Theory More_Tensor
section ‹Further Results on Tensor Products›
theory More_Tensor
imports
Quantum
Tensor
Jordan_Normal_Form.Matrix
Basics
begin
lemma tensor_prod_2 [simp]:
"mult.vec_vec_Tensor (*) [x1::complex,x2] [x3, x4] = [x1 * x3, x1 * x4, x2 * x3, x2 * x4]"
proof -
have "Matrix_Tensor.mult (1::complex) (*)"
by (simp add: Matrix_Tensor.mult_def)
thus "mult.vec_vec_Tensor (*) [x1::complex,x2] [x3,x4] = [x1*x3,x1*x4,x2*x3,x2*x4]"
using mult.vec_vec_Tensor_def[of "(1::complex)" "(*)"] mult.times_def[of "(1::complex)" "(*)"] by simp
qed
lemma list_vec [simp]:
assumes "v ∈ state_qbit 1"
shows "list_of_vec v = [v $ 0, v $ 1]"
proof -
have "Rep_vec v = (fst(Rep_vec v), snd(Rep_vec v))" by simp
also have "… = (2, vec_index v)"
by (metis (mono_tags, lifting) assms dim_vec.rep_eq mem_Collect_eq power_one_right state_qbit_def vec_index.rep_eq)
moreover have "[0..<2::nat] = [0,1]"
by(simp add: upt_rec)
ultimately show ?thesis
by(simp add: list_of_vec_def)
qed
lemma vec_tensor_prod_2 [simp]:
assumes "v ∈ state_qbit 1" and "w ∈ state_qbit 1"
shows "v ⊗ w = vec_of_list [v $ 0 * w $ 0, v $ 0 * w $ 1, v $ 1 * w $ 0, v $ 1 * w $ 1]"
proof -
have "list_of_vec v = [v $ 0, v $ 1]"
using assms by simp
moreover have "list_of_vec w = [w $ 0, w $ 1]"
using assms by simp
ultimately show "v ⊗ w = vec_of_list [v $ 0 * w $ 0, v $ 0 * w $ 1, v $ 1 * w $ 0, v $ 1 * w $ 1]"
by(simp add: tensor_vec_def)
qed
lemma vec_dim_of_vec_of_list [simp]:
assumes "length l = n"
shows "dim_vec (vec_of_list l) = n"
using assms vec_of_list_def by simp
lemma vec_tensor_prod_2_bis [simp]:
assumes "v ∈ state_qbit 1" and "w ∈ state_qbit 1"
shows "v ⊗ w = Matrix.vec 4 (λi. if i = 0 then v $ 0 * w $ 0 else
if i = 3 then v $ 1 * w $ 1 else
if i = 1 then v $ 0 * w $ 1 else v $ 1 * w $ 0)"
proof
define u where "u = Matrix.vec 4 (λi. if i = 0 then v $ 0 * w $ 0 else
if i = 3 then v $ 1 * w $ 1 else
if i = 1 then v $ 0 * w $ 1 else v $ 1 * w $ 0)"
then show f2:"dim_vec (v ⊗ w) = dim_vec u"
using assms by simp
show "⋀i. i < dim_vec u ⟹ (v ⊗ w) $ i = u $ i"
apply (auto simp: u_def)
using assms apply auto[3]
apply (simp add: numeral_3_eq_3)
using u_def vec_of_list_index vec_tensor_prod_2 index_is_2
by (metis (no_types, lifting) One_nat_def assms nth_Cons_0 nth_Cons_Suc numeral_2_eq_2)
qed
lemma index_col_mat_of_cols_list [simp]:
assumes "i < n" and "j < length l"
shows "Matrix.col (mat_of_cols_list n l) j $ i = l ! j ! i"
apply (auto simp: Matrix.col_def mat_of_cols_list_def)
using assms less_le_trans by fastforce
lemma multTensor2 [simp]:
assumes a1:"A = Matrix.mat 2 1 (λ(i,j). if i = 0 then a0 else a1)" and
a2:"B = Matrix.mat 2 1 (λ(i,j). if i = 0 then b0 else b1)"
shows "mult.Tensor (*) (mat_to_cols_list A) (mat_to_cols_list B) = [[a0*b0, a0*b1, a1*b0, a1*b1]]"
proof -
have "mat_to_cols_list A = [[a0, a1]]"
by (auto simp: a1 mat_to_cols_list_def) (simp add: numeral_2_eq_2)
moreover have f2:"mat_to_cols_list B = [[b0, b1]]"
by (auto simp: a2 mat_to_cols_list_def) (simp add: numeral_2_eq_2)
