The Central Idea
A nonzero vector \(v\) is an eigenvector of a square matrix \(A\) if applying \(A\) does not push \(v\) into a new direction.
\[ \boxed{Av=\lambda v} \]The scalar \(\lambda\) is the corresponding eigenvalue.
Along an eigendirection, a multidimensional transformation behaves like simple one-dimensional scaling.
What the Equation Means
The equation
\[ Av=\lambda v \]says that applying \(A\) to \(v\) gives exactly the same result as multiplying \(v\) by the scalar \(\lambda\).
The vector may be stretched, compressed, reversed, or collapsed, but it does not rotate into a genuinely different direction.
The Eigenvector Must Be Nonzero
The zero vector satisfies
\[ A0=\lambda0 \]for every scalar \(\lambda\).
It therefore carries no directional information. By definition,
\[ \boxed{v\neq0}. \]Geometric Meaning of the Eigenvalue
In
\[ Av=\lambda v, \]the value of \(\lambda\) describes what happens along the eigendirection.
- \(\lambda>1\): stretching;
- \(0<\lambda<1\): compression;
- \(\lambda=1\): unchanged;
- \(\lambda=0\): collapsed to zero;
- \(\lambda<0\): scaling with direction reversal.
Example: A Diagonal Matrix
Consider
\[ A= \begin{bmatrix} 3&0\\ 0&2 \end{bmatrix}. \]Then
\[ A \begin{bmatrix} 1\\ 0 \end{bmatrix} = 3 \begin{bmatrix} 1\\ 0 \end{bmatrix}, \]so \(e_1\) is an eigenvector with eigenvalue \(3\).
A Second Eigenvector
Similarly,
\[ A \begin{bmatrix} 0\\ 1 \end{bmatrix} = 2 \begin{bmatrix} 0\\ 1 \end{bmatrix}. \]Thus \(e_2\) is an eigenvector with eigenvalue \(2\).
Not Every Vector Is an Eigenvector
Using the same matrix,
\[ A \begin{bmatrix} 1\\ 1 \end{bmatrix} = \begin{bmatrix} 3\\ 2 \end{bmatrix}. \]The result is not a scalar multiple of \((1,1)^T\), so that vector is not an eigenvector.
How Eigenvalues Are Found
Start with
\[ Av=\lambda v. \]Rewrite:
\[ Av-\lambda Iv=0. \]Factor:
\[ \boxed{(A-\lambda I)v=0}. \]Why the Determinant Appears
An eigenvector requires a nonzero solution of
\[ (A-\lambda I)v=0. \]A homogeneous square system has a nontrivial solution exactly when its coefficient matrix is singular.
Therefore,
\[ \boxed{\det(A-\lambda I)=0}. \]The Characteristic Equation
The equation
\[ \boxed{\det(A-\lambda I)=0} \]is called the characteristic equation.
Its roots are the eigenvalues of \(A\).
The Characteristic Polynomial
The expression
\[ \boxed{p(\lambda)=\det(A-\lambda I)} \]is a characteristic polynomial.
For an \(n\times n\) matrix, it has degree \(n\).
Sign Convention
Some texts use
\[ \det(\lambda I-A) \]instead of
\[ \det(A-\lambda I). \]These differ only by a factor of \((-1)^n\), so they have the same roots.
Example: Find Eigenvalues
Let
\[ A= \begin{bmatrix} 4&1\\ 2&3 \end{bmatrix}. \]Then
\[ A-\lambda I = \begin{bmatrix} 4-\lambda&1\\ 2&3-\lambda \end{bmatrix}. \]Set the determinant equal to zero:
\[ (4-\lambda)(3-\lambda)-2=0. \]Finish the Eigenvalue Calculation
Expand:
\[ \lambda^2-7\lambda+10=0. \]Factor:
\[ (\lambda-5)(\lambda-2)=0. \]Therefore,
\[ \boxed{\lambda_1=5,\qquad\lambda_2=2}. \]How Eigenvectors Are Found
Once an eigenvalue \(\lambda\) is known, solve
\[ \boxed{(A-\lambda I)v=0}. \]Every nonzero solution is an eigenvector associated with \(\lambda\).
Find the Eigenspace for \(\lambda=5\)
Compute
\[ A-5I= \begin{bmatrix} -1&1\\ 2&-2 \end{bmatrix}. \]The equation
\[ -x+y=0 \]gives \(y=x\).
