The Local–Global Principle Using the Hasse–Minkowski Theorem

0
33

Abstract

The Hasse–Minkowski Theorem is a central result in number theory that connects local and global perspectives on quadratic equations. It states that a quadratic equation with rational coefficients has a non-trivial rational solution if and only if it has a solution over the real numbers and over every p-adic field. This idea shows how studying equations locally, one prime at a time, can determine whether they have a global solution. This paper offers an expository overview of the theorem and its mathematical foundation. It begins with quadratic forms and their properties, such as discriminants and isotropy, then introduces local fields and p-adic numbers, which describe completions of \mathbb{Q}. The discussion highlights the Hilbert symbol and the Hilbert Reciprocity Law, which connect local conditions to produce the global result established by the theorem. Several examples and applications are explored, including the representation of integers as sums of squares and the classification of rational solutions. The paper concludes by showing how the local–global principle underlying this theorem continues to shape modern number theory, linking classical results with ongoing research in arithmetic geometry and algebraic structures.

Keywords: Hasse–Minkowski theorem, quadratic forms, p-adic numbers, Hilbert symbol, local–global principle, number theory

Introduction

The Hasse–Minkowski Theorem is a cornerstone in the theory of quadratic forms over number fields, offering a powerful local-global principle. It addresses the question of whether a quadratic equation, such as ax^2 + by^2 + cz^2 = 0, has a non-trivial solution (i.e. not all variables are zero) in rational numbers by checking its solvability in the real numbers and the p-adic numbers, which are completions of the rationals with respect to prime-based metrics1. Similarly, it determines when two quadratic forms are equivalent, meaning one can be transformed into the other via a linear change of variables. Historically, Hermann Minkowski proved the theorem for rational numbers, and Helmut Hasse generalized it to number fields. Its significance lies in its use of p-adic numbers, introduced by Kurt Hensel, to solve arithmetic problems, marking a significant advancement in number theory. This paper focuses on the theorem over \mathbb{Q}, providing a clear exposition suitable for readers familiar with introductory algebra and number theory. The Hasse–Minkowski Theorem emerged from Hermann Minkowski’s work on quadratic forms in the early 20th century, building on Kurt Hensel’s discovery of p-adic numbers in 1897. Helmut Hasse later generalized it to number fields, formalizing the local-global principle. This theorem revolutionized number theory by providing a systematic method for solving Diophantine equations using local fields, thereby establishing it as a cornerstone for modern algebraic geometry and arithmetic. Consider the equation

    \[x^{2}+y^{2}=3z^{2}.\]

Does it admit a non-trivial rational solution? Over \mathbb{R} the form is indefinite, so solutions certainly exist, e.g. (1,1,\sqrt{2/3}). Over \mathbb{Q}_3, we can check for solutions by assuming x, y, z \in \mathbb{Z}_3 and are not all divisible by 3 (by scaling). The equation x^2 + y^2 = 3z^2 modulo 3 becomes x^2 + y^2 \equiv 0 \pmod 3. Since the quadratic residues mod 3 are 0 and 1, this requires x \equiv 0 \pmod 3 and y \equiv 0 \pmod 3. This means x^2 and y^2 are divisible by 9. So x^2 + y^2 = 9x_0^2 + 9y_0^2 = 3z^2, which simplifies to 3(x_0^2 + y_0^2) = z^2. This implies z must also be divisible by 3. Since x, y, and z are all divisible by 3, this contradicts our assumption. Therefore, the only solution in \mathbb{Q}_3 is the trivial one (0,0,0).

Since a local obstruction appears at p=3, there can be no rational solution. This illustrates the “easy” direction of Hasse–Minkowski: any rational solution must survive in every completion, so a failure in \mathbb{Q}_3 blocks global solvability. The deep content of the Hasse–Minkowski Theorem is the converse: if a quadratic form has a non-trivial solution over \mathbb{R} and every \mathbb{Q}_p, then it necessarily has one over \mathbb{Q}. This local-global principle is special to quadratic forms; for higher-degree equations (such as Selmer’s cubic) local solvability does not imply global solvability.

Preliminaries and Definitions

Throughout, F denotes a field of characteristic \neq2. Boldface letters \mathbf{x}, \mathbf{y} represent column vectors.

Quadratic Forms

Definition 1 (Quadratic Form): Let n\ge1. A quadratic form in n variables over F is a homogeneous polynomial of degree 2,

    \[Q(x_1,\dots,x_n) =\sum_{i\le j}c_{ij}x_ix_j =(x_1 x_2 \cdots x_n) A \begin{pmatrix}x_1 x_2 \vdots x_n\end{pmatrix},\]

where A=A^{\top} is an n\times n symmetric matrix over F.

Isotropic vectors and hyperbolic planes

Definition 2 Let Q be a quadratic form on an F-vector space V and \mathbf{x}\in V\setminus\{\mathbf0\}.

\mathbf{x} is isotropic (for Q) if Q(\mathbf{x})=0.

Q (or V) is isotropic if it possesses an isotropic vector; otherwise it is anisotropic.

Definition 3 (Bracket notation for diagonal forms): If a_{1},\dots,a_{n}\in F^{\times}, we write

    \[\langle a_{1},\dots,a_{n}\rangle\]

for the diagonal quadratic form a_{1}x_{1}^{2}+\dots+a_{n}x_{n}^{2}. Thus, for instance, \langle 1,-1\rangle denotes the hyperbolic plane x^{2}-y^{2}.

Definition 4 (Hyperbolic plane): The binary form \mathbb H=\langle 1,-1\rangle is called the hyperbolic plane. Equivalently, \mathbb H=\{(x,y)\in F^{2}\mid Q(x,y)=x^{2}-y^{2}\}. It contains the isotropic vectors (1,1) and (1,-1), which are linearly independent.

Remark 1 Any two-dimensional isotropic subspace of a non-degenerate quadratic space over F is F-isometric to \mathbb H. Hyperbolic planes will be the “building blocks” in Witt decomposition.

Example 1 (Isotropic vs. anisotropic): Over \mathbb{R} the form   Q_1(x,y)=x^{2}-2y^{2} is isotropic because Q_1(1,\tfrac{1}{\sqrt2})=0. In contrast, Q_2(x,y)=x^{2}+2y^{2} is anisotropic over \mathbb{R} (both terms are \ge0 and vanish simultaneously only at (0,0)).

This distinction has a geometric interpretation as shown in Figure 1, which depicts an indefinite quadratic form whose zero set is non-trivial. The surface intersects the plane z=0 along a cone, corresponding to non-zero vectors for which the quadratic form vanishes.

Over \mathbb{Q}_3 the situation reverses: \langle1,2\rangle becomes isotropic because reducing the equation x^2+2y^2=0 modulo 3 gives

    \[x^2+2y^2 \equiv x^2-y^2 \equiv 0 \pmod{3},\]

since 2 \equiv -1 \pmod{3}. This congruence has non-trivial solutions, for example

x \equiv y \not\equiv 0 \pmod{3}

 whereas \langle1,-2\rangle is anisotropic.