ultimately have "mult.Tensor (*) (mat_to_cols_list A) (mat_to_cols_list B) =
mult.Tensor (*) [[a0,a1]] [[b0,b1]]" by simp
thus ?thesis
using mult.Tensor_def[of "(1::complex)" "(*)"] mult.times_def[of "(1::complex)" "(*)"]
by (metis (mono_tags, lifting) append_self_conv list.simps(6) mult.Tensor.simps(2) mult.vec_mat_Tensor.simps(1)
mult.vec_mat_Tensor.simps(2) plus_mult_cpx plus_mult_def tensor_prod_2)
qed
lemma multTensor2_bis [simp]:
assumes a1:"dim_row A = 2" and a2:"dim_col A = 1" and a3:"dim_row B = 2" and a4:"dim_col B = 1"
shows "mult.Tensor (*) (mat_to_cols_list A) (mat_to_cols_list B) =
[[A $$ (0,0) * B $$ (0,0), A $$ (0,0) * B $$ (1,0), A $$ (1,0) * B $$ (0,0), A $$ (1,0) * B $$ (1,0)]]"
proof -
have "mat_to_cols_list A = [[A $$ (0,0), A $$ (1,0)]]"
by (auto simp: a1 a2 mat_to_cols_list_def) (simp add: numeral_2_eq_2)
moreover have f2:"mat_to_cols_list B = [[B $$ (0,0), B $$ (1,0)]]"
by (auto simp: a3 a4 mat_to_cols_list_def) (simp add: numeral_2_eq_2)
ultimately have "mult.Tensor (*) (mat_to_cols_list A) (mat_to_cols_list B) =
mult.Tensor (*) [[A $$ (0,0), A $$ (1,0)]] [[B $$ (0,0), B $$ (1,0)]]" by simp
thus ?thesis
using mult.Tensor_def[of "(1::complex)" "(*)"] mult.times_def[of "(1::complex)" "(*)"]
by (smt append_self_conv list.simps(6) mult.Tensor.simps(2) mult.vec_mat_Tensor.simps(1)
mult.vec_mat_Tensor.simps(2) plus_mult_cpx plus_mult_def tensor_prod_2)
qed
lemma mat_tensor_prod_2_prelim [simp]:
assumes "state 1 v" and "state 1 w"
shows "v ⨂ w = mat_of_cols_list 4
[[v $$ (0,0) * w $$ (0,0), v $$ (0,0) * w $$ (1,0), v $$ (1,0) * w $$ (0,0), v $$ (1,0) * w $$ (1,0)]]"
proof
define u where "u = mat_of_cols_list 4
[[v $$ (0,0) * w $$ (0,0), v $$ (0,0) * w $$ (1,0), v $$ (1,0) * w $$ (0,0), v $$ (1,0) * w $$ (1,0)]]"
then show f1:"dim_row (v ⨂ w) = dim_row u"
using assms state_def mat_of_cols_list_def tensor_mat_def by simp
show f2:"dim_col (v ⨂ w) = dim_col u"
using assms state_def mat_of_cols_list_def tensor_mat_def u_def by simp
show "⋀i j. i < dim_row u ⟹ j < dim_col u ⟹ (v ⨂ w) $$ (i, j) = u $$ (i, j)"
using u_def tensor_mat_def assms state_def by simp
qed
lemma mat_tensor_prod_2_col [simp]:
assumes "state 1 v" and "state 1 w"
shows "Matrix.col (v ⨂ w) 0 = Matrix.col v 0 ⊗ Matrix.col w 0"
proof
show f1:"dim_vec (Matrix.col (v ⨂ w) 0) = dim_vec (Matrix.col v 0 ⊗ Matrix.col w 0)"
using assms vec_tensor_prod_2_bis
by (smt Tensor.mat_of_cols_list_def dim_col dim_row_mat(1) dim_vec mat_tensor_prod_2_prelim state.state_to_state_qbit)
next
show "⋀i. i<dim_vec (Matrix.col v 0 ⊗ Matrix.col w 0) ⟹ Matrix.col (v ⨂ w) 0 $ i = (Matrix.col v 0 ⊗ Matrix.col w 0) $ i"
proof -
have "dim_vec (Matrix.col v 0 ⊗ Matrix.col w 0) = 4"
by (metis (no_types, lifting) assms(1) assms(2) dim_vec state.state_to_state_qbit vec_tensor_prod_2_bis)
moreover have "(Matrix.col v 0 ⊗ Matrix.col w 0) $ 0 = v $$ (0,0) * w $$ (0,0)"
using assms vec_tensor_prod_2 state.state_to_state_qbit col_index_of_mat_col
by (smt nth_Cons_0 power_one_right state_def vec_of_list_index zero_less_numeral)
