01.02.10 · foundations / groups

Tensor product of modules (commutative case)

shipped3 tiersLean: none

Anchor (Master): Whitney 1938 Ann. Math. 39; Bourbaki Algebre Ch. II-III

Intuition [Beginner]

The tensor product takes two mathematical objects and combines them into one bigger object, the way multiplication takes two numbers and produces their product. The defining feature: if you double one input while keeping the other fixed, the output doubles. The same holds if you double the other input instead. The result is "separately linear" in each input — linear in the first slot when the second is held constant, and linear in the second slot when the first is held constant.

Think of multiplying polynomials. When you compute $(1 + 2 x) (3 + 4 x)$ , you pair every term of the first with every term of the second: $1 \times 3$ , $1 \times 4 x$ , $2 x \times 3$ , and $2 x \times 4 x$ . Collecting terms gives $3 + 10 x + 8 x^{2}$ . The tensor product performs exactly this kind of pairing: take every element of the first object, combine it with every element of the second, and then impose the linearity rules so that doubling one factor doubles the result.

Why does this concept exist? The tensor product converts operations that are linear in two inputs separately into ordinary single-input linear operations, which are far simpler to analyse.

Visual [Beginner]

A $2 \times 3$ grid whose rows correspond to basis elements of a 2-dimensional space and whose columns correspond to basis elements of a 3-dimensional space. Each cell in the grid represents one basis element of the tensor product, which has $2 \times 3 = 6$ dimensions in total.

The grid encodes the central idea: the tensor product of an $m$ -dimensional space with an $n$ -dimensional space has $m \times n$ dimensions, one for each pair.

Worked example [Beginner]

Consider the 2-dimensional vector space $R^{2}$ over the real numbers. The tensor product of $R^{2}$ with itself produces a 4-dimensional space.

Take $v = (1, 2)$ and $w = (3, 4)$ .

Step 1. Write each vector using the standard basis: $v = 1 \cdot e_{1} + 2 \cdot e_{2}$ and $w = 3 \cdot e_{1} + 4 \cdot e_{2}$ .

Step 2. Pair every basis element from $v$ with every basis element from $w$ , multiplying their coefficients: $1 \times 3 = 3$ paired with $(e_{1}, e_{1})$ , $1 \times 4 = 4$ paired with $(e_{1}, e_{2})$ , $2 \times 3 = 6$ paired with $(e_{2}, e_{1})$ , and $2 \times 4 = 8$ paired with $(e_{2}, e_{2})$ .

Step 3. The tensor product of $v$ and $w$ is the element with coefficient $3$ for the first basis pair, $4$ for the second, $6$ for the third, and $8$ for the fourth. The four basis pairs span a space of dimension $2 \times 2 = 4$ .

What this tells us: the dimension of the tensor product of an $m$ -dimensional space with an $n$ -dimensional space is $m \times n$ .

Check your understanding [Beginner]

Formal definition [Intermediate+]

Throughout, $R$ is a commutative ring with identity. An $R$ -module $M$ is an abelian group equipped with a scalar multiplication $R \times M \to M$ satisfying the same axioms as a vector space, except that the scalars come from $R$ rather than from a field ^{[Atiyah-Macdonald 1969]}. Every vector space over a field is a module; every abelian group is a $Z$ -module.

Definition (Bilinear map). Let $M$ , $N$ , $P$ be $R$ -modules. A function $f : M \times N \to P$ is $R$ -bilinear if it is $R$ -linear in each argument separately: $$ f(rm + m',, n) = rf(m,n) + f(m',n), \qquad f(m,, rn + n') = rf(m,n) + f(m,n') $$ for all $r \in R$ , $m, m^{'} \in M$ , $n, n^{'} \in N$ .

Definition (Tensor product). The tensor product $M \otimes_{R} N$ is an $R$ -module equipped with an $R$ -bilinear map $\otimes : M \times N \to M \otimes_{R} N$ , written $(m, n) \mapsto m \otimes n$ , satisfying the following universal property: for every $R$ -module $P$ and every $R$ -bilinear map $f : M \times N \to P$ , there exists a unique $R$ -linear map $\overset{ˉ}{f} : M \otimes_{R} N \to P$ such that $f = \overset{ˉ}{f} \circ \otimes$ ^{[Dummit-Foote 2004]}.

