Quantum speed limits answer a precise question: given a Hamiltonian, state and allowed resources, how quickly can quantum evolution move between distinguishable states?
Guide 15 owns quantum metrology and Fisher information; Guide 53 owns optimal control; Guide 77 owns Zeno inhibition. This guide owns lower bounds on evolution time itself: Mandelstam–Tamm, Margolus–Levitin, geometric distances, mixed-state extensions, open-system variants and the difference between a kinematic bound and an actually attainable control time.
Distance between states ÷ maximum dynamical speed = minimum possible evolution time.
1. Orthogonalisation time
Let |ψ(0)⟩ evolve under a time-independent Hamiltonian H. The orthogonalisation time τ⊥ is the smallest positive time for which ⟨ψ(0)|ψ(τ⊥)⟩=0.
A speed limit provides a lower bound on τ⊥. It does not say that every system reaches an orthogonal state, nor that the lower bound is always saturable under the allowed controls.
2. Mandelstam–Tamm bound
For time-independent H and a pure state, the Mandelstam–Tamm orthogonalisation bound is
τ⊥ ≥ πℏ/(2ΔE),
where ΔE=√(⟨H²⟩−⟨H⟩²).
The same energy uncertainty appeared in Guide 77 through the Zeno time. Large ΔE allows faster motion in projective Hilbert space; zero ΔE means the state is an energy eigenstate and its ray does not move at all.
3. Derive the geometric speed
For a pure state evolving under Schrödinger dynamics, the Fubini–Study line element obeys
ds/dt=ΔE/ℏ
under one common normalisation of the projective distance. Orthogonal rays are separated by angle π/2, so the path length must satisfy
∫₀^τ ΔE(t)dt/ℏ ≥ π/2.
For constant ΔE this yields the Mandelstam–Tamm bound.
4. Arbitrary fidelity version
For initial |ψ₀⟩ and final |ψτ⟩, define the Bures/Fubini–Study angle
𝓛=arccos|⟨ψ₀|ψτ⟩|.
Then a time-dependent generalisation is
τ ≥ ℏ𝓛 / overline{ΔE},
where overline{ΔE}=(1/τ)∫₀^τ ΔE(t)dt.
At orthogonality 𝓛=π/2 and the standard bound is recovered.
5. Worked qubit that saturates Mandelstam–Tamm
Take H=ℏΩZ/2 and initial |+⟩. The evolved state has overlap ⟨+|ψ(t)⟩=cos(Ωt/2).
The first orthogonal state occurs at Ωt/2=π/2, so
τ⊥=π/Ω.
The energy distribution is ±ℏΩ/2 with equal probability, so ΔE=ℏΩ/2. The bound gives πℏ/[2(ℏΩ/2)]=π/Ω, exactly the observed time.
6. Margolus–Levitin bound
Let E₀ be the ground-state energy and define mean energy above the ground
Ē=⟨H⟩−E₀.
The Margolus–Levitin orthogonalisation bound is
τ⊥ ≥ πℏ/(2Ē).
Unlike Mandelstam–Tamm, it depends on mean excitation energy rather than variance.
7. Why the ground reference matters
Adding a constant cI to H changes every energy expectation by c but changes no physical state trajectory except a global phase. Therefore a speed limit cannot depend on absolute ⟨H⟩ alone.
Margolus–Levitin uses energy above the ground, which is invariant under a common spectral shift.
8. Same two-level example
For H=ℏΩZ/2, ground energy is −ℏΩ/2. In |+⟩, mean energy is zero, so Ē=ℏΩ/2. Margolus–Levitin gives τ⊥≥π/Ω, again saturated.
This symmetric two-level case saturates both famous bounds simultaneously. Generic states need not.
9. Combined orthogonalisation bound
A common practical statement is
τ⊥ ≥ max{πℏ/(2ΔE), πℏ/(2Ē)}.
The larger lower bound is the active constraint for that state and Hamiltonian.
Recent work has sharpened and extended Margolus–Levitin-type bounds to arbitrary fidelity; these refinements matter when the target is not orthogonal but only partially distinguishable.
10. Quantum speed limits are path-length bounds
A lower bound built from endpoint distance and average speed is tight only if the physical trajectory follows an appropriate geodesic and uses the available speed efficiently.