moreover have "… = Matrix.col (v ⨂ w) 0 $ 0"
using assms by simp
moreover have "(Matrix.col v 0 ⊗ Matrix.col w 0) $ 1 = v $$ (0,0) * w $$ (1,0)"
using assms vec_tensor_prod_2 state.state_to_state_qbit col_index_of_mat_col
by (smt One_nat_def Suc_1 lessI nth_Cons_0 power_one_right state_def vec_index_vCons_Suc
vec_of_list_Cons vec_of_list_index zero_less_numeral)
moreover have "… = Matrix.col (v ⨂ w) 0 $ 1"
using assms by simp
moreover have "(Matrix.col v 0 ⊗ Matrix.col w 0) $ 2 = v $$ (1,0) * w $$ (0,0)"
using assms vec_tensor_prod_2 state.state_to_state_qbit col_index_of_mat_col
by (smt One_nat_def Suc_1 lessI nth_Cons_0 power_one_right state_def vec_index_vCons_Suc
vec_of_list_Cons vec_of_list_index zero_less_numeral)
moreover have "… = Matrix.col (v ⨂ w) 0 $ 2"
using assms by simp
moreover have "(Matrix.col v 0 ⊗ Matrix.col w 0) $ 3 = v $$ (1,0) * w $$ (1,0)"
using assms vec_tensor_prod_2 state.state_to_state_qbit col_index_of_mat_col numeral_3_eq_3
by (smt One_nat_def Suc_1 lessI nth_Cons_0 power_one_right state_def vec_index_vCons_Suc
vec_of_list_Cons vec_of_list_index zero_less_numeral)
moreover have "… = Matrix.col (v ⨂ w) 0 $ 3"
using assms by simp
ultimately show "⋀i. i<dim_vec (Matrix.col v 0 ⊗ Matrix.col w 0) ⟹ Matrix.col (v ⨂ w) 0 $ i = (Matrix.col v 0 ⊗ Matrix.col w 0) $ i"
using index_sl_four by auto
qed
qed
lemma mat_tensor_prod_2 [simp]:
assumes "state 1 v" and "state 1 w"
shows "v ⨂ w = Matrix.mat 4 1 (λ(i,j). if i = 0 then v $$ (0,0) * w $$ (0,0) else
if i = 3 then v $$ (1,0) * w $$ (1,0) else
if i = 1 then v $$ (0,0) * w $$ (1,0) else
v $$ (1,0) * w $$ (0,0))"
proof
define u where "u = Matrix.mat 4 1 (λ(i,j). if i = 0 then v $$ (0,0) * w $$ (0,0) else
if i = 3 then v $$ (1,0) * w $$ (1,0) else
if i = 1 then v $$ (0,0) * w $$ (1,0) else
v $$ (1,0) * w $$ (0,0))"
then show "dim_row (v ⨂ w) = dim_row u"
using assms tensor_mat_def state_def by(simp add: Tensor.mat_of_cols_list_def)
also have "… = 4" by (simp add: u_def)
show "dim_col (v ⨂ w) = dim_col u"
using u_def assms tensor_mat_def state_def Tensor.mat_of_cols_list_def by simp
moreover have "… = 1" by(simp add: u_def)
ultimately show "⋀i j. i < dim_row u ⟹ j < dim_col u ⟹ (v ⨂ w) $$ (i, j) = u $$ (i,j)"
proof -
fix i j::nat
assume a1:"i < dim_row u" and a2:"j < dim_col u"
then have "(v ⨂ w) $$ (i, j) = Matrix.col (v ⨂ w) 0 $ i"
using Matrix.col_def u_def assms by simp
then have f1:"(v ⨂ w) $$ (i, j) = (Matrix.col v 0 ⊗ Matrix.col w 0) $ i"
using assms mat_tensor_prod_2_col by simp
have "(Matrix.col v 0 ⊗ Matrix.col w 0) $ i =
Matrix.vec 4 (λi. if i = 0 then Matrix.col v 0 $ 0 * Matrix.col w 0 $ 0 else
if i = 3 then Matrix.col v 0 $ 1 * Matrix.col w 0 $ 1 else
if i = 1 then Matrix.col v 0 $ 0 * Matrix.col w 0 $ 1 else
Matrix.col v 0 $ 1 * Matrix.col w 0 $ 0) $ i"
using vec_tensor_prod_2_bis assms state.state_to_state_qbit by presburger
thus "(v ⨂ w) $$ (i, j) = u $$ (i,j)"
using u_def a1 a2 assms state_def by simp
qed
qed
lemma mat_tensor_prod_2_bis:
assumes "state 1 v" and "state 1 w"