Fig 1 | indefinite-quad.png – The surface z = x^2 - d y^2 for d > 0, an example of an indefinite quadratic form. Its zero set is a double cone, illustrating the geometric origin of isotropic vectors.

Definition 5 (Equivalence): Quadratic forms Q_1,Q_2 in n variables over F are equivalent over F if some T\in\mathrm{GL}_n(F) (the group of invertible n\times n matrices over F) satisfies Q_2(\mathbf{y})=Q_1(T\mathbf{y}) for every \mathbf{y}\in F^n.

Proposition 1  Every non-degenerate quadratic form over F is equivalent to a diagonal form

    \[Q(x_1,\dots,x_n)=a_1x_1^2+\dots+a_rx_r^2, a_i\in F^\times,\]

where r=\operatorname{rank}(Q) (see Definition 6).

Sketch. Write Q(\mathbf{x})=\mathbf{x}^{\top}A\mathbf{x} with A symmetric. Since \operatorname{char}F\neq2, the element 2 is invertible in F, which allows us to complete the square. If a diagonal coefficient a_{11}\neq 0, then all terms involving x_1can be grouped and rewritten by completing the square. After a suitable linear change of variables, this eliminates all cross terms involving x_1, leaving a single squared term together with a quadratic form in one fewer variable. Repeating this process inductively removes all off-diagonal terms and yields a diagonal form2.

Example 2 (Diagonalising a binary form over \mathbb{Q}): Consider the form

    \[Q(x,y)=3x^{2}+4xy+5y^{2}.\]

Its coefficient matrix is A=\begin{pmatrix}3 & 2 2 & 5\end{pmatrix} So \det A=11\neq0 and Q is non-degenerate. Completing the square gives

    \[Q(x,y)=3\Bigl(x+\tfrac23y\Bigr)^{2}+\tfrac{11}{3}y^{2}.\]

Writing u=x+\tfrac23y, v=y (matrix T=\bigl(\begin{smallmatrix}1&\tfrac23 0&1\end{smallmatrix}\bigr)\in\mathrm{GL}_2(\mathbb{Q})) We obtain the diagonal form Q\cong\langle 3,\tfrac{11}{3}\rangle.

Definition 6 (Rank): For Q(\mathbf{x})=\mathbf{x}^{\top}A\mathbf{x}, the rank of Q is \operatorname{rank}(A), i.e. the dimension of the largest subspace on which Q is not identically 0.

Remark 2 After diagonalisation, \operatorname{rank}(Q) is the number of non-zero diagonal coefficients.

Field extensions

Definition 7 (Field extension): An extension K/F is a pair of fields with F\subseteq K. Its degree is [K:F]=\dim_{F}K as an F-vector space.

Definition 8 (Finite & quadratic extensions): An extension is finite if [K:F]<\infty. If [K:F]=2 it is called quadratic. Every quadratic K/F has the form K=F(\sqrt d) for some d\in F^{\times}\setminus F^{\times2}.

Definition 9 (Separable): For characteristic 0 (in particular, F=\mathbb{Q}) every algebraic extension is automatically separable: every element’s minimal polynomial over F splits into distinct roots in a splitting field.

Remark 3 Finite separable extensions admit well-defined field trace \operatorname{Tr}_{K/F} and norm \operatorname{N}_{K/F}. These appear later when we relate the Hilbert symbol to norm forms.

Example 3 (Quadratic extension of \mathbb Q): Set K=\mathbb Q(\sqrt5). The minimal polynomial of \sqrt5 over \mathbb Q is x^2-5, so [K:\mathbb Q]=2; hence K/\mathbb Q is quadratic (and separable). For \alpha=x+y\sqrt5 (x,y\in\mathbb Q):

    \[\operatorname{Tr}_{K/\mathbb Q}(\alpha)=2x, \operatorname{N}_{K/\mathbb Q}(\alpha)=x^2-5y^2,\]

The latter gives the Pell conic X^2-5Y^2=1.

Fields, absolute values, and completions

Definition 10 (Absolute value): An absolute value on a field K is a map \lvert\cdot\rvert\colon K\to\mathbb{R}_{\ge0} such that

1. \lvert x\rvert=0\iff x=0;

2. \lvert xy\rvert=\lvert x\rvert\lvert y\rvert;

3. \lvert x+y\rvert\le\lvert x\rvert+\lvert y\rvert.

It is non-Archimedean if \lvert x+y\rvert\le\max\{\lvert x\rvert,\lvert y\rvert\}.

On \mathbb{Q} there are, up to equivalence, exactly the usual absolute value \lvert\cdot\rvert_\infty and the p-adic absolute values \lvert\cdot\rvert_p, one for each prime p3.

Definition 11 (Completion): Let (K \lvert\cdot\rvert) be a valued field. Its completion \widehat K is the metric completion of K with respect to d(x,y)=\lvert x-y\rvert. We write

    \[\mathbb{Q}_\infty=\mathbb{R}, \mathbb{Q}_p (p\text{ prime})\]

for the completions of \mathbb{Q}.

Example 4 (Real vs. p-adic magnitude): Take x=\dfrac{14}{75}. The ordinary absolute value is \lvert x\rvert_\infty\approx0.187, but factorising x=5^{-2}\cdot14\cdot3^{-1} gives v_{5}(x)=-2 and \lvert x\rvert_{5}=5^{-(-2)}=25. Thus, a “small” real can be “large” 5-adically.

Bilinear forms, discriminant, Witt decomposition

Definition 12 (Discriminant): Let Q be a non-degenerate quadratic form in n variables over F, represented by a symmetric matrix A. Define

 

    \[\disc(Q) = (-1)^{n(n-1)/2}\det A \in F^{\times}/F^{\times 2}.\]

That is, the discriminant is the determinant of A, corrected by a sign depending on n, taken modulo the subgroup of square elements.

Remark 4 If Q_{1} and Q_{2} are equivalent quadratic forms over F (see Definition 5), then \disc(Q_{1})=\disc(Q_{2}) in F^{\times}/F^{\times 2}. Indeed, if Q_{2}(\mathbf{y}) =Q_{1}(T\mathbf{y}) with T\in\mathrm{GL}_{n}(F), then their coefficient matrices satisfy A_{2}=T^{\top}A_{1}T, so

    \[\det(A_{2}) = \det(T)^{2} \det(A_{1}).\]

Thus, the two determinants differ by a square in F^{\times}, and hence represent the same class in F^{\times}/F^{\times 2}.

Every quadratic form Q yields such a B by B(\mathbf{x},\mathbf{y})=\tfrac12\bigl(Q(\mathbf{x}+\mathbf{y}) -Q(\mathbf{x})-Q(\mathbf{y})\bigr).

Example 5 For the diagonal ternary form \langle 1,1,1\rangle we have

    \[\disc\bigl(\langle 1,1,1\rangle\bigr) =(-1)^{3}= -1 \text{in } \mathbb{Q}^{\times}/\mathbb{Q}^{\times2}.\]

Hence \langle1,1,1\rangle is not \mathbb{Q}-equivalent to \langle1,1,-1\rangle, whose discriminant is +1.

