Tensor theory is a branch of mathematics and has important applications in mechanics.The term tensor originated from mechanics. It was originally used to express the stress state of each point in the elastic medium. Later, tensor theory developed into a powerful mathematical tool in mechanics and physics.Tensor is important because it can satisfy the property that all physical laws must be independent of the choice of coordinate system.The concept of tensor is the generalization of the concept of vector, and vector is a first-order tensor.Tensor is a multilinear function that can be used to express the linear relationship between some vectors, scalars and other tensors.
tensor(Tensor) is defined in some vector spaces andDual spaceThe coordinates of a multilinear mapping on the Cartesian product of are|n|In dimensional space, there are|n|PiecesweightA quantity of, in which each component is coordinatefunctionThese components are also made according to certain rules during coordinate transformationlinear transformation 。R is called theRankorrank(andRank of matrixAnd order).
stayisomorphismThe zero order tensor (r=0) isscalar(Scalar), the first order tensor (r=1) isvector(Vector), the second order tensor (r=2) becomesmatrix(Matrix)。For example, for 3Dimensional spaceThe tensor when r=1 is a vector: (x, y, z). Due to different transformation methods, tensors are divided into three categories: covariant tensor (the indicator is the lower), contravariant tensor (the indicator is the upper), and mixed tensor (the indicator is the upper and the indicator is the lower).
staymathematicsTensor is a geometric entity, or "quantity" in a broad sense.The concept of tensor includes scalar, vector and linear operator.Tensors can be expressed in coordinate systems and recorded as scalar arrays, but they are defined as "independent of the selection of reference systems".Tensors are important in physics and engineering.For example, in diffusion tensor imaging, the tensor of the differential permeability of the expression organ to water in all directions can be used to generate brain scans.Perhaps the most important engineering examples are stress tensors and strain tensors, which are second-order tensors. For general linear materials, the relationship between them is determined by a fourth-order elastic tensor.
Although tensors can be represented by multidimensional arrays of components, the significance of the existence of tensor theory lies in further explaining the meaning of calling a quantity a tensor, not just that it needs a certain number of components with index indexes.In particular, incoordinate transformation The component value of the tensor obeys certain transformation rules.The abstract theory of tensors islinear algebraBranches, now calledMultilinear algebra。
Tensor is the basic language of continuum mechanics, and is the necessary basis for establishing the concept system of mechanics.Tensor is a multilinear functional that inputs several vectors and outputs a number, and the assignment is linear with respect to each vector.[4]
background knowledge
Announce
edit
The word "tensor" was originally used byWilliam Ron Hamiltonstay1846Introduced, but he used the word to refer to what is now calledmodelObject of.The modern meaning of the word isWaldmar Vogtstay1899Started using.
This concept is defined byGregorio Ricci Courbastrostay1890On《Absolute differential geometry》Developed under the title of1900Levy Chiveta's classic article《Absolute differential》(Italian, later published in other translations), which became known to many mathematicians.along with1915aboutEinsteinOfGeneral relativityThe introduction of tensor calculus has gained wider recognition.General relativity is completely expressed by tensor language. Einstein learned a lot of tensor language from Levi Chiveta himself (actually Marcel Grossman, who was Einstein'sZurich Federal Institute of TechnologyHis classmate, a geometer, was also Einstein's mentor and helpful friend in tensor language - see Abraham Pais's Subtle is the Lord, and he learned very hard.But tensors are also used in other fields, such asContinuum mechanics, such asStrain tensor(SeeLinear elasticity)。
Note that the word "tensor" is often usedTensor fieldAnd the tensor field is rightmanifoldEach point of is given a tensor value.To better understand the tensor field, we must first understand the basic ideas of tensors.
regulations
Announce
edit
1. Agreement on summation
It means that in a given item, if there are two indicators that are the same, it means that the indicator is from 1 to the spatial dimensionNSum.For example, in 3D space,
2. Tensor index
Including dummy index and free index.Dumb indicators refer to the same indicators in pairs of the top and bottom indicators.For example, the indicator in the above formulaiIt is a dummy index.Free index refers to the index that appears only once in all terms of the equation.[1]
definition
Announce
edit
There are two ways to define tensors:
1. Defined according to transformation law
If a coordinate systeminQuantityWith another coordinate systeminQuantityMeet the exchange rule between
bebe calledrOrder inversion andsThe components of the order covariant mixed tensor.ifs=0, thenbe calledrThe component of the order inverse tensor.ifr=0, thenbe calledsThe components of the order covariant tensor.The above tensor notation is called component notation.
2. Defined by invariance
Any quantity that can be written in the following invariant form in any coordinate system is defined asr+sOrder tensor:
WhereandCoordinate systemandThe covariant (inverse) basis vector in.The above tensor notation is called invariance notation or dyadic notation.[1]
Basic operation
Announce
edit
1. Addition and subtraction
The sum (difference) of two or more isomorphic tensors of the same order is still the same as their isomorphic tensors.
2. Merging
The union product of two tensors is a new tensor whose order is equal to the sum of the original two tensor orders.