shows "v ⨂ w = |Matrix.vec 4 (λi. if i = 0 then v $$ (0,0) * w $$ (0,0) else
if i = 3 then v $$ (1,0) * w $$ (1,0) else
if i = 1 then v $$ (0,0) * w $$ (1,0) else
v $$ (1,0) * w $$ (0,0))⟩"
using assms ket_vec_def mat_tensor_prod_2 by(simp add: mat_eq_iff)
lemma eq_ket_vec:
fixes u v:: "complex Matrix.vec"
assumes "u = v"
shows "|u⟩ = |v⟩"
using assms by simp
lemma mat_tensor_ket_vec:
assumes "state 1 v" and "state 1 w"
shows "v ⨂ w = |(Matrix.col v 0) ⊗ (Matrix.col w 0)⟩"
proof -
have "v ⨂ w = |Matrix.col v 0⟩ ⨂ |Matrix.col w 0⟩"
using assms state_def by simp
also have "… =
|Matrix.vec 4 (λi. if i = 0 then |Matrix.col v 0⟩ $$ (0,0) * |Matrix.col w 0⟩ $$ (0,0) else
if i = 3 then |Matrix.col v 0⟩ $$ (1,0) * |Matrix.col w 0⟩ $$ (1,0) else
if i = 1 then |Matrix.col v 0⟩ $$ (0,0) * |Matrix.col w 0⟩ $$ (1,0) else
|Matrix.col v 0⟩ $$ (1,0) * |Matrix.col w 0⟩ $$ (0,0))⟩"
using assms mat_tensor_prod_2_bis state_col_ket_vec by simp
also have "… =
|Matrix.vec 4 (λi. if i = 0 then v $$ (0,0) * w $$ (0,0) else
if i = 3 then v $$ (1,0) * w $$ (1,0) else
if i = 1 then v $$ (0,0) * w $$ (1,0) else
v $$ (1,0) * w $$ (0,0))⟩"
using assms mat_tensor_prod_2_bis calculation by auto
also have "… =
|Matrix.vec 4 (λi. if i = 0 then Matrix.col v 0 $ 0 * Matrix.col w 0 $ 0 else
if i = 3 then Matrix.col v 0 $ 1 * Matrix.col w 0 $ 1 else
if i = 1 then Matrix.col v 0 $ 0 * Matrix.col w 0 $ 1 else
Matrix.col v 0 $ 1 * Matrix.col w 0 $ 0)⟩"
apply(rule eq_ket_vec)
apply (rule eq_vecI)
using col_index_of_mat_col assms state_def by auto
finally show ?thesis
using vec_tensor_prod_2_bis assms state.state_to_state_qbit by presburger
qed
text ‹The property of being a state (resp. a gate) is preserved by tensor product.›
lemma tensor_state2 [simp]:
assumes "state 1 u" and "state 1 v"
shows "state 2 (u ⨂ v)"
proof
show "dim_col (u ⨂ v) = 1"
using assms dim_col_tensor_mat state.is_column by presburger
show "dim_row (u ⨂ v) = 2⇧2"
using assms dim_row_tensor_mat state.dim_row
by (metis (mono_tags, lifting) power2_eq_square power_one_right)
show "∥Matrix.col (u ⨂ v) 0∥ = 1"
proof -
define l where d0:"l = [u $$ (0,0) * v $$ (0,0), u $$ (0,0) * v $$ (1,0), u $$ (1,0) * v $$ (0,0), u $$ (1,0) * v $$ (1,0)]"
then have f4:"length l = 4" by simp
also have "u ⨂ v = mat_of_cols_list 4
[[u $$ (0,0) * v $$ (0,0), u $$ (0,0) * v $$ (1,0), u $$ (1,0) * v $$ (0,0), u $$ (1,0) * v $$ (1,0)]]"
using assms by simp
then have "Matrix.col (u ⨂ v) 0 = vec_of_list [u $$ (0,0) * v $$ (0,0), u $$ (0,0) * v $$ (1,0),
u $$ (1,0) * v $$ (0,0), u $$ (1,0) * v $$ (1,0)]"
by (metis One_nat_def Suc_eq_plus1 add_Suc col_mat_of_cols_list list.size(3) list.size(4)
nth_Cons_0 numeral_2_eq_2 numeral_Bit0 plus_1_eq_Suc vec_of_list_Cons zero_less_one_class.zero_less_one)
then have f5:"∥Matrix.col (u ⨂ v) 0∥ = sqrt(∑i<4. (cmod (l ! i))⇧2)"
by (metis d0 f4 One_nat_def cpx_length_of_vec_of_list d0 vec_of_list_Cons)
also have "… = sqrt ((cmod (u $$ (0,0) * v $$ (0,0)))⇧2 + (cmod(u $$ (0,0) * v $$ (1,0)))⇧2 +
(cmod(u $$ (1,0) * v $$ (0,0)))⇧2 + (cmod(u $$ (1,0) * v $$ (1,0)))⇧2)"
