How Intel makes integrated circuit chips?

How Intel makes integrated circuit chips

         

  From sand to circuits

The Intel® Core™2 Duo processor with Intel 45nm High-k metal gate silicon technology

    “ The implementation of High-k and metal materials marks the biggest change in transistor technology since   the introduction of polysilicon gate MOS transistors in the late 1960s.”

Gordon Moore, Intel Co-Founder

Revolutionary

The task of making chips like these—the most complex devices ever manufactured—is no

small feat. A sophisticated chip, such as a microprocessor, can contain hundreds of millions

or even billions of transistors interconnected by fine wires made of copper. These transistors

act as switches, either preventing or allowing electrical current to pass through. A positive

charge fed to a transistor’s gate attracts electrons. This gate creates a channel between the

transistor’s source and drain through which electrical current flows, creating an “on” state.

A negative charge at the gate prevents the current from being able to flow through,

creating an “off” state for the transistor.

Intel uses its advanced manufacturing technology to build several hundred trillion transistors

every day. Intel’s breakthrough 45-nanometer (nm) High-k silicon technology enables the

production of transistors that are so small that 2 million of them would fit into the period

at the end of this sentence.

Intel’s success at reducing transistor size and maximizing performance results in advanced

processor technology that helps drive other innovations in almost all industries. Today, silicon

chips are everywhere—powering the Internet, enabling a revolution in mobile computing,

automating factories, enhancing cell phones, and enriching home entertainment. Silicon is

at the heart of an ever expanding, increasingly connected digital world.

Explore this brochure to learn how Intel makes silicon chips. If you are unfamiliar with a

technical term, see the “Terminology” section at the end of the brochure. It defines the

words that are italicized in the text.

Design

Intel creates a logic description of a new chip.

Simulators take the register transfer level (RTL) code

(shown in the background of this image) to verify the

accuracy of the specification before fabrication begins.

Silicon chip manufacturing starts with a design, or a blueprint.

Intel considers many factors. What type of chip is needed and why?

How many transistors can be built on the chip?

What is the optimal chip size? What technology will be available to create the chip? When does

the chip need to be ready? Where will it be manufactured and tested?

To answer these questions, Intel teams work with customers, software companies, and Intel’s

marketing, manufacturing, and testing staff. Intel design teams take this input and begin the

monumental task of defining a chip’s features and design.

When the specifications for the chip are ready, Intel creates a logic design, an abstract

representation of the millions of transistors and interconnections that control the flow of

electricity through a chip. After this phase is complete, designers create physical representations

of each layer of the chip and its transistors. They then create stencil-like patterns, or masks,

for each layer of the chip. Masks are used with ultraviolet light during a fabrication process

called photolithography.

To complete the design, testing, and simulation of a chip, Intel uses sophisticated computer-aided

design (CAD) tools. CAD helps designers create very complex designs that meet functional and

performance goals.

After extensive modeling, simulation, and verification, the chip is ready for fabrication. It can

take hundreds of engineers working full time for more than two years to design, test, and ready

a new chip design for fabrication.

Moore’s Law

In 1965, Gordon Moore predicted that the number of transistors on a piece of silicon would double every year—an insight later dubbed “Moore’s Law.” Intended as a rule of thumb, it has become the guiding principle for the industry to deliver ever more powerful semiconductor chips at proportionate decreases in cost. In 1975, Moore updated his prediction that the number of transistors that the industry would be able to place on a computer chip would double every couple of years. The original Moore’s Law graph is shown here.

Moore’s Law has been amazingly accurate over
time. In 1971, the Intel 4004 processor held
2,300 transistors. In 2008, the Intel® Core™2
Duo processor holds 410 million transistors.
A dramatic reduction in cost has also occurred.
In 1965, a single transistor cost more than
one dollar. By 1975, the cost was reduced to
one cent, and today Intel can manufacture
transistors that sell for less than 1/10,000 of a
cent each. Intel’s 45nm High-k silicon technology
ensures that Intel will continue to deliver
Moore’s Law into the next decade.

Fabrication

The process of making chips is called fabrication. The factories where
chips are made are called fabrication facilities, or fabs. Intel fabs are
among the most technically advanced manufacturing facilities in the
world. Within these sophisticated fabs, Intel makes chips in a special
area called a cleanroom.

