Theory Poincare_Distance
theory Poincare_Distance
imports Poincare_Lines_Ideal_Points Hyperbolic_Functions
begin
section ‹H-distance in the Poincar\'e model›
text‹Informally, the \emph{h-distance} between the two h-points is defined as the absolute value of
the logarithm of the cross ratio between those two points and the two ideal points.›
abbreviation Re_cross_ratio where "Re_cross_ratio z u v w ≡ Re (to_complex (cross_ratio z u v w))"
definition calc_poincare_distance :: "complex_homo ⇒ complex_homo ⇒ complex_homo ⇒ complex_homo ⇒ real" where
[simp]: "calc_poincare_distance u i1 v i2 = abs (ln (Re_cross_ratio u i1 v i2))"
definition poincare_distance_pred :: "complex_homo ⇒ complex_homo ⇒ real ⇒ bool" where
[simp]: "poincare_distance_pred u v d ⟷
(u = v ∧ d = 0) ∨ (u ≠ v ∧ (∀ i1 i2. ideal_points (poincare_line u v) = {i1, i2} ⟶ d = calc_poincare_distance u i1 v i2))"
definition poincare_distance :: "complex_homo ⇒ complex_homo ⇒ real" where
"poincare_distance u v = (THE d. poincare_distance_pred u v d)"
text‹We shown that the described cross-ratio is always finite,
positive real number.›
lemma distance_cross_ratio_real_positive:
assumes "u ∈ unit_disc" and "v ∈ unit_disc" and "u ≠ v"
shows "∀ i1 i2. ideal_points (poincare_line u v) = {i1, i2} ⟶
cross_ratio u i1 v i2 ≠ ∞⇩h ∧ is_real (to_complex (cross_ratio u i1 v i2)) ∧ Re_cross_ratio u i1 v i2 > 0" (is "?P u v")
proof (rule wlog_positive_x_axis[OF assms])
fix x
assume *: "is_real x" "0 < Re x" "Re x < 1"
hence "x ≠ -1" "x ≠ 1"
by auto
hence **: "of_complex x ≠ ∞⇩h" "of_complex x ≠ 0⇩h" "of_complex x ≠ of_complex (-1)" "of_complex 1 ≠ of_complex x"
"of_complex x ∈ circline_set x_axis"
using *
unfolding circline_set_x_axis
by (auto simp add: of_complex_inj)
have ***: "0⇩h ≠ of_complex (-1)" "0⇩h ≠ of_complex 1"
by (metis of_complex_zero_iff zero_neq_neg_one, simp)
have ****: "- x - 1 ≠ 0" "x - 1 ≠ 0"
using ‹x ≠ -1› ‹x ≠ 1›
by (metis add.inverse_inverse eq_iff_diff_eq_0, simp)
have "poincare_line 0⇩h (of_complex x) = x_axis"
using **
by (simp add: poincare_line_0_real_is_x_axis)
thus "?P 0⇩h (of_complex x)"
using * ** *** ****
using cross_ratio_not_inf[of "0⇩h" "of_complex 1" "of_complex (-1)" "of_complex x"]
using cross_ratio_not_inf[of "0⇩h" "of_complex (-1)" "of_complex 1" "of_complex x"]
using cross_ratio_real[of 0 "-1" x 1] cross_ratio_real[of 0 1 x "-1"]
apply (auto simp add: poincare_line_0_real_is_x_axis doubleton_eq_iff circline_set_x_axis)
apply (subst cross_ratio, simp_all, subst Re_complex_div_gt_0, simp, subst mult_neg_neg, simp_all)+
done
next
fix M u v
let ?Mu = "moebius_pt M u" and ?Mv = "moebius_pt M v"
assume *: "unit_disc_fix M" "u ∈ unit_disc" "v ∈ unit_disc" "u ≠ v"
"?P ?Mu ?Mv"
show "?P u v"
proof safe
fix i1 i2
let ?cr = "cross_ratio u i1 v i2"
assume **: "ideal_points (poincare_line u v) = {i1, i2}"