Thus
\[ \boxed{ E_5= \operatorname{span} \left\{ \begin{bmatrix} 1\\ 1 \end{bmatrix} \right\}}. \]Find the Eigenspace for \(\lambda=2\)
Compute
\[ A-2I= \begin{bmatrix} 2&1\\ 2&1 \end{bmatrix}. \]The equation
\[ 2x+y=0 \]gives \(y=-2x\).
Therefore,
\[ \boxed{ E_2= \operatorname{span} \left\{ \begin{bmatrix} 1\\ -2 \end{bmatrix} \right\}}. \]The Eigenspace
For an eigenvalue \(\lambda\), the eigenspace is
\[ \boxed{ E_\lambda = \ker(A-\lambda I)}. \]It contains the zero vector and all eigenvectors associated with \(\lambda\).
Why an Eigenspace Is a Subspace
Since
\[ E_\lambda=\ker(A-\lambda I), \]and every kernel is a subspace,
\[ \boxed{E_\lambda\text{ is a subspace}}. \]Eigenvectors Come in Families
If \(v\) is an eigenvector and \(c\neq0\), then
\[ A(cv)=cAv=c\lambda v=\lambda(cv). \]Therefore every nonzero scalar multiple of an eigenvector is another eigenvector for the same eigenvalue.
Distinct Eigenvalues Give Independent Eigenvectors
Eigenvectors associated with distinct eigenvalues are linearly independent.
Therefore, if an \(n\times n\) matrix has \(n\) distinct eigenvalues, it has \(n\) linearly independent eigenvectors.
Repeated Eigenvalues
An eigenvalue may appear more than once as a root of the characteristic polynomial.
Repetition does not by itself determine how many independent eigenvectors belong to that eigenvalue.
The eigenspace must be examined.
Algebraic Multiplicity
The algebraic multiplicity of an eigenvalue is its multiplicity as a root of the characteristic polynomial.
If
\[ p(\lambda) = (\lambda-3)^2(\lambda+1), \]then \(\lambda=3\) has algebraic multiplicity \(2\).
Geometric Multiplicity
The geometric multiplicity is the dimension of the eigenspace:
\[ \boxed{ \dim E_\lambda = \dim\ker(A-\lambda I)}. \]Multiplicity Relationship
For every eigenvalue,
\[ \boxed{ 1 \le \text{geometric multiplicity} \le \text{algebraic multiplicity}}. \]This becomes central when testing diagonalizability.
Example: Repeated but Many Eigenvectors
Consider
\[ A= \begin{bmatrix} 2&0\\ 0&2 \end{bmatrix}. \]The eigenvalue \(\lambda=2\) has algebraic multiplicity \(2\).
Since
\[ A-2I=0, \]its eigenspace is all of \(\mathbb{R}^2\), so the geometric multiplicity is also \(2\).
Example: Repeated with One Eigenvector Direction
Consider
\[ A= \begin{bmatrix} 2&1\\ 0&2 \end{bmatrix}. \]Again, \(\lambda=2\) has algebraic multiplicity \(2\).
But
\[ A-2I= \begin{bmatrix} 0&1\\ 0&0 \end{bmatrix}, \]so \(y=0\).
Hence
\[ \boxed{ E_2= \operatorname{span} \left\{ \begin{bmatrix} 1\\ 0 \end{bmatrix} \right\}}. \]Zero as an Eigenvalue
If \(\lambda=0\) is an eigenvalue, then some nonzero vector satisfies
\[ Av=0. \]Thus the null space is nontrivial and \(A\) is singular.
\[ \boxed{ 0\text{ is an eigenvalue} \iff A\text{ is not invertible}}. \]Eigenvalues and the Determinant
If the eigenvalues of an \(n\times n\) matrix are counted with algebraic multiplicity, then
\[ \boxed{ \det(A)=\lambda_1\lambda_2\cdots\lambda_n}. \]This gives another reason that a zero eigenvalue forces determinant zero.
