6.4 Decomposition and Composition

Decomposition and composition are inverse patterns. Decomposition breaks a complex object into simpler parts. Composition builds a complex object from simpler parts. Much of mathematics moves between these two directions.

Decomposition is useful because large objects are often hard to understand directly. A problem becomes easier when its pieces can be studied separately. Composition is useful because it explains how local or partial information forms a larger whole.

A simple example is integer factorization:

The number $60$ is decomposed into prime factors. The prime factors are simpler than the original number, and they determine it uniquely up to order.

This pattern appears throughout mathematics. A vector may be decomposed into basis components. A function may be decomposed into simpler functions. A graph may be decomposed into connected components. A proof may be decomposed into lemmas.

In linear algebra, decomposition is especially visible. If a vector space $V$ is written as a direct sum

then every vector $v \in V$ has a unique expression

with $u \in U$ and $w \in W$.

The decomposition separates the space into independent parts. Computations on $V$ can often be reduced to computations on $U$ and $W$.

Matrix diagonalization follows the same idea. When an operator has enough eigenvectors, the space decomposes into eigenspaces. The operator then acts by simple scaling on each component.

This turns a complicated transformation into a collection of one-dimensional transformations.

Composition works in the opposite direction. Given smaller objects and rules for combining them, one constructs a larger object. A product combines objects side by side. A quotient combines objects by identifying parts. A direct sum combines components while preserving their independence.

For example, if $V_1,\dots,V_n$ are vector spaces, their direct sum

contains tuples with one component in each space. The whole object is built from the parts.

In topology, decomposition may take the form of cutting a space into pieces. A surface can be studied by cutting it along curves. A space can be covered by open sets. The challenge is to understand not only the pieces but also how they are attached.

This attachment data matters. Two spaces may be built from similar pieces but glued differently, producing different global objects.

In graph theory, connected components provide a basic decomposition. A graph decomposes into maximal connected subgraphs. Once the components are known, many questions about reachability reduce to questions inside each component.

For trees, decomposition often occurs by removing an edge or a vertex. This breaks the tree into smaller trees, enabling recursive arguments.

In algebra, decomposition may reveal internal structure. Groups may be built from subgroups and quotients. Modules may decompose into direct sums. Rings may decompose through ideals, idempotents, or localization.

A decomposition is most useful when it satisfies two conditions: the pieces are simpler than the whole, and the method of recombination is controlled.

If the pieces are simple but the recombination is chaotic, little has been gained. If the recombination is simple but the pieces remain as hard as the whole, the decomposition has limited value.

Composition also appears in functions. A complex function may be written as

This says that $h$ acts in stages: first apply $f$, then apply $g$. Understanding the stages can make the whole function easier to analyze.

For example, the chain rule in calculus is a theorem about composition. It explains how rates of change combine:

The derivative of the composite is built from the derivatives of the parts.

Decomposition and composition also structure proofs. A long proof is often made readable by extracting lemmas. Each lemma handles a smaller claim. The final theorem composes these claims into the desired result.

This is not only a writing technique. It is a method of reasoning. A proof becomes manageable when its dependency structure is clear.

The same pattern appears in algorithms. A divide-and-conquer algorithm decomposes a problem into subproblems, solves them, and combines the results. Merge sort decomposes a list into smaller lists, sorts them, and merges them. The correctness of the whole algorithm depends on the correctness of each part and the recombination step.

Decomposition is not always unique. A number may have unique prime factorization, but a vector space may have many choices of basis. A function may have many possible decompositions. A space may be covered by open sets in many ways.

Non-uniqueness can be useful or dangerous. It is useful when different decompositions reveal different features. It is dangerous when conclusions depend on arbitrary choices.

This is why invariance matters. If a result is derived from a decomposition, one should ask whether it depends on the chosen decomposition or only on the original object.

For example, the coordinates of a vector depend on a basis. The dimension of the vector space does not. The diagonal form of a matrix may depend on choices, but the eigenvalues are invariant under change of basis.

A practical workflow is:

First, identify the object to decompose. Second, choose a notion of smaller part. Third, prove that the parts preserve enough information. Fourth, understand how the parts recombine. Finally, check which conclusions are independent of choices.

Decomposition reduces complexity. Composition restores structure. Together they explain how mathematics moves between parts and wholes.