Theory SQ_MTX
section ‹ Square Matrices ›
text ‹ The general solution for affine systems of ODEs involves the exponential function.
Unfortunately, this operation is only available in Isabelle for the type class ``banach''.
Hence, we define a type of square matrices and prove that it is an instance of this class. ›
theory SQ_MTX
imports MTX_Norms
begin
subsection ‹ Definition ›
typedef 'm sq_mtx = "UNIV::(real^'m^'m) set"
morphisms to_vec to_mtx by simp
declare to_mtx_inverse [simp]
and to_vec_inverse [simp]
setup_lifting type_definition_sq_mtx
lift_definition sq_mtx_ith :: "'m sq_mtx ⇒ 'm ⇒ (real^'m)" (infixl "$$" 90) is "($)" .
lift_definition sq_mtx_vec_mult :: "'m sq_mtx ⇒ (real^'m) ⇒ (real^'m)" (infixl "*⇩V" 90) is "(*v)" .
lift_definition vec_sq_mtx_prod :: "(real^'m) ⇒ 'm sq_mtx ⇒ (real^'m)" is "(v*)" .
lift_definition sq_mtx_diag :: "(('m::finite) ⇒ real) ⇒ ('m::finite) sq_mtx" (binder "𝖽𝗂𝖺𝗀 " 10)
is diag_mat .
lift_definition sq_mtx_transpose :: "('m::finite) sq_mtx ⇒ 'm sq_mtx" ("_⇧†") is transpose .
lift_definition sq_mtx_inv :: "('m::finite) sq_mtx ⇒ 'm sq_mtx" ("_⇧-⇧1" [90]) is matrix_inv .
lift_definition sq_mtx_row :: "'m ⇒ ('m::finite) sq_mtx ⇒ real^'m" ("𝗋𝗈𝗐") is row .
lift_definition sq_mtx_col :: "'m ⇒ ('m::finite) sq_mtx ⇒ real^'m" ("𝖼𝗈𝗅") is column .
lemma to_vec_eq_ith: "(to_vec A) $ i = A $$ i"
by transfer simp
lemma to_mtx_ith[simp]:
"(to_mtx A) $$ i1 = A $ i1"
"(to_mtx A) $$ i1 $ i2 = A $ i1 $ i2"
by (transfer, simp)+
lemma to_mtx_vec_lambda_ith[simp]: "to_mtx (χ i j. x i j) $$ i1 $ i2 = x i1 i2"
by (simp add: sq_mtx_ith_def)
lemma sq_mtx_eq_iff:
shows "A = B = (∀i j. A $$ i $ j = B $$ i $ j)"
and "A = B = (∀i. A $$ i = B $$ i)"
by (transfer, simp add: vec_eq_iff)+
lemma sq_mtx_diag_simps[simp]:
"i = j ⟹ sq_mtx_diag f $$ i $ j = f i"
"i ≠ j ⟹ sq_mtx_diag f $$ i $ j = 0"
"sq_mtx_diag f $$ i = axis i (f i)"
unfolding sq_mtx_diag_def by (simp_all add: axis_def vec_eq_iff)
lemma sq_mtx_diag_vec_mult: "(𝖽𝗂𝖺𝗀 i. f i) *⇩V s = (χ i. f i * s$i)"
by (simp add: matrix_vector_mul_diag_mat sq_mtx_diag.abs_eq sq_mtx_vec_mult.abs_eq)
lemma sq_mtx_vec_mult_diag_axis: "(𝖽𝗂𝖺𝗀 i. f i) *⇩V (axis i k) = axis i (f i * k)"
unfolding sq_mtx_diag_vec_mult axis_def by auto
lemma sq_mtx_vec_mult_eq: "m *⇩V x = (χ i. sum (λj. (m $$ i $ j) * (x $ j)) UNIV)"
by (transfer, simp add: matrix_vector_mult_def)
lemma sq_mtx_transpose_transpose[simp]: "(A⇧†)⇧† = A"
by (transfer, simp)
lemma transpose_mult_vec_canon_row[simp]: "(A⇧†) *⇩V (𝖾 i) = 𝗋𝗈𝗐 i A"
by transfer (simp add: row_def transpose_def axis_def matrix_vector_mult_def)
lemma row_ith[simp]: "𝗋𝗈𝗐 i A = A $$ i"
by transfer (simp add: row_def)
lemma mtx_vec_mult_canon: "A *⇩V (𝖾 i) = 𝖼𝗈𝗅 i A"
by (transfer, simp add: matrix_vector_mult_basis)
subsection ‹ Ring of square matrices ›
instantiation sq_mtx :: (finite) ring
begin
lift_definition plus_sq_mtx :: "'a sq_mtx ⇒ 'a sq_mtx ⇒ 'a sq_mtx" is "(+)" .
lift_definition zero_sq_mtx :: "'a sq_mtx" is "0" .
lift_definition uminus_sq_mtx :: "'a sq_mtx ⇒ 'a sq_mtx" is "uminus" .
lift_definition minus_sq_mtx :: "'a sq_mtx ⇒ 'a sq_mtx ⇒ 'a sq_mtx" is "(-)" .
lift_definition times_sq_mtx :: "'a sq_mtx ⇒ 'a sq_mtx ⇒ 'a sq_mtx" is "(**)" .
