Tensors: Definition, Types, Properties, Examples & Applications

A tensor is a mathematical object used to represent quantities and relationships that may depend on multiple directions or dimensions. It extends familiar mathematical concepts such as scalars, vectors, and matrices into a broader framework. A tensor itself is independent of the coordinate system used to describe it, although its components change when the coordinate system changes according to specific transformation rules.
In a particular coordinate system, a tensor can be represented using an array of numbers. However, a tensor should not be confused with the array used to represent it. This distinction becomes important in physics because the physical quantity represented by a tensor remains the same even when its numerical components change with the coordinate system.

Tensors can be classified in several ways. But the most useful classifications are based on rank or order and on the positions of their indices.
I. Types of Tensors Based on RankThe rank or order of a tensor refers to the number of indices needed to represent its components in a chosen basis. In elementary treatments, this gives the following hierarchy:
• Scalar or rank-0 tensor: A scalar has no direction or index and is represented by a single value. Examples include mass, temperature, and energy.
• Vector or rank-1 tensor: A vector has one index and represents a quantity with magnitude and direction. Force and velocity are common examples.
• Second-order tensor: A second-order tensor has two indices and can be represented by a matrix of components. Stress and strain in materials are examples.
• Higher-order tensor: Tensors with three or more indices are called higher-order tensors. They are used when a quantity involves more complex relationships between multiple directions or variables.
II. Types Based on Index PositionIn tensor notation, indices can appear as upper indices or lower indices. This gives three common forms:
• Contravariant tensor: Its components are written with upper indices, such as Tij.
• Covariant tensor: Its components are written with lower indices, such as Tij​.
• Mixed tensor: It contains both upper and lower indices, such as Tij​.
The number of upper and lower indices also describes the tensor's type, commonly written as (p,q), where p and q indicate the numbers of contravariant and covariant indices.

Tensors have specific mathematical properties that distinguish them from ordinary arrays of numbers.
Coordinate IndependenceA tensor represents an underlying mathematical or physical quantity independently of the coordinate system. Its components, however, change when the coordinate system changes. These components must transform according to the tensor's transformation law.
Rank or OrderThe rank indicates how many indices are associated with a tensor in a particular representation. A scalar has rank 0, a vector has rank 1, and a matrix can represent a rank-2 tensor. Higher-rank tensors require more indices to describe their components.
ComponentsA tensor can have multiple components when expressed in a particular basis. The number and arrangement of these components depend on the tensor's rank and the dimension of the underlying space. For example, a rank-2 tensor in three-dimensional space can be represented using a 3×3 array.
Transformation RulesChanging the coordinate system changes the numerical components of a tensor. What makes the object a tensor is that its components transform according to defined rules. This allows the same mathematical or physical quantity to be described using different coordinate systems.
MultilinearityIn the formal mathematical definition, tensors are multilinear objects. This means they are linear in each of their arguments when the other arguments are held fixed. This property allows tensors to represent relationships involving multiple vectors and linear transformations.

Tensors appear in mathematics, physics, engineering and computing. Some important examples include:
• Scalar: Temperature at a point can be represented by a scalar because it has magnitude but no direction.
• Vector: Velocity and force are represented by vectors because their descriptions involve direction as well as magnitude.
• Stress tensor: Stress at a point inside a material depends on the orientation of the surface being considered and is represented using a second-order tensor.
• Strain tensor: Strain describes deformation within a material and is commonly represented as a second-order tensor.
• Moment of inertia tensor: The distribution of mass relative to different directions in a rigid body can be represented using a tensor.
• Stress-energy tensor: In general relativity, the stress-energy tensor describes the distribution and flow of energy and momentum and plays a central role in the theory.

Tensors are useful when a quantity cannot be adequately described by a single number or vector and its relationship with several directions needs to be represented mathematically.
• Physics: Tensors are used in mechanics, elasticity, fluid mechanics, electromagnetism and general relativity. Stress, strain, moment of inertia and the stress-energy tensor are important examples.
• Engineering: Engineers use tensors to describe stresses, strains and other quantities that vary with direction within materials and structures.
• Fluid mechanics: Tensor quantities help describe stresses and other physical properties within moving fluids.
• General relativity: Tensor mathematics provides the framework for expressing quantities such as spacetime curvature and energy-momentum in a coordinate-independent form.
• Machine learning: In computational frameworks such as TensorFlow, tensors are represented as multidimensional arrays with defined shapes and data types. They provide the basic data structure used to store and process numerical data in machine-learning models.

Tensors provide a systematic way to describe quantities involving multiple directions while preserving their mathematical meaning when coordinates change. This makes them especially useful in physics and engineering, where the same physical phenomenon may need to be described from different reference frames. They also provide the mathematical foundation for representing complex multidimensional data in modern computational applications.