Because particles of dust can ruin the complex circuitry on a chip, cleanroom air must be ultra-clean.
Purified air is constantly re-circulated, entering through the ceiling and exiting through floor tiles.
Technicians put on a special suit, commonly called a bunny suit, before they enter a cleanroom.
This helps keep contaminants such as lint and hair off the wafers. In a cleanroom, a cubic foot of
air contains less than one particle measuring about 0.5 micron (millionth of a meter) across. That’s
thousands of times cleaner than a hospital operating room.

Automation also plays a critical role in a fab. Batches of wafers are kept clean and processed quickly
and efficiently by traveling through the fab inside front-opening unified pods (FOUPs) on an overhead
monorail. Each FOUP receives a barcode tag that identifies the recipe that will be used to make
the chips inside. This labeling ensures the correct processing at each step of fabrication. Each FOUP
contains up to 25 wafers and weighs more than 25 pounds. Production automation machinery
allows for this FOUP weight, which is too heavy to be handled manually by technicians.


1 Orange FOUPs carry 300mm wafers in an automated fab.

2 Highly trained technicians monitor each phase of chip fabrication.

3 Purified air enters from the ceiling and exits through perforated floor tiles.

4 A technician holds a 300mm wafer.

Silicon, the principal
ingredient in beach
sand, is a natural
semiconductor and
the most common
element on earth
after oxygen.

The “recipe” for fabricating a chip varies depending on the chip’s proposed use. Intel uses a variety of
ingredients and performs as many as 300 steps with chemicals, gas, or light to complete fabrication.
A sandy start It all starts with silicon, the principal ingredient in common beach sand. Intel builds chips in batches on wafers made of ultra-pure silicon. Silicon is a semiconductor. This means that unlike insulators such as glass (which always resist the passage of electrons) or conductors such as copper (which generally let electrons pass through), silicon can be altered to be a conductor or an insulator. Silicon is a good choice for making wafers because it is abundant, its oxide is a good insulator, and the industry has decades of experience working with it.

To make wafers, silicon is chemically processed so that it becomes 99.9999% pure. The purified silicon is
melted and grown into long, cylindrical ingots. The ingots are then sliced into thin wafers that are polished
until they have flawless, mirror-smooth surfaces. When Intel first started making chips, the company
used 2-inch-diameter wafers. Now the company uses primarily 12-inch, or 300-millimeter (mm) wafers;
larger wafers are more difficult to process, but the result is lower cost per chip.
Layer by layer
Intel uses a photolithographic “printing” process to build a chip layer by layer. Many layers are deposited
across the wafer and then removed in small areas to create transistors and interconnects. Together,
they will form the active (“on/off”) part of the chip’s circuitry plus the connections between them, in a
three-dimensional structure. The process is performed dozens of times on each wafer, with hundreds
or thousands of chips placed grid-like on a wafer and processed simultaneously.

Process

1. Start with a partially processed die on a silicon
wafer. A chip is often referred to as die until final packaging
has been completed.

2. Deposit oxide layer. A thin film of oxide is an electrical
insulator. Like the insulator surrounding household wires,
it is a key component of electronic circuits. Intel “grows”
this layer of oxide on top of the wafer in a furnace at very
high temperatures in the presence of oxygen.

3. Coat with photoresist. A light-sensitive substance called
photoresist prepares the wafer for the removal of sections
of the oxide to create a specific oxide pattern. Photoresist is
sensitive to ultraviolet light, yet it is also resistant to certain
etching chemicals that will be applied later.

4. Position mask and flash ultraviolet light. Masks—
pieces of glass with transparent and opaque regions—are
a result of the design phase and define the circuit pattern
on each layer of a chip. A sophisticated machine called a
stepper aligns the mask to the wafer. The stepper “steps”
across the wafer, stopping briefly at incremental locations to
flash ultraviolet light through the transparent regions of the
mask. This process is called photolithography. The portions
of the photoresist that are exposed to light become soluble.

5. Rinse with solvent. A solvent removes the exposed
portions of photoresist, revealing part of the oxide
layer underneath.
6. Etch with acid. Using an acid in a process called etching,
the exposed oxide is removed. Oxide protected by the
mask remains in place.

7. Remove remaining photoresist. Finally, the
remaining photoresist is removed, leaving the desired
pattern of oxide on the silicon wafer. A new oxide
layer is complete.

Building circuits to form a computer chip is extremely precise and complex. It requires dozens of layers of various materials in specific patterns to simultaneously produce hundreds or thousands of die on each 300mm wafer. The following illustration takes a closer look
at the process of adding one layer—a single patterned oxide film.