have "i1 ≠ u" "i1 ≠ v" "i2 ≠ u" "i2 ≠ v" "i1 ≠ i2"
using ideal_points_different[OF *(2-3), of i1 i2] ** ‹u ≠ v›
by auto
hence "0 < Re (to_complex ?cr) ∧ is_real (to_complex ?cr) ∧ ?cr ≠ ∞⇩h"
using * **
apply (erule_tac x="moebius_pt M i1" in allE)
apply (erule_tac x="moebius_pt M i2" in allE)
apply (subst (asm) ideal_points_poincare_line_moebius[of M u v i1 i2], simp_all)
done
thus "0 < Re (to_complex ?cr)" "is_real (to_complex ?cr)" "?cr = ∞⇩h ⟹ False"
by simp_all
qed
qed
text‹Next we can show that for every different points from the unit disc there is exactly one number
that satisfies the h-distance predicate.›
lemma distance_unique:
assumes "u ∈ unit_disc" and "v ∈ unit_disc"
shows "∃! d. poincare_distance_pred u v d"
proof (cases "u = v")
case True
thus ?thesis
by auto
next
case False
obtain i1 i2 where *: "i1 ≠ i2" "ideal_points (poincare_line u v) = {i1, i2}"
using obtain_ideal_points[OF is_poincare_line_poincare_line] ‹u ≠ v›
by blast
let ?d = "calc_poincare_distance u i1 v i2"
show ?thesis
proof (rule ex1I)
show "poincare_distance_pred u v ?d"
using * ‹u ≠ v›
proof (simp del: calc_poincare_distance_def, safe)
fix i1' i2'
assume "{i1, i2} = {i1', i2'}"
hence **: "(i1' = i1 ∧ i2' = i2) ∨ (i1' = i2 ∧ i2' = i1)"
using doubleton_eq_iff[of i1 i2 i1' i2']
by blast
have all_different: "u ≠ i1" "u ≠ i2" "v ≠ i1" "v ≠ i2" "u ≠ i1'" "u ≠ i2'" "v ≠ i1'" "v ≠ i2'" "i1 ≠ i2"
using ideal_points_different[OF assms, of i1 i2] * ** ‹u ≠ v›
by auto
show "calc_poincare_distance u i1 v i2 = calc_poincare_distance u i1' v i2'"
proof-
let ?cr = "cross_ratio u i1 v i2"
let ?cr' = "cross_ratio u i1' v i2'"
have "Re (to_complex ?cr) > 0" "is_real (to_complex ?cr)"
"Re (to_complex ?cr') > 0" "is_real (to_complex ?cr')"
using False distance_cross_ratio_real_positive[OF assms(1-2)] * **
by auto
thus ?thesis
using **
using cross_ratio_not_zero cross_ratio_not_inf all_different
by auto (subst cross_ratio_commute_24, subst reciprocal_real, simp_all add: ln_div)
qed
qed
next
fix d
assume "poincare_distance_pred u v d"
thus "d = ?d"
using * ‹u ≠ v›
by auto
qed
qed
lemma poincare_distance_satisfies_pred [simp]:
assumes "u ∈ unit_disc" and "v ∈ unit_disc"
shows "poincare_distance_pred u v (poincare_distance u v)"
using distance_unique[OF assms] theI'[of "poincare_distance_pred u v"]
unfolding poincare_distance_def
by blast
lemma poincare_distance_I:
assumes "u ∈ unit_disc" and "v ∈ unit_disc" and "u ≠ v" and "ideal_points (poincare_line u v) = {i1, i2}"
shows "poincare_distance u v = calc_poincare_distance u i1 v i2"
using assms
using poincare_distance_satisfies_pred[OF assms(1-2)]
by simp
lemma poincare_distance_refl [simp]:
assumes "u ∈ unit_disc"
shows "poincare_distance u u = 0"
using assms
using poincare_distance_satisfies_pred[OF assms assms]
by simp