Example 6 (Hyperbolic plane): \mathbb{H}=\langle1,-1\rangle is isotropic, since (1,1) is a non-trivial zero. Any isotropic plane is F-isometric to \mathbb{H}.

Definition 13 (Orthogonal sum): If Q_{1} and Q_{2} are quadratic forms over F, their orthogonal sum is written

    \[Q_{1} \perp Q_{2},\]

and defined on the direct sum of the underlying vector spaces by

    \[(Q_{1} \perp Q_{2})(v_{1},v_{2}) = Q_{1}(v_{1})+Q_{2}(v_{2}).\]

More generally, Q^{\perp r} denotes the orthogonal sum of r copies of Q.

Theorem 1 (Witt decomposition): Every non-degenerate quadratic form Q over F decomposes uniquely (up to isometry) as

    \[Q \cong \mathbb{H}^{\perp r} \perp Q_{\mathrm{an}},\]

where \mathbb{H} is the hyperbolic plane and Q_{\mathrm{an}} is anisotropic. The integer r is the Witt index of Q. See4.

Example 7 (Witt decomposition of a quaternary form): Let

    \[Q = \langle 1,1,-1,-1\rangle = x^{2} + y^{2} - z^{2} - w^{2}.\]

Choose isotropic vectors \mathbf{v}_1 = (1,0,1,0) and \mathbf{v}_2 = (0,1,0,1) with B_Q(\mathbf{v}_1, \mathbf{v}_2) = 0. They span a hyperbolic plane \mathbb{H}; repeating with another pair yields

    \[Q \cong \mathbb{H} \perp \mathbb{H},\]

so the Witt index is 2 and the anisotropic part is 0.

Norms and traces in quadratic extensions

For K=F(\sqrt{d}) and \alpha=x+y\sqrt{d}, put

    \[\operatorname{Tr}_{K/F}(\alpha)=2x,\operatorname{N}_{K/F}(\alpha)=x^{2}-dy^{2}.\]

The norm form \operatorname{N}_{K/F} is a basic 2-variable quadratic form whose local isotropy answers norm-related questions.

Local Fields and Completions

Classical Diophantine problems over \mathbb{Q} can often be understood one prime at a time. This idea is made precise by working in the \emph{completions} of \mathbb{Q}—the real field \mathbb{R} at the infinite place and the p-adic fields \mathbb{Q}_p at each finite place p.

Ostrowski’s classification of norms on \mathbb{Q}

Proposition 2 (Product formula): For every x\in\mathbb Q^{\times}, \displaystyle \prod_{p\le\infty} |x|_p = 1.

Theorem 2 (Ostrowski, 1916): Every non-trivial absolute value on \mathbb{Q} is equivalent to either

    \[|\cdot|_\infty \text{(the usual Archimedean norm)} \text{or} |\cdot|_p (p \text{ a prime}).\]

No other inequivalent norms exist.

Idea of proof. For any x\in\mathbb{Q}^{\times} write x=\pm p_{1}^{k_{1}}\cdots p_{r}^{k_{r}}. If |\cdot| is non-Archimedean one shows |x|=\rho^{k_j} for a single prime p_j; rescaling makes it |\cdot|_{p_j}. If |\cdot| is Archimedean, Kronecker’s lemma implies it coincides (up to equivalence) with the usual absolute value. A more detailed proof can be found at5.

Remark 5 Ostrowski’s theorem shows that, up to equivalence, the only non-trivial completions of \mathbb{Q} are the real field \mathbb{R} (corresponding to |\cdot|_\infty) and the p-adic fields \mathbb{Q}_p for primes p. Hence, when we say a statement holds “at every completion of \mathbb{Q},” we really mean: it holds over \mathbb{R} and over \mathbb{Q}_p for each prime p. Verifying local conditions at these places is therefore exhaustive in the Hasse–Minkowski theorem.

Example 8 (Product formula sanity check): For n=30=2\cdot3\cdot5 one has

    \[|30|_\infty = 30, |30|_{2}=2^{-1}, |30|_{3}=3^{-1}, |30|_{5}=5^{-1}, |30|_{p}=1 (p\neq2,3,5).\]

Hence \displaystyle |30|_\infty\prod_{p}|30|_{p}=30\cdot 2^{-1}3^{-1}5^{-1}=1, illustrating the global product formula used implicitly in Ostrowski’s proof.

p-adic valuation and norm

Definition 14 (p-adic valuation): For a prime p and a non-zero rational x\in\mathbb{Q}^\times, write x=p^{k} a/b with a,b\in\mathbb{Z} not divisible by p. Set v_p(x)=k and v_p(0)=\infty.

Definition 15 (p-adic norm): The p-adic norm is

    \[|x|_{p}=p^{-v_{p}(x)}, x\in\mathbb{Q}.\]

It satisfies the non-Archimedean inequality |x+y|_{p}\le\max\{|x|_{p},|y|_{p}\}.

Completions and the field Qp

Completing (\mathbb{Q},|\cdot|_{p}) in the metric d_{p}(x,y)=|x-y|_{p} yields the p-adic field \mathbb{Q}_{p}. For the usual absolute value, we obtain \mathbb{Q}_{\infty}=\mathbb{R}. Elements of \mathbb{Q}_{p} can be written as series \sum_{n\ge k} a_{n}p^{ n} with digits a_{n}\in\{0,\dots ,p-1\}.

Constructing p-adic Numbers

As mentioned above, the p-adic numbers \mathbb{Q}_p arise as the completion of \mathbb{Q} with respect to the p-adic norm. For example, in \mathbb{Q}_5, the number \frac{1}{1-5} = -\frac{1}{4} can be written as a 5-adic series:

    \[\frac{1}{1-5} = \sum_{n=0}^\infty 5^n = 1 + 5 + 5^2 + \cdots.\]

This series converges in \mathbb{Q}_5 because |5^n|_5 = 5^{-n} \to 0 as n \to \infty. Figure 2 illustrates the hierarchical structure of the 5-adic integers \mathbb{Z}_5, showing how residue classes modulo higher powers of 5 nest within one another.

Fig 2 | 5-adic.png – A visualization of the 5-adic integers \mathbb{Z}_5, showing the nested structure of residue classes modulo 5 and their further refinement modulo 25, 125, and so on

Figure 3 illustrates how the partial sums converge 5-adically even though their absolute size grows in the real numbers.

Fig 3 | 5-adic expansion of -\frac{1}{4} = 1 + 5 + 5^2 + \cdots in \mathbb{Q}_5. Each dot represents a_n = 1.

Such series make \mathbb{Q}_p a complete field, enabling tools like Hensel’s lemma for solving equations.

Topology of \mathbb{Q}_p

Each \mathbb{Q}_p is locally compact (meaning every point has a compact neighborhood) and totally disconnected (meaning the only connected subsets are single points). The compact open unit group \mathbb{Z}_p^\times (the p-adic integers with no factor of p) is pro-cyclic for p\neq2 (it is the limit of cyclic groups).