Equivalently, $M \otimes_{R} N$ is the quotient of the free $R$ -module on the set $M \times N$ by the submodule generated by all elements of the form $$ (m+m',n)-(m,n)-(m',n), \quad (m,n+n')-(m,n)-(m,n'), \quad (rm,n)-r(m,n), \quad (m,rn)-r(m,n) $$ for $m, m^{'} \in M$ , $n, n^{'} \in N$ , $r \in R$ . The image of $(m, n)$ in this quotient is $m \otimes n$ .

Counterexamples to common slips [Intermediate+]

The tensor product of two nonzero modules can be zero. For instance, $Z /2 Z \otimes_{Z} Z /3 Z = 0$ because $1 \otimes 1 = 3 (1 \otimes 1) - 2 (1 \otimes 1) = 1 \otimes 3 - 2 \otimes 1 = 0 - 0 = 0$ , using $r (m \otimes n) = (r m) \otimes n = m \otimes (r n)$ and the fact that $3 = 0$ in $Z /3 Z$ , $2 = 0$ in $Z /2 Z$ .
Not every element of $M \otimes_{R} N$ is a pure tensor $m \otimes n$ . General elements are finite sums $\sum_{i} m_{i} \otimes n_{i}$ . For example, in $R^{2} \otimes_{R} R^{2}$ , the element $e_{1} \otimes e_{1} + e_{2} \otimes e_{2}$ cannot be written as a single pure tensor.
Injective maps need not remain injective after tensoring. The map $Z \times 2 Z$ is injective, but after tensoring with $Z /2 Z$ the induced map $Z /2 Z \to Z /2 Z$ sends $\overset{ˉ}{1} \mapsto \overline{2} = \overset{ˉ}{0}$ , which is the zero map.

Key theorem with proof [Intermediate+]

Theorem (Universal property — existence and uniqueness). Let $R$ be a commutative ring and let $M$ , $N$ be $R$ -modules. The tensor product $M \otimes_{R} N$ exists and is unique up to unique isomorphism.

(i) Existence. The quotient of the free $R$ -module on $M \times N$ by the bilinearity relations is a tensor product of $M$ and $N$ .

(ii) Uniqueness. If $T_{1}$ and $T_{2}$ are both tensor products of $M$ and $N$ (with bilinear maps $\otimes_{1}$ , $\otimes_{2}$ ), there is a unique isomorphism $φ : T_{1} \to T_{2}$ satisfying $φ (m \otimes_{1} n) = m \otimes_{2} n$ for all $m \in M$ , $n \in N$ .

Proof. Existence. Let $F$ be the free $R$ -module on the set $M \times N$ . Let $K$ be the submodule of $F$ generated by all elements $$ (m+m',n)-(m,n)-(m',n), \quad (m,n+n')-(m,n)-(m,n'), \quad (rm,n)-r(m,n), \quad (m,rn)-r(m,n). $$ Set $T = F / K$ and write $m \otimes n$ for the image of $(m, n)$ in $T$ . By construction, the relations enforcing $R$ -bilinearity hold in $T$ .

Given an $R$ -bilinear map $f : M \times N \to P$ , define $\overset{ˉ}{f} : T \to P$ by $\overset{ˉ}{f} ((m, n) + K) = f (m, n)$ . This is well-defined because each generator of $K$ maps to zero: for instance, $$ \bar{f}\bigl((m+m',n)-(m,n)-(m',n)\bigr) = f(m+m',n) - f(m,n) - f(m',n) = 0 $$ by bilinearity of $f$ . The map $\overset{ˉ}{f}$ is $R$ -linear on $F$ and descends to $T = F / K$ . Uniqueness of $\overset{ˉ}{f}$ follows because the elements $m \otimes n$ generate $T$ as an $R$ -module, so any $R$ -linear map from $T$ is determined by its values on these generators.