text‹Unit disc preserving Möbius transformations preserve h-distance. ›
lemma unit_disc_fix_preserve_poincare_distance [simp]:
assumes "unit_disc_fix M" and "u ∈ unit_disc" and "v ∈ unit_disc"
shows "poincare_distance (moebius_pt M u) (moebius_pt M v) = poincare_distance u v"
proof (cases "u = v")
case True
have "moebius_pt M u ∈ unit_disc" "moebius_pt M v ∈ unit_disc"
using unit_disc_fix_iff[OF assms(1), symmetric] assms
by blast+
thus ?thesis
using assms ‹u = v›
by simp
next
case False
obtain i1 i2 where *: "ideal_points (poincare_line u v) = {i1, i2}"
using ‹u ≠ v›
by (rule obtain_ideal_points[OF is_poincare_line_poincare_line[of u v]])
let ?Mu = "moebius_pt M u" and ?Mv = "moebius_pt M v" and ?Mi1 = "moebius_pt M i1" and ?Mi2 = "moebius_pt M i2"
have **: "?Mu ∈ unit_disc" "?Mv ∈ unit_disc"
using assms
using unit_disc_fix_iff
by blast+
have ***: "?Mu ≠ ?Mv"
using ‹u ≠ v›
by simp
have "poincare_distance u v = calc_poincare_distance u i1 v i2"
using poincare_distance_I[OF assms(2-3) ‹u ≠ v› *]
by auto
moreover
have "unit_circle_fix M"
using assms
by simp
hence ++: "ideal_points (poincare_line ?Mu ?Mv) = {?Mi1, ?Mi2}"
using ‹u ≠ v› assms *
by simp
have "poincare_distance ?Mu ?Mv = calc_poincare_distance ?Mu ?Mi1 ?Mv ?Mi2"
by (rule poincare_distance_I[OF ** *** ++])
moreover
have "calc_poincare_distance ?Mu ?Mi1 ?Mv ?Mi2 = calc_poincare_distance u i1 v i2"
using ideal_points_different[OF assms(2-3) ‹u ≠ v› *]
unfolding calc_poincare_distance_def
by (subst moebius_preserve_cross_ratio[symmetric], simp_all)
ultimately
show ?thesis
by simp
qed
text‹Knowing ideal points for x-axis, we can easily explicitly calculate distances.›
lemma poincare_distance_x_axis_x_axis:
assumes "x ∈ unit_disc" and "y ∈ unit_disc" and "x ∈ circline_set x_axis" and "y ∈ circline_set x_axis"
shows "poincare_distance x y =
(let x' = to_complex x; y' = to_complex y
in abs (ln (Re (((1 + x') * (1 - y')) / ((1 - x') * (1 + y'))))))"
proof-
obtain x' y' where *: "x = of_complex x'" "y = of_complex y'"
using inf_or_of_complex[of x] inf_or_of_complex[of y] ‹x ∈ unit_disc› ‹y ∈ unit_disc›
by auto
have "cmod x' < 1" "cmod y' < 1"
using ‹x ∈ unit_disc› ‹y ∈ unit_disc› *
by (metis unit_disc_iff_cmod_lt_1)+
hence **: "x' ≠ 1" "x' ≠ 1" "y' ≠ -1" "y' ≠ 1"
by auto
have "1 + y' ≠ 0"
using **
by (metis add.left_cancel add_neg_numeral_special(7))
show ?thesis
proof (cases "x = y")
case True
thus ?thesis
using assms(1-2)
using unit_disc_iff_cmod_lt_1[of "to_complex x"] * ** ‹1 + y' ≠ 0›
by auto
next
case False
hence "poincare_line x y = x_axis"
using poincare_line_x_axis[OF assms]
by simp
hence "ideal_points (poincare_line x y) = {of_complex (-1), of_complex 1}"
by simp
hence "poincare_distance x y = calc_poincare_distance x (of_complex (-1)) y (of_complex 1)"
using poincare_distance_I assms ‹x ≠ y›
by auto
