Floating-gate MOSFET

Non-volatile memory was born when engineers learned how to trap electrons on purpose. The floating-gate MOSFET, proposed by Dawon Kahng and Simon Sze at Bell Labs in 1967, took the already powerful `mosfet` and gave it memory. Add one electrically isolated gate inside the oxide stack, inject charge onto it, and the transistor's threshold voltage shifts even after power disappears. That sounds like a small structural tweak. In practice it created the physical basis for programmable semiconductor storage.

The adjacent possible depended on the ordinary MOSFET first becoming ordinary. A floating gate only works if the surrounding oxide leaks slowly enough to hold charge for useful periods, if fabrication is clean enough to make threshold shifts predictable, and if engineers understand how fields, tunneling, and trapped charge interact inside a silicon device. Bell Labs had spent the early 1960s turning surface passivation and MOS processing from fragile laboratory tricks into reliable device physics. Once that foundation existed, the next question became obvious: if the gate is insulated so well, why not use that insulation to store information?

Kahng and Sze answered with a structure that looked almost philosophical in its simplicity. Instead of using the gate only as a momentary control electrode, they treated it as a sealed reservoir of charge. Electrons placed on the floating gate changed how much external voltage the transistor needed to switch on. Readout therefore became non-destructive; the device could reveal its state by how it conducted, not by draining away the stored charge. The memory time in the first paper was short by later standards, but the principle was decisive. A transistor no longer had to forget when the power rail went dark.

That shift shows `niche-construction` at the device level. The `mosfet` had already made dense integrated electronics practical. The floating-gate version changed the environment inside chip design by giving those circuits a compact, rewritable, non-volatile storage element. Once that element existed, engineers could imagine firmware that survived shutdown, calibration data that stayed with a sensor, and program code that no longer had to be hard-wired at the factory. The invention did not arrive because one market had neatly asked for it. It arrived because oxide quality, planar processing, and memory demand had converged hard enough that stored charge became a design resource.

The first major descendant was `eprom`. Intel's Dov Frohman turned the Bell Labs concept into a commercially useful ultraviolet-erasable memory cell in 1971, solving a practical problem in development and manufacturing: code could now be programmed, tested, erased, and programmed again without replacing the chip. From there the logic extended outward. EEPROM kept the floating gate but made erasure electrical. Toshiba then reorganized the same principle into flash memory, erasing blocks rather than bytes. What began as one transistor with a hidden gate became a whole lineage of storage architectures.

That is where `path-dependence` enters. Once firmware, BIOS chips, microcontrollers, cameras, phones, and solid-state drives were built around floating-gate descendants, the rest of the electronics stack adapted to their strengths and limits. Designers assumed that some memory should be dense, silent, shock-resistant, and persistent. Manufacturing lines assumed they would need to program chips late in assembly. Consumers came to expect devices that woke up remembering everything. Later charge-trap memories challenged parts of the floating-gate family, but they still inherited the non-volatile logic that floating gates had normalized.

The commercial story spread well beyond Bell Labs. `intel` turned the concept into EPROM and then into a core piece of personal-computing infrastructure. `toshiba` pushed floating-gate ideas into flash memory, which made portable digital storage cheap enough to become universal. `samsung-electronics` scaled that memory into the mass-market era, supplying the phones, cards, and drives that put non-volatile semiconductor storage in billions of hands. That diffusion created a `trophic-cascades`: persistent code enabled embedded systems, embedded systems enabled ubiquitous electronics, and ubiquitous electronics changed how people store, move, and trust information.

The floating-gate MOSFET therefore mattered less as a lone component than as a new answer to an old engineering problem. Magnetic media could remember without power, but it was mechanical. Logic transistors were fast, but they forgot. The floating gate joined persistence to semiconductor scaling. Once that happened, memory stopped being a separate machine and became something a transistor could do for itself.