Applications and Case Studies

What would make a quantum calculation useful for a molecule, a communication link or a sensor? Each application starts with an ordinary scientific or practical question. The quantum method is one part of answering it. These case studies connect the desired output to the physical model, the experiment and its cost.

A small demonstration can establish control of an important operation. A resource estimate can describe what a larger calculation would require if its assumptions hold. Reading each result at that level helps you recognize progress without confusing it with a completed application.

Helpful background. Why benchmarking is hard explains how tasks and input distributions affect comparisons. Each case study links the algorithms and physical concepts used in its examples.

Choose a question

• What are a molecule’s or material’s properties? Start with chemistry for molecular energies, active spaces and eigensolvers. Continue to materials simulation for spin and lattice models, correlations and the limits of finite systems.
• Can a difficult search be performed more efficiently? Optimization compares acceptable solution quality, repeated attempts and classical methods on specified families of instances.
• Can information be communicated securely or between remote devices? Cryptography distinguishes key distribution, cryptanalytic algorithms and attack resource estimates. Networks follows entangled pairs from generation to a usable remote service.
• Can quantum control improve precision or protect information? Sensing compares measurements of the same physical quantity. Error correction examines syndrome measurements, logical memory and the further operations needed for computation.

Use the questions for evaluating claims while reading. Negative results and limitations explains what can be learned when a method fails to improve the chosen task.

A faster attempt can give a slower answer

Suppose a solver produces a verifiably acceptable solution with probability $p$ per independent attempt. After $k$ attempts, the probability of no success is $(1-p)^k$. Thus a comparison should include the repetitions needed to reach the same success probability.

Solver	One attempt	Attempts for ≥95% success	Time budget
A	1 ms; $p=0.1$	29	29 ms
B	4 ms; $p=0.5$	5	20 ms

A finishes one attempt sooner, while B reaches the chosen reliability with a smaller time budget. The table assumes known, constant success probabilities, independent repetitions and negligible extra cost. These are calculated examples, not experimental comparisons.

The stated time is a budget allowing all $k$ attempts. If success can be recognized immediately, stopping at the first success lowers the average time. A full comparison also includes setup, compilation, measurements, classical processing and verification. The algorithmic benchmarking page derives the relevant repetition and timing formulas.

Rønnow and colleagues explain why quantum speedup depends on the comparison being made. A faster implementation than one classical program does not establish superiority over every classical method, and a small-instance result does not determine asymptotic scaling.

Accurate for which question?

An accurate solution of an approximate model is useful only insofar as the model answers the intended question. A molecular energy can agree with exact diagonalization while retaining basis-set error. A finite lattice calculation still needs a size extrapolation before it predicts an infinite system. A sensor can have low statistical noise and a large calibration bias.

Likewise, performance has several dimensions. High fidelity among accepted entangled pairs can coexist with a very low delivery rate. A long-lived logical memory does not establish the accuracy of every logical gate. Read the measured quantity and its conditions before interpreting a single headline number.

When a mechanism is unfamiliar, follow the explanations in quantum simulation, communication and networks, sensing or error correction. Reporting standards shows how to describe a result precisely. Dated assessments of changing research questions belong in Research; the examples here explain how to interpret particular studies.

Try the comparisons

1. Include setup time

In the table, B requires an additional 12 ms setup that can be reused across a batch, while A needs no setup. Which budget is smaller for one instance? For two instances?

Answer

For one instance, A needs 29 ms and B needs $12+20=32$ ms. For two, A needs 58 ms and B needs $12+40=52$ ms. Reuse changes the ranking. This assumes setup remains valid for both instances; the ≥95% guarantee applies to each instance separately, not automatically to simultaneous success on the whole batch.

2. Interpret an energy agreement

A quantum calculation matches the exact ground energy of a chosen small Hamiltonian. Has it established a useful chemistry advantage?

Answer

It supports accuracy for that Hamiltonian under the stated measurement assumptions. Chemical usefulness also depends on whether its basis, active space and neglected effects adequately describe the molecule and observable. A computational advantage additionally needs a comparison with an appropriate classical calculation at matched accuracy and cost. The small exact reference helps validate this instance; it does not settle scaling to harder instances.

3. Compare delivered pairs

Link A accepts 100 entangled pairs per second with fidelity 0.95; link B accepts one per second with fidelity 0.99. Which is better?

Answer

The intended task must specify an acceptable fidelity, timing and other requirements. If fidelity at least 0.94 suffices and accepted rate is the objective, A is better under these assumptions. If at least 0.98 is required without further processing, only B qualifies. Purification, storage and additional noise could change either comparison and require their own costs.

References

• Rønnow, T. F., Wang, Z., Job, J., Boixo, S., Isakov, S. V., Wecker, D., Martinis, J. M., Lidar, D. A., and Troyer, M. “Defining and detecting quantum speedup.” Science 345, 420–424 (2014). doi:10.1126/science.1252319; arXiv:1401.2910.