declare plus_sq_mtx.rep_eq [simp]
and minus_sq_mtx.rep_eq [simp]
instance apply intro_classes
by(transfer, simp add: algebra_simps matrix_mul_assoc matrix_add_rdistrib matrix_add_ldistrib)+
end
lemma sq_mtx_zero_ith[simp]: "0 $$ i = 0"
by (transfer, simp)
lemma sq_mtx_zero_nth[simp]: "0 $$ i $ j = 0"
by transfer simp
lemma sq_mtx_plus_eq: "A + B = to_mtx (χ i j. A$$i$j + B$$i$j)"
by transfer (simp add: vec_eq_iff)
lemma sq_mtx_plus_ith[simp]:"(A + B) $$ i = A $$ i + B $$ i"
unfolding sq_mtx_plus_eq by (simp add: vec_eq_iff)
lemma sq_mtx_uminus_eq: "- A = to_mtx (χ i j. - A$$i$j)"
by transfer (simp add: vec_eq_iff)
lemma sq_mtx_minus_eq: "A - B = to_mtx (χ i j. A$$i$j - B$$i$j)"
by transfer (simp add: vec_eq_iff)
lemma sq_mtx_minus_ith[simp]:"(A - B) $$ i = A $$ i - B $$ i"
unfolding sq_mtx_minus_eq by (simp add: vec_eq_iff)
lemma sq_mtx_times_eq: "A * B = to_mtx (χ i j. sum (λk. A$$i$k * B$$k$j) UNIV)"
by transfer (simp add: matrix_matrix_mult_def)
lemma sq_mtx_plus_diag_diag[simp]: "sq_mtx_diag f + sq_mtx_diag g = (𝖽𝗂𝖺𝗀 i. f i + g i)"
by (subst sq_mtx_eq_iff) (simp add: axis_def)
lemma sq_mtx_minus_diag_diag[simp]: "sq_mtx_diag f - sq_mtx_diag g = (𝖽𝗂𝖺𝗀 i. f i - g i)"
by (subst sq_mtx_eq_iff) (simp add: axis_def)
lemma sum_sq_mtx_diag[simp]: "(∑n<m. sq_mtx_diag (g n)) = (𝖽𝗂𝖺𝗀 i. ∑n<m. (g n i))" for m::nat
by (induct m, simp, subst sq_mtx_eq_iff, simp_all)
lemma sq_mtx_mult_diag_diag[simp]: "sq_mtx_diag f * sq_mtx_diag g = (𝖽𝗂𝖺𝗀 i. f i * g i)"
by (simp add: matrix_mul_diag_diag sq_mtx_diag.abs_eq times_sq_mtx.abs_eq)
lemma sq_mtx_mult_diagl: "(𝖽𝗂𝖺𝗀 i. f i) * A = to_mtx (χ i j. f i * A $$ i $ j)"
by transfer (simp add: matrix_mul_diag_matl)
lemma sq_mtx_mult_diagr: "A * (𝖽𝗂𝖺𝗀 i. f i) = to_mtx (χ i j. A $$ i $ j * f j)"
by transfer (simp add: matrix_matrix_mul_diag_matr)
lemma mtx_vec_mult_0l[simp]: "0 *⇩V x = 0"
by (simp add: sq_mtx_vec_mult.abs_eq zero_sq_mtx_def)
lemma mtx_vec_mult_0r[simp]: "A *⇩V 0 = 0"
by (transfer, simp)
lemma mtx_vec_mult_add_rdistr: "(A + B) *⇩V x = A *⇩V x + B *⇩V x"
unfolding plus_sq_mtx_def
apply(transfer)
by (simp add: matrix_vector_mult_add_rdistrib)
lemma mtx_vec_mult_add_rdistl: "A *⇩V (x + y) = A *⇩V x + A *⇩V y"
unfolding plus_sq_mtx_def
apply transfer
by (simp add: matrix_vector_right_distrib)
lemma mtx_vec_mult_minus_rdistrib: "(A - B) *⇩V x = A *⇩V x - B *⇩V x"
unfolding minus_sq_mtx_def by(transfer, simp add: matrix_vector_mult_diff_rdistrib)
lemma mtx_vec_mult_minus_ldistrib: "A *⇩V (x - y) = A *⇩V x - A *⇩V y"
by (metis (no_types, lifting) add_diff_cancel diff_add_cancel
matrix_vector_right_distrib sq_mtx_vec_mult.rep_eq)
lemma sq_mtx_times_vec_assoc: "(A * B) *⇩V x = A *⇩V (B *⇩V x)"
by (transfer, simp add: matrix_vector_mul_assoc)
lemma sq_mtx_vec_mult_sum_cols: "A *⇩V x = sum (λi. x $ i *⇩R 𝖼𝗈𝗅 i A) UNIV"
by(transfer) (simp add: matrix_mult_sum scalar_mult_eq_scaleR)
subsection ‹ Real normed vector space of square matrices ›
instantiation sq_mtx :: (finite) real_normed_vector
begin
definition norm_sq_mtx :: "'a sq_mtx ⇒ real" where "∥A∥ = ∥to_vec A∥⇩o⇩p"
lift_definition scaleR_sq_mtx :: "real ⇒ 'a sq_mtx ⇒ 'a sq_mtx" is scaleR .