Eigenvalues and the Trace
The trace of a square matrix is
\[ \operatorname{tr}(A) = a_{11}+\cdots+a_{nn}. \]It also equals the sum of the eigenvalues counted with algebraic multiplicity:
\[ \boxed{ \operatorname{tr}(A) = \lambda_1+\cdots+\lambda_n}. \]Quick Check for a \(2\times2\) Matrix
For
\[ A= \begin{bmatrix} a&b\\ c&d \end{bmatrix}, \]the characteristic polynomial may be written
\[ \lambda^2-\operatorname{tr}(A)\lambda+\det(A). \]Therefore the eigenvalues satisfy
\[ \lambda_1+\lambda_2=\operatorname{tr}(A) \]and
\[ \lambda_1\lambda_2=\det(A). \]Triangular Matrices
For an upper or lower triangular matrix, the eigenvalues are the diagonal entries.
This follows because
\[ \det(A-\lambda I) \]is the product of the diagonal entries of \(A-\lambda I\).
Example: Triangular Matrix
Let
\[ A= \begin{bmatrix} 4&2&1\\ 0&-1&3\\ 0&0&5 \end{bmatrix}. \]Its eigenvalues are immediately
\[ \boxed{4,\ -1,\ 5}. \]Identity Matrix Eigenvalues
For
\[ I_nv=v, \]every nonzero vector is an eigenvector with eigenvalue \(1\).
\[ \boxed{\lambda=1}. \]Scalar Matrices
If
\[ A=cI, \]then
\[ Av=cv \]for every vector \(v\).
Thus every nonzero vector is an eigenvector with eigenvalue \(c\).
Projection Eigenvalues
An orthogonal projection matrix \(P\) satisfies
\[ P^2=P. \]If
\[ Pv=\lambda v, \]then
\[ P^2v=\lambda^2v \]but also
\[ P^2v=Pv=\lambda v. \]Therefore
\[ \lambda^2=\lambda, \]so the only possible eigenvalues are
\[ \boxed{0\text{ and }1}. \]Reflection Eigenvalues
A reflection preserves directions lying in the mirror subspace and reverses directions perpendicular to it.
Thus typical reflection eigenvalues are
\[ \boxed{1\text{ and }-1}. \]Rotation and Real Eigenvectors
A nontrivial planar rotation generally has no real eigenvectors.
For example, a \(90^\circ\) rotation sends every nonzero real vector into a different direction.
Over the complex numbers, however, such rotations do have eigenvalues and eigenvectors.
Complex Eigenvalues
A real matrix can have complex eigenvalues.
When the characteristic polynomial has real coefficients, nonreal roots occur in complex-conjugate pairs.
Thus if
\[ a+bi \]is an eigenvalue of a real matrix, then
\[ a-bi \]is also an eigenvalue.
Invariant Directions
An eigendirection is invariant under the transformation.
If
\[ v\in E_\lambda, \]then
\[ Av\in E_\lambda. \]The matrix never pushes a vector in that eigenspace outside the same eigenspace.
Invariant Subspaces
More generally, a subspace \(W\) is invariant under \(A\) if
\[ \boxed{A(W)\subseteq W}. \]Every eigenspace is an invariant subspace.
Invariant subspaces help decompose complicated transformations into smaller structural pieces.
Eigenvectors and Matrix Powers
If
\[ Av=\lambda v, \]then
\[ A^2v = A(\lambda v) = \lambda Av = \lambda^2v. \]Continuing,
\[ \boxed{A^kv=\lambda^kv}. \]Why Matrix Powers Matter
Repeated matrix multiplication appears in:
- discrete dynamical systems;
- population models;
- network processes;
- Markov-type transitions;
- iterative numerical methods;
- long-term growth models.
Eigenvectors turn repeated multidimensional action into scalar powers.
Discrete Dynamical Systems
Consider
\[ x_{k+1}=Ax_k. \]If the initial state is an eigenvector,
\[ x_0=v, \]then
\[ x_k=A^kv=\lambda^kv. \]The entire evolution is controlled by one scalar \(\lambda\).
Growth and Decay
In a discrete system:
- \(|\lambda|>1\): magnitude grows;
- \(|\lambda|<1\): magnitude decays;
- \(|\lambda|=1\): magnitude may persist;
- \(\lambda<0\): signs alternate while magnitude scales.
Dominant Eigenvalues
When a vector contains components in several eigendirections, repeated multiplication often magnifies the component corresponding to the eigenvalue with largest absolute value.
Such an eigenvalue is often called a dominant eigenvalue.
