Holographic Complexity Proposals and Diagnostics

Holographic complexity asks whether aspects of a boundary computational cost are encoded by bulk volume, gravitational action, optimized path integrals, or tensor networks. This chapter treats those as inequivalent proposals. It teaches how to compute each object, remove or compare its divergences, test formation and growth, and state exactly what switchbacks, bounds, and counterexamples do—and do not—show.

Helpful background. What Task Does Complexity Answer? fixes the computational object before a bulk proxy is chosen. Regulator Dependence and Continuum Complexity explains why continuum costs require a cutoff and comparison scheme. Two-Sided Black Holes and Thermofield-Double States supplies the principal state family; Shockwaves, OTOCs, and Scrambling supplies the independent chaos observables; Holographic Quantum Error Correction helps separate a useful encoding model from a proved microscopic dictionary.

Enter the comparison

Begin by writing a boundary task as a state, unitary, channel, or operator problem with a reference, allowed gates, cost, tolerance, symmetry constraints, and ultraviolet regulator. Only then choose a bulk candidate. CV extremizes a codimension-one volume Stanford and Susskind 2014, §§ 2–3; CA evaluates the fully regulated action of a codimension-zero Wheeler–DeWitt patch Brown et al. 2016, §§ II–III. Path-integral optimization varies a Euclidean preparation cost, while tensor networks assign a cost only after an architecture, bond data, and equivalence moves are specified. Similar qualitative behavior is not an equality among these objects.

The controlling limits must accompany every conclusion. A classical bulk normally assumes large $N$, strong enough coupling to suppress $\alpha'/L^2$ corrections, $G_N/L^{d-1}\ll1$, and curvature below the ultraviolet completion scale. Kaluza–Klein modes can be truncated only when their gap exceeds every energy and inverse length used in the calculation. Finite-$N$, string-loop, higher-derivative, and singularity-sensitive contributions can change a proposal rather than merely perturb its answer.

Route through the chapter

Order	Page	Use it to answer
1	Boundary Complexity Inputs: Tasks, Reference States, and Gate Sets	Which boundary optimization problem is actually defined, and what data change its value?
2	Complexity Equals Volume Proposals	What maximal slice is used, how is it anchored, and where does the arbitrary length normalization enter?
3	Complexity Equals Action Proposals	Which Wheeler–DeWitt region and complete set of spacelike, timelike, null, joint, and counterterm contributions define CA?
4	Path-Integral and Tensor-Network Complexity	What Euclidean cost or discrete network is optimized, and why is neither automatically CV, CA, or circuit complexity?
5	Complexity of Formation and Time Growth	Which matched subtraction makes formation finite, and what early- and late-time rates follow?
6	Divergences, Counterterms, and Scheme Dependence	Which divergent coefficients are fixed, which finite pieces are scheme dependent, and which comparisons survive?
7	Shockwaves, Switchbacks, and Scrambling Diagnostics	How does a precursor produce a switchback delay, and why is that not the same observable as an OTOC?
8	Proposed Complexity Bounds and Their Counterexamples	Which energy and charge convention enters a proposed rate bound, and what do quantum black holes falsify?
9	Operational Meaning, Nonuniqueness, and Evidence Status	What is supported through the stated evidence cutoff, and what boundary dictionary remains missing?

The order is deliberate. Pages 1–4 define the competing tasks and functionals. Pages 5–7 extract quantities that can be compared without erasing their differences. Pages 8–9 stress-test the strongest interpretations against normalization changes, finite terms, quantum corrections, and absent operational maps.

A common comparison protocol

For any proposed calculation, record the following before evaluating it:

1. Boundary problem: target and reference states, gate set or path/network class, cost, tolerance, regulator, and symmetry sector.
2. Bulk problem: spacetime, matter content, state or ensemble, anchoring times, radial cutoff, and whether the geometry is classical, semiclassical, or quantum corrected.
3. Prescription data: CV length $\ell_{\mathrm V}$; or CA normalization, null-generator convention, joints, counterterms, and singularity treatment; or the path/network cost and allowed variations.
4. Controlled hierarchy: $N$, $g_s$, $\alpha'/L^2$, curvature invariants, Kaluza–Klein gap, backreaction parameter, and order of the cutoff, late-time, weak-coupling, and large-$N$ limits.
5. Observable: bare cost, vacuum-subtracted formation cost, time derivative, shock response, or proposed bound. Do not compare different rows as if they were the same quantity.
6. Falsifier: a reference-state change, a shared finite counterterm, a second geometry with matched thermodynamics, a quantum-corrected limit, or a prediction not used to calibrate normalization.

This protocol prevents a common circularity: choosing a free normalization to fit one result, changing the scheme for a second result, and then reporting both as independent confirmation.

Synthesis problem

Choose one eternal AdS black-hole family and a single regulated thermofield-double boundary task. Compute or source the leading CV and CA time dependence, including the full regulator and boundary-term conventions. Define an optimized Euclidean preparation and a tensor-network cost for the same initial state without identifying them with CV or CA. Compare four outputs: the leading divergence, formation subtraction, late-time slope, and shockwave delay.

