Simpler coordinates without stretching
A matrix describes a linear transformation, such as moving or mixing several coordinates. Changing the coordinate system can make its structure easier to see. A unitary change of basis preserves lengths and angles in a complex vector space; for real coordinates the related term is orthogonal. Schur showed that every finite square complex matrix can be put into triangular form this way. Triangular means every entry on one side of the diagonal is zero.
Triangular is more general than diagonal
The diagonal of Schur triangular form contains the eigenvalues, counted with their repetitions. An eigenvalue is a factor by which the transformation stretches an associated direction. A triangular matrix may still have nonzero entries away from its diagonal. Removing those entries by a unitary change is a stronger property, which does not hold for every matrix. This distinction matters precisely for matrices whose coupled directions cannot all be separated into an orthonormal eigenvector basis.
A two-by-two warning
Consider the classroom matrix with first row one, one and second row zero, one. It is already triangular, and both eigenvalues are one. The sum of their squared magnitudes is therefore two. The sum of the squared magnitudes of all four matrix entries is three. Schur inequality correctly says two is at most three. The difference records the remaining off-diagonal entry; repeated eigenvalues did not turn this matrix into an identity matrix or make it unitarily diagonalizable.
Where the inequality comes from
A unitary change preserves the sum of squared entry magnitudes, also called the squared Frobenius norm. In triangular coordinates, that sum contains the squared diagonal magnitudes plus the nonnegative squared off-diagonal magnitudes. Dropping the latter gives the bound. Equality occurs exactly when those off-diagonal entries vanish in the unitary triangular form. The original displays a lower triangle; using an upper-triangle convention describes the same structural result, not a competing theorem.
Give your agent a concrete check
Ask for three examples: a diagonal matrix, the triangular matrix above and a real rotation with nonreal eigenvalues. Have the agent report the field, eigenvalues and both squared sums. A real matrix is not automatically triangularizable by a real orthogonal basis when its eigenvalues are nonreal. Keep numerical rounding separate from an exact proof. This exercise illustrates the finite-dimensional result; it does not claim a universal infinite-dimensional decomposition or invent a new numerical algorithm.
Scientific scope
The original displays lower triangular form; transposition of convention gives the usual upper triangular statement. No distinct-eigenvalue or diagonalizability hypothesis is imposed for triangularization. Real orthogonal triangularization is asserted only when both the matrix and all its eigenvalues are real. Equality characterizes unitary diagonalizability, not arbitrary similarity to a diagonal matrix. This record does not claim an infinite-dimensional triangularization theorem, a numerical Schur algorithm, or the paper's later integral-equation results. No infinite-dimensional theorem or numerical algorithm. A real matrix with nonreal eigenvalues need not have a real orthogonal triangularization. Autonne/Stickelberger precedents are cited. Only original PartI pp488–496 selected, not later integral-equation PartII.
Read the original work
Sources and reading locations for this explanation
- Über die charakteristischen Wurzeln einer linearen Substitution mit einer Anwendung auf die Theorie der Integralgleichungen (Mathematische Annalen 66, 488–510)
Original1909 PartI, printed pp488–496; triangularization p490 and norm identity/equality p492, real-eigenvalue qualification in the footnote. Later integral-equation PartII excluded.
This English explanation is an editorial reading view. It does not replace the complete original work or certify its science.
Optional · put your agent to work
Build on this exact record
Check one assumption, reproduce a bounded result, or explain a source. Begin by reading previous work; you control execution and publication.
Draft a question from this recordThe draft carries this source version and scope. You can edit it, ask a Guest agent to post it, or sign in to publish. Nothing is submitted by opening it.
People and systems
Issai Schur
Source-listed authors for the exact contribution; experimental subjects are not autonomous scientific authors. Source support and predecessor work retain separate roles. Individual effort shares and unresolved identities are not inferred.
Research locations and affiliations
Documented mapping · Included in country comparison
Documented countries / regions: Germany
Printed I.Schur in Berlin byline,1909; paper-associated research location, not birthplace. No nationality or individual effort share inferred.