A system can possess a large ΔE and still take much longer than the speed limit because its Hamiltonian drives it along a curved path through state space.
Thus “the speed limit is 10 ns” means no allowed trajectory under the assumptions can be faster than 10 ns. It does not guarantee a 10 ns control pulse exists.
11. Bures angle for mixed states
For density matrices, fidelity is
F(ρ,σ)=[Tr√(√ρ σ √ρ)]².
The Bures angle is
𝓛(ρ,σ)=arccos√F(ρ,σ).
For pure states this reduces to arccos|⟨ψ|φ⟩|.
Many mixed-state and open-system speed limits replace the pure-state projective angle by this or a related contractive distance.
12. Quantum Fisher information connection
For unitary parameter evolution generated by H, pure-state quantum Fisher information is
F_Q=4(ΔH)²/ℏ²
when the parameter is time.
The infinitesimal Bures metric is proportional to F_Q. Quantum metrology and quantum speed limits therefore share the same local geometry: Fisher information quantifies how rapidly neighbouring parameter values become distinguishable.
13. Time-dependent Hamiltonians
When H(t) varies, energy uncertainty varies too. A Mandelstam–Tamm-type bound uses the time-averaged uncertainty along the actual trajectory.
This matters for pulse shaping: a brief high-energy pulse can increase the average speed, but hardware amplitude and bandwidth constraints limit which H(t) are admissible.
14. Control time versus kinematic speed limit
Suppose a qubit can be driven by H(t)=u_x(t)X+u_y(t)Y+u_z(t)Z with amplitude bound √(u_x²+u_y²+u_z²)≤u_max.
The minimal control time depends on the allowed control directions, drift Hamiltonian and amplitude constraints. A general energy-based speed limit may be weaker than a control-specific reachability bound.
Guide 53 owns the optimisation problem; this guide supplies the universal lower-bound layer beneath it.
15. Worked bounded-control qubit rotation
Suppose only H=uX is available with |u|≤u_max and initial |0⟩. Evolution is exp(−iuXt)|0⟩. Reaching |1⟩ requires u t=π/2.
Therefore the exact minimum time is τ=π/(2u_max), achieved by constant maximum-amplitude control.
Here the speed limit and an admissible control trajectory coincide. In higher-dimensional constrained systems, this equality is much less automatic.
16. Open-system speed limits
For nonunitary dynamics, energy variance alone no longer describes the full motion. Dissipation can move a state through density-operator space even when H=0.
Open-system speed limits use norms of the generator, the time derivative of the state or quantum-information metrics to bound the time needed to reach a target distinguishability.
The exact bound depends on which norm and distance are used. Different bounds can order the same process differently.
17. Purity loss is motion too
A pure state undergoing depolarisation may remain centred at the same Bloch direction while moving radially inward toward the maximally mixed state.
A pure-state projective angle cannot describe this radial motion. Density-matrix geometry is required.
This is one reason Bures and trace-distance-based bounds are useful beyond closed-system unitary dynamics.
18. Non-Markovian memory can alter the instantaneous speed
In a non-Markovian process, information or excitation can return from the environment, increasing the magnitude of the state derivative over some intervals.
This can reduce certain open-system speed-limit estimates relative to a monotonic Markovian reference. It does not imply that memory universally accelerates every task.
Guide 76 owns the memory criteria themselves.
19. Zeno effect and speed limits constrain opposite goals
Guide 77 repeatedly projects before the state travels far, suppressing departure. Quantum-control design often seeks the opposite: move as far as possible as quickly as allowed.
Both are governed by the same local state-space geometry. Energy variance sets the initial departure rate from a ray.
20. Orthogonality is not always the operational target
Quantum algorithms may only need a target fidelity 0.999 rather than an exactly orthogonal or exact target state.
State discrimination may require a particular trace distance. Error correction may require crossing a code-space threshold. Control engineering should use the distance relevant to the task rather than defaulting to orthogonality.
21. Resource normalisation is essential
Without an energy or control-amplitude constraint, a Hamiltonian can be multiplied by an arbitrarily large constant and the same unitary path traversed arbitrarily quickly.
A meaningful speed limit therefore states the resource being held fixed: ΔE, mean energy above ground, operator norm, control amplitude, bandwidth or another physical constraint.
