. Microwave receivers needed rectifiers that worked at frequencies beyond the comfortable range of vacuum-tube diodes. Programs in the United States and Britain improved silicon and germanium crystals, while Bell Labs researchers examined how trace impurities changed conductivity.

Russell Ohl’s 1940 silicon sample exposed a more useful structure. Part of the sample behaved as p-type material, in which mobile positive charge carriers called holes dominate. Another part behaved as n-type material, with electrons as the majority carriers. Where the two regions met, carriers diffused and left a depletion region containing fixed charged atoms. The resulting internal electric field allowed current to pass much more readily in one direction than the other.

That p-n junction was not yet a transistor, but it established a controllable electrical boundary inside one crystal. Junctions became the working core of rectifier diodes and solar cells. Combining two junctions produced bipolar transistors, while later integrated circuits also used reverse-biased junctions to isolate neighboring components.

Purity, Crystal Growth, and Deliberate Doping

Ordinary elemental silicon is far too impure and structurally irregular for predictable transistors. A stray atom can donate an electron, accept one, trap a carrier, or create a leakage path. The process must first remove unwanted contaminants and then add chosen dopants in controlled places and amounts.

Removing the background impurities

William Pfann’s zone-refining method moved a narrow molten region through a solid bar. Many impurities preferred the liquid phase, so the moving zone swept them toward the end of the bar. Henry Theurer adapted the idea to silicon with float-zone refining. Because the molten section was suspended by surface tension and did not rest in a crucible, it avoided contamination from a container at silicon’s high melting temperature.

For large-scale wafer supply, silicon is also produced as high-purity polycrystalline material and then grown into a single-crystal ingot. The Czochralski method, adapted to semiconductor crystals from a technique Jan Czochralski devised decades earlier, draws a rotating seed slowly from molten silicon. Atoms follow the seed’s lattice, forming a cylindrical crystal that can be sliced and polished into wafers.

Adding the right atoms back

Purification makes the starting point reproducible; doping gives it a function. Dopants with five valence electrons, such as phosphorus or arsenic, can supply mobile electrons and create n-type regions. Dopants with three valence electrons, commonly boron, create holes and p-type regions. The dopant concentration and depth influence resistance, junction voltage, switching speed, and breakdown behavior.

Early grown-junction transistors introduced dopant pellets while a crystal was being pulled. Diffusion later allowed dopant atoms to enter selected surface areas at high temperature. Ion implantation accelerated dopant ions into the wafer and gave manufacturers finer control of dose and location. Heat treatments then repaired crystal damage and moved dopants into electrically active lattice sites.

The 1954 Silicon Transistor Was Two Milestones

By the early 1950s, transistor theory was ahead of silicon manufacturing. The transistor discovered by John Bardeen and Walter Brattain in 1947 and the junction transistor developed from William Shockley’s work were based first on germanium. Large, uniform crystals and well-controlled junctions were still difficult to make in silicon.

Morris Tanenbaum solved enough of those problems to fabricate a grown-junction silicon transistor at Bell Labs on January 26, 1954. It was a working laboratory device, but Bell Labs judged that process unattractive for production and did not turn it into a commercial line.

Texas Instruments reached a different outcome a few months later. Gordon Teal, who had previously worked on crystal growth at Bell Labs, organized a semiconductor research group there. A team led by Willis Adcock made a working n-p-n silicon transistor on April 14. On May 10, Teal told an Institute of Radio Engineers conference that silicon transistors were in production and available. Texas Instruments then sold the first commercial devices.

How a Silicon Transistor Controls Current

A transistor uses a small electrical input to control a larger current. Silicon supplies a crystal lattice whose charge-carrier population can be shaped by doping and electric fields. Two device families illustrate the change from early discrete components to modern chips.

Bipolar junction transistors

An n-p-n bipolar transistor places a thin p-type base between n-type emitter and collector regions. A small base-emitter input injects carriers into the base. Because the base is thin and lightly doped, many carriers cross it and are collected by the collector. The input at the base therefore controls a larger collector current. A p-n-p device reverses the material polarities and carrier roles.