Spectral Radius
The spectral radius of \(A\) is
\[ \boxed{ \rho(A) = \max_i|\lambda_i|}. \]It records the largest eigenvalue magnitude and often helps describe long-term behavior under repeated multiplication.
Continuous Dynamical Systems
For a differential system
\[ x'(t)=Ax(t), \]an eigenvector direction produces solutions of the form
\[ \boxed{ x(t)=e^{\lambda t}v}. \]The sign of the real part of \(\lambda\) then influences growth or decay.
Eigenvalues and Stability
In many continuous linear systems:
- negative real parts correspond to decaying modes;
- positive real parts correspond to growing modes;
- purely imaginary parts produce oscillatory behavior;
- mixed complex values can produce oscillation with growth or decay.
Eigenvalues and Diagonalization
If a matrix has enough linearly independent eigenvectors to form a basis, then the matrix can be rewritten in that eigenbasis as a diagonal matrix.
This produces
\[ \boxed{A=PDP^{-1}}, \]where the columns of \(P\) are eigenvectors and the diagonal entries of \(D\) are the matching eigenvalues.
Why Distinct Eigenvalues Help
An \(n\times n\) matrix with \(n\) distinct eigenvalues automatically has \(n\) independent eigenvectors.
Therefore,
\[ \boxed{ n\text{ distinct eigenvalues} \Longrightarrow \text{diagonalizable}}. \]Repeated Eigenvalues and Diagonalization
Repeated eigenvalues do not automatically prevent diagonalization.
The key question is whether the eigenspaces collectively provide enough independent eigenvectors to form a basis.
Symmetric Matrices
Real symmetric matrices have especially well-behaved eigenstructure.
They have real eigenvalues and admit an orthonormal basis of eigenvectors.
Therefore every real symmetric matrix can be orthogonally diagonalized.
Orthogonality of Symmetric-Matrix Eigenvectors
For a real symmetric matrix, eigenvectors associated with distinct eigenvalues are orthogonal.
This fact connects eigenstructure directly to inner products, orthogonality, and the spectral theorem.
Eigenvalues of Similar Matrices
If
\[ B=P^{-1}AP, \]then \(A\) and \(B\) are similar.
Similar matrices have the same characteristic polynomial and therefore the same eigenvalues.
Eigenvalues and Change of Basis
Similar matrices may have very different entries, but they represent the same linear transformation in different coordinate systems.
Eigenvalues therefore describe structure that does not depend on the choice of basis.
Eigenvectors Do Depend on Coordinates
The geometric eigendirections belong to the transformation itself, but their coordinate vectors change when the basis changes.
This mirrors the broader distinction between an abstract vector and its coordinate representation.
Characteristic Polynomial Coefficients
The characteristic polynomial contains important invariants.
For a \(2\times2\) matrix,
\[ p(\lambda) = \lambda^2 - \operatorname{tr}(A)\lambda + \det(A). \]The trace and determinant therefore appear directly as symmetric combinations of eigenvalues.
A Reliable Procedure for Finding Eigenvalues
- Confirm that \(A\) is square.
- Form \(A-\lambda I\).
- Compute \(\det(A-\lambda I)\).
- Set the determinant equal to zero.
- Solve the characteristic equation.
- Record repeated roots with their algebraic multiplicities.
A Reliable Procedure for Finding Eigenvectors
- Choose one eigenvalue \(\lambda\).
- Form \(A-\lambda I\).
- Solve \((A-\lambda I)v=0\).
- Write the null space in parametric or span form.
- Exclude the zero vector when naming eigenvectors.
- Repeat for each eigenvalue.
A Reliable Procedure for Checking a Proposed Eigenpair
- Compute \(Av\).
- Compute \(\lambda v\).
- Compare the two vectors.
- If they agree and \(v\neq0\), then \((\lambda,v)\) is an eigenpair.
Common Error: Allowing \(v=0\)
The zero vector is never called an eigenvector.
The defining condition requires
\[ \boxed{v\neq0}. \]Common Error: Solving \(\det(A)=0\)
To find eigenvalues, do not solve
\[ \det(A)=0. \]Instead solve
\[ \boxed{\det(A-\lambda I)=0}. \]Common Error: Subtracting \(\lambda\) from Every Entry
In
\[ A-\lambda I, \]the scalar \(\lambda\) is subtracted only from the diagonal entries because \(\lambda I\) is diagonal.