3. Collapse
The operation of making a superscript and a subscript of a tensor the same, the result is a new tensor of order 2 lower than the original tensor.
4. Dot product
The joint operation of union product and contraction union between two tensors.For example, in the polar decomposition theorem, three second-order tensorsR、UandVIntermediate primary dot productR·UandV·RThe result of is a second order tensorF。
5. Symmetrization and antisymmetry
For tensorednIndexn1 Different permutationsnThe operation of the arithmetic mean of a new tensor is called symmetrization.The operation of calculating the arithmetic mean after taking the inverse sign of the new tensor of the index through odd permutation is called inverse symmetry.
6. Additive decomposition
Any second order tensor can be uniquely decomposed into the sum of the symmetric part and the antisymmetric part.For example, velocity gradientCan be decomposed into, whereandAreThe symmetric and antisymmetric parts of theand。
1. Law of Quotient
The law affirming the tensority of certain quantities.[1]
Special tensor
Announce
edit
There are four kinds of special tensors:
① The result of the dot product of two base vectors of a metric tensor.andThey are called covariant and inverse metric tensors respectively, while mixed metric tensors, here(or written as)ForKronecker Symbol, which is defined as:
② A staggered tensor or an Eddington tensor can be defined as, hereRepresentation elementIs determinant, but the replacement symbolexpress(Is an even permutation of (1,2,3), - 1(Is odd permutation of (1,2,3), 0 (other cases)
③ Transposed tensor versus arbitrary second order tensorThe result of component index replacement of is recorded as。
④ Orthogonal tensor A second order tensor that preserves the length of the image.
Covariant vectorAnd inverter vectoraboutThe covariant derivatives of are defined as:and。The above results can be extended to the covariant derivatives of higher-order tensors.
2. Invariant differential operator
The concept of vector analysis is extended to any tensor fieldTThere are four kinds of invariant differential operators, namely gradient ▽T, divergence ▽·T, curl ▽×TAnd Laplacian ▽twoT。
In the rectangular coordinate system, the difference between covariant and contravariant disappears, so it can be specified that all indicators are written as subscripts. In addition, since the Christophel symbol is zero, the covariant derivative becomes a common partial derivative.[1]
example
Announce
edit
A tensor can be expressed as a sequence of valuesDefine FieldsAnd a scalar value.these ones hereDefine FieldsThe vector in isNatural numberAnd these numbers are calledindex。For example, a third-order tensor can have dimensions 2, 5, and 7.Here, indicators range from<1,1,1,>to<2,5,7>.Tensors can have one value in the indicator<1,1,1>, another value in the indicator<1,1,2>, and a total of 70 values.(Similarly, a vector can be expressed as a sequence of valuesDefine FieldsAnd a scalar value. The number in the definition field isNatural number, called indicators, and the number of different indicators is sometimes called vectordimension。)
A tensorsiteIs onEuclidean spaceEach point in is given a tensor value.This is not like the simple example above, which has 70 values. For a third-order tensor, the dimension is<2,5,7>, and every point in the space has 70 values related to it.In other words, a tensor field represents a tensor valued function whose domain of definition is Euclidean space.Not all functions work -- see for more details on these requirementsTensor field。
Not all relationships in nature are linear, but many areDifferentiableSo it can be used locallyMultilinear mappingTo local approximation.In this way, most quantities in physics can be expressed in terms of tensors.
As a simple example, consider a boat in the water.If you want to describe its response to the force, the force is a vector, while the ship's response is an acceleration, which is also a vector.Generally, the acceleration is not in the same direction as the force, because of the specific shape of the hull.However, the relationship between this force and acceleration is actuallylinearOf.Such a relationship can be represented by a tensor of type (1,1) (that is, it turns one vector into another vector).This tensor can be usedmatrixThat is, when it multiplies one vector, it will get another as the result.When the coordinate system changes, the number representing a vector will change, as will the number representing the tensor matrix.
In engineering,rigid bodyorfluidInternalstressIt is also expressed by a tensor;The Latin word for "tensor" means a certain stretch that causes tension.If a specific surface element in the material is selected, the material on one side of the surface will exert a force on the other side.Normally, this force is not orthogonal to the surface, but it will be linearly dependent on the orientation of the surface.This can be accurately described with a tensor of type (2,0), or more precisely, with a tensor of type (2,0)siteBecause tensors may be different in each.
geometryAnd physical quantities can be consideredfreedomTo classify.Scalars are those that can be represented by a number——rate,quality,temperature, etc.There are some vector type quantities, such aspower, it needs a list of numbers to express.Finally, a quantity such as a quadratic form needs to be represented by a multidimensional array.These latter quantities can only be regarded as tensors.
In fact, the concept of tensor is quite extensive and can be used in all the above examples;Scalars and vectors are special cases of tensors.The distinguishing feature between scalars and vectors and between these two and more general tensors is the number of indicators representing their arrays.This number is called tensorrank。Thus, a scalar is a tensor of order 0 (no index is required), and a vector is a tensor of order 1.