Strong approximation

The strong approximation theorem states that for any finite set S of primes, the diagonal embedding \mathbb{Q}\hookrightarrow\prod_{v\in S}\mathbb{Q}_v is dense. This means any set of p-adic numbers in these fields can be simultaneously approximated by a single rational number. This property is key, as it lets us patch local solutions into a global one once obstructions vanish.

Why bother with completions?

• Analytic control. Limits exist in a completion, so Newton iteration and Hensel’s lemma can lift solutions of congruences to genuine p-adic (hence rational) solutions.

• Local–global philosophy. Many arithmetic statements are true over \mathbb{Q} exactly when they hold in every completion; Hasse–Minkowski for quadratic forms is the prime example.

Hensel’s Lemma

Lemma 1 (Hensel’s Lemma, simple form): Let p be a prime and f(x)\in\mathbb{Z}_{p}[x]. Assume there exists a_{0}\in\mathbb{Z}_{p} such that

    \[f(a_{0})\equiv 0 \pmod{p} \text{and} f'(a_{0})\not\equiv 0 \pmod{p}.\]

Then there is a unique \tilde{a}\in\mathbb{Z}_{p} satisfying

    \[f(\tilde{a})=0 \text{and} \tilde{a}\equiv a_{0}\pmod{p}.\]

Equivalently, any root modulo p with non-vanishing derivative lifts to a unique root in the entire p-adic field \mathbb{Q}_{p}.

Example: lifting a square root of 2 from F7 to Q7

We illustrate Hensel’s lemma with the congruence x^{2}\equiv2\pmod7.

1. In \mathbb{F}_{7}, 3^{2}=9\equiv2, so x_{0}=3 is a root modulo 7.

2. Let f(x)=x^{2}-2. Because f'(x_{0})=2x_{0}=6\not\equiv0\pmod7, Hensel’s lemma applies.

3. One Newton–Hensel step (working mod 49).

    \[x_{1}=x_{0}-\frac{f(x_{0})}{f'(x_{0})} =3-\frac{9-2}{6} =3-\frac{7}{6}.\]

We need 6^{-1}\pmod{49}. Since 6\cdot41=246=49\cdot5+1, we have 6^{-1}\equiv41\pmod{49}.

Hence

    \[\frac{7}{6}\equiv 7\cdot41=287\equiv42\pmod{49}, \text{so } x_{1}\equiv 3-42\equiv-39\equiv10\pmod{49}.\]

4. Verification. 10^{2}=100=49\cdot 2+2\equiv 2\pmod{49}. Thus x_{1}=10 is a root modulo 49. Repeating the process (or invoking Hensel directly) yields a unique \tilde{x}\in\mathbb{Z}_{7} with \tilde{x}\equiv 10\pmod{49} and \tilde{x}^{2}=2.

Remark 6: The condition f'(\tilde{x})\not\equiv 0\pmod{p} is essential for Hensel’s lemma: it ensures the lift exists and is unique.

Isotropy over local fields: quick examples

We record three illustrative calculations that foreshadow the local analysis in the Hasse–Minkowski theorem.

Example 9 (A binary form anisotropic over \mathbb{R} but isotropic over \mathbb{Q}_3): Take B=\langle 1,2\rangle = x^{2}+2y^{2}.

• Over \mathbb{R} both terms are non-negative and vanish simultaneously only at (0,0); B is anisotropic.

• In \mathbb{F}_{3} we have 2\equiv-1, and x^{2}-y^{2}=0 has solutions (1,1), (1,2). Hensel’s lemma lifts either to a 3-adic isotropic vector, so B is isotropic over \mathbb{Q}_{3}.

Example 10 (Hyperbolic plane everywhere locally): The form \mathbb H=\langle 1,-1\rangle is isotropic over \mathbb{R} (obvious) and over every \mathbb{Q}_{p}, since x^{2}\equiv y^{2}\pmod{p} always has non-trivial solutions.

These computations illustrate that local isotropy can vary wildly with the place v, highlighting the necessity of checking all completions in the Hasse–Minkowski criterion.

Legendre Symbol and Quadratic Residues

Basic definitions

Definition 16 (Quadratic residue modulo p): Let p be an odd prime. An integer a is a quadratic residue modulo p if the congruence

    \[x^{2} \equiv a \pmod{p}\]

has a solution x\in\mathbb{Z}. Otherwise, a is a quadratic non-residue.

Definition 17 (Legendre symbol): For an odd prime p and any integer a, define

    \[\left(\frac{a}{p}\right) = \begin{cases} 0 &\text{if } p \mid a, 1 &\text{if } a \text{ is a quadratic residue }(\bmod p), -1 &\text{if } a \text{ is a quadratic non-residue }(\bmod p). \end{cases}\]

The map \bigl(\tfrac{\cdot}{p}\bigr):\mathbb{Z}\to\{-1,0,1\} descends to a group homomorphism \bigl(\tfrac{\cdot}{p}\bigr):(\mathbb{Z}/p\mathbb{Z})^{\times} \longrightarrow\{\pm1\}.

Example 11 (Computing a Legendre symbol): Compute \bigl(\tfrac{7}{19}\bigr).Because 19 is prime,

 

    \[\Bigl(\tfrac{7}{19}\Bigr) = 7^{ (19-1)/2}\bmod 19 = 7^{ 9}\bmod 19.\]

Fast exponentiation:

    \[7^{2}=49\equiv11, 7^{4}\equiv11^{2}=121\equiv7, 7^{8}\equiv7^{2}=11.\]

Hence 7^{9}=7^{8}\cdot7\equiv11\cdot7=77\equiv1\pmod{19}, so \bigl(\tfrac{7}{19}\bigr)=+1.. Thus 7 is a quadratic residue modulo 19.

Quadratic Reciprocity

Theorem 3 (Quadratic Reciprocity Law): For distinct odd primes p and q,

    \[\left(\frac{q}{p}\right) \left(\frac{p}{q}\right) = (-1)^{\frac{p-1}{2}\cdot\frac{q-1}{2}}.\]

Equivalently,

 

    \[\left(\frac{q}{p}\right) = \begin{cases} \left(\dfrac{p}{q}\right) &\text{if } p\equiv1 \pmod4 \text{ or } q\equiv1 \pmod4, -\left(\dfrac{p}{q}\right) &\text{if } p\equiv q\equiv3 \pmod4. \end{cases}\]

Together with the supplementary laws \bigl(\tfrac{-1}{p}\bigr)=(-1)^{(p-1)/2} and \bigl(\tfrac{2}{p}\bigr)=(-1)^{(p^{2}-1)/8}, this completely determines all Legendre symbols.

Example 12 Compute \bigl(\tfrac{7}{19}\bigr). Since 7\equiv3\pmod4 and 19\equiv3\pmod4,

    \[\Bigl(\tfrac{7}{19}\Bigr) =-\Bigl(\tfrac{19}{7}\Bigr) =-\Bigl(\tfrac{5}{7}\Bigr).\]

Because 5^{3}=125\equiv-1\pmod7 we have \bigl(\tfrac{5}{7}\bigr)=-1, hence \bigl(\tfrac{7}{19}\bigr)=+1.

