How Do Quantum Computers Work? (Superconducting Qubits, Josephson Junctions & Surface Codes)

Macroscopic Quantum Coherence and the Superconducting State

Classical microprocessors represent information through binary voltages governed by billions of semiconductor field-effect transistors [1,2]. Quantum computers replace classical bits with quantum two-level systems (qubits) capable of occupying arbitrary coherent superpositions $|\psi\rangle = \alpha |0\rangle + \beta |1\rangle$, where $\alpha, \beta \in \mathbb{C}$ and $|\alpha|^2 + |\beta|^2 = 1$ [1,3].

In superconducting quantum computing, macroscopic circuits fabricated from aluminum or niobium are cooled below their critical temperature ($T_c \approx 1.2\text{ K}$ for aluminum) [1,4]. At these cryogenic temperatures, conduction electrons bind via phonon-mediated interactions into Cooper pairs, forming a macroscopic Bose-Einstein condensate described by a single collective wave function $\Psi = \sqrt{n_s} e^{i\theta}$ [1,4]. Electrical resistance drops strictly to zero, allowing macroscopic electrical currents to exhibit quantum mechanical coherence [1,4].

"Superconducting qubits elevate quantum mechanics from single subatomic particles to macroscopic millimeter-scale electronic circuits."

The Josephson Junction and Transmon Anharmonicity

A simple LC circuit made of an inductor $L$ and capacitor $C$ behaves as a harmonic oscillator with equidistant energy levels $E_n = \hbar \omega_0 (n + 1/2)$, where $\omega_0 = 1/\sqrt{LC}$ [1,4,5]. In a harmonic oscillator, applying a resonant microwave pulse at $\omega_0$ induces transitions between $|0\rangle \rightarrow |1\rangle$, but immediately excites $|1\rangle \rightarrow |2\rangle$ and higher energy states, preventing selective qubit control [1,4].

To isolate a computational two-level subspace ${|0\rangle, |1\rangle}$, physicists replace the linear inductor with a Josephson junction—a nanometer-thin insulating barrier (typically $\text{Al}_2\text{O}_3$) sandwiched between two superconducting leads [1,4,5]. The non-linear Josephson inductance $L_J(\Phi) = \Phi_0 / (2\pi I_c \cos \delta)$ introduces anharmonicity, causing higher energy level transitions ($E_{1\rightarrow 2}$) to deviate in frequency from the primary qubit transition frequency ($E_{0\rightarrow 1} \approx 4\text{--}6\text{ GHz}$) by several hundred megahertz [1,4,5].

The modern standard design, the transmon qubit, shunts the junction with a large capacitor to operate in the regime $E_J / E_C \gg 50$, exponentially suppressing sensitivity to charge noise while maintaining sufficient anharmonicity for nanosecond microwave gate operations [4,5].

Cryogenics, Microwave Control, and Quantum Decoherence

Because the energy spacing between transmon $|0\rangle$ and $|1\rangle$ states corresponds to microwave photons with energies of only $\approx 20\text{--}30\ \mu\text{eV}$ ($h \cdot 5\text{ GHz}$), thermal fluctuations at room temperature ($k_B T \approx 25\text{ meV}$) would instantly destroy quantum states [1,4].

Quantum processors reside at the lowest stage of a dilution refrigerator, continuously cooled to $\approx 15\text{ millikelvin}$ (0.015 K) using the endothermic phase-separation mixing enthalpy of helium-3 ($^3\text{He}$) and helium-4 ($^4\text{He}$) isotopes [1,4]. Qubit states are manipulated via coax cables transmitting shaped microwave pulses at gigahertz frequencies, while state measurement is performed via dispersive readout using coupled coplanar waveguide resonators [1,4,6].

The primary operational limitation is quantum decoherence, quantified by energy relaxation time $T_1$ (spontaneous decay $|1\rangle \rightarrow |0\rangle$) and dephasing time $T_2$ (loss of phase coherence), driven by dielectric loss, quasiparticle poisoning, and flux noise [1,4,6].

"At 15 millikelvin, thermal energy is orders of magnitude lower than a single microwave photon, protecting delicate superpositions."

Fault-Tolerance and Topological Surface Codes

Because physical quantum gates have error rates on the order of $10^{-3}$, executing algorithms like Shor's prime factorization requires Quantum Error Correction (QEC) [1,3,7]. Quantum information cannot be cloned (the No-Cloning Theorem), nor can qubits be directly measured without collapsing their superposition [3,7].

To overcome this, quantum processors employ topological surface codes, entangling a two-dimensional grid of physical data qubits with interleaved ancilla qubits [1,7]. The ancillas measure multi-qubit Pauli parity stabilizers ($X_1 X_2 X_3 X_4$ and $Z_1 Z_2 Z_3 Z_4$) without reading out individual data states, pinpointing bit-flip ($X$) and phase-flip ($Z$) errors as topological syndromes [1,7]. When physical gate error rates fall below the surface code threshold ($approx 1\%$, practically requiring $< 0.1\%$), increasing code distance exponentially suppresses logical error rates, paving the way for thousands of fault-tolerant logical qubits [1,7].