definition sgn_sq_mtx :: "'a sq_mtx ⇒ 'a sq_mtx"
where "sgn_sq_mtx A = (inverse (∥A∥)) *⇩R A"
definition dist_sq_mtx :: "'a sq_mtx ⇒ 'a sq_mtx ⇒ real"
where "dist_sq_mtx A B = ∥A - B∥"
definition uniformity_sq_mtx :: "('a sq_mtx × 'a sq_mtx) filter"
where "uniformity_sq_mtx = (INF e∈{0<..}. principal {(x, y). dist x y < e})"
definition open_sq_mtx :: "'a sq_mtx set ⇒ bool"
where "open_sq_mtx U = (∀x∈U. ∀⇩F (x', y) in uniformity. x' = x ⟶ y ∈ U)"
instance apply intro_classes
unfolding sgn_sq_mtx_def open_sq_mtx_def dist_sq_mtx_def uniformity_sq_mtx_def
prefer 10
apply(transfer, simp add: norm_sq_mtx_def op_norm_triangle)
prefer 9
apply(simp_all add: norm_sq_mtx_def zero_sq_mtx_def op_norm_eq_0)
by (transfer, simp add: norm_sq_mtx_def op_norm_scaleR algebra_simps)+
end
lemma sq_mtx_scaleR_eq: "c *⇩R A = to_mtx (χ i j. c *⇩R A $$ i $ j)"
by transfer (simp add: vec_eq_iff)
lemma scaleR_to_mtx_ith[simp]: "c *⇩R (to_mtx A) $$ i1 $ i2 = c * A $ i1 $ i2"
by transfer (simp add: scaleR_vec_def)
lemma sq_mtx_scaleR_ith[simp]: "(c *⇩R A) $$ i = (c *⇩R (A $$ i))"
by (unfold scaleR_sq_mtx_def, transfer, simp)
lemma scaleR_sq_mtx_diag: "c *⇩R sq_mtx_diag f = (𝖽𝗂𝖺𝗀 i. c * f i)"
by (subst sq_mtx_eq_iff, simp add: axis_def)
lemma scaleR_mtx_vec_assoc: "(c *⇩R A) *⇩V x = c *⇩R (A *⇩V x)"
unfolding scaleR_sq_mtx_def sq_mtx_vec_mult_def apply simp
by (simp add: scaleR_matrix_vector_assoc)
lemma mtx_vec_scaleR_commute: "A *⇩V (c *⇩R x) = c *⇩R (A *⇩V x)"
unfolding scaleR_sq_mtx_def sq_mtx_vec_mult_def apply(simp, transfer)
by (simp add: vector_scaleR_commute)
lemma mtx_times_scaleR_commute: "A * (c *⇩R B) = c *⇩R (A * B)" for A::"('n::finite) sq_mtx"
unfolding sq_mtx_scaleR_eq sq_mtx_times_eq
apply(simp add: to_mtx_inject)
apply(simp add: vec_eq_iff fun_eq_iff)
by (simp add: semiring_normalization_rules(19) vector_space_over_itself.scale_sum_right)
lemma le_mtx_norm: "m ∈ {∥A *⇩V x∥ |x. ∥x∥ = 1} ⟹ m ≤ ∥A∥"
using cSup_upper[of _ "{∥(to_vec A) *v x∥ | x. ∥x∥ = 1}"]
by (simp add: op_norm_set_proptys(2) op_norm_def norm_sq_mtx_def sq_mtx_vec_mult.rep_eq)
lemma norm_vec_mult_le: "∥A *⇩V x∥ ≤ (∥A∥) * (∥x∥)"
by (simp add: norm_matrix_le_mult_op_norm norm_sq_mtx_def sq_mtx_vec_mult.rep_eq)
lemma bounded_bilinear_sq_mtx_vec_mult: "bounded_bilinear (λA s. A *⇩V s)"
apply (rule bounded_bilinear.intro, simp_all add: mtx_vec_mult_add_rdistr
mtx_vec_mult_add_rdistl scaleR_mtx_vec_assoc mtx_vec_scaleR_commute)
by (rule_tac x=1 in exI, auto intro!: norm_vec_mult_le)
lemma norm_sq_mtx_def2: "∥A∥ = Sup {∥A *⇩V x∥ |x. ∥x∥ = 1}"
unfolding norm_sq_mtx_def op_norm_def sq_mtx_vec_mult_def by simp
lemma norm_sq_mtx_def3: "∥A∥ = (SUP x. (∥A *⇩V x∥) / (∥x∥))"
unfolding norm_sq_mtx_def onorm_def sq_mtx_vec_mult_def by simp
lemma norm_sq_mtx_diag: "∥sq_mtx_diag f∥ = Max {¦f i¦ |i. i ∈ UNIV}"
unfolding norm_sq_mtx_def apply transfer
by (rule op_norm_diag_mat_eq)
lemma sq_mtx_norm_le_sum_col: "∥A∥ ≤ (∑i∈UNIV. ∥𝖼𝗈𝗅 i A∥)"
using op_norm_le_sum_column[of "to_vec A"]
apply(simp add: norm_sq_mtx_def)
by(transfer, simp add: op_norm_le_sum_column)
lemma norm_le_transpose: "∥A∥ ≤ ∥A⇧†∥"
unfolding norm_sq_mtx_def by transfer (rule op_norm_le_transpose)
lemma norm_eq_norm_transpose[simp]: "∥A⇧†∥ = ∥A∥"
using norm_le_transpose[of A] and norm_le_transpose[of "A⇧†"] by simp
lemma norm_column_le_norm: "∥A $$ i∥ ≤ ∥A∥"
using norm_vec_mult_le[of "A⇧†" "𝖾 i"] by simp
subsection ‹ Real normed algebra of square matrices ›
instantiation sq_mtx :: (finite) real_normed_algebra_1
begin
lift_definition one_sq_mtx :: "'a sq_mtx" is "to_mtx (mat 1)" .
