The Peirce Basis and the Corners at $n=2$
Introduction
A rank-one idempotent $e$, one nonzero element $\upsilon_{+}$ of the corner $A_{12}$, and the partner $\upsilon_{-}$ fixed by $\upsilon_{+}\upsilon_{-}=e$ form a Peirce basis $(e,\upsilon_{+},\upsilon_{-},1-e)$ in which the multiplication is the multiplication of the matrix units.
The article builds the basis, states the eight nonzero products that define it, proves that the gauge freedom is exactly one reciprocal rescaling $\upsilon_{+}\mapsto\lambda\upsilon_{+}$, $\upsilon_{-}\mapsto\lambda^{-1}\upsilon_{-}$ with $\lambda\in\mathbb{C}^{\times}$, reads the corners in the basis, and identifies the basis with the system of matrix units $E_{11}$, $E_{12}$, $E_{21}$, $E_{22}$. It assumes The Peirce Decomposition in Two Dimensions and Lagrange Idempotents of a $2\times2$ Matrix with Simple Spectrum, and it takes the nilpotent background from The Minimal Polynomial and the Two Jordan Regimes of $M_2(\mathbb{C})$.
It defers the frame in general dimension to The Frame, the Simple Nilpotents and the Gauge of $M_n(\mathbb{C})$ and the recursion to The Peirce Decomposition and the Corner Recursion of $M_n(\mathbb{C})$.
The Frame
Definition. A frame at order two is a pair of orthogonal idempotents $e$ and $f=1-e$, each of rank one.
Proposition. The pair is determined by its first member, and any two frames are conjugate.
Proof. The second member is the complement of the first, and any two rank-one idempotents are conjugate because they have the same rank, so a conjugating matrix carries one frame to the other.
Corollary. Every rank-one idempotent is conjugate to $E_{11}$, so the study of the frame reduces to the standard one.
Proof. The rank of an idempotent is the dimension of its image, and a matrix of full rank is invertible and conjugates the images.
The frame is the matrix form of the corpus's root of one. The corpus builds the frame from a root $\tilde{Ζ}$ of the identity and the pair $\tilde\Pi_{\pm}=\tfrac12(e_t\pm\tilde{Ζ})$; on the matrix side the root is the involution $\zeta$ of Lagrange Idempotents of a $2\times2$ Matrix with Simple Spectrum and the pair is $\rho_{\pm}$.
Proof. The identities $\tilde{Ζ}^{2}=e_t$ and $\tilde\Pi_{\pm}^{2}=\tilde\Pi_{\pm}$ are the identities $\zeta^{2}=I$ and $\rho_{\pm}^{2}=\rho_{\pm}$ proved there.
The Nilpotent Pair
Definition. The companion pair of the frame is a pair $(\upsilon_{+},\upsilon_{-})$ with $\upsilon_{+}\in A_{12}$, $\upsilon_{-}\in A_{21}$ and $\upsilon_{+}\upsilon_{-}=e$.
Proposition. The companion pair exists, and for every frame and every choice of $\upsilon_{+}\ne0$ in $A_{12}$ there is exactly one partner.
Proof. At the standard frame the pair $E_{12},E_{21}$ works. In general the corner $A_{12}$ is one-dimensional, spanned by some $\upsilon_{+}$; the product $\upsilon_{+}A_{21}\subseteq A_{11}$ is the whole line of $e$ when $\upsilon_{+}\ne0$, so there is a unique $\upsilon_{-}$ in $A_{21}$ with $\upsilon_{+}\upsilon_{-}=e$.
Corollary. The pair is automatically nilpotent and satisfies $\upsilon_{+}^{2}=0$, $\upsilon_{-}^{2}=0$ and $\upsilon_{-}\upsilon_{+}=f$.
Proof. The mixed spaces are square-zero by The Peirce Decomposition in Two Dimensions, and $\upsilon_{-}\upsilon_{+}$ lies in $A_{00}$, which is the line of $f$; its trace is $\operatorname{tr}(\upsilon_{-}\upsilon_{+})=\operatorname{tr}(\upsilon_{+}\upsilon_{-})=\operatorname{tr}e=1$, so it is $f$.