Uniqueness. Suppose $T_{1}$ and $T_{2}$ are both tensor products with bilinear maps $\otimes_{1}$ and $\otimes_{2}$ . By the universal property of $T_{1}$ , there is a unique $R$ -linear map $φ : T_{1} \to T_{2}$ with $φ (m \otimes_{1} n) = m \otimes_{2} n$ . By the universal property of $T_{2}$ , there is a unique $R$ -linear map $ψ : T_{2} \to T_{1}$ with $ψ (m \otimes_{2} n) = m \otimes_{1} n$ . Then $ψ \circ φ$ and $φ \circ ψ$ are $R$ -linear endomorphisms that satisfy the same identity as the respective identity maps. By uniqueness in the universal property, $ψ \circ φ = id_{T_{1}}$ and $φ \circ ψ = id_{T_{2}}$ . $□$

Bridge. The universal property builds toward 01.01.15 (bilinear and quadratic forms), where every bilinear form on $V \times W$ corresponds to a linear functional on $V \otimes W$ , and this is exactly the mechanism that identifies $Bil (V, W; R)$ with $Hom (V \otimes W, R)$ . The foundational reason is that the tensor product is the universal receptacle for bilinear maps, and the central insight is that bilinearity on a product decomposes into linearity on the tensor product. The bridge is between the bilinear world of two-variable maps and the linear world of single-variable maps, and the pattern appears again in 01.01.05 (rank-nullity) through the tensor-hom adjunction.

Exercises [Intermediate+]

Exercise 3 (medium, symbolic).

Prove that the map $φ : M \to M \otimes_{R} R$ defined by $φ (m) = m \otimes 1$ is an isomorphism of $R$ -modules, with inverse $ψ : M \otimes_{R} R \to M$ given by $ψ (m \otimes r) = r m$ .

Hint

Show $ψ$ is well-defined by verifying it descends from a bilinear map $M \times R \to M$ , then show $ψ \circ φ = id_{M}$ and $φ \circ ψ = id_{M \otimes R}$ .

Answer

The map $M \times R \to M$ sending $(m, r) \mapsto r m$ is $R$ -bilinear, so it induces a well-defined $R$ -linear map $ψ : M \otimes_{R} R \to M$ with $ψ (m \otimes r) = r m$ .

For $m \in M$ : $ψ (φ (m)) = ψ (m \otimes 1) = 1 \cdot m = m$ , so $ψ \circ φ = id_{M}$ .

For $\sum_{i} m_{i} \otimes r_{i} \in M \otimes_{R} R$ : each term satisfies $m_{i} \otimes r_{i} = r_{i} m_{i} \otimes 1$ , so $\sum_{i} m_{i} \otimes r_{i} = \sum_{i} r_{i} m_{i} \otimes 1 = (\sum_{i} r_{i} m_{i}) \otimes 1$ . Then $φ (ψ (\sum_{i} m_{i} \otimes r_{i})) = φ (\sum_{i} r_{i} m_{i}) = (\sum_{i} r_{i} m_{i}) \otimes 1 = \sum_{i} m_{i} \otimes r_{i}$ . Hence $φ \circ ψ = id$ . Therefore $φ$ is an isomorphism.

Exercise 6 (medium, symbolic).

Prove that $M \otimes_{R} (N_{1} \oplus N_{2}) ≅ (M \otimes_{R} N_{1}) \oplus (M \otimes_{R} N_{2})$ as $R$ -modules.

Hint

Define maps in both directions using the universal property. Use the inclusion and projection maps for the direct sum.

Answer

Define $φ : M \otimes_{R} (N_{1} \oplus N_{2}) \to (M \otimes_{R} N_{1}) \oplus (M \otimes_{R} N_{2})$ by $φ (m \otimes (n_{1}, n_{2})) = (m \otimes n_{1}, m \otimes n_{2})$ , induced by the bilinear map $M \times (N_{1} \oplus N_{2}) \to (M \otimes N_{1}) \oplus (M \otimes N_{2})$ sending $(m, (n_{1}, n_{2})) \mapsto (m \otimes n_{1}, m \otimes n_{2})$ .

Define $ψ$ in the reverse direction by $ψ ((m \otimes n_{1}, m^{'} \otimes n_{2})) = m \otimes (n_{1}, 0) + m^{'} \otimes (0, n_{2})$ . On summands: $ψ (m \otimes n_{1}, 0) = m \otimes (n_{1}, 0)$ and $ψ (0, m \otimes n_{2}) = m \otimes (0, n_{2})$ .