lemma sq_mtx_one_idty: "1 * A = A" "A * 1 = A" for A :: "'a sq_mtx"
by(transfer, transfer, unfold mat_def matrix_matrix_mult_def, simp add: vec_eq_iff)+
lemma sq_mtx_norm_1: "∥(1::'a sq_mtx)∥ = 1"
unfolding one_sq_mtx_def norm_sq_mtx_def
apply(simp add: op_norm_def)
apply(subst cSup_eq[of _ 1])
using ex_norm_eq_1 by auto
lemma sq_mtx_norm_times: "∥A * B∥ ≤ (∥A∥) * (∥B∥)" for A :: "'a sq_mtx"
unfolding norm_sq_mtx_def times_sq_mtx_def by(simp add: op_norm_matrix_matrix_mult_le)
instance
apply intro_classes
apply(simp_all add: sq_mtx_one_idty sq_mtx_norm_1 sq_mtx_norm_times)
apply(simp_all add: to_mtx_inject vec_eq_iff one_sq_mtx_def zero_sq_mtx_def mat_def)
by(transfer, simp add: scalar_matrix_assoc matrix_scalar_ac)+
end
lemma sq_mtx_one_ith_simps[simp]: "1 $$ i $ i = 1" "i ≠ j ⟹ 1 $$ i $ j = 0"
unfolding one_sq_mtx_def mat_def by simp_all
lemma of_nat_eq_sq_mtx_diag[simp]: "of_nat m = (𝖽𝗂𝖺𝗀 i. m)"
by (induct m) (simp, subst sq_mtx_eq_iff, simp add: axis_def)+
lemma mtx_vec_mult_1[simp]: "1 *⇩V s = s"
by (auto simp: sq_mtx_vec_mult_def one_sq_mtx_def
mat_def vec_eq_iff matrix_vector_mult_def)
lemma sq_mtx_diag_one[simp]: "(𝖽𝗂𝖺𝗀 i. 1) = 1"
by (subst sq_mtx_eq_iff, simp add: one_sq_mtx_def mat_def axis_def)
abbreviation "mtx_invertible A ≡ invertible (to_vec A)"
lemma mtx_invertible_def: "mtx_invertible A ⟷ (∃A'. A' * A = 1 ∧ A * A' = 1)"
apply (unfold sq_mtx_inv_def times_sq_mtx_def one_sq_mtx_def invertible_def, clarsimp, safe)
apply(rule_tac x="to_mtx A'" in exI, simp)
by (rule_tac x="to_vec A'" in exI, simp add: to_mtx_inject)
lemma mtx_invertibleI:
assumes "A * B = 1" and "B * A = 1"
shows "mtx_invertible A"
using assms unfolding mtx_invertible_def by auto
lemma mtx_invertibleD[simp]:
assumes "mtx_invertible A"
shows "A⇧-⇧1 * A = 1" and "A * A⇧-⇧1 = 1"
apply (unfold sq_mtx_inv_def times_sq_mtx_def one_sq_mtx_def)
using assms by simp_all
lemma mtx_invertible_inv[simp]: "mtx_invertible A ⟹ mtx_invertible (A⇧-⇧1)"
using mtx_invertibleD mtx_invertibleI by blast
lemma mtx_invertible_one[simp]: "mtx_invertible 1"
by (simp add: one_sq_mtx.rep_eq)
lemma sq_mtx_inv_unique:
assumes "A * B = 1" and "B * A = 1"
shows "A⇧-⇧1 = B"
by (metis (no_types, lifting) assms mtx_invertibleD(2)
mtx_invertibleI mult.assoc sq_mtx_one_idty(1))
lemma sq_mtx_inv_idempotent[simp]: "mtx_invertible A ⟹ A⇧-⇧1⇧-⇧1 = A"
using mtx_invertibleD sq_mtx_inv_unique by blast
lemma sq_mtx_inv_mult:
assumes "mtx_invertible A" and "mtx_invertible B"
shows "(A * B)⇧-⇧1 = B⇧-⇧1 * A⇧-⇧1"
by (simp add: assms matrix_inv_matrix_mul sq_mtx_inv_def times_sq_mtx_def)
lemma sq_mtx_inv_one[simp]: "1⇧-⇧1 = 1"
by (simp add: sq_mtx_inv_unique)
definition similar_sq_mtx :: "('n::finite) sq_mtx ⇒ 'n sq_mtx ⇒ bool" (infixr "∼" 25)
where "(A ∼ B) ⟷ (∃ P. mtx_invertible P ∧ A = P⇧-⇧1 * B * P)"
lemma similar_sq_mtx_matrix: "(A ∼ B) = similar_matrix (to_vec A) (to_vec B)"
apply(unfold similar_matrix_def similar_sq_mtx_def, safe)
apply (metis sq_mtx_inv.rep_eq times_sq_mtx.rep_eq)
by (metis UNIV_I sq_mtx_inv.abs_eq times_sq_mtx.abs_eq to_mtx_inverse to_vec_inverse)
lemma similar_sq_mtx_refl[simp]: "A ∼ A"
by (unfold similar_sq_mtx_def, rule_tac x="1" in exI, simp)
lemma similar_sq_mtx_simm: "A ∼ B ⟹ B ∼ A"
apply(unfold similar_sq_mtx_def, clarsimp)
apply(rule_tac x="P⇧-⇧1" in exI, simp add: mult.assoc)
by (metis mtx_invertibleD(2) mult.assoc mult.left_neutral)
lemma similar_sq_mtx_trans: "A ∼ B ⟹ B ∼ C ⟹ A ∼ C"
unfolding similar_sq_mtx_matrix using similar_matrix_trans by blast
lemma power_sq_mtx_diag: "(sq_mtx_diag f)^n = (𝖽𝗂𝖺𝗀 i. f i^n)"
by (induct n, simp_all)
lemma power_similiar_sq_mtx_diag_eq:
assumes "mtx_invertible P"
and "A = P⇧-⇧1 * (sq_mtx_diag f) * P"
shows "A^n = P⇧-⇧1 * (𝖽𝗂𝖺𝗀 i. f i^n) * P"
proof(induct n, simp_all add: assms)
fix n::nat
have "P⇧-⇧1 * sq_mtx_diag f * P * (P⇧-⇧1 * (𝖽𝗂𝖺𝗀 i. f i ^ n) * P) =
P⇧-⇧1 * sq_mtx_diag f * (𝖽𝗂𝖺𝗀 i. f i ^ n) * P"
by (metis (no_types, lifting) assms(1) mtx_invertibleD(2) mult.assoc mult.right_neutral)
also have "... = P⇧-⇧1 * (𝖽𝗂𝖺𝗀 i. f i * f i ^ n) * P"
by (simp add: mult.assoc)
finally show "P⇧-⇧1 * sq_mtx_diag f * P * (P⇧-⇧1 * (𝖽𝗂𝖺𝗀 i. f i ^ n) * P) =
P⇧-⇧1 * (𝖽𝗂𝖺𝗀 i. f i * f i ^ n) * P" .
