Search and review plan
Extension requested 2026-09-14
Extend main coverage to 1985-01-01, searching the disjoint interval through
1999-12-31 with the same core and related families. Preserve both the frozen
ten-year and 2000 backfill registers. Record new screening under
reviews/backfill-1985/ and provider requests under
searches/2026-09-14-backfill-1985/. Existing topic exclusions remain decisions;
reconsider date-only exclusions when the new window supplies evidence. Audit
native-ID coverage and previously accepted foundations independently of new IDs.
Extension requested 2026-09-11
Backfill 2000-01-01 through 2016-09-08 with the same core and related
query families. The combined corpus therefore starts in 2000; the original
2016-09-09–2026-09-09 register remains a frozen snapshot. Use a separate search
directory, review register, validation record, and coverage report for the
extension. Pre-2000 foundations remain explicitly distinguished.
Preserve previous review history. Revisit an earlier exclusion only when the
expanded date window removes its stated exclusion reason, with a new evidence
record. Do not treat every previously excluded record as a new candidate.
Scope
Exclude conference abstracts, presentations, proceedings contributions, and
proceedings volumes, even when topically relevant. A preprint of the same
conference contribution does not make it eligible. Retain an independently
identified journal publication or standalone manuscript; defer conflicting
version evidence for review. Check all manifestations before excluding a mixed
record. Journal titles containing “Proceedings” (for example PNAS and Proceedings
of the IEEE) are not by themselves evidence of conference publication.
This user-requested policy supersedes earlier conference inclusions. Preserve
those judgments in the historical registers and review-event history.
Include literature in which Andreev processes are a substantive subject,
mechanism, probe, or device resource. A passing mention or an author's surname
alone is insufficient. The original completed collection used a rolling ten-year window:
2016-09-09 through 2026-09-09. Use initial arXiv submission dates for arXiv
queries and publication dates for OpenAlex queries; retain the date provenance
when a work has multiple manifestations. Retain earlier foundational papers and
directly relevant predecessors with an earlier-foundation label.
Initial query families to explore include Andreev reflection, multiple Andreev
reflection, Andreev bound states, Andreev spectroscopy, Andreev qubits, and
crossed Andreev reflection. Expand related Josephson and superconducting
proximity-effect searches where seed papers expose a coverage gap. These are
starting hypotheses for discovery, not a fixed taxonomy or automatic inclusion
rules.
Include work on superconducting impurity states (including Yu–Shiba–Rusinov
states), Cooper-pair splitting, Josephson spectra and dynamics, quasiparticle
parity, microwave readout, and superconducting semiconductor devices when the
connection to Andreev mechanisms or accepted device platforms is substantive.
Majorana papers qualify when they analyze subgap transport, compare ordinary
Andreev states, or develop a directly relevant detection or control method.
Early gatemon and hybrid microwave circuits qualify as device predecessors;
label them related-device and explain the connection.
General topology, superconductivity, Kondo physics, or quantum computing alone
does not establish relevance. For broad books, identify the relevant sections.
Substantive formal bosonic analogues qualify as related-analogue; record the
connection explicitly. Keep
agent relevance decisions distinct from endorsement of a paper's scientific claims.
Andreev–Bashkin entrainment and Andreev–Lifshitz supersolidity are different
eponyms: they do not qualify without a separate substantive connection to
reflection, particle–hole conversion, or bound states. Exclude surname matches.
Keep research articles about relevant computational methods, but exclude
software releases, datasets, figure deposits, and supplementary files as separate
literature entries. Retain corrections and comments as distinct, labeled works.
Collection loop
- Record representative seed papers and reviews, including more than one
subtopic and time period.
- Save provider-native search specs under
searches/ and run bounded batches
through OpenAlex and arXiv. Split by date or query family where useful.
- Attach new works to the named corpus as deferred candidates, preserving
existing decisions. Raw fetch and graph expansion results do not automatically
become corpus members.
- Screen titles and abstracts, then inspect original full text for uncertain
cases. Record decisions, topical labels, reasons, and supporting evidence.
- Expand references and citing works around accepted seeds. Inspect unresolved
source links and possible preprint/article duplication.
- Compare new batches with accepted and rejected works; revisit search terms
and date windows where known relevant papers are missing.
Useful initial review labels are andreev-reflection, multiple-andreev-reflection,
andreev-bound-states, spectroscopy, qubits, crossed-andreev-reflection,
theory, experiment, review, earlier-foundation, and needs-fulltext.
Labels can overlap and evolve as the corpus grows.
Coverage record
Keep searches/ledger.csv alongside the exact request specs. For each search,
record provider, date window, run time, retrieval count, new corpus candidates,
continuation state, and any limitations. Reference local pack-run paths in notes
when they hold the detailed evidence. Status values are planned, partial,
exhausted, and failed; only use exhausted when the provider's pagination
ended for the recorded request.
PatchScholar's convenient corpus discover requests at most 100 results per
provider and starts from the beginning on each call. Use raw provider operations
for larger searches. Raw search results now expose pagination.next_spec and
provider exhaustion evidence. Follow these continuations and save each request
and result explicitly; persistent automatic topic runners are not required.
The first bounded collection is recorded in the coverage report. Before
claiming broad coverage, check query completion, year and subtopic balance,
recovery of known relevant papers, citation-expansion results, and unresolved
full-text or identity gaps. None alone establishes that every relevant paper has
been found.
Completion criteria for the ten-year collection
Exhaust the recorded core arXiv title/abstract and OpenAlex title and
title/abstract queries,
exhaust a complementary search for related papers without the Andreev keyword,
and screen every attached candidate. Check identifier duplicates, year coverage,
known-paper recovery, and citation neighborhoods across distinct subtopics.
Resolve review deferrals from available abstract or full-text evidence; retain
explicit exclusions and any external access limitations. Treat citation expansion
as a coverage audit rather than recursively including every background reference.
Report the exact completed queries and remaining discovery limitations; completion
of this protocol does not prove that no relevant paper exists outside the providers.
Author identity quality cycle
Use the systematic author workflow for identity
cleanup. Scope is the full registered corpus; user-reported duplicates are
regression tests only. Luna owns semantic decisions, and a completed review
batch must not be reported as complete identity cleanup.