Check: $ψ \circ φ$ sends $m \otimes (n_{1}, n_{2})$ to $m \otimes (n_{1}, 0) + m \otimes (0, n_{2}) = m \otimes (n_{1}, n_{2})$ . The reverse composite sends each summand back to itself. Both composites are identities on generators, so $φ$ is an isomorphism.

Exercise 7 (hard, symbolic).

Prove the tensor-hom adjunction: for $R$ -modules $M$ , $N$ , $P$ , there is a canonical isomorphism of $R$ -modules $$ \operatorname{Hom}_R(M \otimes_R N,, P) \cong \operatorname{Hom}_R(M,, \operatorname{Hom}_R(N, P)). $$

Hint

Given $g : M \otimes_{R} N \to P$ , define $Φ (g) : M \to Hom (N, P)$ by $Φ (g) (m) (n) = g (m \otimes n)$ . Verify that both $Φ$ and its inverse are well-defined and $R$ -linear.

Answer

Define $Φ : Hom_{R} (M \otimes_{R} N, P) \to Hom_{R} (M, Hom_{R} (N, P))$ as follows. Given $g \in Hom_{R} (M \otimes_{R} N, P)$ , set $Φ (g) (m) (n) = g (m \otimes n)$ .

For fixed $m$ , the map $n \mapsto g (m \otimes n)$ is $R$ -linear because $g$ is $R$ -linear and $n \mapsto m \otimes n$ is $R$ -linear. For fixed $n$ , the map $m \mapsto g (m \otimes n)$ is $R$ -linear. So $Φ (g) \in Hom_{R} (M, Hom_{R} (N, P))$ .

Define the inverse $Ψ : Hom_{R} (M, Hom_{R} (N, P)) \to Hom_{R} (M \otimes_{R} N, P)$ by: given $f : M \to Hom_{R} (N, P)$ , the map $M \times N \to P$ sending $(m, n) \mapsto f (m) (n)$ is $R$ -bilinear, so by the universal property it induces $Ψ (f) : M \otimes_{R} N \to P$ with $Ψ (f) (m \otimes n) = f (m) (n)$ .

Check: $Ψ (Φ (g)) (m \otimes n) = Φ (g) (m) (n) = g (m \otimes n)$ , so $Ψ \circ Φ = id$ . Similarly, $Φ (Ψ (f)) (m) (n) = Ψ (f) (m \otimes n) = f (m) (n)$ , so $Φ \circ Ψ = id$ . Both maps are $R$ -linear. Therefore $Φ$ is an isomorphism.

Exercise 8 (hard, symbolic).

Let $F$ be a free $R$ -module with basis ${f_{i}}_{i \in I}$ . Prove that the functor $F \otimes_{R} -$ is exact: if $0 \to A α B β C \to 0$ is a short exact sequence of $R$ -modules, then $0 \to F \otimes_{R} A id \otimes α F \otimes_{R} B id \otimes β F \otimes_{R} C \to 0$ is also exact.

Hint

Use the isomorphism $F \otimes_{R} M ≅ ⨁_{i \in I} M$ for any $R$ -module $M$ , then verify exactness componentwise.

Answer

Since $F$ is free with basis ${f_{i}}$ , the map $F \otimes_{R} M \to ⨁_{i \in I} M$ sending $f_{i} \otimes m \mapsto (0, \dots, m, \dots)$ (with $m$ in the $i$ -th position) is an isomorphism, natural in $M$ . Under this identification, the map $id \otimes α$ becomes $⨁_{i \in I} α : ⨁_{I} A \to ⨁_{I} B$ , and $id \otimes β$ becomes $⨁_{i \in I} β : ⨁_{I} B \to ⨁_{I} C$ .

Since direct sums preserve exactness, the sequence $0 \to ⨁_{I} A ⨁ α ⨁_{I} B ⨁ β ⨁_{I} C \to 0$ is exact. The naturality of the isomorphism $F \otimes_{R} M ≅ ⨁_{I} M$ gives a commutative diagram, so the original tensor-product sequence is exact as well.

