1 Definition and basic properties
An orthonormal basis is a basis for a vector space with an inner product in which every basis vector has length 1 and any two distinct basis vectors are orthogonal. In finite-dimensional Euclidean spaces, such a basis provides a particularly simple coordinate system because vector components can be read directly from inner products.
Orthonormal bases are central in linear algebra because they simplify calculations involving norms, angles, and projections. They also make many proofs and algorithms more transparent, especially in spaces where geometry and algebra interact closely.
1.1 Basis in an inner product space
In an inner product space, a basis is a set of vectors that spans the space and is linearly independent. For an orthonormal basis, these spanning vectors also satisfy the inner product structure of the space in a especially convenient way. Every vector in the space can be written uniquely as a linear combination of the basis vectors.
The presence of an inner product gives meaning to length and angle, which are not available in arbitrary vector spaces. This additional structure allows orthonormality to be defined and used effectively.
1.2 Orthogonality
Two vectors are orthogonal when their inner product is zero. Orthogonality expresses a form of geometric independence, since orthogonal vectors contribute in separate directions. In an orthonormal basis, all different basis vectors are orthogonal to one another.
Orthogonality is useful because it eliminates cross terms in many calculations. This property makes sums of squared coordinates behave predictably and supports efficient decomposition of vectors.
1.3 Normalization
A vector is normalized when it has unit length. In an orthonormal basis, each basis vector is normalized as well as orthogonal to the others. The combination of these two conditions produces a basis with especially simple algebraic behavior.
Normalization ensures that the coefficient of a basis vector in an expansion is obtained directly from an inner product. Without unit length, a scaling factor would have to be included.
1.4 Equivalent characterizations
Several conditions are equivalent to a set of vectors forming an orthonormal basis. One equivalent statement is that the vectors are orthonormal and span the whole space. Another is that the matrix formed by placing the basis vectors into columns is an orthogonal or unitary matrix, depending on the field.
In finite-dimensional spaces, a set of orthonormal vectors with the correct number of elements automatically forms a basis. This equivalence is one reason orthonormal systems are so widely used in computation and theory.
2 Coordinate representation
Coordinates with respect to an orthonormal basis are especially simple. The coefficient of each basis vector can be recovered by taking an inner product with the vector being expanded. This makes orthonormal bases ideal for expressing and analyzing vectors numerically and symbolically.
2.1 Expansion of vectors
If a vector space has an orthonormal basis, every vector can be written as a sum of basis vectors multiplied by scalar coefficients. The resulting representation is unique. In geometric terms, the vector is decomposed into independent perpendicular components.
This expansion mirrors the familiar decomposition of a point in Euclidean space into coordinates along perpendicular axes. The simplicity of the formula is a direct consequence of orthonormality.
2.2 Coefficients by inner products
For an orthonormal basis, the coefficient of a basis vector in the expansion of a given vector equals the inner product of the vector with that basis vector. This works because orthogonality removes contributions from all other basis elements, and normalization removes the need to divide by a length factor.
This rule is one of the most useful properties of orthonormal bases. It turns coordinate extraction into a straightforward projection computation.
2.3 Parseval-type identities
Orthonormal bases lead to identities connecting the norm of a vector with the sum of the squares of its coefficients. In finite-dimensional settings, this is a direct consequence of the Pythagorean theorem. In infinite-dimensional settings, related formulas appear as Parseval's identity or its generalizations.
Such identities show that orthonormal coordinates preserve energy or magnitude in a decomposed form. They are foundational in analysis, approximation theory, and signal representation.
3 Construction of orthonormal bases
Orthonormal bases are often built from arbitrary spanning sets by an orthonormalization process. These constructions are important both theoretically and computationally, since many natural bases are not orthogonal at the outset.
3.1 Gram-Schmidt process
The Gram-Schmidt process converts a linearly independent set into an orthonormal set spanning the same subspace. It is one of the standard methods for constructing orthonormal bases in inner product spaces. The process works by removing from each new vector the components already accounted for by earlier orthonormal vectors.
3.1.1 Step-by-step procedure
Given vectors in sequence, the first vector is normalized. Each later vector is made orthogonal to the previous orthonormal vectors by subtracting its projections onto them, and the result is then normalized. Repeating this step produces an orthonormal set.
The method preserves the span of the original vectors, provided they are linearly independent. It is therefore a systematic way to replace an inconvenient basis with a more useful one.