also have "... = abs (ln (Re (((x' + 1) * (y' - 1)) / ((x' - 1) * (y' + 1)))))"
using * ‹cmod x' < 1› ‹cmod y' < 1›
by (simp, transfer, transfer, auto)
finally
show ?thesis
using *
by (metis (no_types, lifting) add.commute minus_diff_eq minus_divide_divide mult_minus_left mult_minus_right to_complex_of_complex)
qed
qed
lemma poincare_distance_zero_x_axis:
assumes "x ∈ unit_disc" and "x ∈ circline_set x_axis"
shows "poincare_distance 0⇩h x = (let x' = to_complex x in abs (ln (Re ((1 - x') / (1 + x')))))"
using assms
using poincare_distance_x_axis_x_axis[of "0⇩h" x]
by (simp add: Let_def)
lemma poincare_distance_zero:
assumes "x ∈ unit_disc"
shows "poincare_distance 0⇩h x = (let x' = to_complex x in abs (ln (Re ((1 - cmod x') / (1 + cmod x')))))" (is "?P x")
proof (cases "x = 0⇩h")
case True
thus ?thesis
by auto
next
case False
show ?thesis
proof (rule wlog_rotation_to_positive_x_axis)
show "x ∈ unit_disc" "x ≠ 0⇩h" by fact+
next
fix φ u
assume "u ∈ unit_disc" "u ≠ 0⇩h" "?P (moebius_pt (moebius_rotation φ) u)"
thus "?P u"
using unit_disc_fix_preserve_poincare_distance[of "moebius_rotation φ" "0⇩h" u]
by (cases "u = ∞⇩h") (simp_all add: Let_def)
next
fix x
assume "is_real x" "0 < Re x" "Re x < 1"
thus "?P (of_complex x)"
using poincare_distance_zero_x_axis[of "of_complex x"]
by simp (auto simp add: circline_set_x_axis cmod_eq_Re complex_is_Real_iff)
qed
qed
lemma poincare_distance_zero_opposite [simp]:
assumes "of_complex z ∈ unit_disc"
shows "poincare_distance 0⇩h (of_complex (- z)) = poincare_distance 0⇩h (of_complex z)"
proof-
have *: "of_complex (-z) ∈ unit_disc"
using assms
by auto
show ?thesis
using poincare_distance_zero[OF assms]
using poincare_distance_zero[OF *]
by simp
qed
subsection‹Distance explicit formula›
text‹Instead of the h-distance itself, very frequently its hyperbolic cosine is analyzed.›
abbreviation "cosh_dist u v ≡ cosh (poincare_distance u v)"
lemma cosh_poincare_distance_cross_ratio_average:
assumes "u ∈ unit_disc" "v ∈ unit_disc" "u ≠ v" "ideal_points (poincare_line u v) = {i1, i2}"
shows "cosh_dist u v =
((Re_cross_ratio u i1 v i2) + (Re_cross_ratio v i1 u i2)) / 2"
proof-
let ?cr = "cross_ratio u i1 v i2"
let ?crRe = "Re (to_complex ?cr)"
have "?cr ≠ ∞⇩h" "is_real (to_complex ?cr)" "?crRe > 0"
using distance_cross_ratio_real_positive[OF assms(1-3)] assms(4)
by simp_all
then obtain cr where *: "cross_ratio u i1 v i2 = of_complex cr" "cr ≠ 0" "is_real cr" "Re cr > 0"
using inf_or_of_complex[of "cross_ratio u i1 v i2"]
by (smt to_complex_of_complex zero_complex.simps(1))
thus ?thesis
using *
using assms cross_ratio_commute_13[of v i1 u i2]
unfolding poincare_distance_I[OF assms] calc_poincare_distance_def cosh_def
by (cases "Re cr ≥ 1")
(auto simp add: ln_div[of 0] exp_minus field_simps Re_divide power2_eq_square complex.expand)
