GOLD STANDARD · TOPOLOGY · HOW X WORKS
Understand how a map can look null on every finite subcomplex while remaining globally nontrivial
A phantom map exposes a sharp limit of finite-stage testing. For a CW complex X, a map f from X to Y is phantom when its restriction to every finite subcomplex of X is null-homotopic, even though f itself may be non-null. The phenomenon is not a licence to distrust every finite calculation. It is a reason to state the source, finiteness hypotheses, tower model and inverse-limit defect that connect local tests to the infinite complex.
This guide connects the Gold Standard of Topology, Gold Standard of Education, and How X Works hub.
Chapter contents and reading routes
Choose the route closest to your present task. Every teaching chapter remains expanded, so the article also reads naturally from start to finish.
Define the phenomenon
See the finite-stage gap
Detect and control the defect
Transfer and qualify
Use FAQs, sources and connected routes
CHAPTER 1 OF 12 · DEFINE THE PHANTOM
1. Null on every finite subcomplex is not automatically null on X
Let X be a CW complex and f:X→Y a map. The classical phantom condition asks that the restriction f|K be null-homotopic for every finite subcomplex K of X. If X itself is finite, taking K=X shows that every phantom map from X is null.
The interesting case therefore needs an infinite source. The global map may encode a compatibility defect that no one finite subcomplex can see.
CHAPTER 2 OF 12 · DEFINE THE PHANTOM
2. Keep three claims separate
Stagewise nullity, trivial action on familiar invariants and global null-homotopy are not identical statements. A phantom map is defined through finite subcomplex restrictions. Consequences for homology or homotopy depend on the spaces and theorem being used.
| Claim | What is tested | What it does not alone prove |
|---|---|---|
| Finite restriction null | f|Xn is null for each finite stage | One compatible global null-homotopy |
| Induced invariant vanishes | A chosen functor sees zero | Global map is null in all settings |
| Global map null | One homotopy contracts f on X | Automatically follows from stages only with extra hypotheses |
| Phantom class | All finite restrictions vanish | Its classification without a tower theorem |
| Finite source | X equals a finite subcomplex | No nontrivial classical phantom from X |
| Infinite source | X is a union of finite stages | Compatibility still needs analysis |
Specify based or unbased homotopy and the relevant category. Low-dimensional or non-simple cases can require additional care.
CHAPTER 3 OF 12 · BUILD THE TOWER
3. Filter the source by finite subcomplexes
Choose an increasing sequence X1⊂X2⊂··· whose union is X when such a countable filtration is available. Each restriction gives a class in [Xn,Y], and restriction maps form an inverse system.
A phantom class maps to the zero element at every displayed stage. The question is whether the global mapping set injects into the inverse limit of those stage sets. A derived-limit term can obstruct that injection in suitable settings.
CHAPTER 4 OF 12 · SEE THE GAP
4. Independent null-homotopies need not be coherent
For each n, choose a null-homotopy of f|Xn. The choice on Xn+1 may not restrict to the choice on Xn. Replacing the choices stage by stage can create an endless compatibility problem.
The defect is not visible from the statement “every stage is null”. It lives in how the spaces of choices and their transition maps fit together.
CHAPTER 5 OF 12 · WORK THE LOGIC
5. Work a finite-stage audit without overclaiming a concrete example
Suppose X is filtered by finite subcomplexes Xn and a map f restricts to the null class on the first ten stages. A computation then proves the same for every n. This establishes stagewise nullity, but a global conclusion still needs a comparison theorem.
| Audit item | Result | Logical status |
|---|---|---|
| Finite-stage classes | [f|Xn]=0 for every n | Phantom condition established |
| Chosen null-homotopies | Exist separately at each stage | Coherence not established |
| Inverse-limit image | Zero compatible family | Stage invariant sees no difference |
| Derived-limit contribution | Not computed | Possible global ambiguity remains |
| Finiteness theorem | Not yet invoked | Cannot conclude f=0 globally |
| Safe conclusion | f is phantom | Nontriviality or triviality needs further evidence |
A genuine example must additionally prove that the global class is nonzero. Conversely, a vanishing theorem must state hypotheses that force every phantom class in the setting to be null.
CHAPTER 6 OF 12 · DETECT THE DEFECT
6. Use mapping towers and derived limits carefully
Under standard hypotheses, mapping information from an increasing CW filtration can be organised into a tower, and a Milnor-type exact sequence may place a lim¹ term beside the inverse limit of finite-stage homotopy classes. Phantom phenomena are closely related to the kernel detected by that derived term.
Do not transfer a group-valued exact sequence blindly to unbased mapping sets or low degrees. Name the theorem, basepoints, connectedness and finiteness conditions.
CHAPTER 7 OF 12 · CONTROL THE DEFECT
7. Look for finiteness and stabilisation hypotheses
Mittag–Leffler behaviour, finite-type assumptions, nilpotence, target restrictions or other compactness conditions can make a derived-limit defect vanish or make phantom classes classifiable. The exact condition depends on the category and mapping problem.
Prove the condition for the actual tower. Stabilising group ranks do not necessarily mean stabilising images.
CHAPTER 8 OF 12 · TRANSFER THE METHOD
8. Changed-context transfer: shapes, pro-objects and completions
Shape theory, completion towers and pro-spaces also compare infinite objects through finite or stagewise data. The same question appears: does the chosen tower detect the global morphism, and what defect remains?
Transfer the finite-versus-global checklist, not the assertion that every invisible morphism is a phantom map in the classical sense.
CHAPTER 9 OF 12 · QUALIFY THE CLAIM
9. State exactly what finite tests cover
List the finite subcomplexes or compact probes, the invariant or mapping set evaluated, and the comparison map to the global object. If only a bounded range of skeleta was checked, say so; that is not the phantom condition.
When computations use a replacement tower, also prove that the replacement preserves the mapping and derived-limit information needed for the claim.
CHAPTER 10 OF 12 · QUALIFY THE CLAIM
10. Limits and common misconceptions
A map that is zero on homology is not automatically phantom. A phantom map is not automatically nontrivial; the null map is phantom. Checking many finite stages is not the same as checking every finite subcomplex.
Do not call the phenomenon a measurement error or a paradox. It is a precise failure of a chosen finite-stage detection map to be injective under particular infinite constructions.
CHAPTER 11 OF 12 · CHECK AND CONTINUE
11. Questions readers often ask
Can a finite CW complex support a nontrivial phantom map as the source?
Not under the classical definition above. The whole source is a finite subcomplex, so the phantom condition forces the map itself to be null-homotopic.
Does lim¹ always classify phantom maps?
No universal statement should be made without hypotheses. Milnor-type sequences and phantom classifications depend on the mapping tower, basepoints, finiteness and algebraic structure.
CHAPTER 12 OF 12 · CHECK AND CONTINUE
12. Evidence, connected routes and the next check
Schwass on phantom maps and finiteness conditions gives a primary treatment of phantom maps and the role of source and target finiteness. Ghienne on phantom maps and natural filtrations on lim¹ sets studies the algebraic connection between phantom phenomena and derived inverse-limit data. The Stacks Project derived-limit exact sequence displays how an R¹lim term can sit in the kernel of comparison with stagewise Hom data.
Continue through tower-model replacement and derived-limit invariance, obstruction-tower convergence and infinite-stage realisation, Whitehead towers, and Moore–Postnikov factorisations. The next check is to identify the exact comparison map from global mapping classes to the inverse limit of finite-stage classes.