Then repeat the comparison under one reference-state change and one shared finite counterterm choice. Finally examine a quantum-corrected BTZ family while holding the boundary clock and energy origin fixed. Your conclusion must say which observations are invariant, which depend on a prescription, which fail in the quantum example, and what operational relation would still need to be proved. Emparan, Frassino, and Sasieta’s quantum-BTZ calculation is an instructive direct test because generalized CV has a controlled classical limit while generalized CA is singularity sensitive Emparan, Frassino, and Sasieta 2022, §§ 4–6.

Review the chapter

A satisfactory answer should meet all of these criteria:

• It names the boundary task, reference, gate/path/network class, tolerance, and regulator before quoting a complexity.
• It keeps CV, CA, path-integral optimization, tensor-network size, circuit distance, and Krylov growth distinct.
• It derives at least one extremization or action-growth result rather than citing a slogan.
• It declares the CV length scale or the complete CA action, including null-boundary and counterterm choices.
• It removes divergences by a matched comparison and labels every remaining finite scheme dependence.
• It distinguishes switchback behavior from the OTOC that diagnoses scrambling.
• It states energy, charge, ground-state, time, and normalization conventions before testing a Lloyd-type rate.
• It treats an adverse quantum-BTZ result as a counterexample to the frozen hypothesis, not as permission to refit the target.
• It gives the $N$, $g_s$, $\alpha'$, curvature, Kaluza–Klein, decoupling, and truncation limits relevant to the claimed bulk regime.
• It ends at the evidence ceiling: qualitative semiclassical regularities do not establish a unique operational boundary dictionary.

Exits

For the definition and rigorous analysis of state, unitary, channel, operator, preparation, and continuum complexity tasks, continue to Quantum Information in QFT. For the standards required of exact mathematical structures and status claims, use Source Authority, Dated Status, and Specialist Review. New calculations, proposal comparisons, and changing confidence judgments require a dated holography research dossier rather than an undated textbook verdict.

Chapter-scale structure and validity checks

The chapter-scale structure map locates this page’s result inside the full reasoning chain. Follow the solid arrows through the declared inputs and checks; the dashed final arrow marks the point where an additional inference would be required.

CV, CA, path-integral, and tensor-network prescriptions are inequivalent conjectures whose divergences and operational meaning must be compared. The diagram is an original schematic, is not to scale, and uses the dashed final arrow to mark the claim boundary.

Accessible figure data (JSON)

The companion validity map turns three common overclaims into explicit failure tests. Read each row from its declared object to the diagnostic, then compare the licensed conclusion with the dashed “not” endpoint.

CV, CA, path-integral, and tensor-network prescriptions are inequivalent conjectures whose divergences and operational meaning must be compared. Each row pairs a diagnostic with the strongest supported conclusion and an explicitly unsupported promotion. The diagram is an original schematic and is not to scale.

Accessible figure data (JSON)

Claim-domain comparison

The table below gives a screen-reader-friendly comparison of three representative claims. It keeps the required declaration, approximation status, evidence timing, counterevidence, falsifier, failure condition, and licensed conclusion in one reading order.

Representative claim domains and validity boundaries for Holographic Complexity Proposals and Diagnostics

Claim object	State, ensemble, and conventions	Approximation, status, and evidence timing	Uncertainty and counterevidence	Falsifier	Failure condition	Licensed conclusion
complexity equals volume	Declare slice, length scale, and subtraction; use the volume conventions unless the page states a local replacement.	Proposal or conditional construction. Control chain: boundary complexity task → CV, CA, or path proposal → regulator and counterterms → growth and switchback tests → proposal comparison. Sources are cited on the destination page; literature checked through 10 August 2026.	Track omitted corrections, alternate branches, and competing definitions. A failed “late-growth and formation checks” check is counterevidence to the promoted claim.	late-growth and formation checks	the same observable as action	one geometric proposal
complexity equals action	Declare WdW patch and null-boundary terms; use the volume conventions unless the page states a local replacement.	Proposal or conditional construction. Control chain: boundary complexity task → CV, CA, or path proposal → regulator and counterterms → growth and switchback tests → proposal comparison. Sources are cited on the destination page; literature checked through 10 August 2026.	Track omitted corrections, alternate branches, and competing definitions. A failed “joint, counterterm, and switchback check” check is counterevidence to the promoted claim.	joint, counterterm, and switchback check	scheme-independent complexity	one action-based proposal
complexity bound	Declare gate set or bulk prescription; use the volume conventions unless the page states a local replacement.	Proposal or conditional construction. Control chain: boundary complexity task → CV, CA, or path proposal → regulator and counterterms → growth and switchback tests → proposal comparison. Sources are cited on the destination page; literature checked through 10 August 2026.	Track omitted corrections, alternate branches, and competing definitions. A failed “counterexample and normalization test” check is counterevidence to the promoted claim.	counterexample and normalization test	a universal operational theorem	a bound in the stated model

Download the structured table data (JSON).

References

• Brown, A. R., Roberts, D. A., Susskind, L., Swingle, B., and Zhao, Y. (2016). “Complexity, Action, and Black Holes.” Physical Review D 93, 086006. DOI.
• Emparan, R., Frassino, A. M., and Sasieta, B. (2022). “Holographic Complexity in Quantum Black Holes.” Journal of High Energy Physics 2022(2), 204. DOI.
• Stanford, D., and Susskind, L. (2014). “Complexity and Shock Wave Geometries.” Physical Review D 90, 126007. DOI.