qed
lemma power_similar_sq_mtx_diag:
assumes "A ∼ (sq_mtx_diag f)"
shows "A^n ∼ (𝖽𝗂𝖺𝗀 i. f i^n)"
using assms power_similiar_sq_mtx_diag_eq
unfolding similar_sq_mtx_def by blast
subsection ‹ Banach space of square matrices ›
lemma Cauchy_cols:
fixes X :: "nat ⇒ ('a::finite) sq_mtx"
assumes "Cauchy X"
shows "Cauchy (λn. 𝖼𝗈𝗅 i (X n))"
proof(unfold Cauchy_def dist_norm, clarsimp)
fix ε::real assume "ε > 0"
then obtain M where M_def:"∀m≥M. ∀n≥M. ∥X m - X n∥ < ε"
using ‹Cauchy X› unfolding Cauchy_def by(simp add: dist_sq_mtx_def) metis
{fix m n assume "m ≥ M" and "n ≥ M"
hence "ε > ∥X m - X n∥"
using M_def by blast
moreover have "∥X m - X n∥ ≥ ∥(X m - X n) *⇩V 𝖾 i∥"
by(rule le_mtx_norm[of _ "X m - X n"], force)
moreover have "∥(X m - X n) *⇩V 𝖾 i∥ = ∥X m *⇩V 𝖾 i - X n *⇩V 𝖾 i∥"
by (simp add: mtx_vec_mult_minus_rdistrib)
moreover have "... = ∥𝖼𝗈𝗅 i (X m) - 𝖼𝗈𝗅 i (X n)∥"
by (simp add: mtx_vec_mult_minus_rdistrib mtx_vec_mult_canon)
ultimately have "∥𝖼𝗈𝗅 i (X m) - 𝖼𝗈𝗅 i (X n)∥ < ε"
by linarith}
thus "∃M. ∀m≥M. ∀n≥M. ∥𝖼𝗈𝗅 i (X m) - 𝖼𝗈𝗅 i (X n)∥ < ε"
by blast
qed
lemma col_convergence:
assumes "∀i. (λn. 𝖼𝗈𝗅 i (X n)) ⇢ L $ i"
shows "X ⇢ to_mtx (transpose L)"
proof(unfold LIMSEQ_def dist_norm, clarsimp)
let ?L = "to_mtx (transpose L)"
let ?a = "CARD('a)" fix ε::real assume "ε > 0"
hence "ε / ?a > 0" by simp
hence "∀i. ∃ N. ∀n≥N. ∥𝖼𝗈𝗅 i (X n) - L $ i∥ < ε/?a"
using assms unfolding LIMSEQ_def dist_norm convergent_def by blast
then obtain N where "∀i. ∀n≥N. ∥𝖼𝗈𝗅 i (X n) - L $ i∥ < ε/?a"
using finite_nat_minimal_witness[of "λ i n. ∥𝖼𝗈𝗅 i (X n) - L $ i∥ < ε/?a"] by blast
also have "⋀i n. (𝖼𝗈𝗅 i (X n) - L $ i) = (𝖼𝗈𝗅 i (X n - ?L))"
unfolding minus_sq_mtx_def by(transfer, simp add: transpose_def vec_eq_iff column_def)
ultimately have N_def:"∀i. ∀n≥N. ∥𝖼𝗈𝗅 i (X n - ?L)∥ < ε/?a"
by auto
have "∀n≥N. ∥X n - ?L∥ < ε"
proof(rule allI, rule impI)
fix n::nat assume "N ≤ n"
hence "∀ i. ∥𝖼𝗈𝗅 i (X n - ?L)∥ < ε/?a"
using N_def by blast
hence "(∑i∈UNIV. ∥𝖼𝗈𝗅 i (X n - ?L)∥) < (∑(i::'a)∈UNIV. ε/?a)"
using sum_strict_mono[of _ "λi. ∥𝖼𝗈𝗅 i (X n - ?L)∥"] by force
moreover have "∥X n - ?L∥ ≤ (∑i∈UNIV. ∥𝖼𝗈𝗅 i (X n - ?L)∥)"
using sq_mtx_norm_le_sum_col by blast
moreover have "(∑(i::'a)∈UNIV. ε/?a) = ε"
by force
ultimately show "∥X n - ?L∥ < ε"
by linarith
qed
thus "∃no. ∀n≥no. ∥X n - ?L∥ < ε"
by blast
qed
instance sq_mtx :: (finite) banach
proof(standard)
fix X :: "nat ⇒ 'a sq_mtx"
assume "Cauchy X"
hence "⋀i. Cauchy (λn. 𝖼𝗈𝗅 i (X n))"
using Cauchy_cols by blast
hence obs: "∀i. ∃! L. (λn. 𝖼𝗈𝗅 i (X n)) ⇢ L"
using Cauchy_convergent convergent_def LIMSEQ_unique by fastforce
define L where "L = (χ i. lim (λn. 𝖼𝗈𝗅 i (X n)))"
hence "∀i. (λn. 𝖼𝗈𝗅 i (X n)) ⇢ L $ i"