Corollary. The four elements $e$, $\upsilon_{+}$, $\upsilon_{-}$, $f$ are linearly independent, hence a basis.
Proof. They lie in four different Peirce spaces, whose direct sum is the whole space.
The Structure Constants
Theorem. In the basis $(e,\upsilon_{+},\upsilon_{-},f)$ the eight nonzero products are
$$ e^{2}=e,\quad e\upsilon_{+}=\upsilon_{+},\quad \upsilon_{+}f=\upsilon_{+},\quad \upsilon_{+}\upsilon_{-}=e,\quad \upsilon_{-}e=\upsilon_{-},\quad f\upsilon_{-}=\upsilon_{-},\quad \upsilon_{-}\upsilon_{+}=f,\quad f^{2}=f, $$
and every other product of two basis elements vanishes.
Proof. Each product is read from the index rule of The Peirce Decomposition in Two Dimensions together with the normalisation $\upsilon_{+}\upsilon_{-}=e$ and the identity $\upsilon_{-}\upsilon_{+}=f$; the products with a wrong middle index vanish by the second Peirce rule.
Corollary. The table is the table of the matrix units, with $e$ playing $E_{11}$, $\upsilon_{+}$ playing $E_{12}$, $\upsilon_{-}$ playing $E_{21}$ and $f$ playing $E_{22}$.
Proof. The eight products listed are exactly the eight nonzero products of the four matrix units.
Corollary. The algebra has no room for a second relation: the table determines the product of any two elements from their coordinates in the basis.
Proof. The product is bilinear, so it is determined on the basis, and the table gives it there.
The Gauge
Theorem. The pair $(\upsilon_{+},\upsilon_{-})$ is determined by $\upsilon_{+}$ up to the reciprocal rescaling
$$ \upsilon_{+}\mapsto\lambda\upsilon_{+},\qquad \upsilon_{-}\mapsto\lambda^{-1}\upsilon_{-},\qquad \lambda\in\mathbb{C}^{\times}. $$
Proof. The product of the rescaled pair is $\lambda\lambda^{-1}\upsilon_{+}\upsilon_{-}=e$, so every such rescaling gives a companion pair; conversely, if $(\upsilon_{+}',\upsilon_{-}')$ is another pair then $\upsilon_{+}'=\lambda\upsilon_{+}$ for some $\lambda\ne0$ because the corner is a line, and the normalisation forces $\upsilon_{-}'=\lambda^{-1}\upsilon_{-}$.
Corollary. The gauge is one complex parameter, and the frame itself has no gauge freedom once $e$ is fixed.
Proof. The idempotent $e$ is fixed and the complement is determined by it; only the choice of the generator of the mixed line and its reciprocal partner is free.
Corollary. The element $\upsilon_{+}\upsilon_{-}$ and the element $\upsilon_{-}\upsilon_{+}$ are gauge invariant, and the two products $\upsilon_{+}^{2}$ and $\upsilon_{-}^{2}$ vanish in every gauge.
Proof. The first two products are $e$ and $f$ independently of $\lambda$, and the squares vanish because the mixed spaces are square-zero.
The gauge is what remains of the diagonal matrix of the diagonalisation. A matrix that diagonalises an element is determined up to the diagonal rescalings, and at order two those rescalings act on the two mixed lines reciprocally.
Proof. Conjugating by a diagonal matrix $\operatorname{diag}(\lambda,\mu)$ multiplies $E_{12}$ by $\lambda\mu^{-1}$ and $E_{21}$ by $\mu\lambda^{-1}$; the normalisation $\upsilon_{+}\upsilon_{-}=e$ then forces $\mu=1$, leaving the single parameter $\lambda$.
The System of Matrix Units
Theorem. The assignment $E_{11}\mapsto e$, $E_{12}\mapsto\upsilon_{+}$, $E_{21}\mapsto\upsilon_{-}$, $E_{22}\mapsto f$ extends to an algebra isomorphism $M_2(\mathbb{C})\to M_2(\mathbb{C})$, namely the conjugation by the matrix whose columns are the images of the standard basis.
