Superposition, Entanglement & Quantum Gates Explained with 2×2 Matrices
By Neil Harwani • 1 min read • Published 13 Nov 2026
Quantum amplitude estimation needs on the order of 1/ε samples where classical Monte Carlo needs 1/ε² — a quadratic speed-up for pricing and risk, once fault-tolerant hardware exists.
Student lens
A qubit is a vector
A classical bit is 0 or 1. A qubit's state is a unit vector
Measuring gives 0 with probability $|\alpha|^2$ and 1 with probability $|\beta|^2$.
Gates are matrices
- X (NOT) = [[0, 1], [1, 0]] turns $|0\rangle$ into $|1\rangle$.
- H (Hadamard) = (1/√2)[[1, 1], [1, −1]] turns $|0\rangle$ into an equal superposition $(|0\rangle + |1\rangle)/\sqrt{2}$.
Applying H twice returns you to where you started — try it in the quantum lab.
Entanglement
Apply H to the first of two qubits, then a CNOT. The result $(|00\rangle + |11\rangle)/\sqrt{2}$ cannot be written as two separate qubit states: measuring one tells you the other.
Study next: tensor products and the Bloch sphere.
Young-pro lens
Where the speed-up comes from
Pricing a path-dependent option or estimating Value-at-Risk often relies on Monte Carlo simulation. Halving the error needs four times as many samples, because error falls as $1/\sqrt{N}$.
Quantum amplitude estimation reaches error $\varepsilon$ with roughly $1/\varepsilon$ queries instead of $1/\varepsilon^2$. That is a quadratic, not exponential, advantage — useful, but it only pays once error-corrected hardware can run deep circuits.
The nearer-term priority: cryptography
Shor's algorithm would break RSA and elliptic-curve cryptography on a large fault-tolerant machine. Data recorded today could be decrypted later ("harvest now, decrypt later"). NIST has published post-quantum standards, so the practical work now is:
- an inventory of where public-key cryptography is used,
- crypto-agility in systems so algorithms can be swapped,
- vendor road-maps for post-quantum TLS and signatures.
Treat claims of near-term quantum advantage in finance with care; ask which hardware and error rates the result assumes.