using obs theI_unique[of "λL. (λn. 𝖼𝗈𝗅 _ (X n)) ⇢ L" "L $ _"] by (simp add: lim_def)
thus "convergent X"
using col_convergence unfolding convergent_def by blast
qed
lemma exp_similiar_sq_mtx_diag_eq:
assumes "mtx_invertible P"
and "A = P⇧-⇧1 * (𝖽𝗂𝖺𝗀 i. f i) * P"
shows "exp A = P⇧-⇧1 * exp (𝖽𝗂𝖺𝗀 i. f i) * P"
proof(unfold exp_def power_similiar_sq_mtx_diag_eq[OF assms])
have "(∑n. P⇧-⇧1 * (𝖽𝗂𝖺𝗀 i. f i ^ n) * P /⇩R fact n) =
(∑n. P⇧-⇧1 * ((𝖽𝗂𝖺𝗀 i. f i ^ n) /⇩R fact n) * P)"
by simp
also have "... = (∑n. P⇧-⇧1 * ((𝖽𝗂𝖺𝗀 i. f i ^ n) /⇩R fact n)) * P"
apply(subst suminf_multr[OF bounded_linear.summable[OF bounded_linear_mult_right]])
unfolding power_sq_mtx_diag[symmetric] by (simp_all add: summable_exp_generic)
also have "... = P⇧-⇧1 * (∑n. (𝖽𝗂𝖺𝗀 i. f i ^ n) /⇩R fact n) * P"
apply(subst suminf_mult[of _ "P⇧-⇧1"])
unfolding power_sq_mtx_diag[symmetric]
by (simp_all add: summable_exp_generic)
finally show "(∑n. P⇧-⇧1 * (𝖽𝗂𝖺𝗀 i. f i ^ n) * P /⇩R fact n) =
P⇧-⇧1 * (∑n. sq_mtx_diag f ^ n /⇩R fact n) * P"
unfolding power_sq_mtx_diag by simp
qed
lemma exp_similiar_sq_mtx_diag:
assumes "A ∼ sq_mtx_diag f"
shows "exp A ∼ exp (sq_mtx_diag f)"
using assms exp_similiar_sq_mtx_diag_eq
unfolding similar_sq_mtx_def by blast
lemma suminf_sq_mtx_diag:
assumes "∀i. (λn. f n i) sums (suminf (λn. f n i))"
shows "(∑n. (𝖽𝗂𝖺𝗀 i. f n i)) = (𝖽𝗂𝖺𝗀 i. ∑n. f n i)"
proof(rule suminfI, unfold sums_def LIMSEQ_iff, clarsimp simp: norm_sq_mtx_diag)
let ?g = "λn i. ¦(∑n<n. f n i) - (∑n. f n i)¦"
fix r::real assume "r > 0"
have "∀i. ∃no. ∀n≥no. ?g n i < r"
using assms ‹r > 0› unfolding sums_def LIMSEQ_iff by clarsimp
then obtain N where key: "∀i. ∀n≥N. ?g n i < r"
using finite_nat_minimal_witness[of "λi n. ?g n i < r"] by blast
{fix n::nat
assume "n ≥ N"
obtain i where i_def: "Max {x. ∃i. x = ?g n i} = ?g n i"
using cMax_finite_ex[of "{x. ∃i. x = ?g n i}"] by auto
hence "?g n i < r"
using key ‹n ≥ N› by blast
hence "Max {x. ∃i. x = ?g n i} < r"
unfolding i_def[symmetric] .}
thus "∃N. ∀n≥N. Max {x. ∃i. x = ?g n i} < r"
by blast
qed
lemma exp_sq_mtx_diag: "exp (sq_mtx_diag f) = (𝖽𝗂𝖺𝗀 i. exp (f i))"
apply(unfold exp_def, simp add: power_sq_mtx_diag scaleR_sq_mtx_diag)
apply(rule suminf_sq_mtx_diag)
using exp_converges[of "f _"]
unfolding sums_def LIMSEQ_iff exp_def by force
lemma exp_scaleR_diagonal1:
assumes "mtx_invertible P" and "A = P⇧-⇧1 * (𝖽𝗂𝖺𝗀 i. f i) * P"
shows "exp (t *⇩R A) = P⇧-⇧1 * (𝖽𝗂𝖺𝗀 i. exp (t * f i)) * P"
proof-
have "exp (t *⇩R A) = exp (P⇧-⇧1 * (t *⇩R sq_mtx_diag f) * P)"
using assms by simp
also have "... = P⇧-⇧1 * (𝖽𝗂𝖺𝗀 i. exp (t * f i)) * P"
by (metis assms(1) exp_similiar_sq_mtx_diag_eq exp_sq_mtx_diag scaleR_sq_mtx_diag)
finally show "exp (t *⇩R A) = P⇧-⇧1 * (𝖽𝗂𝖺𝗀 i. exp (t * f i)) * P" .