proof -
have "(∑i<4. (cmod (l ! i))⇧2) = (cmod (l ! 0))⇧2 + (cmod (l ! 1))⇧2 + (cmod (l ! 2))⇧2 +
(cmod (l ! 3))⇧2"
using sum_insert
by (smt One_nat_def empty_iff finite.emptyI finite.insertI insertE lessThan_0 lessThan_Suc
numeral_2_eq_2 numeral_3_eq_3 numeral_plus_one one_plus_numeral_commute plus_1_eq_Suc semiring_norm(2)
semiring_norm(8) sum.empty)
also have "… = (cmod (u $$ (0,0) * v $$ (0,0)))⇧2 + (cmod(u $$ (0,0) * v $$ (1,0)))⇧2 +
(cmod(u $$ (1,0) * v $$ (0,0)))⇧2 + (cmod(u $$ (1,0) * v $$ (1,0)))⇧2"
using d0 by simp
finally show ?thesis by(simp add: f5)
qed
moreover have "… =
sqrt ((cmod (u $$ (0,0)))⇧2 * (cmod (v $$ (0,0)))⇧2 + (cmod(u $$ (0,0)))⇧2 * (cmod (v $$ (1,0)))⇧2 +
(cmod(u $$ (1,0)))⇧2 * (cmod (v $$ (0,0)))⇧2 + (cmod(u $$ (1,0)))⇧2 * (cmod(v $$ (1,0)))⇧2)"
by (simp add: norm_mult power_mult_distrib)
moreover have "… = sqrt ((((cmod(u $$ (0,0)))⇧2 + (cmod(u $$ (1,0)))⇧2)) *
(((cmod(v $$ (0,0)))⇧2 + (cmod(v $$ (1,0)))⇧2)))"
by (simp add: distrib_left mult.commute)
ultimately have f6:"∥Matrix.col (u ⨂ v) 0∥⇧2 = (((cmod(u $$ (0,0)))⇧2 + (cmod(u $$ (1,0)))⇧2)) *
(((cmod(v $$ (0,0)))⇧2 + (cmod(v $$ (1,0)))⇧2))"
by (simp add: f4)
also have f7:"… = (∑i< 2. (cmod (u $$ (i,0)))⇧2) * (∑i<2. (cmod (v $$ (i,0)))⇧2)"
by (simp add: numeral_2_eq_2)
also have f8:"… = (∑i< 2.(cmod (Matrix.col u 0 $ i))⇧2) * (∑i<2.(cmod (Matrix.col v 0 $ i))⇧2)"
using assms index_col state_def by simp
finally show ?thesis
proof -
have f1:"(∑i< 2.(cmod (Matrix.col u 0 $ i))⇧2) = 1"
using assms(1) state_def cpx_vec_length_def by auto
have f2:"(∑i< 2.(cmod (Matrix.col v 0 $ i))⇧2) = 1"
using assms(2) state_def cpx_vec_length_def by auto
thus ?thesis
using f1 f8 f5 f6 f7
by (simp add: ‹sqrt (∑i<4. (cmod (l ! i))⇧2) = sqrt ((cmod (u $$ (0, 0) * v $$ (0, 0)))⇧2 +
(cmod (u $$ (0, 0) * v $$ (1, 0)))⇧2 + (cmod (u $$ (1, 0) * v $$ (0, 0)))⇧2 + (cmod (u $$ (1, 0) * v $$ (1, 0)))⇧2)›)
qed
qed
qed
lemma sum_prod:
fixes f::"nat ⇒ complex" and g::"nat ⇒ complex"
shows "(∑i<a*b. f(i div b) * g(i mod b)) = (∑i<a. f(i)) * (∑j<b. g(j))"
proof (induction a)
case 0
then show ?case by simp
next
case (Suc a)
have "(∑i<(a+1)*b. f (i div b) * g (i mod b)) = (∑i<a*b. f (i div b) * g (i mod b)) +
(∑i∈{a*b..<(a+1)*b}. f (i div b) * g (i mod b))"
apply (auto simp: algebra_simps)
by (smt add.commute mult_Suc sum.lessThan_Suc sum.nat_group)
also have "… = (∑i<a. f(i)) * (∑j<b. g(j)) + (∑i∈{a*b..<(a+1)*b}. f (i div b) * g (i mod b))"
by (simp add: Suc.IH)
also have "… = (∑i<a. f(i)) * (∑j<b. g(j)) + (∑i∈{a*b..<(a+1)*b}. f (a) * g(i-a*b))" by simp
also have "… = (∑i<a. f(i)) * (∑j<b. g(j)) + f(a) * (∑i∈{a*b..<(a+1)*b}. g(i-a*b))"
by(simp add: sum_distrib_left)
also have "… = (∑i<a. f(i)) * (∑j<b. g(j)) + f(a) * (∑i∈{..<b}. g(i))"
using sum_of_index_diff[of "g" "(a*b)" "b"] by (simp add: algebra_simps)
ultimately show ?case by (simp add: semiring_normalization_rules(1))
qed
lemma tensor_state [simp]:
assumes "state m u" and "state n v"
shows "state (m + n) (u ⨂ v)"
proof
show c1:"dim_col (u ⨂ v) = 1"