Proof. The assignment preserves the eight nonzero products, so it preserves every product by bilinearity, and an algebra map of a simple algebra to itself with nonzero image is injective and hence bijective.
Corollary. Every frame is the image of the standard frame under an automorphism, and the automorphisms of the algebra are the inner ones.
Proof. The map constructed is an automorphism, and every automorphism of a full matrix algebra is inner.
Corollary. The corners are the four lines of the basis, and their products are the products of the corners of the standard frame.
Proof. The isomorphism carries corners to corners, and at the standard frame the corners are the four lines.
The Coordinates
Proposition. In the Peirce coordinates $x=x_{0}e+x_{+}\upsilon_{+}+x_{-}\upsilon_{-}+x_{1}f$ the product is
$$ (xy)_{0}=x_{0}y_{0}+x_{+}y_{-},\quad (xy)_{+}=x_{0}y_{+}+x_{+}y_{1},\quad (xy)_{-}=x_{-}y_{0}+x_{1}y_{-},\quad (xy)_{1}=x_{-}y_{+}+x_{1}y_{1}. $$
Proof. Multiply the two coordinate expressions and collect the four basis elements using the structure constants.
Corollary. The trace is $x_{0}+x_{1}$ and the determinant is $x_{0}x_{1}-x_{+}x_{-}$.
Proof. The trace is the sum of the two idempotent coordinates because $\operatorname{tr}\upsilon_{\pm}=0$, and the determinant is the product of the diagonal coordinates minus the product of the two mixed ones, as at the standard frame.
Corollary. The element is invertible exactly when $x_{0}x_{1}\ne x_{+}x_{-}$, and the inverse has the coordinates $\dfrac{1}{x_{0}x_{1}-x_{+}x_{-}}(x_{1},-x_{+},-x_{-},x_{0})$.
Proof. The inverse of a two-by-two matrix in the matrix-unit coordinates is the adjugate divided by the determinant, and the Peirce coordinates are the matrix-unit coordinates.
Summary
A frame at order two is a pair of orthogonal rank-one idempotents $e$ and $f=1-e$. Adjoining a nonzero $\upsilon_{+}$ in the corner $A_{12}$ and the unique partner $\upsilon_{-}$ with $\upsilon_{+}\upsilon_{-}=e$ gives the Peirce basis $(e,\upsilon_{+},\upsilon_{-},f)$, in which the eight nonzero products are those of the matrix units.
The pair $(\upsilon_{+},\upsilon_{-})$ is determined up to the reciprocal rescaling by $\lambda\in\mathbb{C}^{\times}$, and that is the only freedom of the frame. The coordinates in the basis carry the trace $x_{0}+x_{1}$ and the determinant $x_{0}x_{1}-x_{+}x_{-}$, and the assignment of the basis to the matrix units is an inner automorphism.
Summary of Notation
| Symbol | Meaning |
|---|---|
| $e,f$ | the two idempotents of the frame, $f=1-e$ |
| $\upsilon_{+},\upsilon_{-}$ | the companion pair, $\upsilon_{+}\upsilon_{-}=e$ |
| $A_{12},A_{21}$ | the mixed corners |
| $\lambda$ | the gauge parameter in $\mathbb{C}^{\times}$ |
| $x_{0},x_{+},x_{-},x_{1}$ | the Peirce coordinates |
Further Reading
- Richard S. Pierce, Associative Algebras (Springer, 1982), for the Peirce basis and the system of matrix units.
- Kevin McCrimmon, A Taste of Jordan Algebras (Springer, 2004), for the frame and the gauge in the Peirce language.
- The Peirce Decomposition in Two Dimensions, for the four spaces that the basis linearises.
- The Peirce Basis of the Biquaternion Algebra, for the same basis in the biquaternion language.
- The Frame, the Simple Nilpotents and the Gauge of $M_n(\mathbb{C})$, for the frame carried to the general order.