Common Error: Stopping After Finding Eigenvalues
Eigenvalues alone do not identify the preserved directions.
For each eigenvalue, solve
\[ (A-\lambda I)v=0 \]to find the corresponding eigenspace.
Common Error: Reporting Only One Eigenvector
An eigenspace usually contains infinitely many eigenvectors.
A single nonzero vector may serve as a convenient representative, but the complete structural answer is the eigenspace.
Common Error: Confusing Algebraic and Geometric Multiplicity
Algebraic multiplicity comes from the characteristic polynomial.
Geometric multiplicity comes from
\[ \dim\ker(A-\lambda I). \]They are related but not identical.
Common Error: Assuming Every Matrix Has Real Eigenvalues
Real matrices can have complex eigenvalues.
A planar rotation is a standard example.
Common Error: Assuming Every Vector in an Eigenspace Is an Eigenvector
The eigenspace includes the zero vector because it is a subspace.
The actual eigenvectors are the
\[ \boxed{\text{nonzero vectors in the eigenspace}}. \]Example: Full Eigenvalue-Eigenvector Workflow
Consider
\[ A= \begin{bmatrix} 6&2\\ 2&3 \end{bmatrix}. \]Form
\[ A-\lambda I = \begin{bmatrix} 6-\lambda&2\\ 2&3-\lambda \end{bmatrix}. \]Then
\[ \det(A-\lambda I) = (6-\lambda)(3-\lambda)-4. \]Solve the Characteristic Equation
Expand:
\[ \lambda^2-9\lambda+14=0. \]Factor:
\[ (\lambda-7)(\lambda-2)=0. \]Hence
\[ \boxed{\lambda_1=7,\qquad\lambda_2=2}. \]Find the Eigenspace for \(\lambda=7\)
Compute
\[ A-7I= \begin{bmatrix} -1&2\\ 2&-4 \end{bmatrix}. \]Solve
\[ -x+2y=0. \]Thus \(x=2y\), so
\[ \boxed{ E_7= \operatorname{span} \left\{ \begin{bmatrix} 2\\ 1 \end{bmatrix} \right\}}. \]Find the Eigenspace for \(\lambda=2\)
Compute
\[ A-2I= \begin{bmatrix} 4&2\\ 2&1 \end{bmatrix}. \]Solve
\[ 2x+y=0. \]Therefore,
\[ \boxed{ E_2= \operatorname{span} \left\{ \begin{bmatrix} 1\\ -2 \end{bmatrix} \right\}}. \]Check Orthogonality in the Symmetric Example
The matrix
\[ \begin{bmatrix} 6&2\\ 2&3 \end{bmatrix} \]is symmetric.
Its eigenvectors
\[ \begin{bmatrix} 2\\ 1 \end{bmatrix} \quad\text{and}\quad \begin{bmatrix} 1\\ -2 \end{bmatrix} \]satisfy
\[ 2(1)+1(-2)=0. \]They are orthogonal, as expected for distinct eigenvalues of a real symmetric matrix.
Why Eigenstructure Matters in Multidimensional Systems
Eigenvalues and eigenvectors identify the natural modes of a linear transformation.
They reveal which directions remain structurally intact, how strongly those directions grow or decay, and which coordinate systems can simplify the transformation.
These ideas appear in vibration modes, stability analysis, population models, differential equations, Markov processes, network analysis, principal directions, quantum mechanics, control systems, and numerical algorithms.
Practice 1: Check an Eigenpair
Let
\[ A= \begin{bmatrix} 3&0\\ 0&2 \end{bmatrix}, \qquad v= \begin{bmatrix} 1\\ 0 \end{bmatrix}. \]Is \(v\) an eigenvector? If so, identify its eigenvalue.
Practice 2: Characteristic Equation
Find the characteristic equation of
\[ A= \begin{bmatrix} 2&1\\ 0&4 \end{bmatrix}. \]Practice 3: Triangular Eigenvalues
Find the eigenvalues of
\[ A= \begin{bmatrix} 5&2&1\\ 0&-3&4\\ 0&0&7 \end{bmatrix}. \]Practice 4: Zero Eigenvalue
What does the presence of eigenvalue \(0\) tell you about a square matrix?