using assms dim_col_tensor_mat state.is_column by presburger
show c2:"dim_row (u ⨂ v) = 2^(m + n)"
using assms dim_row_tensor_mat state.dim_row by (metis power_add)
have "(∑i<2^(m + n). (cmod (u $$ (i div 2 ^ n, 0) * v $$ (i mod 2 ^ n, 0)))⇧2) =
(∑i<2^(m + n). (cmod (u $$ (i div 2 ^ n, 0)))⇧2 * (cmod (v $$ (i mod 2 ^ n, 0)))⇧2)"
by (simp add: sqr_of_cmod_of_prod)
also have "… = (∑i<2^m. (cmod (u $$ (i, 0)))⇧2) * (∑i<2^n. (cmod (v $$ (i, 0)))⇧2)"
proof-
have "… = (∑i<2^(m + n).complex_of_real((cmod (u $$ (i div 2^n,0)))⇧2) * complex_of_real((cmod (v $$ (i mod 2^n,0)))⇧2))"
by simp
moreover have "(∑i<2^m. (cmod (u $$ (i, 0)))⇧2) = (∑i<2^m. complex_of_real ((cmod (u $$ (i,0)))⇧2))" by simp
moreover have "(∑i<2^n. (cmod (v $$ (i, 0)))⇧2) = (∑i<2^n. complex_of_real ((cmod (v $$ (i, 0)))⇧2))" by simp
ultimately show ?thesis
using sum_prod[of "λi. (cmod (u $$ (i , 0)))⇧2" "2^n" "λi. (cmod (v $$ (i , 0)))⇧2" "2^m"]
by (smt of_real_eq_iff of_real_mult power_add)
qed
also have "… = 1"
proof-
have "(∑i<2^m. (cmod (u $$ (i, 0)))⇧2) = 1"
using assms(1) state_def cpx_vec_length_def by auto
moreover have "(∑i<2^n. (cmod (v $$ (i, 0)))⇧2) = 1"
using assms(2) state_def cpx_vec_length_def by auto
ultimately show ?thesis by simp
qed
ultimately show "∥Matrix.col (u ⨂ v) 0∥ = 1"
using c1 c2 assms state_def by (auto simp add: cpx_vec_length_def)
qed
lemma dim_row_of_tensor_gate:
assumes "gate m G1" and "gate n G2"
shows "dim_row (G1 ⨂ G2) = 2^(m+n)"
using assms dim_row_tensor_mat gate.dim_row by (simp add: power_add)
lemma tensor_gate_sqr_mat:
assumes "gate m G1" and "gate n G2"
shows "square_mat (G1 ⨂ G2)"
using assms gate.square_mat by simp
lemma dim_row_of_one_mat_less_pow:
assumes "gate m G1" and "gate n G2" and "i < dim_row (1⇩m(dim_col G1 * dim_col G2))"
shows "i < 2^(m+n)"
using assms gate_def by (simp add: power_add)
lemma dim_col_of_one_mat_less_pow:
assumes "gate m G1" and "gate n G2" and "j < dim_col (1⇩m(dim_col G1 * dim_col G2))"
shows "j < 2^(m+n)"
using assms gate_def by (simp add: power_add)
lemma index_tensor_gate_unitary1:
assumes "gate m G1" and "gate n G2" and "i < dim_row (1⇩m(dim_col G1 * dim_col G2))" and
"j < dim_col (1⇩m(dim_col G1 * dim_col G2))"
shows "((G1 ⨂ G2)⇧† * (G1 ⨂ G2)) $$ (i, j) = 1⇩m(dim_col G1 * dim_col G2) $$ (i, j)"
proof-
have "⋀k. k<2^(m+n) ⟹ cnj((G1 ⨂ G2) $$ (k,i)) =
cnj(G1 $$ (k div 2^n, i div 2^n)) * cnj(G2 $$ (k mod 2^n, i mod 2^n))"
using assms(1-3) by (simp add: gate_def power_add)
moreover have "⋀k. k<2^(m+n) ⟹ (G1 ⨂ G2) $$ (k,j) =
G1 $$ (k div 2^n, j div 2^n) * G2 $$ (k mod 2^n, j mod 2^n)"
using assms(1,2,4) by (simp add: gate_def power_add)
ultimately have "⋀k. k<2^(m+n) ⟹ cnj((G1 ⨂ G2) $$ (k,i)) * ((G1 ⨂ G2) $$ (k,j)) =
cnj(G1 $$ (k div 2^n, i div 2^n)) * G1 $$ (k div 2^n, j div 2^n) *
cnj(G2 $$ (k mod 2^n, i mod 2^n)) * G2 $$ (k mod 2^n, j mod 2^n)" by simp
then have "(∑k<2^(m+n). cnj((G1 ⨂ G2) $$ (k,i)) * ((G1 ⨂ G2) $$ (k,j))) =
(∑k<2^(m+n). cnj(G1 $$ (k div 2^n, i div 2^n)) * G1 $$ (k div 2^n, j div 2^n) *