Another example of tensors isGeneral relativityInRiemannian curvature tensor, which is a fourth order tensor with dimensions<4,4,4>(3 spatial dimensions+time dimensions=4 dimensions).It can be regarded as a matrix (or vector) of 256 components (256=4 × 4 × 4 × 4), which is actually 4dimensionGroup).The fact that only 20 components are independent of each other can greatly simplify its actual expression.
Tensor density
Announce
edit
Tensor fieldThere can also be a Density.Density isrThe tensor of is the same coordinate transformation as the ordinary tensor, but it has to be multiplied byJacobian matrix OfdeterminantValuerPower.The best explanation for this may be to useVector bundle: where,CutterplexThe determinant bundle of is aLinear plexus, can be used to 'twist' other clumpsrTimes.
Tensor correlation
Announce
edit
1. The theoretical source of tensor.
Arthur Cayley's researchInvariant theory(variant theory)Matrix theoryThe establishment ofdeterminantThe algebraic expression of, which becomesProjective geometryImportant tools.Kelley'sInvariant theoryBritain, which emerged in the first half of the 19th century, focused onAlgebraAnd the application of algebra in geometry.Matrix theory introduces the algebraic definition of vector into the study of linear transformation, which is the precursor of the concept of tensor.
On the other hand, Georg Friedrich Bernhard Riemann proposedN-dimensional manifoldThe concept ofAlgebraFormal subject.RiemanniangeometryThought is expandinggeometryWhile improvingAlgebraThe degree of abstraction in representing geometric objects.After RiemannChristophelWith the efforts of Ricci, Levi Chevita and others, the tensor analysisMathematical method,Riemannian geometry It has also been established.
2. Definition, nature and application value of tensor
Algebraically speaking, it is the generalization of vector.As you know,vectorCan be regarded as one-dimensional“form”(that is, the components are arranged in a row in order),matrixyestwo-dimensionalThe n-order tensor is the so-called n-dimensional "table".The strict definition of tensor is to useLinear mappingTo describe.Similar to vector, the definition is defined by severalCoordinate systemThe set of ordinal numbers satisfying certain coordinate transformation relations when changing is a tensor.
Scalar can be regarded as a tensor of order 0,vectorIt can be regarded as a tensor of order 1.There are many special forms in tensor, such as symmetric tensor, antisymmetric tensor and so on.
Sometimes, people directlyCoordinate systemThe tensor is represented by several numbers (called components), and certain transformation rules should be met between components in different coordinate systems (seecovariantlaw,ContravariantLaws), such as matrices, multivariable linear forms, etc.somephysical quantityFor example, the stress and strain of elastic body and theenergy, momentum, etc. need to be expressed in terms of tensors.staydifferential geometry In the development of C.FGaussian、B.Riemann、E.B.ChristophelThe concept of tensor was introduced by GRichieAnd his student T. Levitzivita developed into tensor analysis, AEinsteinIn itsGeneral relativityTensors are widely used in.
Second, a series of reasoning can be carried out in the new axiom system to obtain a series of newtheoremTo form a new theory.
RochegeometryThe difference between the axiom system of Euclidean geometry and Euclidean geometry is thatAxiom of parallelismChange to: From a point outside the line, you can make at least two lines parallel to this line.Riemannian Geometry And Roche geometricAxiom of parallelismContrary: When passing through a point outside the straight line, the straight line cannot be parallel to the known straight line.in other words,Riemannian Geometry It is stipulated that any two straight lines in the same plane have common points,Riemannian geometry Non recognition of existenceParallel line。Naturally, there is another common assumption: a straight line can be extended to any length, but the length is limited, which can be compared to asphere。Riemannian Geometry Yesdifferential geometry So it is fundamentally different from Roche geometry.
Riemannian geometry The axiom system of Riemann introduced a curved geometric space (which can be described by the curvilinear coordinate system introduced by Lame), and Riemann tried to establish a correspondingalgebraic structure 。Riemann himself failed to achieve this goal, but along the road he opened up, Christophel and Ritchie completed the newgeometryConstruction of.In other words, tensor analysis constitutesRiemannian geometry The core content of.
2. The metric of Riemannian space is expressed by metric tensor;
3. Parallelism in Riemannian space is defined asScalar productRemains unchanged (i.eincluded angleRemains unchanged), relying on the Christophel symbol;
Because of the tool of tensor analysis,Riemannian Geometry Just got something likeCalculusThe same computing function, thus getting rid of the constraints on the level of logical structuredifferential geometry It has realized inheritance and the progress of differential geometry from linear coordinate system to curve coordinate system, makinggeometryAndAlgebraConnect more closely.
In a word, tensor analysis is a generalization of vector analysis on the one hand, anddifferential geometry Development.Tensor Analysis andRiemannian Geometry Develop and promote each other in interweaving.
The input of a color image is a three-dimensional tensor whose size is composed of the width, height and depth of the image[2]。
From the perspective of data storage form, data can be divided into vector representation, matrix representation and tensor representation[3]。