Advanced results [Master]

Theorem 1 (Associativity). For $R$ -modules $M$ , $N$ , $P$ , there is a canonical isomorphism $(M \otimes_{R} N) \otimes_{R} P ≅ M \otimes_{R} (N \otimes_{R} P)$ sending $(m \otimes n) \otimes p \mapsto m \otimes (n \otimes p)$ .

Both sides satisfy the universal property for $R$ -trilinear maps $M \times N \times P \to Q$ . The isomorphism identifies the two universal receptacles ^{[Eisenbud 1995]}.

Theorem 2 (Commutativity). For $R$ -modules $M$ , $N$ , there is a canonical isomorphism $M \otimes_{R} N ≅ N \otimes_{R} M$ sending $m \otimes n \mapsto n \otimes m$ .

This is an isomorphism, not an equality: $m \otimes n$ and $n \otimes m$ live in different modules. Commutativity of $R$ is essential; the map $(m, n) \mapsto n \otimes m$ is bilinear because scalars commute in $R$ .

Theorem 3 (Distributivity over direct sums). $M \otimes_{R} (⨁_{i \in I} N_{i}) ≅ ⨁_{i \in I} (M \otimes_{R} N_{i})$ .

Tensor products commute with arbitrary direct sums in the second argument. The isomorphism sends $m \otimes (n_{i})_{i \in I}$ to $(m \otimes n_{i})_{i \in I}$ .

Theorem 4 (Tensor product of algebras). If $A$ and $B$ are commutative $R$ -algebras, then $A \otimes_{R} B$ is a commutative $R$ -algebra with multiplication $(a \otimes b) (a^{'} \otimes b^{'}) = a a^{'} \otimes b b^{'}$ .

This equips the tensor product of algebras with a ring structure. The polynomial ring $R [x] \otimes_{R} R [y] ≅ R [x, y]$ is the canonical example: the variables from each factor become independent ^{[Atiyah-Macdonald 1969]}.

Theorem 5 (Extension of scalars). Let $f : R \to S$ be a homomorphism of commutative rings. For any $R$ -module $M$ , the tensor product $S \otimes_{R} M$ is an $S$ -module with $s \cdot (s^{'} \otimes m) = (s s^{'}) \otimes m$ . The functor $S \otimes_{R} -$ is left adjoint to the restriction functor from $S$ -modules to $R$ -modules.

Extension of scalars converts an $R$ -module into an $S$ -module. The canonical example: if $V$ is an $R$ -vector space, then $C \otimes_{R} V$ is a complex vector space (complexification).

Theorem 6 (Tensor-hom adjunction). For $R$ -modules $M$ , $N$ , $P$ , there is a canonical isomorphism of $R$ -modules: $$ \operatorname{Hom}_R(M \otimes_R N,, P) \cong \operatorname{Hom}_R(M,, \operatorname{Hom}_R(N, P)). $$ The isomorphism sends $g \mapsto [m \mapsto [n \mapsto g (m \otimes n)]]$ .

This is the defining adjunction of the tensor product: the functor $- \otimes_{R} N$ is left adjoint to $Hom_{R} (N, -)$ . As with any adjunction, the tensor product preserves colimits (in particular, direct sums and cokernels) ^{[Eisenbud 1995]}.

Theorem 7 (Flatness criterion). An $R$ -module $M$ is flat if the functor $M \otimes_{R} -$ is exact. Equivalently, $M$ is flat if for every injective $R$ -linear map $N^{'} \to N$ , the induced map $M \otimes_{R} N^{'} \to M \otimes_{R} N$ is injective.

Free modules are flat. Projective modules are flat. Finitely generated flat modules over a Noetherian local ring are free. The failure of flatness is measured by the Tor functor: $M$ is flat if and only if $Tor_{n}^{R} (M, -) = 0$ for all $n \geq 1$ .

Synthesis. The tensor-hom adjunction is the foundational reason that the tensor product governs the passage between multilinear and linear algebra across all of commutative algebra. The central insight is that the six structural properties — associativity, commutativity, distributivity, the unit $R$ , the tensor-hom adjunction, and extension of scalars — equip the category of $R$ -modules with the structure of a symmetric monoidal closed category. Putting these together with the flatness criterion, which detects when $M \otimes_{R} -$ preserves exact sequences, the bridge is between the homological algebra of Tor functors and the geometric problem of how modules change base ring.