Example: deciding local solvability with (·/p)

Determine whether

    \[x^{2} \equiv 5 \pmod{11}\]

has a solution.

    \begin{align*} \left(\frac{5}{11}\right) &= \left(\frac{11}{5}\right) (-1)^{\frac{5-1}{2}\cdot\frac{11-1}{2}} &&\text{(Quadratic Reciprocity)} &= \left(\frac{1}{5}\right) (-1)^{2\cdot5} = 1. \end{align*}

Hence 5 is a quadratic residue mod 11, so the congruence is solvable. Indeed x\equiv4 or x\equiv7 works.

Remark 7 Legendre symbols (and their higher-power generalisation, the Jacobi symbol) give a quick local test at each prime. In later sections, we will combine these local conditions with Hensel’s lemma and completions to analyse global solvability of quadratic forms.

Hilbert Symbol and Local Quadratic Forms

The Hilbert symbol is a key invariant for classifying quadratic forms over local fields.

Definition 18 (Hilbert Symbol): For a local field K (e.g., \mathbb{Q}_p or \mathbb{R}) and a, b \in K^\times, the Hilbert symbol (a, b)_K is defined as:

    \[(a, b)_K = \begin{cases} 1 & \text{if } x^2 - a y^2 - b z^2 = 0 \text{ has a non-trivial solution in } K, -1 & \text{otherwise}. \end{cases}\]

Proposition 3 (Explicit formula for p odd): Let p be an odd prime. Write a=p^{\alpha}u, b=p^{\beta}v with \alpha,\beta\in\mathbb Z and u,v\in\mathbb Z_p^\times. Then

    \[(a,b)_p = (-1)^{\alpha\beta \frac{p-1}{2}} \left(\frac{u}{p}\right)^{\beta} \left(\frac{v}{p}\right)^{\alpha},\]

where \bigl(\tfrac{\cdot}{p}\bigr) is the Legendre symbol. In particular, if p\nmid ab then (a,b)_p=1 unless both \bigl(\tfrac{u}{p}\bigr)=\bigl(\tfrac{v}{p}\bigr)=-1.

Remark 8 (Real place): Over \mathbb{R} one has (a,b)_\infty=-1 iff a<0 and b<0, otherwise +1.

Example 13 (Computing (2,3)_5): Over \mathbb{Q}_5, 2=5^0\cdot 2, 3=5^0\cdot 3, so \alpha=\beta=0. By Proposition 3,

    \[(2,3)_5=(-1)^{0}\left(\tfrac{2}{5}\right)^{0}\left(\tfrac{3}{5}\right)^{0}=1.\]

Thus z^2=2x^2+3y^2 has a nontrivial 5-adic solution.

Proposition 4 (Basic properties of the Hilbert symbol): Let K be a local field with \operatorname{char}K\ne2 and a,b,c\in K^{\times}. Then

1. Symmetry: (a,b)_{K}=(b,a)_{K}.

2. Bilinearity in the first slot: (ab,c)_{K}=(a,c)_{K} (b,c)_{K} (and hence also in the second by (1)).

3. (a,-a)_{K}=1.

4. If a\ne1 then (a,1-a)_{K}=1.

5. (a,b)_{K}=1 for every b iff a is a square in K (for K\ne\mathbb{C}).

Proof. Write Q_{a,b}(x,y,z)=x^{2}-ay^{2}-bz^{2}. (1) follows because Q_{a,b} and Q_{b,a} are isometric via (x,y,z)\mapsto(x,z,y). For (2) observe Q_{ab,c}(x,y,z)=x^{2}-ab y^{2}-cz^{2} splits into the direct orthogonal sum of Q_{a,c} and Q_{b,c} on suitable 2-planes, so the symbol multiplies. Property (3) is immediate from x^{2}-ay^{2}+az^{2}=0 with (x,y,z)=(\sqrt{a},1,1). For (4) note x^{2}-ay^{2}-(1-a)z^{2}=0 has the rational solution (1,a,1). Finally, (5) is a restatement of the fact that the 1-dimensional quadratic form \langle a\rangle is isotropic over K exactly when a is a square6.

Example 14 (Bilinearity check): Over \mathbb{Q}_{3}: (30,7)_{3} =(6\cdot5,7)_{3} =(6,7)_{3}(5,7)_{3}, matching Proposition 4. Explicit calculation confirms each factor.

Theorem 4 (Hilbert Reciprocity): For a,b\in\mathbb{Q}^{\times} one has

    \[\prod_{v}(a,b)_{v}=1,\]

where the product ranges over all places v\colon\mathbb{Q}\hookrightarrow\mathbb{R} or \mathbb{Q}_{p}.

Sketch of proof. Let K=\mathbb{Q}( \sqrt{a}) and write {\rm N}_{K/\mathbb{Q}}( \cdot ) for the norm. A classical argument shows

    \[(a,b)_{v}=1 \Longleftrightarrow b\text{ is a norm from }K\otimes_{\mathbb{Q}}\mathbb{Q}_{v}.\]

Class field theory (or the product formula for global Hilbert symbols) asserts that an element of \mathbb{Q}^{\times} is a global norm iff it is a local norm everywhere and the product of all local Hilbert symbols equals 1. Applying this to both b and an auxiliary n\in\mathbb{Q}^{\times} chosen so that (a,n)_{v}=(a,b)_{v} except at one place, one deduces \prod_{v}(a,b)_{v}=1. See7.

Example 15 Take a=3, b=5. Direct computation shows

    \[(3,5)_{\infty}=1, (3,5)_{2}=1, (3,5)_{3}=-1, (3,5)_{5}=-1,\]

and (3,5)_p=1 for all other p, so \prod_{v}(3,5)_v=1 as predicted by reciprocity.

For \mathbb{R}, (a, b)_R = -1 if and only if a < 0 and b < 0. For \mathbb{Q}_p, the Hilbert symbol can be computed using the Legendre symbol and local invariants. The Hasse invariant of a quadratic form, defined using Hilbert symbols, helps classify forms over local fields. Hilbert reciprocity states that for a, b \in \mathbb{Q}^\times, \prod_v (a, b)_v = 1, where v runs over all places.

Hasse-Minkowski Theorem and Proof

Theorem 5 (Hasse–Minkowski Theorem): Let Q be a quadratic form over \mathbb{Q}. Then Q(\mathbf{x}) = 0 has a non-trivial solution over \mathbb{Q} if and only if it has a non-trivial solution over \mathbb{R} and over \mathbb{Q}_p for every prime p. Moreover, two quadratic forms over \mathbb{Q} are equivalent if and only if they are equivalent over \mathbb{R} and every \mathbb{Q}_p.