Bipolar transistors offered useful amplification and fast switching, and they powered early transistorized computers and integrated logic. Their operation involves both electrons and holes, which is why they are called bipolar. They also require input current and can consume more steady power than a MOS switch in dense digital circuits.

MOS field-effect transistors

A MOSFET separates its gate electrode from the silicon by a very thin insulating layer. A voltage on the gate creates an electric field that changes the carrier population beneath it. Once the field is strong enough, a conducting channel connects source and drain. Removing or reversing the gate voltage closes that channel.

Since the gate is insulated, an ideal MOSFET needs almost no steady gate current. Complementary MOS, or CMOS, combines n-channel and p-channel transistors so that a settled logic gate draws little static current apart from leakage. Most energy is used while charging and discharging capacitances during switching. That behavior made CMOS suited to dense logic and memory.

1. Gate voltage arrivesAn electrical potential is applied to the gate electrode while the oxide keeps it insulated from the silicon.
2. Electric field formsThe gate field reaches through the dielectric and changes the distribution of carriers near the silicon surface.
3. Channel appearsAbove the threshold voltage, a conductive path forms between source and drain in the channel region.
4. Current is controlledDrain current flows through the channel, with its amount set by terminal voltages and device geometry.
5. Channel closesWhen the gate no longer sustains inversion, the conductive path disappears and the device returns toward its off state.

Silicon Dioxide Changed the Manufacturing Problem

The surface of a semiconductor directly affects device behavior. Broken atomic bonds and contamination create electronic states that can trap charge. In early field-effect experiments, these states screened the applied field, so it could not control a surface channel as theory predicted.

Mohamed Atalla led Bell Labs research into thermally grown silicon dioxide and silicon surface stabilization. The work showed that a carefully prepared oxide interface greatly reduced the density of interfering surface states. Atalla and Dawon Kahng then fabricated the first successful silicon MOS field-effect transistor in 1959 and demonstrated it in 1960.

The oxide also changed production. It could remain on the wafer as a protective film, serve as an insulator, and act as a mask against diffusion in selected areas. Photoresist and optical masks defined windows in the oxide, allowing dopants or contacts to reach only the exposed silicon. Repeating patterns across a wafer processed many devices in parallel.

Planar Processing Made Silicon Repeatable

Fairchild Semiconductor’s early mesa transistors were etched into raised structures. Their p-n junctions reached exposed edges, where contamination and moisture could produce leakage. Jean Hoerni proposed a different geometry: keep the silicon surface flat, cover it with oxide, and process selected regions through patterned openings.

Hoerni recorded the idea in December 1957, wrote a patent disclosure in January 1959, and demonstrated a working planar transistor that March. Fairchild introduced a commercial planar transistor in 1960. Protecting the junction edges improved stability, while the flat surface made repeated photographic patterning and metal deposition more practical.

The planar process improved discrete transistors and provided a common wafer sequence for creating many components, insulating their surfaces, and adding connections above them. That production logic remains visible in modern fabrication even though transistor shapes, dielectric materials, and interconnect stacks have changed.

From One Transistor to a Circuit on One Die

The integrated circuit also has more than one defensible origin claim. Jack Kilby demonstrated a working circuit at Texas Instruments on September 12, 1958. Its transistor, capacitor, and resistive elements were formed in one piece of germanium, but thin external wires connected the parts. It proved that circuit elements could share semiconductor material.

Robert Noyce recognized that Hoerni’s planar silicon process offered a production method for monolithic circuits. Components could be formed in one wafer, electrically isolated, and connected by deposited metal paths running over the insulating oxide. Noyce filed his patent application in July 1959. Under Jay Last, a Fairchild team produced working planar monolithic circuits in 1960 and announced commercial Micrologic devices in 1961.

This distinction explains why Kilby and Noyce are both associated with the integrated circuit. Kilby built the earlier working integrated device. Noyce described the metal-over-oxide interconnection and planar silicon structure that led directly to practical batch manufacture. Neither event alone describes the entire route from idea to dependable product.