Performing More Fabrication Steps

Laying down an oxide layer is just one part of the fabrication
process. Other steps include the following.
Adding more layers
Additional materials such as polysilicon, which conducts
electricity, are deposited on the wafer through further
film deposition, masking, and etching steps. Each layer of
material has a unique pattern.

Doping

The doping operation bombards the exposed areas of
the silicon wafer with various chemical impurities, altering
the way the silicon in these areas conducts electricity.
Doping is what turns silicon into silicon transistors, enabling
the switching between the two states, on and off, that
represent binary 1s and 0s, which provide the basis
for representing information in a computer.

Metallization

Multiple layers of metal are applied to form the electrical
connections between the transistors. Intel uses eight or
more patterned layers of copper because of its low resistance
and because it can be cost-effectively integrated into the
manufacturing process. Interconnects between layers, called
contacts, are made of tungsten. The specific patterns of
these metals are also formed using photolithography, as
described previously.

Completing the wafer

A completed wafer contains millions or even billions of
transistors connected by a multi-layer maze of metal
“wires.” Finally, the wafer is coated with a passivation
layer to help protect it from contamination and increase
its electrical stability.

Testing and Packaging

After creating layers on the wafers, Intel performs wafer sort, and a computer completes a series of tests to ensure
that chip circuits meet specifications to perform as designed.
Intel sends the approved wafers to an Intel assembly facility, where a precision saw
separates each wafer into individual rectangular chips, called die. Each functioning
die is assembled into a package that, in addition to protecting the die, delivers critical
power and electrical connections from the main circuit board on a computer. It is
this final “package” that is placed directly on a computer circuit board or in other
devices such as cell phones and personal digital assistants (PDAs).
As processor technologies advance, the demands on packaging to support and
optimize the technologies increase. Because Intel makes chips that have many
different applications, the company uses a variety of packaging technologies.

High-performance packages

Flip-chip packaging. Flip-chip packaging is an example of one of the advanced
packages that Intel uses. To package the die, Intel begins by attaching tiny metal
bumps on the die surface to the supporting base, or substrate of the package,
completing an electrical connection from the chip to the package. This method is
called “flip chip” because the silicon die are “flipped” to their front side for attachment,
compared to other types of packaging that attach to the back of the die. Intel uses
an organic or polymer substrate to enable higher performance copper electrical
interconnections from the die to the circuit board. A compliant material is then added
between the substrate and the die to manage mechanical stress. In the last step,
Intel attaches a structure called a heat spreader to help disperse the heat generated
by the chip during normal use.

Wire bond for stacked-chip packaging. Stacked-chip packaging technologies
result in packages that are only slightly larger than the multiple silicon die that they
contain. Intel stacks multiple memory and logic die in a single package to increase
performance and minimize the use of space, which are critical in today’s small handheld
devices. When attaching the die, Intel uses a special material that is optimized for
mechanical, thermal, and electrical performance to “glue” the first die to the substrate.
The other die are then stacked and “glued” to each other to create a combination of
chips that meet product performance goals.
After the die are attached, sophisticated tools bond extremely fine wires from each
die to the substrate. This process, called wire bonding, is repeated for each die included
in the stack until all die are electrically connected to the same package. The die are
then encapsulated with a molding process and a protective coating that flows into
the narrow spaces between the die and the package. Lastly, Intel attaches specialized
alloy “balls” to the bottom of the package to electrically connect the package to the
circuit board.

One more check

Intel performs reliability and electrical “tests” on each completed unit. The company
verifies that the chips are functional and perform at their designed speed across a
variety of temperatures. Because chips may end up in items ranging from automobile
engines to spacecraft and laptops, they must be able to withstand many different
environmental stresses. Chips are also tested for long-term reliability to ensure that
they will continue to perform as specified. Upon approval, chips are electrically coded,
visually inspected, and packaged in protective materials for shipment to Intel customers.

Innovation

Intel’s processor technologies offer exciting advancements, including

state-of-the-art chips with multiple cores, or “brains.”

These brains enable the efficient execution of parallel tasks, such as when a computer simultaneously

performs word processing, plays music, prints a file, and checks for viruses. Multi-core processor architectures significantly improve performance while increasing energy efficiency, which is an important consideration in today’s high-performance products.

Intel has a long history of translating technology leaps into tangible benefits. It’s not just about making technology faster, smarter, and cheaper—it’s about using that technology to make life better and our experiences richer.