Proof. Let Q = \langle a_1, \ldots, a_n \rangle be a non-degenerate quadratic form over \mathbb{Q}. The goal is to show:

    \begin{align*} & Q(\mathbf{x}) = 0 \text{ has nontrivial } \mathbf{x} \in \mathbb{Q}^n \Leftrightarrow Q \text{ is isotropic over } \mathbb{R} \text{ and all } \mathbb{Q}_p \forall p, & Q \sim Q' \text{ over } \mathbb{Q} \Leftrightarrow Q \sim Q' \text{ over } \mathbb{R} \text{ and } \mathbb{Q}_p \forall p. \end{align*}

For any local field F, the local invariants are: dimension, discriminant d_F(Q) = (-1)^{n(n-1)/2} \det(Q) \in F^\times/F^{\times 2}, and Hasse invariant

    \[\epsilon_F(Q) = \prod_{1 \leq i < j \leq n} (a_i, a_j)_F,\]

where (a, b)_F is the Hilbert symbol. Quadratic forms over F are equivalent iff these three invariants agree.

Necessity: Immediate: a rational solution persists in all completions.

Sufficiency: Diagonalize Q over \mathbb{Q}, Q = \langle a_1, \dots, a_n \rangle, so Q(\mathbf{x}) = \sum_{i=1}^n a_i x_i^2. Assume Q is isotropic in all \mathbb{Q}_p and \mathbb{R}.

Consider cases by n:

Case n = 1:

Q(x) = a_1 x^2. The only nontrivial solution in \mathbb{Q} requires a_1 = 0, which is ruled out for non-degeneracy.

Case n = 2:

Let Q(x,y)=a_1x^2+a_2y^2. Then Q is isotropic over a field F if and only if -a_1/a_2 is a square in F. Thus, isotropy over \mathbb{R} and over every \mathbb{Q}_p implies that -a_1/a_2 is a square in all completions of \mathbb{Q}. By the local–global principle for squares8, an element of \mathbb{Q}^\times that is a square in \mathbb{R} and in every \mathbb{Q}_p is already a square in \mathbb{Q}. Hence Q is isotropic over \mathbb{Q}.

Case n = 3 (Proof sketch):

Let Q(x,y,z)=ax^2+by^2+cz^2. After scaling, assume c=1. Then Q is isotropic over a field F if and only if the Hilbert symbol (a,b)_F=1. Hence, isotropy at every completion \mathbb{Q}_v is equivalent to (a,b)_v=1 for all places v. By Hilbert reciprocity,

    \[\prod_v (a,b)_v = 1,\]

and the condition (a,b)_v=1 for all v is equivalent to b being a norm from the quadratic extension \mathbb{Q}(\sqrt{a})/\mathbb{Q}9. Thus there exist u,v\in\mathbb{Q} with

    \[b = u^2 - a v^2,\]

and substituting (x,y,z)=(v,0,u) gives a nontrivial rational zero of Q.

Case n = 4 (Proof sketch):

Let Q=\langle a,b,c,d\rangle and scale so that abcd \in \mathbb{Q}^{\times 2}. In dimension 4, isotropy of Q over a field F is equivalent to the splitting of an associated quaternion algebra (a,b)_F, and local isotropy at a place v is therefore equivalent to the splitting of (a,b)_{\mathbb{Q}_v}. By the global reciprocity law for quaternion algebras, a quaternion algebra over \mathbb{Q} splits if and only if it splits at every completion. Hence if Q is isotropic over \mathbb{R} and over every \mathbb{Q}_p, the associated quaternion algebra splits globally, and therefore Q is isotropic over \mathbb{Q}10.

Case n \geq 5 (Proof sketch):

In dimension at least 5, a non-degenerate quadratic form over a number field that is isotropic over every completion is isotropic over the field itself. This follows from the structure theory of quadratic forms and the local–global principle for isotropy in sufficiently large dimension11. Hence, since Q is isotropic over \mathbb{R} and over every \mathbb{Q}_p, it is isotropic over \mathbb{Q}. Equivalently, Q splits off a hyperbolic plane,

    \[Q \cong \mathbb{H} \perp Q',\]

and we conclude by induction on the dimension.While this principle guarantees existence, the

quantitative problem of bounding the size of nontrivial isotropic vectors is more subtle; Diet-

Mann12 gives explicit bounds for small solutions of quadratic Diophantine equations.

Equivalence: Suppose Q, Q' are quadratic forms of same dimension, equivalent at every completion. Then d_{\mathbb{Q}}(Q) = d_{\mathbb{Q}}(Q') and \epsilon_v(Q) = \epsilon_v(Q') for all v. By the product formula for Hilbert symbols and Hasse invariants (Hilbert reciprocity), and the matching local invariants everywhere, the forms must also be equivalent over \mathbb{Q}. Indeed, patch local isometries at each place to get a global isometry, as quadratic form equivalence over global fields is determined by the aggregate of local invariants. □

Applications

We highlight three classical consequences of the Hasse–Minkowski theorem and sketch the underlying proofs.

Sums of squares

Theorem 6 (Lagrange, 1770): Every nonnegative integer is the sum of four integer squares.

Proof. Let n \in \mathbb{N}. We seek x_1, x_2, x_3, x_4 \in \mathbb{Z} such that n = x_1^2 + x_2^2 + x_3^2 + x_4^2.

Step 1: Consider the difference form

    \[Q_n(x_1, x_2, x_3, x_4, z) = x_1^2 + x_2^2 + x_3^2 + x_4^2 - n z^2,\]

which is a quadratic form over \mathbb{Q} in five variables. By the local-global principle, Q_n is isotropic over \mathbb{Q} if and only if it is isotropic over \mathbb{R} and every \mathbb{Q}_p. Over \mathbb{R}, Q_n is isotropic because there are both positive and negative coefficients. Over each \mathbb{Q}_p, it is a classical fact that any u\in\mathbb{Q}_p^\times is a sum of four squares for p\ne 2. For p=2, one checks explicitly or uses Hensel’s lemma to lift solutions from \mathbb{F}_2. So, by Hasse–Minkowski, there exist x_1,x_2,x_3,x_4,z\in\mathbb{Q}, z\ne 0, so that

    \[n = \frac{x_1^2 + x_2^2 + x_3^2 + x_4^2}{z^2}\]

and thus n is a sum of four rational squares.