3.1.2 Numerical stability issues
In exact arithmetic, Gram-Schmidt gives the desired result. In floating-point computation, however, rounding errors can accumulate, especially when the starting vectors are nearly dependent. This can lead to loss of orthogonality in the computed basis.
Modified versions of the algorithm are often used to improve stability. These variants are especially important in numerical linear algebra, where accurate orthonormalization is required.
3.2 Orthonormalization from a spanning set
A spanning set may contain redundant vectors and need not be linearly independent. By selecting independent vectors and applying orthonormalization, one can obtain an orthonormal basis for the same subspace. This process is common in practice when data or functional bases are given in nonorthogonal form.
The resulting orthonormal basis may differ from the original set, but it preserves the essential geometric information. The span remains unchanged.
3.3 Extension to a full basis
If an orthonormal set spans only a subspace, it may sometimes be extended to an orthonormal basis of a larger ambient space. In finite-dimensional spaces, this can be done by adding vectors and orthonormalizing them. In infinite-dimensional spaces, extension depends on the structure of the space and the meaning of completeness.
This idea is useful when one wants a basis for an entire space but initially has only a basis for a subspace. It also appears in constructions of complementary subspaces.
4 Matrix and operator viewpoints
Orthonormal bases have a clean matrix interpretation. They correspond to transformations that preserve inner products, and they simplify the representation of linear maps. This viewpoint is fundamental in both pure and applied linear algebra.
4.1 Orthonormal matrices
A matrix whose columns form an orthonormal basis has special algebraic properties. In the real case, such a matrix is orthogonal; in the complex case, it is unitary. Its inverse is given by its transpose or conjugate transpose.
These matrices preserve lengths and angles. As a result, they represent rigid transformations in Euclidean geometry and analogous transformations in complex inner product spaces.
4.2 Change of basis matrices
When changing from one orthonormal basis to another, the change of basis matrix has orthogonal or unitary form. This makes conversion between coordinate systems particularly simple and stable. The new coordinates can be obtained through inner products with the new basis vectors.
Because orthonormal bases preserve the inner product structure, the transition matrix does not distort norms. This is a major advantage in computational settings.
4.3 Orthogonal and unitary operators
A linear operator is orthogonal or unitary if it preserves inner products. Such operators send orthonormal bases to orthonormal bases. They play a key role in geometry, differential equations, and quantum theory.
Operators of this kind are often easier to analyze because their spectral and norm properties are well controlled. They also model rotations, reflections, and other structure-preserving transformations.
4.4 Diagonalization in orthonormal bases
Some operators can be represented by diagonal matrices with respect to an orthonormal basis. When this occurs, the operator acts independently on each basis direction. Such representations are especially valuable because they turn a complicated transformation into a collection of simple scalar actions.
Symmetric and Hermitian operators are classic examples of operators that admit orthonormal eigenbases under suitable conditions. This fact underlies much of spectral theory.
5 Examples
Concrete examples help illustrate how orthonormal bases work in familiar settings. The simplest examples come from standard coordinate systems, while more advanced ones arise in harmonic analysis and complex vector spaces.
5.1 Standard basis in Euclidean space
In \(\mathbb{R}^n\), the standard basis consists of vectors with a single 1 and all other entries 0. These vectors are mutually orthogonal and each has length 1, so they form an orthonormal basis.
This basis underlies ordinary Cartesian coordinates. Its simplicity makes it the reference point for most finite-dimensional calculations.
5.2 Orthonormal bases in R^n
Many different orthonormal bases exist in \(\mathbb{R}^n\). Any rotation or reflection of the standard basis is again orthonormal. More generally, any set of \(n\) mutually orthogonal unit vectors forms such a basis.
These bases are often chosen to suit a specific problem. For instance, an orthonormal basis may align with a geometric object or simplify a matrix.
5.3 Orthonormal bases in complex vector spaces
In complex vector spaces, orthonormality is defined using a complex inner product. The vectors must satisfy the same orthogonality and unit-length conditions, but the inner product typically involves complex conjugation.
Orthonormal bases in complex spaces are central in quantum mechanics, signal analysis, and the theory of unitary matrices. They allow complex-valued vectors to be handled with the same geometric intuition as real vectors.
5.4 Fourier basis examples
Fourier bases are built from sines, cosines, or complex exponential functions. These systems form orthonormal bases or orthonormal sets in suitable function spaces. They allow periodic functions to be decomposed into frequency components.
Such bases are essential in harmonic analysis and many areas of applied mathematics. They provide a bridge between geometry, analysis, and frequency-domain methods.