22. Quantum computation and gate times
A lower bound on one gate does not directly become a lower bound on an entire algorithm without accounting for parallelism, locality, communication and error correction.
Logical gates may also be much slower than physical speed limits because fault-tolerant protocols require syndrome extraction, lattice surgery or magic-state consumption.
23. Many-body locality adds another speed concept
Lieb–Robinson bounds limit how quickly information can propagate through local many-body systems. They concern spatial causal cones, whereas Mandelstam–Tamm and Margolus–Levitin concern state-space evolution.
A many-body task can be limited simultaneously by local propagation speed and global state-space speed.
24. Common misconceptions
“A speed limit predicts the actual runtime.” It gives a lower bound, often not saturable under realistic controls.
“More energy always means proportionally faster computation.” The distribution of energy, allowed control directions and hardware constraints matter.
“The bound is one universal formula.” Different physical assumptions produce different valid bounds.
“Open-system dissipation simply slows evolution.” Dissipation can itself move states rapidly through density-operator space; whether that helps a task is separate.
25. Worked synthesis problem
A qubit has H=ℏΩZ/2 and begins in |+⟩.
ΔE=ℏΩ/2.- The ground energy is −ℏΩ/2 and ⟨H⟩=0, so
Ē=ℏΩ/2. - Mandelstam–Tamm gives
τ≥π/Ω. - Margolus–Levitin gives
τ≥π/Ω. - Exact overlap is
cos(Ωt/2), whose first zero is att=π/Ω.
This state follows the geodesic efficiently enough to saturate both bounds.
26. Practice set
- Define orthogonalisation time.
- State the Mandelstam–Tamm bound.
- State the Margolus–Levitin bound.
- Why is Ē measured above the ground?
- Define the pure-state Bures/Fubini–Study angle.
- What quantity connects quantum Fisher information to the pure-state speed?
- Why can a speed limit be unattainable?
- Why does an open-system limit need more than ΔE?
- What resource must always accompany a meaningful minimum-time statement?
- How does this guide connect to the Zeno effect?
Answers
- The earliest time at which the evolved state is orthogonal to the initial state.
πℏ/(2ΔE).πℏ/(2Ē)with Ē=⟨H⟩−E₀.- A constant energy shift changes no dynamics, so the bound must be shift invariant.
arccos|⟨ψ|φ⟩|.- Energy variance; for pure unitary time encoding,
F_Q=4(ΔH)²/ℏ². - The allowed Hamiltonians may not generate the geodesic or may be subject to amplitude/locality constraints.
- Dissipative motion changes mixed-state geometry independently of Hamiltonian energy spread.
- An energy, norm, amplitude, bandwidth or comparable dynamical constraint.
- Both arise from the short-time geometry controlled by ΔH; Zeno dynamics interrupts motion while speed limits bound how fast motion can proceed.
Sources and further study
[1] L. Mandelstam and I. Tamm, The uncertainty relation between energy and time in non-relativistic quantum mechanics (1945). Foundational energy-uncertainty speed bound.
[2] Norman Margolus and Lev B. Levitin, The maximum speed of dynamical evolution, Physica D 120, 188–195 (1998). Mean-energy-above-ground speed limit.
[3] Sebastian Deffner and Steve Campbell, Quantum speed limits: from Heisenberg’s uncertainty principle to optimal quantum control, Journal of Physics A 50, 453001 (2017). Review connecting geometry, open systems and control.
[4] Ole Andersson and Hoshang Heydari, Quantum speed limits and optimal Hamiltonians for driven systems in mixed states. Geometric treatment of driven mixed-state evolution.
[5] Niklas Hörnedal and Ole Sönnerborn, Margolus-Levitin quantum speed limit for an arbitrary fidelity. Modern arbitrary-fidelity extension and saturation analysis.
Continue through Quantum Mathematics — Batch 20
Guide 77: Quantum Zeno Effect, Anti-Zeno Effect, Survival Probabilities and Zeno Subspaces develops repeated-intervention limits. Guide 79: Quantum Quasiprobabilities, Wigner Functions, Negativity and Discrete Phase Space develops phase-space geometry. Guide 80: Quantum Trajectories, Monte Carlo Wave Functions, Quantum Jumps and Diffusive Unravellings develops stochastic open-system evolution.