Step 2: Proof (Descent from rationals to integers): Suppose n admits a rational representation

    \[n = \left(\frac{a_1}{m}\right)^2 + \left(\frac{a_2}{m}\right)^2 + \left(\frac{a_3}{m}\right)^2 + \left(\frac{a_4}{m}\right)^2, a_i,m\in\Z, m>0.\]

Clearing denominators gives

 

    \[n m^2 = a_1^2 + a_2^2 + a_3^2 + a_4^2.\]

We show by descent on m that one can reach m=1. Choose integers a_i,m with \gcd(a_1,\dots,a_4,m)=1 such that n m^2=\sum a_i^2 and m minimal. If m=1 we are done. Let p be a prime dividing m. Reduce the congruence \sum a_i^2\equiv0\pmod p. Since not all a_i vanish mod p (minimality of m), this gives a nontrivial quadruple mod p satisfying \sum x_i^2\equiv0\pmod p. It is a classical fact that every prime p divides a sum of four squares, so there exist integers r_1,r_2,r_3,r_4 with r_1^2+r_2^2+r_3^2+r_4^2=p^2 and at least one r_i divisible by p. Using Euler’s identity

    \[(x_1^2+x_2^2+x_3^2+x_4^2) (y_1^2+y_2^2+y_3^2+y_4^2) =(x_1y_1-x_2y_2-x_3y_3-x_4y_4)^2+\cdots+(x_4y_1+x_3y_2-x_2y_3+x_1y_4)^2,\]

multiply the two representations

    \[(a_1^2+\cdots+a_4^2)(r_1^2+\cdots+r_4^2) = (n m^2)(p^2).\]

Because one r_i is divisible by p, all new coordinates produced by Euler’s formula are divisible by p; dividing by p^2 yields another representation

    \[n (m')^2 = b_1^2 + b_2^2 + b_3^2 + b_4^2\]

with m' = m/p < m. This contradicts the minimality of m unless m=1.

By repeated application of this reduction for each prime dividing m, we eventually reach m=1, producing integers x_i with n = x_1^2+x_2^2+x_3^2+x_4^2. Hence, any rational representation can be converted into an integer one.

Alternatively, Gauss’s reduction algorithm applied to the lattice of quadruples representing n (see classic number theory references) guarantees one can always find integer solutions. This yields Lagrange’s integer result from the Hasse–Minkowski rational step combined with integrality tools.

More generally, modern work has extended this perspective to classify which quadratic forms represent all positive integers. Most notably, the 290-Theorem of Bhargava and Hanke13 shows that a positive-definite integral quadratic form is universal if and only if it represents each integer in a specific finite test set contained among the integers up to 290. Related classification results for forms representing all odd integers were established by Rouse14.

The same reasoning with the ternary difference form x^2+y^2+z^2-nw^2 recovers Legendre’s three-square criterion n\not\equiv 0,4,7\pmod8.

Application to Sums of Three Squares

Legendre’s three-square theorem states that a positive integer n can be represented as a sum of three squares if and only if it is not of the form 4^{k} (8m + 7). Using Hasse–Minkowski, we check local conditions at infinity (real positive definite fails for negative, but difference form is indefinite) and at p=2 (anisotropic for forbidden forms mod 8). This local failure at p=2 or infinity explains the criterion.

Table 1 shows that the integers 1, 2, 3, and 9 can be written as sums of three squares, but 7 cannot. This illustrates the exceptional case identified by Legendre’s three-square theorem and demonstrates how a local obstruction leads to global non-representability.

nRepresentation
11^2+0^2+0^2
21^2+1^2+0^2
31^2+1^2+1^2
7No Representation
93^2+0^2+0^2

Table 1 | First few positives and three-square representations.

Classification of quadratic forms over \mathbb{Q}

Theorem 7: Two non-degenerate quadratic forms Q,Q' over \mathbb{Q} are equivalent over \mathbb{Q} if and only if

    \[\dim Q=\dim Q', \disc(Q)=\disc(Q')\in\mathbb{Q}^{\times}/\mathbb{Q}^{\times2}, (a_i,a_j)_v=(a'_i,a'_j)_v \text{ for all }v,\]

i.e. they have the same dimension, the same discriminant, and matching Hilbert invariants at every place.

Idea. Over each completion \mathbb{Q}_{v}, the triple (\dim,\disc,\epsilon_{v}) with \epsilon_{v}(Q)=\prod_{i<j}(a_{i},a_{j})_{v} classifies quadratic forms15. If the three data agree globally, then Q\cong Q' locally everywhere. The second part of Theorem 5 (equivalence) then upgrades these local isometries to a rational isometry.

Algorithmic test for \mathbb{Q}-equivalence

A practical version of Theorem 7 is the Cassels–Ehrlich algorithm. Given two non-degenerate forms Q,Q' in the same number of variables, it decides (in polynomial time for fixed dimension) whether they are \mathbb Q-equivalent.

1. Diagonalise. Use Proposition 1 to write Q\cong\langle a_1,\dots,a_n\rangle and Q'\cong\langle a'_1,\dots,a'_n\rangle.

2. Match discriminants. If

\disc(Q)\neq\disc(Q')\in\mathbb{Q}^{\times}/\mathbb{Q}^{\times2} return No.

3. Compute local symbols. For each finite set of primes dividing 2\disc(Q)\disc(Q') and for v=\infty: evaluate (a_i,a_j)_v and (a'_i,a'_j)_v. If any place disagrees, return No.

4. Solve a gluing problem. Having identical local invariants, construct an explicit isometry matrix T\in\mathrm{GL}_n(\mathbb{Q}) by CRT-patching the local isometries; see15.

5. Return. Output T (or Yes) if the gluing succeeds, otherwise No.

While the Cassels-Ehrlich algorithm handles equivalence, determining specific small solutions or enumerating representations requires more specialized techniques. Simon16 and Kirschmer and Voight17 have developed advanced algorithms for solving quadratic equations in dimension 4 and higher, optimizing the computational complexity of finding explicit solutions.

Example 16 (Two equivalent quaternary forms): Let Q = \langle 1,1,1,-1\rangle, Q'=\langle 2,2,2,-2\rangle.

Step 1: already diagonal.
Step 2: \disc(Q)=\disc(Q')=+1. Step 3: for every place v, (1,1)_v=(2,2)_v=1 and the mixed symbols coincide, so local data match. Step 4: a CRT construction gives T=\operatorname{diag}(1,1,1,\tfrac12) \in\mathrm{GL}_4(\mathbb{Q}) with Q'(x)=Q(Tx). Hence, the algorithm outputs Yes.

This result may be viewed as the global analogue of Witt’s local classification and is a template for more sophisticated adelic invariants in higher-degree forms.

Rational points on conics

Let a,b\in\mathbb{Q}^{\times}. The projective conic

    \[C_{a,b}\colon ax^{2}+by^{2}=z^{2}\]

has a \mathbb{Q}-rational point [x:y:z]\neq[0:0:0] if and only if the following local conditions hold:

1. Real place: a and b are not both negative.

2. p-adic places: (a,b)_p=1 for every prime p\mid2ab.

Proof. Write the associated ternary quadratic form Q_{a,b}(x,y,z)=ax^{2}+by^{2}-z^{2}. A rational point on C_{a,b} corresponds to an isotropic vector for Q_{a,b}. Condition (a) is exactly isotropy over \mathbb{R}. Condition (b) is equivalent to isotropy over each \mathbb{Q}_{p} by Definition 18. Applying Theorem 5 yields the desired equivalence.

Remark 9: Once a single rational point is known, one obtains all rational solutions by a standard line-through-a-point parameterisation.

A cubic counter-example: Selmer’s form

The Hasse–Minkowski theorem is special to quadratic forms. For higher-degree equations, local solvability need not imply global solvability.