6 Applications
Orthonormal bases are used wherever vector decomposition, approximation, or stable computation is needed. Their influence extends across numerical analysis, signal processing, and physics.
6.1 Least squares approximation
In least squares problems, one seeks the vector in a subspace that best approximates a target vector. An orthonormal basis makes this task easier because the projection coefficients are simple inner products. The approximation error can also be analyzed cleanly.
This is one reason orthonormal bases are common in regression, data fitting, and numerical algorithms. They reduce the complexity of solving overdetermined systems.
6.2 Projections onto subspaces
Projection onto a subspace is especially straightforward when the subspace has an orthonormal basis. The projection is obtained by summing the components along each basis vector. This produces the closest vector in the subspace to the original vector.
The geometric meaning is clear: the projected vector is the shadow cast onto the subspace. Orthogonality ensures that the residual error is perpendicular to the subspace.
6.3 Signal processing
In signal processing, orthonormal bases are used to represent signals in a way that separates different features or frequencies. Fourier bases, wavelet bases, and related systems provide compact and efficient representations. They can improve compression, filtering, and noise analysis.
The orthonormal structure helps preserve signal energy across representations. This is valuable when comparing or modifying signals without introducing unwanted distortion.
6.4 Quantum mechanics
In quantum mechanics, states are modeled by vectors in complex Hilbert spaces, and orthonormal bases are used to describe measurement outcomes and state expansions. The probabilities associated with measurement are linked to inner products with basis states.
The orthonormal basis formalism allows observables and state vectors to be handled in a precise linear algebraic framework. It is a standard language for describing finite and infinite quantum systems.
7 Orthonormal bases in infinite-dimensional spaces
The concept of an orthonormal basis extends beyond finite-dimensional vector spaces. In infinite-dimensional settings, such as function spaces, orthonormal bases must be understood together with completeness and convergence.
7.1 Hilbert spaces
A Hilbert space is an inner product space that is complete with respect to the norm induced by the inner product. Orthonormal bases in Hilbert spaces are fundamental because they allow vectors to be represented by convergent series rather than finite sums.
This setting includes many spaces of functions used in analysis and physics. The geometry of orthonormal systems remains essential, even though the basis may be infinite.
7.2 Completeness and closure
In infinite-dimensional spaces, an orthonormal set is not necessarily an orthonormal basis. To qualify as a basis in the stronger Hilbert-space sense, it must be complete, meaning that no nonzero vector is orthogonal to every vector in the set. Equivalently, its linear span must be dense in the space.
Closure is needed because infinite sums may converge to vectors not in the finite span. Completeness ensures that the orthonormal system captures the whole space.
7.3 Countable orthonormal bases
Many important Hilbert spaces have countable orthonormal bases. Such bases permit expansions as infinite series with coefficients determined by inner products. This gives a powerful way to analyze functions and operators.
Countable orthonormal bases are particularly useful because they combine the manageability of sequences with the richness of infinite-dimensional analysis. They are a cornerstone of functional analysis.
8 Related concepts
Several nearby notions are closely connected to orthonormal bases. These concepts help clarify what orthonormality adds to the general idea of a basis or spanning set.
8.1 Orthogonal basis
An orthogonal basis consists of mutually orthogonal nonzero vectors, but the vectors are not necessarily of unit length. It is a weaker notion than orthonormal basis. One can obtain an orthonormal basis by normalizing each vector in an orthogonal basis.
Orthogonal bases still offer many of the same computational advantages, though coefficients must be adjusted by vector lengths.
8.2 Orthonormal set
An orthonormal set is a collection of vectors that are mutually orthogonal and each have unit length, but it need not span the whole space. Such a set may be extended to an orthonormal basis if the ambient space allows it.
These sets often arise as intermediate results in orthonormalization procedures or as partial systems in analysis.
8.3 Orthogonal projection
An orthogonal projection maps a vector to the closest vector in a subspace, with the difference being orthogonal to that subspace. Orthonormal bases make this projection easy to compute explicitly. The notion is central to approximation and decomposition.
The projection operator is closely tied to least squares methods and geometric interpretations of distance.
8.4 Orthonormal coordinates
Orthonormal coordinates are coordinates taken with respect to an orthonormal basis. They are especially convenient because the coordinates correspond directly to inner products with basis vectors. This makes them stable and geometrically meaningful.
In many settings, orthonormal coordinates preserve the structure of the original space while simplifying calculation and interpretation.