Practice 5: Eigenspace
If
\[ A-\lambda I= \begin{bmatrix} 1&-2\\ 0&0 \end{bmatrix}, \]describe the corresponding eigenspace.
Practice 6: Determinant from Eigenvalues
A \(3\times3\) matrix has eigenvalues
\[ 2,\quad -1,\quad 5. \]Find its determinant.
Practice 7: Trace from Eigenvalues
Using the same eigenvalues,
\[ 2,\quad -1,\quad 5, \]find the trace.
Practice 8: Matrix Powers
If
\[ Av=3v, \]find \(A^4v\).
Practice 9: Algebraic Multiplicity
If
\[ p(\lambda) = (\lambda-2)^3(\lambda+4), \]what is the algebraic multiplicity of \(\lambda=2\)?
Practice 10: Distinct Eigenvalues
A \(4\times4\) matrix has four distinct eigenvalues. What can you conclude about its eigenvectors?
Practice 11: Projection Eigenvalues
What are the only possible eigenvalues of a projection matrix?
Practice 12: Identify the Error
A student solves
\[ \det(A)=0 \]to find the eigenvalues of \(A\). Explain the error.
Practice Answers
Practice 1
\[ Av= \begin{bmatrix} 3\\ 0 \end{bmatrix} = 3v. \]Therefore,
\[ \boxed{v\text{ is an eigenvector with }\lambda=3}. \]Practice 2
\[ \det(A-\lambda I) = \begin{vmatrix} 2-\lambda&1\\ 0&4-\lambda \end{vmatrix} = (2-\lambda)(4-\lambda). \]Thus
\[ \boxed{(2-\lambda)(4-\lambda)=0}. \]Practice 3
The matrix is triangular, so its eigenvalues are its diagonal entries:
\[ \boxed{5,\ -3,\ 7}. \]Practice 4
A zero eigenvalue means some nonzero vector satisfies
\[ Av=0. \]Therefore,
\[ \boxed{A\text{ is singular and not invertible}}. \]Practice 5
Solve
\[ x-2y=0. \]Hence \(x=2y\), so
\[ \boxed{ E_\lambda= \operatorname{span} \left\{ \begin{bmatrix} 2\\ 1 \end{bmatrix} \right\}}. \]Practice 6
\[ \det(A) = 2(-1)(5) = \boxed{-10}. \]Practice 7
\[ \operatorname{tr}(A) = 2-1+5 = \boxed{6}. \]Practice 8
\[ A^4v=3^4v=\boxed{81v}. \]Practice 9
\[ \boxed{3}. \]Practice 10
Eigenvectors corresponding to distinct eigenvalues are independent.
Therefore the matrix has
\[ \boxed{4\text{ linearly independent eigenvectors}}. \]Practice 11
\[ \boxed{0\text{ and }1}. \]Practice 12
The determinant of \(A\) tests whether \(0\) is an eigenvalue. To find all eigenvalues, one must solve
\[ \boxed{\det(A-\lambda I)=0}. \]What to Remember
The defining eigenvalue equation is
\[ \boxed{Av=\lambda v,\qquad v\neq0}. \]- Eigenvectors are directions preserved by the transformation.
- Eigenvalues are the scaling factors along those directions.
- Eigenvalues satisfy \(\det(A-\lambda I)=0\).
- The eigenspace is \(E_\lambda=\ker(A-\lambda I)\).
- Distinct eigenvalues have linearly independent eigenvectors.
- Algebraic multiplicity comes from the characteristic polynomial.
- Geometric multiplicity is the dimension of the eigenspace.
- Zero is an eigenvalue exactly when the matrix is singular.
- The determinant is the product of eigenvalues.
- The trace is the sum of eigenvalues.
- If \(Av=\lambda v\), then \(A^kv=\lambda^kv\).
- Eigenstructure reveals natural modes of multidimensional systems.
The core structural question is:
\[ \boxed{ \text{which directions does the transformation preserve, and by how much?}} \]Where We Go Next
Once enough independent eigenvectors are available, they can be used as a basis for the entire space.
In that eigenbasis, the transformation becomes diagonal and acts through independent coordinate scalings.
The next lesson, Diagonalization, develops the factorization
\[ \boxed{A=PDP^{-1}} \]and shows how eigenbases simplify matrix powers, dynamical systems, and multidimensional analysis.