This is exactly the structure that generalises from $R$ -modules to quasi-coherent sheaves on a scheme, where the tensor product computes fibre products of affine schemes. The pattern recurs in 01.01.15 (bilinear and quadratic forms), where the tensor product encodes the universal bilinear pairing, and appears again in 01.01.03 (vector space), where extension of scalars transforms a real vector space into a complex one.

Full proof set [Master]

Proposition 1 (Tensor-hom adjunction). For $R$ -modules $M$ , $N$ , $P$ , the map $Φ : Hom_{R} (M \otimes_{R} N, P) \to Hom_{R} (M, Hom_{R} (N, P))$ defined by $Φ (g) (m) (n) = g (m \otimes n)$ is an isomorphism of $R$ -modules.

Proof. For fixed $g \in Hom_{R} (M \otimes_{R} N, P)$ , the map $Φ (g) : M \to Hom_{R} (N, P)$ is $R$ -linear: $Φ (g) (r m + m^{'}) (n) = g ((r m + m^{'}) \otimes n) = g (r m \otimes n + m^{'} \otimes n) = r g (m \otimes n) + g (m^{'} \otimes n) = r Φ (g) (m) (n) + Φ (g) (m^{'}) (n)$ .

The inverse $Ψ$ sends $f \in Hom_{R} (M, Hom_{R} (N, P))$ to the map $Ψ (f) : M \otimes_{R} N \to P$ induced by the bilinear map $(m, n) \mapsto f (m) (n)$ . Bilinearity of $(m, n) \mapsto f (m) (n)$ follows from $R$ -linearity of $f (m)$ in $n$ and $R$ -linearity of $f$ in $m$ .

One verifies $Ψ (Φ (g)) (m \otimes n) = g (m \otimes n)$ and $Φ (Ψ (f)) (m) (n) = f (m) (n)$ on generators, so both composites are identities. $□$

Proposition 2 (Extension of scalars is left adjoint to restriction). Let $f : R \to S$ be a ring homomorphism. The functor $S \otimes_{R} -$ from $R$ -modules to $S$ -modules is left adjoint to the restriction functor $U$ from $S$ -modules to $R$ -modules.

Proof. For an $R$ -module $M$ and an $S$ -module $N$ , define $$ \Phi\colon \operatorname{Hom}_S(S \otimes_R M, N) \to \operatorname{Hom}_R(M, U(N)), \quad \Phi(g)(m) = g(1 \otimes m). $$ This is well-defined: $g (1 \otimes m) \in N$ and the map $m \mapsto g (1 \otimes m)$ is $R$ -linear because $g$ is $S$ -linear and $1 \otimes (r m) = f (r) \otimes m$ , so $g (1 \otimes r m) = g (f (r) \otimes m) = f (r) g (1 \otimes m) = r \cdot g (1 \otimes m)$ .

The inverse sends $h \in Hom_{R} (M, U (N))$ to the unique $S$ -linear map $Ψ (h) : S \otimes_{R} M \to N$ with $Ψ (h) (s \otimes m) = s \cdot h (m)$ . This is induced by the $R$ -bilinear map $S \times M \to N$ sending $(s, m) \mapsto s \cdot h (m)$ .

Check: $Ψ (Φ (g)) (s \otimes m) = s \cdot Φ (g) (m) = s \cdot g (1 \otimes m) = g (s \otimes m)$ , and $Φ (Ψ (h)) (m) = Ψ (h) (1 \otimes m) = 1 \cdot h (m) = h (m)$ . Both composites are identities. $□$

Proposition 3 (Associativity). The map $α : (M \otimes_{R} N) \otimes_{R} P \to M \otimes_{R} (N \otimes_{R} P)$ defined by $(m \otimes n) \otimes p \mapsto m \otimes (n \otimes p)$ is a canonical isomorphism.

Proof. For fixed $p \in P$ , the map $M \times N \to M \otimes_{R} (N \otimes_{R} P)$ sending $(m, n) \mapsto m \otimes (n \otimes p)$ is $R$ -bilinear, so it induces $α_{p} : M \otimes_{R} N \to M \otimes_{R} (N \otimes_{R} P)$ with $α_{p} (m \otimes n) = m \otimes (n \otimes p)$ .