cnj(G2 $$ (k mod 2^n, i mod 2^n)) * G2 $$ (k mod 2^n, j mod 2^n))" by simp
also have "… =
(∑k<2^m. cnj(G1 $$ (k, i div 2^n)) * G1 $$ (k, j div 2^n)) *
(∑k<2^n. cnj(G2 $$ (k, i mod 2^n)) * G2 $$ (k, j mod 2^n))"
using sum_prod[of "λx. cnj(G1 $$ (x, i div 2^n)) * G1 $$ (x, j div 2^n)" "2^n"
"λx. cnj(G2 $$ (x, i mod 2^n)) * G2 $$ (x, j mod 2^n)" "2^m"]
by (metis (no_types, lifting) power_add semigroup_mult_class.mult.assoc sum.cong)
also have "((G1 ⨂ G2)⇧† * (G1 ⨂ G2)) $$ (i, j) = (∑k<2^(m+n).cnj((G1 ⨂ G2) $$ (k,i)) * ((G1 ⨂ G2) $$ (k,j)))"
using assms index_matrix_prod[of "i" "(G1 ⨂ G2)⇧†" "j" "(G1 ⨂ G2)"] dagger_def
dim_row_of_tensor_gate tensor_gate_sqr_mat by simp
ultimately have "((G1 ⨂ G2)⇧† * (G1 ⨂ G2)) $$ (i,j) =
(∑k1<2^m. cnj(G1 $$ (k1, i div 2^n)) * G1 $$ (k1, j div 2^n)) *
(∑k2<2^n. cnj(G2 $$ (k2, i mod 2^n)) * G2 $$ (k2, j mod 2^n))" by simp
moreover have "(∑k<2^m. cnj(G1 $$ (k, i div 2^n))* G1 $$ (k, j div 2^n)) = (G1⇧† * G1) $$ (i div 2^n, j div 2^n)"
using assms gate_def dagger_def index_matrix_prod[of "i div 2^n" "G1⇧†" "j div 2^n" "G1"]
by (simp add: less_mult_imp_div_less power_add)
moreover have "… = 1⇩m(2^m) $$ (i div 2^n, j div 2^n)"
using assms(1,2) gate_def unitary_def by simp
moreover have "(∑k<2^n. cnj(G2 $$ (k, i mod 2^n))* G2 $$ (k, j mod 2^n)) = (G2⇧† * G2) $$ (i mod 2^n, j mod 2^n)"
using assms(1,2) gate_def dagger_def index_matrix_prod[of "i mod 2^n" "G2⇧†" "j mod 2^n" "G2"] by simp
moreover have "… = 1⇩m(2^n) $$ (i mod 2^n, j mod 2^n)"
using assms(1,2) gate_def unitary_def by simp
ultimately have "((G1 ⨂ G2)⇧† * (G1 ⨂ G2)) $$ (i,j) = 1⇩m (2^m) $$ (i div 2^n, j div 2^n) * 1⇩m (2^n) $$ (i mod 2^n, j mod 2^n)"
by simp
thus ?thesis
using assms assms(3,4) gate_def index_one_mat_div_mod[of "i" "m" "n" "j"] by(simp add: power_add)
qed
lemma tensor_gate_unitary1 [simp]:
assumes "gate m G1" and "gate n G2"
shows "(G1 ⨂ G2)⇧† * (G1 ⨂ G2) = 1⇩m(dim_col G1 * dim_col G2)"
proof
show "dim_row ((G1 ⨂ G2)⇧† * (G1 ⨂ G2)) = dim_row (1⇩m(dim_col G1 * dim_col G2))" by simp
show "dim_col ((G1 ⨂ G2)⇧† * (G1 ⨂ G2)) = dim_col (1⇩m(dim_col G1 * dim_col G2))" by simp
fix i j assume "i < dim_row (1⇩m(dim_col G1 * dim_col G2))" and "j < dim_col (1⇩m(dim_col G1 * dim_col G2))"
thus "((G1 ⨂ G2)⇧† * (G1 ⨂ G2)) $$ (i, j) = 1⇩m(dim_col G1 * dim_col G2) $$ (i, j)"
using assms index_tensor_gate_unitary1 by simp
qed
lemma index_tensor_gate_unitary2 [simp]:
assumes "gate m G1" and "gate n G2" and "i < dim_row (1⇩m (dim_col G1 * dim_col G2))" and
"j < dim_col (1⇩m (dim_col G1 * dim_col G2))"
shows "((G1 ⨂ G2) * ((G1 ⨂ G2)⇧†)) $$ (i, j) = 1⇩m(dim_col G1 * dim_col G2) $$ (i, j)"
proof-
have "⋀k. k<2^(m+n) ⟹ (G1 ⨂ G2) $$ (i,k) =
G1 $$ (i div 2^n, k div 2^n) * G2 $$ (i mod 2^n, k mod 2^n)"
using assms(1-3) by (simp add: gate_def power_add)
moreover have "⋀k. k<2^(m+n) ⟹ cnj((G1 ⨂ G2) $$ (j,k)) =
cnj(G1 $$ (j div 2^n, k div 2^n)) * cnj(G2 $$ (j mod 2^n, k mod 2^n))"
using assms(1,2,4) by (simp add: gate_def power_add)
ultimately have "⋀k. k∈{..<2^(m+n)} ⟹ (G1 ⨂ G2) $$ (i,k) * cnj((G1 ⨂ G2) $$ (j,k)) =
