The physics of devices: quantum, semiconductors, electronics, photonics.Figures checked

The physics of devices, with the numbers.

How matter, charge and light behave at the scale of a device: the definitions that are right, the figures with their sources, and the calculators that give the number.

One energy window, four subjects. The quantities these pages compute, on a single scale from the thermal energy of a transistor to the ionisation of hydrogen.

CODATA 2018, materials library
10 meV100 meV1 eV10 eV20 eV17.9 meVkT ln 2, Landauer limit25.9 meVkT at 300 K0.80 eV1550 nm photon1.12 eVSi band gap1.42 eV GaAsvisible light1.65 to 3.26 eV3.62 eVper e-h pair in Si13.6 eVhydrogen ionisation

Instruments

Ten calculators, free, one answer at a time. Every result prints its source and a citation.

Also in Radiation: Cosmic radiation and flight dose calculator, X-ray tube spectrum and filtration calculator, Radiation units converter. All ten instruments

Start here

The first article of each pillar’s reading path. Each path goes from the first concept to the current state of the field in five or six pieces.

Latest

Across the four pillars, newest first.

ElectronicsHow does a CCD sensor work? A pixel’s charge can travel 4,096 steps before it is readYour phone reads each pixel where it sits. A CCD passes every pixel's charge, step by step, to an amplifier at the edge of the chip.ElectronicsAn H100 operation costs about 100 million Landauer limitsPhysics sets a floor for erasing one bit. At peak, an H100 spends about 100 million times that energy per operation: a scale, not a like-for-like test.PhotonicsWhy data in optical fibre is 32% slower than lightData in fibre is often said to move at the speed of light. In standard fibre it moves at 204,107 km/s, 68.1% of it, because the glass slows the pulse.ElectronicsCosmic Rays and Silent Data Corruption in Data CentresCosmic-ray neutrons are real, but field studies point to hardware defects behind most DRAM faults and the silent core errors operators have traced.SemiconductorsMoS2 Radio Survives Nine Months in Orbit: What It ShowsA 2026 orbital MoS2 radio experiment ran for nine months. Its measured error rate supports a promising result, with mission qualification still ahead.Quantum physicsQuantum Error Correction Threshold Theorem: Why Scale HelpsQuantum error correction threshold theorem: why Google's Willow gets more reliable as it scales, while near-threshold chips don't.ElectronicsIntel Glass Core Substrate: Why It’s Not ShippingIntel's glass core substrate is not in Clearwater Forest, which uses organic substrate and EMIB. What ECTC 2026 showed, and why glass is years away.Quantum physicsOne-Dimensional Anyons: Breaking Statistics Without BraidingOne-dimensional anyons ditch braiding entirely — cold-atom experiments show head-on collisions alone tune quantum statistics continuously.Quantum physicsMagnon Qubit Quantum Computing: Wrong Job Title?Magnon qubit quantum computing headlines miss the point: real experiments show magnons acting as quantum memory, not stand-alone qubits.Electronics3D Chip Stacking’s Thermal Limit Splits Cache From Cores3D chip stacking's thermal limit lets cache stack almost free, but logic-on-logic still cooks itself — a split that won't close for years.ElectronicsData Center SSD Shortage 2026: Why NAND Lost to DRAMThe 2026 NAND and SSD shortage comes from fabs moving lines to DRAM and HBM. NAND contract prices rise 15 to 20% in 4Q26; plan for tight supply into 2027.PhotonicsSiPM vs PMT: Why Silicon Hasn’t Won Everywhere YetSiPM vs PMT: silicon wins on efficiency, voltage and magnetic immunity, but tubes still win on collecting area per euro. The numbers and a verdict.

Unsubscribe in one click. Published every Monday only if something was published.