How a Silicon Wafer Becomes a Chip

Modern fabrication is far more elaborate than the first planar process, but it retains its repeated pattern-and-process logic. A polished single-crystal wafer carries many copies of a circuit. Thin films are grown or deposited, light transfers patterns into photoresist, selected material is etched away, and dopants alter chosen silicon regions. Measurement is interleaved with production because tiny deviations can affect performance or yield.

1. Prepare the waferElectronic-grade silicon is grown as a single crystal, sliced into wafers, polished, cleaned, and checked for defects and contamination.
2. Form thin filmsOxides, dielectrics, semiconductors, and conductors are grown or deposited to supply transistor regions, insulation, gates, and wiring.
3. Pattern with lightPhotoresist is coated, exposed through a reticle in a lithography system, and developed to leave a temporary pattern.
4. Etch or modifyOpen areas are etched, implanted with dopant ions, or filled with another material while protected regions remain unchanged.
5. Repeat and alignMany patterned layers are aligned over earlier ones to construct transistors, contacts, and a multi-level interconnect network.
6. Test and packageCompleted dies are electrically tested on the wafer, separated, attached to packages, connected to terminals, and tested again.

Lithography does not carve an entire finished chip in one exposure. It defines one pattern in a much longer sequence. Deposition, etching, implantation, cleaning, heat treatment, polishing, and metrology all have to work together. Modern chips may also use silicon-germanium regions, metal gates, high-permittivity dielectrics, copper or other interconnect metals, and many insulating materials while retaining a silicon wafer and silicon transistor channels for much of the circuitry.

Why MOS Technology Won the Density Race

The first MOSFET did not immediately displace bipolar transistors. It was slow, and charges or mobile ions in the oxide could shift its behavior. Bell Labs did not see an urgent telephone-system use for the early device. Researchers at Fairchild, RCA, and other companies pursued it because its structure promised simpler isolation, smaller devices, and low gate power.

Commercial MOS transistors appeared in 1964. Improvements in oxide cleanliness, interface control, photolithography, and process repeatability then made MOS integrated circuits practical. Silicon-gate technology in the late 1960s aligned the gate with source and drain regions more accurately and reduced parasitic overlap. CMOS later became the dominant digital circuit method because it combined high density with low static power.

Scaling required repeated changes rather than endless shrinking of the original flat transistor. Very thin silicon-dioxide gate layers eventually leaked through quantum tunneling, prompting high-permittivity gate dielectrics and metal gates in advanced logic. Flat channels gave way to fin-shaped and gate-all-around structures for tighter electrostatic control. The device architecture evolved, but the industrial knowledge accumulated around silicon wafers kept silicon at the center of production.

Where Silicon Is Not the Best Semiconductor

Silicon’s dominance does not make it ideal for every electronic task. It has an indirect band gap, so it converts electrical energy into light inefficiently. Compound semiconductors such as gallium arsenide and gallium nitride are better suited to many light-emitting and high-frequency applications. Silicon photonics can guide and modulate light on a chip, but lasers are often supplied by another material.

Silicon carbide and gallium nitride have wider band gaps and can support high electric fields, high switching frequencies, or elevated operating temperatures. They are used where power conversion losses and device size justify their higher material and manufacturing costs. Silicon still serves a broad range of power devices, and it remains the main platform for logic, memory, sensors, analog circuits, and mixed-signal systems.

The practical comparison is therefore application-specific. Silicon combines adequate electrical performance with mature crystal growth, oxide chemistry, lithography, huge production capacity, detailed device models, and established packaging. A rival material must offer more than a better laboratory property; it must also support reliable devices and an economical manufacturing chain.

Silicon’s Enduring Role

Silicon became the semiconductor foundation because its electrical properties could be engineered alongside its surface and production process. Purity made doping predictable. Junctions made charge directional. Silicon dioxide protected surfaces and enabled insulated gates. Planar processing turned device structures into repeatable wafer patterns, and monolithic integration placed whole circuits behind a common set of manufacturing steps.

Later chips changed nearly every detail of those early devices, yet they kept the central method: control matter region by region and layer by layer on a silicon wafer. The history is therefore not a straight line from one inventor to the microprocessor. It is a sequence in which materials science, device physics, chemistry, equipment, measurement, and factory practice solved one another’s limits.