Undisputed Leadership

Unwavering commitment to moving technology forward

Terminology

Intel 45nm High-k metal gate silicon technology: One of

the biggest advances in fundamental transistor design. Intel’s

innovative combination of metal gates and High-k gate dielectrics

reduces electrical current leakage as transistors get ever smaller.

Binary: Having two parts. The binary number system that

computers use is composed of the digits 0 and 1.

Channel: The region under the gate of a transistor where current

flows when the transistor is in the “on” state.

Chip: A tiny, thin square or rectangle that contains integrated

electronic circuitry. Die are built in batches on wafers of silicon.

A chip is a packaged die. See also “Microprocessor.”

Circuit: A network of transistors interconnected by wires in a

specific configuration to perform a function.

Cleanroom: The ultra-clean room where chips are fabricated.

The air in a cleanroom is thousands of times cleaner than that

in a typical hospital operating room.

Computer-aided design (CAD): Sophisticated computerized

workstations and software that Intel uses to design integrated

circuits.

Die: Alternate name for a chip, usually before it is packaged.

See also “Chip.”

Doping: A wafer fabrication process in which exposed areas

of silicon are bombarded with chemical impurities to alter the

way the silicon in those regions conducts electricity.

Drain: A highly doped region adjacent to a transistor’s

current-carrying channel that transports electrons from the

transistor to the next circuit element or conductor.

Etching: The removal of selected portions of materials to

define patterned layers on chips.

Fab: A shortened term for “fabrication facility,” where Intel

manufactures silicon chips.

Fabrication: The process of making chips.

Flip-chip packaging: A type of chip package in which a die is

“flipped” to its front side and attached to the package, compared

to packaging such as wirebond that attaches the back of the die

to the package.

Front-opening unified pod (FOUP): A container that holds and

carries wafers as part of an automated system in a fab.

Gate: The input control region of a transistor where a negative or

positive charge is applied.

Gate dielectric: A thin layer underneath the gate that isolates

the gate from the channel.

High-k material: A material that can replace silicon dioxide as

a gate dielectric. It has good insulating properties and creates a

high field effect between the gate and channel. Both are desirable

properties for high-performance transistors. Also, because High-k

materials can be thicker than silicon dioxide, while retaining the same

desirable properties, they greatly reduce current leakage.

Mask: A stencil-like pattern used during fabrication to “print” layered

circuit patterns on a wafer.

Microprocessor: The “brain” of a computer. Multiple microprocessors

working together are the “hearts” of servers, communications

products, and other digital devices. See also “Chip.”

Multi-core processor: A chip with two or more processing cores,

or “brains.”

Nanometer: One billionth of a meter.

Oxide: An insulating layer that is formed on a wafer during chip

fabrication. Silicon dioxide is one example.

Passivation: The process of coating a silicon chip with an oxide

layer to help protect it from contamination and increase its

electrical stability.

Photolithography: The process of creating a specific pattern of

material onto a silicon wafer by using ultraviolet light and a mask

to define the desired pattern.

Photoresist: A substance that becomes soluble when exposed

to ultraviolet light. Analogous to photographic film, it is sensitive to

ultraviolet light but is also resistant to certain etching chemicals.

Used to help define circuit patterns during chip fabrication.

Polysilicon: A shortened term for “polycrystalline silicon,” or silicon

made up of many crystals. This conductive material is used as an

interconnect layer on a chip, and as the gates of transistors.

Register Transfer Level (RTL) code: A computer language that

processor designers use to create a functional description of the

chip. RTL is used to define, simulate, and test processor functionality,

before actually producing the processor.

Semiconductor: A material (such as silicon) that can be altered

to conduct electrical current or block its passage.

Silicon: The principal ingredient in common beach sand and the

element used to make the wafers upon which chips are fabricated.

It is a natural semiconductor and is the most common element

on earth after oxygen.

Silicon ingot: A cylinder formed of 99.9999% pure silicon.

Source: The region of a transistor where electrons move into

the channel.

Stacked-chip packaging: A type of chip package that contains

multiple die stacked in a single package.

Transistor: A type of switch that controls the flow of electricity.

A chip may contain millions or billions of transistors.

Wafer: A thin silicon disc sliced from a cylindrical crystal ingot.

Used as the base material for building integrated circuits.

Wafer sort: An electrical test procedure that identifies the chips

on a wafer that are not fully functional.

Wire bonding: The process of connecting extremely thin wires

from a chip’s bond pads to leads on a package.