A classical counterexample is Selmer’s18 cubic equation

    \[3x^3 + 4y^3 + 5z^3 = 0.\]

This equation has nontrivial solutions over {R} and over {Q}_p for every prime p, but no nontrivial solution over {Q}. Thus, it is locally solvable everywhere but not globally solvable, showing that the local–global principle fails for cubic equations.

Proof. S(x,y,z)=3x^{3}+4y^{3}+5z^{3}=0 is locally solvable everywhere, yet has no nontrivial rational solution.

(1) Local solvability:

Over \R:  We have S(1,1,0)=7>0 and S(1,-1,0)=-1<0. By the Intermediate Value Theorem19, there exists (x,y,0)\in\mathbb{R}^3 with S=0.

Reformulation as an elliptic curve. To study solutions over \mathbb{Q}_p, it is convenient to pass to an equivalent projective model. After a projective change of variables, one obtains

    \[E:\quad X^{3}+Y^{3}+60Z^{3}=0,\]

an elliptic curve over \mathbb{Q}. The coefficient 60 arises from Selmer’s normalization, which concentrates the arithmetic difficulty at the primes 2,3,5 and makes the 3-descent computation tractable18.

Over \mathbb{Q}_p: For primes p\neq2,3,5, the curve has good reduction. By the Hasse bound, E(\mathbb{F}_p) has nonsingular points, which lift via Hensel’s lemma to points in E(\mathbb{Q}_p). Local solvability at p=2,3,5 was established by Selmer. Thus, solutions exist over every \mathbb{Q}_p.

Therefore, the equation S(x,y,z)=0 is locally solvable at every place of \mathbb{Q}.

(2) Global failure: Selmer proved using 3-descent that E(\mathbb{Q}) has no nontrivial

rational points18. Here E(\{Q}) denotes the set of rational points on the elliptic curve E, and 3-descent is a standard method for proving the nonexistence of such points; see20

Thus, the “if and only if” statement of Hasse–Minkowski fails in degree 3: although S=0 has real and p-adic solutions for every prime p, it has no rational solution.

Remark 10 Selmer’s cubic marks the first explicit failure of the Hasse principle. Modern language interprets the obstruction via the non-trivial element of the Tate–Shafarevich group of the associated elliptic curve.

Beyond cubic curves, the failure of the Hasse principle is a major area of study in arithmetic geometry. Colliot-Thelene and Xu21 and Poonen and Voloch22 have explored these failures using the Brauer-Manin obstruction, while Hassett and Várilly-Alvarado23 have extended these inquiries to K3 surfaces, showing that the local-global failure is not unique to simple cubic equations.

Conclusion

The Hasse–Minkowski Theorem demonstrates that quadratic forms over \mathbb{Q} satisfy a local–global principle: questions of rational solvability and equivalence can be resolved entirely by examining the real completion and the p-adic fields. By developing the theory of quadratic forms alongside local fields, Hilbert symbols, and reciprocity laws, this paper showed how local invariants combine to determine global behavior. The applications discussed illustrate the power of this principle, from classical results on sums of squares and rational points on conics to the classification of quadratic forms and algorithmic equivalence tests. At the same time, the contrast with higher-degree equations such as Selmer’s cubic highlights the exceptional nature of quadratic forms, for which local information is sufficient to guarantee global conclusions.

Overall, the Hasse–Minkowski Theorem provides a unifying framework that connects local arithmetic data with global structure. Its influence extends beyond classical number theory, continuing to inform modern research in arithmetic geometry and algorithmic classification problems.

References

  1. Wang, X. (2019). An Introduction to p-adic Numbers and the Hasse Principle. University of Chicago REU Paper. []
  2. Martin, K. (2009). Number Theory II: Chapter 7 — Quadratic Forms in n Variables. University of Oklahoma Lecture Notes. []
  3. Serre, J.-P. (1973). A Course in Arithmetic. Graduate Texts in Mathematics, Vol. 7. Springer. []
  4. O’Meara, O. T. (1971). Introduction to Quadratic Forms. Springer. []
  5. Ruiter, J. (2022). A Proof of Ostrowski’s Theorem. Michigan State University. []
  6. Conrad, K. (2011). Notes from Jack Thorne’s Course on Quadratic Forms. University of Connecticut. []
  7. Cassels, J. W. S., Fröhlich, A. (1967). Algebraic Number Theory. Academic Press. []
  8. Lam, T. Y. (2005; reprinted and widely used in current editions). Introduction to Quadratic Forms over Fields. Graduate Studies in Mathematics, Vol. 67. American Mathematical Society []
  9. Elman, R., Karpenko, N., Merkurjev, A. (2015). The Algebraic and Geometric Theory of Quadratic Forms. AMS Colloquium Publications []
  10. Voight, J. (2021). Quaternion algebras over global fields. In Recent Advances in Algebra (Springer), a chapter on Quaternion Algebras []
  11. Elman, R., Karpenko, N., Merkurjev, A. (2015). The Algebraic and Geometric Theory of Quadratic Forms. AMS Colloquium Publications []
  12. Dietmann, R. (2015). Small solutions of quadratic Diophantine equations. Proceedings of the London Mathematical Society, 110(3), 579–601 []
  13. Bhargava, M., and Hanke, J. (2014). Universal quadratic forms and the 290-theorem. Inventiones Mathematicae, 197, 427–463. []
  14. Rouse, J. (2014). Quadratic forms representing all odd positive integers. American Journal of Mathematics, 136(6), 1693–1745. []
  15. Cassels, J. W. S. (1978). Rational Quadratic Forms. London Mathematical Society Monographs, Vol. 13. Academic Press. [] []
  16. Simon, D. (2005). Quadratic equations in dimension 4, 5 and more. Journal de Theorie des Nombres de Bordeaux, 17(3), 909–923. []
  17. Kirschmer, M., & Voight, J. (2010). Algorithmic enumeration of ideal classes for quaternion orders. SIAM Journal on Computing, 39(5), 1714–1747. []
  18. Selmer, E. S. (1951). The Diophantine equation ax^{3} + by^{3} + cz^{3} = 0. Acta Mathematica, 85, 203–362. [] [] []
  19. Just, W. (2021). Lecture notes on the Intermediate Value Theorem. Ohio University MATH4/530. []
  20. Silverman, J. H. (2009). The Arithmetic of Elliptic Curves. Graduate Texts in Mathematics, Vol. 106, 2nd ed. Springer. []
  21. Colliot-Thelene, J. L., and Xu, F. (2009). Brauer-Manin obstruction for integral points of homogeneous spaces and representation by integral quadratic forms. Compositio Mathematica, 145(2), 309-363. []
  22. Poonen, B., and Voloch, J. F. (2010). The Brauer-Manin obstruction for subvarieties of abelian varieties over number fields. Annals of Mathematics, 171(1), 511-532. []
  23. Hassett, B., and Varilly-Alvarado, A. (2010). Failure of the Hasse principle on general K3 surfaces. Journal of Algebraic Geometry, 19(4), 601-611. []

LEAVE A REPLY

Please enter your comment!
Please enter your name here