ConceptBeginnerQuantum computing fundamentalsLesson 9

Probability Amplitudes

A probability amplitude is a complex number attached to each basis state; its squared modulus gives the outcome probability. Amplitudes, not probabilities, add and interfere.

In the state $|\psi\rangle=\sum_x\alpha_x|x\rangle$, the coefficient $\alpha_x\in\mathbb{C}$ is called the probability amplitude of the basis state $|x\rangle$. The Born rule links the amplitudes to experiment: $p(x)=|\alpha_x|^2=\alpha_x^*\alpha_x$. The normalisation condition guarantees that the probabilities sum to 1.

An amplitude has a magnitude $|\alpha_x|$ and a phase $\arg\alpha_x$. The magnitude alone determines the measurement probability in this basis, while the phase does not appear directly. Yet the relative phase matters a great deal: the states $|+\rangle=\tfrac{1}{\sqrt2}(|0\rangle+|1\rangle)$ and $|-\rangle=\tfrac{1}{\sqrt2}(|0\rangle-|1\rangle)$ have the same probabilities in the $Z$ basis, but they are fully orthogonal and can be distinguished with certainty by measuring in the $X$ basis. So a list of probabilities does not describe the quantum state; the amplitudes are needed.

Amplitudes evolve linearly: if a unitary matrix $U$ is applied, the new amplitude is $\alpha'_y=\sum_x U_{yx}\alpha_x$, i.e. the sum of the contributions of all paths $x\to y$. This linearity is what makes interference possible. Gates do not act linearly on probabilities in the same way.

In practice we do not see the amplitudes on a real device; we estimate the probabilities from the frequencies of the outcomes over $N$ shots. The estimate $\hat p=k/N$ is unbiased, with standard error $\sqrt{p(1-p)/N}$. That is why the counts shown by the lab may differ slightly from the theoretical probabilities, and the difference shrinks as the number of shots grows. Displaying the state vector with its complex amplitudes is a feature of classical simulation.

Note: to compare two states we use the overall overlap amplitude $\langle\phi|\psi\rangle$, whose squared magnitude is the probability that $|\psi\rangle$ passes a test for the state $|\phi\rangle$; this is the basis of the concept of fidelity and of quantum learning kernels.

Finally, note that the Born rule makes probability a quadratic function of the amplitudes. So if two paths reach the same outcome with amplitudes $a$ and $b$, the probability is $|a+b|^2=|a|^2+|b|^2+2\,\mathrm{Re}(a^*b)$; the last term, the interference term, can be positive or negative depending on the phase difference between the paths. It is completely absent from classical probability theory, which adds only $|a|^2+|b|^2$.

This is why amplitudes are sometimes called "square roots of probabilities", with the caveat that the description is approximate, because an amplitude carries a phase that a real square root does not, and that phase is the heart of the difference.

Mathematical formula

$$p(x)=|\alpha_x|^2=\alpha_x^{*}\alpha_x$$

Worked example

Let $|\psi\rangle=\tfrac{1+i}{2}|0\rangle+\tfrac{1}{\sqrt2}|1\rangle$.

$|\alpha_0|^2=\tfrac{(1+i)(1-i)}{4}=\tfrac{2}{4}=0.5$ and $|\alpha_1|^2=0.5$. The state is normalised.

With $N=1000$ shots, the standard error of the estimate of $P(0)$ is $\sqrt{0.5\cdot0.5/1000}\approx0.0158$, so values such as 0.487 or 0.516 are entirely normal. Halving the error needs four times as many shots.

The relative phase here is $\arg\alpha_1-\arg\alpha_0=0-\pi/4=-\pi/4$; it is not visible in a $Z$ measurement, but it changes the outcome of an $X$ measurement.

Test your understanding

The quiz needs JavaScript.

Related topics

References

  1. Michael A. Nielsen; Isaac L. Chuang (2010). Quantum Computation and Quantum Information (10th Anniversary Edition). Cambridge University Press doi:10.1017/CBO9780511976667The standard reference for quantum computing; the Crossref record gives 2012 for the Cambridge Core online edition.
  2. John Watrous / IBM Quantum (2025). Basics of Quantum Information (course). IBM Quantum Learning quantum.cloud.ibm.com
  3. Microsoft (2025). The Qubit in Quantum Computing. Azure Quantum documentation learn.microsoft.com