qed
definition poincare_distance_formula' :: "complex ⇒ complex ⇒ real" where
[simp]: "poincare_distance_formula' u v = 1 + 2 * ((cmod (u - v))⇧2 / ((1 - (cmod u)⇧2) * (1 - (cmod v)⇧2)))"
text‹Next we show that the following formula expresses h-distance between any two h-points (note
that the ideal points do not figure anymore).›
definition poincare_distance_formula :: "complex ⇒ complex ⇒ real" where
[simp]: "poincare_distance_formula u v = arcosh (poincare_distance_formula' u v)"
lemma blaschke_preserve_distance_formula [simp]:
assumes "of_complex k ∈ unit_disc" "u ∈ unit_disc" "v ∈ unit_disc"
shows "poincare_distance_formula (to_complex (moebius_pt (blaschke k) u)) (to_complex (moebius_pt (blaschke k) v)) =
poincare_distance_formula (to_complex u) (to_complex v)"
proof (cases "k = 0")
case True
thus ?thesis
by simp
next
case False
obtain u' v' where *: "u' = to_complex u" "v' = to_complex v"
by auto
have "cmod u' < 1" "cmod v' < 1" "cmod k < 1"
using assms *
using inf_or_of_complex[of u] inf_or_of_complex[of v]
by auto
obtain nu du nv dv d kk ddu ddv where
**: "nu = u' - k" "du = 1 - cnj k *u'" "nv = v' - k" "dv = 1 - cnj k * v'"
"d = u' - v'" "ddu = 1 - u'*cnj u'" "ddv = 1 - v'*cnj v'" "kk = 1 - k*cnj k"
by auto
have d: "nu*dv - nv*du = d*kk"
by (subst **)+ (simp add: field_simps)
have ddu: "du*cnj du - nu*cnj nu = ddu*kk"
by (subst **)+ (simp add: field_simps)
have ddv: "dv*cnj dv - nv*cnj nv = ddv*kk"
by (subst **)+ (simp add: field_simps)
have "du ≠ 0"
proof (rule ccontr)
assume "¬ ?thesis"
hence "cmod (1 - cnj k * u') = 0"
using ‹du = 1 - cnj k * u'›
by auto
hence "cmod (cnj k * u') = 1"
by auto
thus False
using ‹cmod k < 1› ‹cmod u' < 1›
using mult_strict_mono[of "cmod k" 1 "cmod u'" 1]
by (simp add: norm_mult)
qed
have "dv ≠ 0"
proof (rule ccontr)
assume "¬ ?thesis"
hence "cmod (1 - cnj k * v') = 0"
using ‹dv = 1 - cnj k * v'›
by auto
hence "cmod (cnj k * v') = 1"
by auto
thus False
using ‹cmod k < 1› ‹cmod v' < 1›
using mult_strict_mono[of "cmod k" 1 "cmod v'" 1]
by (simp add: norm_mult)
qed
have "kk ≠ 0"
proof (rule ccontr)
assume "¬ ?thesis"
hence "cmod (1 - k * cnj k) = 0"
using ‹kk = 1 - k * cnj k›
by auto
hence "cmod (k * cnj k) = 1"
by auto
thus False
using ‹cmod k < 1›
using mult_strict_mono[of "cmod k" 1 "cmod k" 1]
using complex_mod_sqrt_Re_mult_cnj by auto
qed
note nz = ‹du ≠ 0› ‹dv ≠ 0› ‹kk ≠ 0›
have "nu / du - nv / dv = (nu*dv - nv*du) / (du * dv)"
using nz
by (simp add: field_simps)
hence "(cmod (nu/du - nv/dv))⇧2 = cmod ((d*kk) / (du*dv) * (cnj ((d*kk) / (du*dv))))" (is "?lhs = _")
unfolding complex_mod_mult_cnj[symmetric]
by (subst (asm) d) simp
also have "... = cmod ((d*cnj d*kk*kk) / (du*cnj du*dv*cnj dv))"
by (simp add: field_simps norm_mult norm_divide)
finally have 1: "?lhs = cmod ((d*cnj d*kk*kk) / (du*cnj du*dv*cnj dv))" .
