Dot

a.b.c or Dot[a,b,c]

gives products of vectors, matrices, and tensors.

Details

a.b gives an explicit result when a and b are lists with appropriate dimensions. It contracts the last index in a with the first index in b.
Various applications of Dot:

	{a₁,a₂}.{b₁,b₂}	scalar product of vectors
	{a₁,a₂}.{{m₁₁,m₁₂},{m₂₁,m₂₂}}
		product of a vector and a matrix
	{{m₁₁,m₁₂},{m₂₁,m₂₂}}.{a₁,a₂}
		product of a matrix and a vector
	{{m₁₁,m₁₂},{m₂₁,m₂₂}}.{{n₁₁,n₁₂},{n₂₁,n₂₂}}
		product of two matrices

The result of applying Dot to two tensors and is the tensor . Applying Dot to a rank tensor and a rank tensor gives a rank tensor. »
Dot can be used on SparseArray and structured array objects. It will return an object of the same type as the input when possible. »
Dot is linear in all arguments. » It does not define a complex (Hermitian) inner product on vectors. »
When its arguments are not lists or sparse arrays, Dot remains unevaluated. It has the attribute Flat.

Examples

open allclose all

Basic Examples (4)

Scalar product of vectors in three dimensions:

Scalar product of vectors in two dimensions:

Vectors are perpendicular if their inner product is zero:

Visualize the vectors:

The product of a matrix and a vector:

The product of a vector and a matrix:

The product of a matrix and two vectors:

The product of two matrices:

Multiply in the other order:

Use rectangular matrices:

Scope (26)

Dot Products of Vectors (7)

Scalar product of machine-precision vectors:

Dot product of exact vectors:

Inner product of symbolic vectors:

The dot product of arbitrary-precision vectors:

Dot allows complex inputs, but does not conjugate any of them:

To compute the complex or Hermitian inner product, apply Conjugate to one of the inputs:

Some sources, particularly in the mathematical literature, conjugate the second argument:

Compute the norm of u using the two inner products:

Verify the result using Norm:

Dot product of sparse vectors:

Compute the scalar product of two QuantityArray vectors:

Matrix-Vector Multiplication (5)

Define a rectangular matrix of dimensions :

Define a 2-vector and a 3-vector:

The matrix can be multiplied by the 2-vector only on the left:

Multiplying in the opposite order produces an error message due to the incompatible shapes:

Similarly, the matrix can be multiplied by the 3-vector only on the right:

Multiply the matrix on both sides at once:

Define a square matrix and a compatible vector:

The products m.v and v.m return different vectors:

The product v.m.v is a scalar:

Define a column and row matrices c and r with the same numerical entries as v:

Products involving m, c and r have the same entries as those involving m and v, but are all matrices:

Moreover, the products must be done in an order that respects the matrices' shapes:

Define a matrix and two vectors:

Since is a vector, is an allowed product:

Note that it is effectively multiplying on the left side of the matrix, not the right:

The product of a sparse matrix and sparse vector is a sparse vector:

Format the result as a row matrix:

The product of a sparse matrix and an ordinary vector is a normal vector:

The product of a structured matrix with a vector will retain the structure if possible:

The product of a normal matrix with a structured vector may have the structure of the vector:

Matrix-Matrix Multiplication (9)

Multiply real machine-number matrices:

Product of complex matrices:

Product of exact matrices:

Multiply in the other order:

Visualize the input and output matrices:

Multiply arbitrary-precision matrices:

Since , these matrices cannot be multiplied in the opposite order:

Product of symbolic matrices:

The product of sparse matrices is another sparse matrix:

Format the result:

The product of structured matrices preserves the structure if possible:

Format the result:

Cube a matrix:

Compare with MatrixPower:

Raise a matrix to the tenth power using Dot in combination with Apply (@@) and ConstantArray:

Verify the result:

Efficiently multiply large matrices:

Higher-Rank Arrays (5)

Dot works for arrays of any rank:

The dimensions of the result are those of the input with the common dimension collapsed:

Any combination is allowed as long as products are done with a common dimension:

Create a rank-three array with three equal dimensions:

Create three vectors of the same dimension:

is the complete contraction that pairs with 's last level and with its first:

is the different contraction that pairs with 's first level and with its last:

Contract both levels of m with the second and third levels of a, respectively:

Dot of two sparse arrays is generally another sparse array:

Dot of a sparse array and an ordinary list may be another sparse array or an ordinary list:

Format the rank-three array:

The product of two SymmetrizedArray objects is generally another symmetrized array:

The symmetry of the new array may be much more complicated than the symmetry of either input:

Applications (16)

Projections and Bases (6)

Project the vector on the line spanned by the vector :

Visualize and its projection onto the line spanned by :

Project the vector on the plane spanned by the vectors and :

First, replace with a vector in the plane perpendicular to :

The projection in the plane is the sum of the projections onto and :

Find the component perpendicular to the plane:

Confirm the result by projecting onto the normal to the plane:

Visualize the plane, the vector and its parallel and perpendicular components:

Apply the Gram–Schmidt process to construct an orthonormal basis from the following vectors:

The first vector in the orthonormal basis, , is merely the normalized multiple :

For subsequent vectors, components parallel to earlier basis vectors are subtracted prior to normalization:

Confirm the answers using Orthogonalize:

Define a basis for :

Verify that the basis is orthonormal:

Find the components of a general vector with respect to this new basis:

Verify the components with respect to the :

Define a basis for :

Verify that it is a basis by showing that the matrix formed by the vectors has nonzero determinant:

The change of basis matrix is the inverse of the matrix whose columns are the :

A vector whose coordinates are in the standard bases will have coordinates with respect to :

Verify that these coordinates give back the vector :

The Frenet–Serret system encodes every space curve's properties in a vector basis and scalar functions. Consider the following curve:

Construct an orthonormal basis from the first three derivatives by subtracting parallel projections:

Ensure that the basis is right-handed:

Compute the curvature, , and torsion, , which quantify how the curve bends:

Verify the answers using FrenetSerretSystem:

Visualize the curve and the associated moving basis, also called a frame:

Matrices and Linear Operators (6)

A matrix is orthogonal of . Show that a rotation matrix is orthogonal:

Confirm using OrthogonalMatrixQ:

A matrix is unitary of . Show that Pauli matrices are unitary:

Confirm with UnitaryMatrixQ:

A matrix is normal if . Show that the following matrix is normal:

Confirm using NormalMatrixQ:

Normal matrices include many other types of matrices as special cases. Unitary matrices are normal:

Hermitian or self-adjoint matrices for which are also normal, as the matrix shows:

However, the matrix is not a named type of normal matrix such as unitary or Hermitian:

In quantum mechanics, systems with finitely many states are represented by unit vectors and physical quantities by matrices that act on them. Consider a spin-1/2 particle such as an electron. It might be in a state such as the following:

The angular momentum in the direction is given by the following matrix:

The angular momentum of this state is :

The uncertainty in the angular momentum of this state is :

The uncertainty in the direction is computed analogously:

The uncertainty principle gives a lower bound on the product of uncertainties, :

Consider a linear mapping :

Get the matrix representation for :

Create a vector to be acted upon:

Apply the linear mapping to the vector using different methods:

Using with Dot is the faster method:

The application of a single matrix to multiple vectors can be computed as ${v_i}.TemplateBox[{m}, Transpose]$ :

The matrix method is significantly faster than repeated application:

Matrices and Arrays with Symmetry (4)

A real symmetric matrix gives a quadratic form $q:TemplateBox[{}, Reals]^n->TemplateBox[{}, Reals]$ by the formula :

Quadratic forms have the property that :

Equivalently, they define a homogeneous quadratic polynomial in the variables of $TemplateBox[{}, Reals]^n$ :

The range of the polynomial can be , , or . In this case it is :

Visualize the polynomial:

A positive-definite, real symmetric matrix or metric defines an inner product by :

Being positive-definite means that the associated quadratic form is negative for :

Note that Dot itself is the inner product associated with the identity matrix:

Apply the Gram–Schmidt process to the standard basis to obtain an orthonormal basis:

Confirm that this basis is orthonormal with respect to the inner product :

An antisymmetric matrix for which defines a Hamiltonian 2-form :

Verify the conditions on :

is identically zero:

However, the form is nondegenerate, meaning implies :

Construct six vectors in dimension six:

Construct the totally antisymmetric array in dimension six using LeviCivitaTensor:

Compute the complete contraction $sum_( i_1... i_6) epsilon_(i_1,i_2,...,i_6)TemplateBox[{a, {i, _, 1}}, Superscript] TemplateBox[{b, {i, _, 2}}, Superscript]... TemplateBox[{f, {i, _, 6}}, Superscript]$ :