qed
lemma exp_scaleR_diagonal2:
assumes "mtx_invertible P" and "A = P * (𝖽𝗂𝖺𝗀 i. f i) * P⇧-⇧1"
shows "exp (t *⇩R A) = P * (𝖽𝗂𝖺𝗀 i. exp (t * f i)) * P⇧-⇧1"
apply(subst sq_mtx_inv_idempotent[OF assms(1), symmetric])
apply(rule exp_scaleR_diagonal1)
by (simp_all add: assms)
subsection ‹ Examples ›
definition "mtx A = to_mtx (vector (map vector A))"
lemma vector_nth_eq: "(vector A) $ i = foldr (λx f n. (f (n + 1))(n := x)) A (λn x. 0) 1 i"
unfolding vector_def by simp
lemma mtx_ith_eq[simp]: "mtx A $$ i $ j = foldr (λx f n. (f (n + 1))(n := x))
(map (λl. vec_lambda (foldr (λx f n. (f (n + 1))(n := x)) l (λn x. 0) 1)) A) (λn x. 0) 1 i $ j"
unfolding mtx_def vector_def by (simp add: vector_nth_eq)
subsubsection ‹ 2x2 matrices ›
lemma mtx2_eq_iff: "(mtx
([a1, b1] #
[c1, d1] # []) :: 2 sq_mtx) = mtx
([a2, b2] #
[c2, d2] # []) ⟷ a1 = a2 ∧ b1 = b2 ∧ c1 = c2 ∧ d1 = d2"
apply(simp add: sq_mtx_eq_iff, safe)
using exhaust_2 by force+
lemma mtx2_to_mtx: "mtx
([a, b] #
[c, d] # []) =
to_mtx (χ i j::2. if i=1 ∧ j=1 then a
else (if i=1 ∧ j=2 then b
else (if i=2 ∧ j=1 then c
else d)))"
apply(subst sq_mtx_eq_iff)
using exhaust_2 by force
abbreviation diag2 :: "real ⇒ real ⇒ 2 sq_mtx"
where "diag2 ι⇩1 ι⇩2 ≡ mtx
([ι⇩1, 0] #
[0, ι⇩2] # [])"
lemma diag2_eq: "diag2 (ι 1) (ι 2) = (𝖽𝗂𝖺𝗀 i. ι i)"
apply(simp add: sq_mtx_eq_iff)
using exhaust_2 by (force simp: axis_def)
lemma one_mtx2: "(1::2 sq_mtx) = diag2 1 1"
apply(subst sq_mtx_eq_iff)
using exhaust_2 by force
lemma zero_mtx2: "(0::2 sq_mtx) = diag2 0 0"
by (simp add: sq_mtx_eq_iff)
lemma scaleR_mtx2: "k *⇩R mtx
([a, b] #
[c, d] # []) = mtx
([k*a, k*b] #
[k*c, k*d] # [])"
by (simp add: sq_mtx_eq_iff)
lemma uminus_mtx2: "-mtx
([a, b] #
[c, d] # []) = (mtx
([-a, -b] #
[-c, -d] # [])::2 sq_mtx)"
by (simp add: sq_mtx_uminus_eq sq_mtx_eq_iff)
lemma plus_mtx2: "mtx
([a1, b1] #
[c1, d1] # []) + mtx
([a2, b2] #
[c2, d2] # []) = ((mtx
([a1+a2, b1+b2] #
[c1+c2, d1+d2] # []))::2 sq_mtx)"
by (simp add: sq_mtx_eq_iff)
lemma minus_mtx2: "mtx
([a1, b1] #
[c1, d1] # []) - mtx
([a2, b2] #
[c2, d2] # []) = ((mtx
([a1-a2, b1-b2] #
[c1-c2, d1-d2] # []))::2 sq_mtx)"
by (simp add: sq_mtx_eq_iff)
lemma times_mtx2: "mtx
([a1, b1] #
[c1, d1] # []) * mtx
([a2, b2] #
[c2, d2] # []) = ((mtx
([a1*a2+b1*c2, a1*b2+b1*d2] #
[c1*a2+d1*c2, c1*b2+d1*d2] # []))::2 sq_mtx)"
unfolding sq_mtx_times_eq UNIV_2
by (simp add: sq_mtx_eq_iff)
subsubsection ‹ 3x3 matrices ›
lemma mtx3_to_mtx: "mtx
