Research Directions

A useful first research calculation answers a small question precisely. You might recover an observable from two descriptions, find where an approximation fails, or determine which geometric data a supersymmetric quantity can depend on. The articles in this volume begin with an established result that you can reproduce, then develop one controlled variation. Their later reading suggestions are starting points for further study, rather than claims that a proposed variation is new.

Start with a result you already understand

Begin with From reproduction to a research question. You need one completed capstone from a theory or observable route, not every volume on this site. Its worked period integral shows why agreement at a few numerical points is weaker than a checked expansion, and why convergence of a function need not justify differentiating its limit.

Choose the physical quantity before choosing a computational method. Write down its theory, background, normalization and regime. Then reproduce a coefficient or limit whose value is known. Only after that comparison should you change a mass, radius, boundary condition, gauge group, or approximation order.

The sidebar organizes the available articles by topic. Each article gives its own preparation and a dated source discussion. Use the learning paths when an entry requires a lesson you have not yet studied; the four-dimensional routes remain independent of the three-dimensional sphere route.

Match a project to your preparation

Your starting point	A first project	What you will check
Rigid supersymmetry and geometric background data	Which geometric data does the partition function see?	Follow a torus deformation while identifying the complex data held fixed.
Three-dimensional localization and charged matter	The three-dimensional A-twist	Compare a flux-dependent mode count with a regulated determinant before extending to Seifert partition functions and fibering operators.
A duality and its background-source conventions	Reproduce a duality with its contact terms	Retain the complex phase and source terms when comparing two descriptions.
Four-dimensional N=4 local operators	Operator mixing, integrability, and finite N	Compare explicit operator mixing with a spin-chain calculation, keeping its planar assumptions visible.
Protected Wilson loops	N=4 loop and defect responses or ABJM defect observables	Extract a prescribed response while retaining the operator normalization, regulator and framing appropriate to each calculation.
CFT operator products and shortening	Enter the superconformal bootstrap	Recover a protected central-charge constraint with its higher-spin assumption stated explicitly.
ABJM localization and source responses	Deformed spheres and CFT data	Recover a fourth mass derivative and distinguish a fixed-source derivative from a tuned complex path.
The ABJM Airy approximation	Spectral methods and nonperturbative corrections	Resolve a level-four correction while separating omitted physical sectors from numerical errors.
Seiberg–Witten periods and stable BPS particles	BPS spectra and wall crossing	Compare a central-charge alignment with a protected-index identity in a specified chamber.

The choices in the table are entrances, not an order that everyone must follow. A partition-function project may lead to global gauge groups, boundary gluing, or orbifold backgrounds. Each changes a different piece of the physical problem; none can be reduced to changing a parameter in a formula without checking the new assumptions.

Two spectral continuations make that lesson particularly concrete. Instantons and quantum periods matches a gauge-theory coefficient to a differential equation with a specified return phase. Compactified Coulomb branches sums a light particle’s finite-radius corrections while keeping the gauge lattice, angular periods and omitted heavy states distinct. In both cases, reproducing one local result does not establish a complete spectrum or a global metric.

Make the comparison capable of failing

A second calculation is useful when it could reveal a different error. Examples include a direct integral versus a differential equation, a local field equation versus a global patching condition, or a finite matrix result versus an expansion with a controlled remainder. Raising the precision of the same expression checks arithmetic; it does not by itself check the expression’s physical normalization.

For your first report, retain three pieces of evidence:

1. A recovered result. State the exact source version, translated conventions and value or identity that you reproduced.
2. One changed input. Explain what remains fixed, why the changed problem is well defined, and which prediction you are testing.
3. An error test. Separate omitted physical effects from truncation, discretization and rounding. Include a limit or deliberate sign error that the calculation can detect.

Suppose a discrepancy shrinks when you increase a cutoff but stops changing when you increase arithmetic precision. That behavior points toward a truncation error. If it persists under both checks, compare charge units, source derivatives, phases and boundary conditions before interpreting it as physics. The individual articles provide worked versions of these tests and complete exercise solutions.

Your first checkpoint

Choose one article and write a single sentence naming its quantity, model and regime. Recover its benchmark and one independent check. Then describe a variation whose result could distinguish two plausible explanations of a discrepancy.

You are ready to extend the calculation when another reader can recover that benchmark from your stated inputs and understand what a failed comparison would mean. Before calling the extension an open problem or a new result, inspect the relevant primary literature again. The dated discussions here support a specific reproduction; they do not certify the present frontier of an entire subject.