Contextuality as Resource

Contextuality is a resource when a specified task can be performed better with contextual correlations than with an explicitly defined noncontextual model. The connection is conditional on the operational scenario and free operations. A contextuality proof alone does not establish computational speedup or a practical device advantage.

Helpful background. Contextuality distinguishes operational noncontextuality from assignments to sharp observables.

From an impossibility proof to a task

A behavior assigns probabilities to outcomes in each compatible measurement context. A noncontextual model supplies one consistent underlying assignment, or the appropriate probabilistic generalization, across those contexts.

To make this a resource theory, specify operations that map every free behavior to another free behavior. These may include relabeling, mixing with classical randomness, and particular allowed wirings. Allowing an operation that secretly prepares the desired contextual behavior would erase the distinction.

For a linear witness $W$, suppose noncontextual behaviors obey $W\leq W_{\mathrm{NC}}$, while all behaviors under consideration obey $W\leq W_{\max}$. If

$$p=(1-\lambda)p_{\mathrm{NC}}+\lambda p_{\mathrm{other}},$$

then linearity implies

$$\lambda\geq \frac{W(p)-W_{\mathrm{NC}}}{W_{\max}-W_{\mathrm{NC}}}$$

whenever the observed witness exceeds the noncontextual bound. This lower bound quantifies a necessary nonfree component; it is not automatically a rate of distillation or an algorithmic speedup.

Computation needs a restricted baseline

In measurement-based computation with restricted classical side processing, contextual correlations can supply nonlinear Boolean processing unavailable to the baseline. Raussendorf gives precise assumptions under which sufficiently successful nonlinear computations are contextual.

In stabilizer-based models, the Clifford operations and stabilizer states form a restricted computational subtheory. Additional “magic” states can extend it. Howard and collaborators connect contextuality with the relevant nonclassical resource in specified finite-dimensional settings. Qubits require care because state-independent contextuality complicates a direct identification with computational magic.

The free computational model, dimension, measurement set, and target task must accompany any statement that contextuality “powers computation.” Otherwise one can confuse a necessary structural feature with a sufficient, efficiently exploitable resource.

Experimental interpretation

A noisy contextuality test also depends on operational equivalences or compatibility assumptions. A violation derived under ideal identities cannot be applied to imperfect measurements without justifying how those identities are established or relaxed.

For device-independent reasoning, spatial separation and input independence provide a different route to constraining models. The relation between contextuality and nonlocality is useful, but their experimental assumptions are not identical.

Exercises

A witness has $W_{\mathrm{NC}}=2$, $W_{\max}=4$, and observed value $2.6$. What follows about the decomposition above?

Solution

Every such decomposition requires $\lambda\geq0.3$. This is a witness-dependent lower bound on the nonfree weight. It does not prove that 30 percent of the experimental runs carried a preexisting “contextual substance,” or that the apparatus achieves a particular computation.

References

• Howard, Mark, Joel Wallman, Victor Veitch, and Joseph Emerson, “Contextuality supplies the ‘magic’ for quantum computation,” Nature 510, 351–355 (2014). Paper.
• Raussendorf, Robert, “Contextuality in measurement-based quantum computation,” Physical Review A 88, 022322 (2013). Paper.
• Spekkens, Robert W., “Contextuality for preparations, transformations, and unsharp measurements,” Physical Review A 71, 052108 (2005). Paper.