([a⇩1⇩1, a⇩1⇩2, a⇩1⇩3] #
[a⇩2⇩1, a⇩2⇩2, a⇩2⇩3] #
[a⇩3⇩1, a⇩3⇩2, a⇩3⇩3] # []) =
to_mtx (χ i j::3. if i=1 ∧ j=1 then a⇩1⇩1
else (if i=1 ∧ j=2 then a⇩1⇩2
else (if i=1 ∧ j=3 then a⇩1⇩3
else (if i=2 ∧ j=1 then a⇩2⇩1
else (if i=2 ∧ j=2 then a⇩2⇩2
else (if i=2 ∧ j=3 then a⇩2⇩3
else (if i=3 ∧ j=1 then a⇩3⇩1
else (if i=3 ∧ j=2 then a⇩3⇩2
else a⇩3⇩3))))))))"
apply(simp add: sq_mtx_eq_iff)
using exhaust_3 by force
abbreviation diag3 :: "real ⇒ real ⇒ real ⇒ 3 sq_mtx"
where "diag3 ι⇩1 ι⇩2 ι⇩3 ≡ mtx
([ι⇩1, 0, 0] #
[0, ι⇩2, 0] #
[0, 0, ι⇩3] # [])"
lemma diag3_eq: "diag3 (ι 1) (ι 2) (ι 3) = (𝖽𝗂𝖺𝗀 i. ι i)"
apply(simp add: sq_mtx_eq_iff)
using exhaust_3 by (force simp: axis_def)
lemma one_mtx3: "(1::3 sq_mtx) = diag3 1 1 1"
apply(subst sq_mtx_eq_iff)
using exhaust_3 by force
lemma zero_mtx3: "(0::3 sq_mtx) = diag3 0 0 0"
by (simp add: sq_mtx_eq_iff)
lemma scaleR_mtx3: "k *⇩R mtx
([a⇩1⇩1, a⇩1⇩2, a⇩1⇩3] #
[a⇩2⇩1, a⇩2⇩2, a⇩2⇩3] #
[a⇩3⇩1, a⇩3⇩2, a⇩3⇩3] # []) = mtx
([k*a⇩1⇩1, k*a⇩1⇩2, k*a⇩1⇩3] #
[k*a⇩2⇩1, k*a⇩2⇩2, k*a⇩2⇩3] #
[k*a⇩3⇩1, k*a⇩3⇩2, k*a⇩3⇩3] # [])"
by (simp add: sq_mtx_eq_iff)
lemma plus_mtx3: "mtx
([a⇩1⇩1, a⇩1⇩2, a⇩1⇩3] #
[a⇩2⇩1, a⇩2⇩2, a⇩2⇩3] #
[a⇩3⇩1, a⇩3⇩2, a⇩3⇩3] # []) + mtx
([b⇩1⇩1, b⇩1⇩2, b⇩1⇩3] #
[b⇩2⇩1, b⇩2⇩2, b⇩2⇩3] #
[b⇩3⇩1, b⇩3⇩2, b⇩3⇩3] # []) = (mtx
([a⇩1⇩1+b⇩1⇩1, a⇩1⇩2+b⇩1⇩2, a⇩1⇩3+b⇩1⇩3] #
[a⇩2⇩1+b⇩2⇩1, a⇩2⇩2+b⇩2⇩2, a⇩2⇩3+b⇩2⇩3] #
[a⇩3⇩1+b⇩3⇩1, a⇩3⇩2+b⇩3⇩2, a⇩3⇩3+b⇩3⇩3] # [])::3 sq_mtx)"
by (subst sq_mtx_eq_iff) simp
lemma minus_mtx3: "mtx
([a⇩1⇩1, a⇩1⇩2, a⇩1⇩3] #
[a⇩2⇩1, a⇩2⇩2, a⇩2⇩3] #
[a⇩3⇩1, a⇩3⇩2, a⇩3⇩3] # []) - mtx
([b⇩1⇩1, b⇩1⇩2, b⇩1⇩3] #
[b⇩2⇩1, b⇩2⇩2, b⇩2⇩3] #
[b⇩3⇩1, b⇩3⇩2, b⇩3⇩3] # []) = (mtx
([a⇩1⇩1-b⇩1⇩1, a⇩1⇩2-b⇩1⇩2, a⇩1⇩3-b⇩1⇩3] #
[a⇩2⇩1-b⇩2⇩1, a⇩2⇩2-b⇩2⇩2, a⇩2⇩3-b⇩2⇩3] #
[a⇩3⇩1-b⇩3⇩1, a⇩3⇩2-b⇩3⇩2, a⇩3⇩3-b⇩3⇩3] # [])::3 sq_mtx)"
by (simp add: sq_mtx_eq_iff)
lemma times_mtx3: "mtx
([a⇩1⇩1, a⇩1⇩2, a⇩1⇩3] #
[a⇩2⇩1, a⇩2⇩2, a⇩2⇩3] #
[a⇩3⇩1, a⇩3⇩2, a⇩3⇩3] # []) * mtx
([b⇩1⇩1, b⇩1⇩2, b⇩1⇩3] #
[b⇩2⇩1, b⇩2⇩2, b⇩2⇩3] #
[b⇩3⇩1, b⇩3⇩2, b⇩3⇩3] # []) = (mtx
([a⇩1⇩1*b⇩1⇩1+a⇩1⇩2*b⇩2⇩1+a⇩1⇩3*b⇩3⇩1, a⇩1⇩1*b⇩1⇩2+a⇩1⇩2*b⇩2⇩2+a⇩1⇩3*b⇩3⇩2, a⇩1⇩1*b⇩1⇩3+a⇩1⇩2*b⇩2⇩3+a⇩1⇩3*b⇩3⇩3] #
[a⇩2⇩1*b⇩1⇩1+a⇩2⇩2*b⇩2⇩1+a⇩2⇩3*b⇩3⇩1, a⇩2⇩1*b⇩1⇩2+a⇩2⇩2*b⇩2⇩2+a⇩2⇩3*b⇩3⇩2, a⇩2⇩1*b⇩1⇩3+a⇩2⇩2*b⇩2⇩3+a⇩2⇩3*b⇩3⇩3] #
[a⇩3⇩1*b⇩1⇩1+a⇩3⇩2*b⇩2⇩1+a⇩3⇩3*b⇩3⇩1, a⇩3⇩1*b⇩1⇩2+a⇩3⇩2*b⇩2⇩2+a⇩3⇩3*b⇩3⇩2, a⇩3⇩1*b⇩1⇩3+a⇩3⇩2*b⇩2⇩3+a⇩3⇩3*b⇩3⇩3] # [])::3 sq_mtx)"
unfolding sq_mtx_times_eq
unfolding UNIV_3 by (simp add: sq_mtx_eq_iff)
end