G1 $$ (i div 2^n, k div 2^n) * cnj(G1 $$ (j div 2^n, k div 2^n)) *
G2 $$ (i mod 2^n, k mod 2^n) * cnj(G2 $$ (j mod 2^n, k mod 2^n))" by simp
then have "(∑k<2^(m+n). (G1 ⨂ G2) $$ (i,k) * cnj((G1 ⨂ G2) $$ (j,k))) =
(∑k<2^(m+n). G1 $$ (i div 2^n, k div 2^n) * cnj(G1 $$ (j div 2^n, k div 2^n)) *
G2 $$ (i mod 2^n, k mod 2^n) * cnj(G2 $$ (j mod 2^n, k mod 2^n)))" by simp
also have "… =
(∑k<2^m. G1 $$ (i div 2^n, k) * cnj(G1 $$ (j div 2^n, k))) *
(∑k<2^n. G2 $$ (i mod 2^n, k) * cnj(G2 $$ (j mod 2^n, k)))"
using sum_prod[of "λk. G1 $$ (i div 2^n, k) * cnj(G1 $$ (j div 2^n, k))" "2^n"
"λk. G2 $$ (i mod 2^n, k) * cnj(G2 $$ (j mod 2^n, k))" "2^m"]
by (metis (no_types, lifting) power_add semigroup_mult_class.mult.assoc sum.cong)
also have "((G1 ⨂ G2) * ((G1 ⨂ G2)⇧†)) $$ (i, j) = (∑k<2^(m+n). (G1 ⨂ G2) $$ (i,k) * cnj((G1 ⨂ G2) $$ (j,k)))"
using assms index_matrix_prod[of "i" "(G1 ⨂ G2)" "j" "(G1 ⨂ G2)⇧†"] dagger_def
dim_row_of_tensor_gate tensor_gate_sqr_mat by simp
ultimately have "((G1 ⨂ G2) * ((G1 ⨂ G2)⇧†)) $$ (i,j) =
(∑k<2^m. G1 $$ (i div 2^n, k) * cnj(G1 $$ (j div 2^n, k))) *
(∑k<2^n. G2 $$ (i mod 2^n, k) * cnj(G2 $$ (j mod 2^n, k)))" by simp
moreover have "(∑k<2^m. G1 $$ (i div 2^n, k) * cnj(G1 $$ (j div 2^n, k))) = (G1 * (G1⇧†)) $$ (i div 2^n, j div 2^n)"
using assms gate_def dagger_def index_matrix_prod[of "i div 2^n" "G1" "j div 2^n" "G1⇧†"]
by (simp add: less_mult_imp_div_less power_add)
moreover have "… = 1⇩m(2^m) $$ (i div 2^n, j div 2^n)"
using assms(1,2) gate_def unitary_def by simp
moreover have "(∑k<2^n. G2 $$ (i mod 2^n, k) * cnj(G2 $$ (j mod 2^n, k))) = (G2 * (G2⇧†)) $$ (i mod 2^n, j mod 2^n)"
using assms(1,2) gate_def dagger_def index_matrix_prod[of "i mod 2^n" "G2" "j mod 2^n" "G2⇧†"] by simp
moreover have "… = 1⇩m(2^n) $$ (i mod 2^n, j mod 2^n)"
using assms(1,2) gate_def unitary_def by simp
ultimately have "((G1 ⨂ G2) * ((G1 ⨂ G2)⇧†)) $$ (i,j) = 1⇩m(2^m) $$ (i div 2^n, j div 2^n) * 1⇩m(2^n) $$ (i mod 2^n, j mod 2^n)"
by simp
thus ?thesis
using assms gate_def index_one_mat_div_mod[of "i" "m" "n" "j"] by(simp add: power_add)
qed
lemma tensor_gate_unitary2 [simp]:
assumes "gate m G1" and "gate n G2"
shows "(G1 ⨂ G2) * ((G1 ⨂ G2)⇧†) = 1⇩m(dim_col G1 * dim_col G2)"
proof
show "dim_row ((G1 ⨂ G2) * ((G1 ⨂ G2)⇧†)) = dim_row(1⇩m (dim_col G1 * dim_col G2))"
using assms gate_def by simp
show "dim_col ((G1 ⨂ G2) * ((G1 ⨂ G2)⇧†)) = dim_col (1⇩m(dim_col G1 * dim_col G2))"
using assms gate_def by simp
fix i j assume "i < dim_row (1⇩m (dim_col G1 * dim_col G2))" and "j < dim_col (1⇩m (dim_col G1 * dim_col G2))"
thus "((G1 ⨂ G2) * ((G1 ⨂ G2)⇧†)) $$ (i, j) = 1⇩m(dim_col G1 * dim_col G2) $$ (i, j)"
using assms index_tensor_gate_unitary2 by simp
qed
lemma tensor_gate [simp]:
assumes "gate m G1" and "gate n G2"
shows "gate (m + n) (G1 ⨂ G2)"
proof
show "dim_row (G1 ⨂ G2) = 2^(m+n)"
using assms dim_row_tensor_mat gate.dim_row by (simp add: power_add)
show "square_mat (G1 ⨂ G2)"
using assms gate.square_mat by simp
thus "unitary (G1 ⨂ G2)"
using assms unitary_def by simp
qed
end