The map $(M \otimes_{R} N) \times P \to M \otimes_{R} (N \otimes_{R} P)$ sending $(t, p) \mapsto α_{p} (t)$ is $R$ -bilinear: for $r \in R$ , $α_{r p} (m \otimes n) = m \otimes (n \otimes r p) = m \otimes r (n \otimes p) = r (m \otimes (n \otimes p)) = r α_{p} (m \otimes n)$ . So it induces $α : (M \otimes_{R} N) \otimes_{R} P \to M \otimes_{R} (N \otimes_{R} P)$ .

The inverse $β : M \otimes_{R} (N \otimes_{R} P) \to (M \otimes_{R} N) \otimes_{R} P$ is constructed symmetrically. One verifies $β (α ((m \otimes n) \otimes p)) = (m \otimes n) \otimes p$ and $α (β (m \otimes (n \otimes p))) = m \otimes (n \otimes p)$ on generators. $□$

Connections [Master]

Bilinear and quadratic forms 01.01.15. The tensor product provides the universal receptacle for bilinear maps: every bilinear form on $V \times W$ factors through $V \otimes W$ as a linear functional. This is exactly the bridge between bilinear algebra and the linear-algebraic formalism of quadratic forms, and the identification $Bil (V, W; R) ≅ (V \otimes W)^{*}$ is the starting point for the classification of bilinear forms.
Linear transformation, rank-nullity 01.01.05. The tensor-hom adjunction $Hom (M \otimes N, P) ≅ Hom (M, Hom (N, P))$ is the natural generalisation of the rank-nullity framework: the functor $- \otimes_{R} N$ is left adjoint to $Hom_{R} (N, -)$ , and the exactness properties of this adjunction control how rank behaves under extension of scalars.
Vector space 01.01.03. Modules generalise vector spaces by replacing the field of scalars with a ring. The tensor product of vector spaces over a field is the simplest case: dimensions multiply, every element decomposes as a sum of pure tensors, and flatness is automatic because vector spaces are free. The foundational reason that the module-theoretic tensor product reduces to the vector-space tensor product when $R$ is a field is that vector spaces are free $R$ -modules.

Historical & philosophical context [Master]

Whitney 1938 introduced the tensor product of abelian groups in the context of cohomology ^{[Whitney 1938]}, defining $\otimes$ as a functor on the category of abelian groups that converts bilinear pairings into homomorphisms. The modern algebraic formulation crystallised with Bourbaki's Algebre in the 1940s, which systematised the universal property and the quotient construction for modules over a commutative ring ^{[Bourbaki 1940s]}.

The tensor-hom adjunction was recognised as the structural centrepiece by Cartan and Eilenberg 1956 in Homological Algebra, and extension of scalars became the primary tool in algebraic geometry through Grothendieck's 1960s reformulation of scheme theory in terms of quasi-coherent sheaves and their tensor products. Flatness, introduced by Serre 1956 in the context of algebraic geometry, measures when the tensor product preserves exact sequences, and its characterisation via the Tor functor closes the loop between the commutative-algebraic and homological-algebraic perspectives.

Bibliography [Master]

@article{Whitney1938,
  author = {Whitney, Hassler},
  title = {Tensor products of abelian groups},
  journal = {Annals of Mathematics},
  volume = {39},
  year = {1938},
  pages = {657--665},
}

@book{Bourbaki1989,
  author = {Bourbaki, Nicolas},
  title = {Algebra I: Chapters 1--3},
  publisher = {Springer},
  year = {1989},
}

@book{AtiyahMacdonald1969,
  author = {Atiyah, Michael F. and Macdonald, Ian G.},
  title = {Introduction to Commutative Algebra},
  publisher = {Addison-Wesley},
  year = {1969},
}

@book{DummitFoote2004,
  author = {Dummit, David S. and Foote, Richard M.},
  title = {Abstract Algebra},
  edition = {3rd},
  publisher = {Wiley},
  year = {2004},
}

@book{Eisenbud1995,
  author = {Eisenbud, David},
  title = {Commutative Algebra with a View Toward Algebraic Geometry},
  publisher = {Springer},
  year = {1995},
  series = {Graduate Texts in Mathematics 150},
}