capacitor

A capacitor is a two-terminal component that stores electrical energy in an electric field between two conductive plates separated by an insulator, called the dielectric. Charge builds up on each plate in opposite polarity whenever a voltage is applied, and the amount of charge stored per volt — the capacitance, measured in farads — depends entirely on the plates' area, how far apart they are, and what insulating material sits between them. Outside a chip, capacitors are discrete parts on a board; inside a chip, they're built directly into the silicon and metal layers as stacked plates or interdigitated metal fingers, sized by geometry the same way an on-chip resistor is. How a capacitor actually gets built and used, from raw structure to a finished circuit role, looks like this:

Plate structure chosen
Parallel metal plates, MOS gate stack, or interdigitated fingers
On-chip capacitor formationCapacitance set by geometry and dielectric, not a discrete part
Dielectric deposited between plates
Thinner dielectric or higher permittivity raises capacitance
Plate area and spacing set by layout
Capacitance = permittivity × area ÷ plate spacing
Contacts formed on each plate
Connects the capacitor into the surrounding circuit
↻ next capacitor on the die
Deployed for decoupling, filtering, or charge storage
Same physics, different circuit role

A few things are worth knowing about why a capacitor's value comes from geometry and dielectric choice, not from a single fixed ingredient:

A capacitor's value comes from a simple relationship: permittivity times plate area divided by the distance between the plates, which is why shrinking that gap matters more than almost anything else. Permittivity describes how well the dielectric material supports an electric field, plate area is simply how much the two conductors overlap, and plate spacing is the thickness of the dielectric separating them. Because spacing appears in the denominator, halving the distance between plates doubles the capacitance — which is exactly why advanced chip processes chase ever-thinner dielectrics to pack more capacitance into the same silicon area, and why the gate oxide in a transistor, itself a tiny capacitor, has been thinned down to just a few atomic layers.

Three Ways to Build a Capacitor On-Chip same charge-storage physics, three different plate geometries MIM (Metal-Insulator-Metal) two stacked metal plates, thin dielectric between high density, low leakage, used in analog and RF blocks MOS Capacitor gate electrode over thin gate oxide, over silicon same structure as a transistor's gate stack, reused as storage Interdigitated Fingers comb-like metal fingers on the same metal layer no extra process steps, but lower density than MIM capacitance = permittivity × (area ÷ spacing) — thinner dielectric or more overlap means more stored charge

A capacitor resists sudden voltage changes but passes current freely as voltage changes, which is the opposite behavior from a resistor and is exactly why the two are used together everywhere. Current through a capacitor is proportional to how fast the voltage across it is changing, not to the voltage itself — a steady DC voltage eventually draws no current at all, while a rapidly switching signal passes through almost unimpeded. This frequency-dependent behavior is why capacitors are the standard tool for filtering: placed in parallel with a power supply, a capacitor smooths out fast voltage spikes while leaving the steady DC level untouched, and placed in series with a signal line, it blocks DC while letting an AC signal pass.

Decoupling capacitors exist specifically to answer a problem that a power supply alone can't solve: delivering current fast enough when a transistor switches. When billions of transistors on a chip switch simultaneously, the sudden current demand can be far faster than the power delivery network can supply from the board-level supply alone, causing the local voltage to sag right when it's needed most. Small capacitors placed as close as possible to the switching circuits — sometimes built directly into unused space between standard cells — act as local reservoirs of charge, releasing current instantly during a switching event and recharging more slowly from the main supply in between. Without adequate decoupling capacitance, voltage noise from switching can corrupt the very signals the chip is trying to compute.

Capacitor value trades off directly against leakage, size, and voltage rating, so choosing a capacitor type is choosing which of those three to sacrifice. A MOS capacitor packs the most capacitance per unit area because its dielectric is extremely thin, but that same thinness means it leaks charge over time and can't tolerate high voltage without breaking down. A MIM capacitor trades some density for much lower leakage and better voltage tolerance, which is why precision analog and RF circuits use MIM structures even though they take more process steps to build. An interdigitated finger capacitor needs no extra masks at all because it reuses existing metal layers, but needs far more die area to reach the same capacitance — a cost that's often worth paying when process simplicity matters more than density.

Read a capacitor as stored charge with a job, not a fixed component: on a chip its value comes from plate area, plate spacing, and dielectric choice combined through the same geometry logic as a resistor, its leakage and voltage tolerance are deliberate tradeoffs picked by which structure was chosen, and its role in the circuit — decoupling a power rail, filtering a signal, or storing charge in a memory cell — is what actually explains why that particular structure and value were selected in the first place.

Go deeper with CFSGPT

Get AI-powered deep-dives, save terms, and run advanced simulations — free account.

Create Free Account