have "(1 - ((cmod nu) / (cmod du))⇧2)*(1 - ((cmod nv) / (cmod dv))⇧2) =
(1 - cmod((nu * cnj nu) / (du * cnj du)))*(1 - cmod((nv * cnj nv) / (dv * cnj dv)))" (is "?rhs = _")
by (metis norm_divide complex_mod_mult_cnj power_divide)
also have "... = cmod(((du*cnj du - nu*cnj nu) / (du * cnj du)) * ((dv*cnj dv - nv*cnj nv) / (dv * cnj dv)))"
proof-
have "u' ≠ 1 / cnj k" "v' ≠ 1 / cnj k"
using ‹cmod u' < 1› ‹cmod v' < 1› ‹cmod k < 1›
by (auto simp add: False norm_divide)
moreover
have "cmod k ≠ 1"
using ‹cmod k < 1›
by linarith
ultimately
have "cmod (nu/du) < 1" "cmod (nv/dv) < 1"
using **(1-4)
using unit_disc_fix_discI[OF blaschke_unit_disc_fix[OF ‹cmod k < 1›] ‹u ∈ unit_disc›] ‹u' = to_complex u›
using unit_disc_fix_discI[OF blaschke_unit_disc_fix[OF ‹cmod k < 1›] ‹v ∈ unit_disc›] ‹v' = to_complex v›
using inf_or_of_complex[of u] ‹u ∈ unit_disc› inf_or_of_complex[of v] ‹v ∈ unit_disc›
using moebius_pt_blaschke[of k u'] using moebius_pt_blaschke[of k v']
by auto
hence "(cmod (nu/du))⇧2 < 1" "(cmod (nv/dv))⇧2 < 1"
by (simp_all add: cmod_def)
hence "cmod (nu * cnj nu / (du * cnj du)) < 1" "cmod (nv * cnj nv / (dv * cnj dv)) < 1"
by (metis complex_mod_mult_cnj norm_divide power_divide)+
moreover
have "is_real (nu * cnj nu / (du * cnj du))" "is_real (nv * cnj nv / (dv * cnj dv))"
using eq_cnj_iff_real[of "nu * cnj nu / (du * cnj du)"]
using eq_cnj_iff_real[of "nv * cnj nv / (dv * cnj dv)"]
by (auto simp add: mult.commute)
moreover
have "Re (nu * cnj nu / (du * cnj du)) ≥ 0" "Re (nv * cnj nv / (dv * cnj dv)) ≥ 0"
using ‹du ≠ 0› ‹dv ≠ 0›
unfolding complex_mult_cnj_cmod
by simp_all
ultimately
have "1 - cmod (nu * cnj nu / (du * cnj du)) = cmod (1 - nu * cnj nu / (du * cnj du))"
"1 - cmod (nv * cnj nv / (dv * cnj dv)) = cmod (1 - nv * cnj nv / (dv * cnj dv))"
by (simp_all add: cmod_def)
thus ?thesis
using nz
by (simp add: diff_divide_distrib norm_mult)
qed
also have "... = cmod(((ddu * kk) / (du * cnj du)) * ((ddv * kk) / (dv * cnj dv)))"
by (subst ddu, subst ddv, simp)
also have "... = cmod((ddu*ddv*kk*kk) / (du*cnj du*dv*cnj dv))"
by (simp add: field_simps)
finally have 2: "?rhs = cmod((ddu*ddv*kk*kk) / (du*cnj du*dv*cnj dv))"
.
have "?lhs / ?rhs =
cmod ((d*cnj d*kk*kk) / (du*cnj du*dv*cnj dv)) / cmod((ddu*ddv*kk*kk) / (du*cnj du*dv*cnj dv))"
by (subst 1, subst 2, simp)
also have "... = cmod ((d*cnj d)/(ddu*ddv))"
using nz by (simp add: norm_mult norm_divide)
also have "... = (cmod d)⇧2 / ((1 - (cmod u')⇧2)*(1 - (cmod v')⇧2))"
proof-
have "(cmod u')⇧2 < 1" "(cmod v')⇧2 < 1"
using ‹cmod u' < 1› ‹cmod v' < 1›
by (simp_all add: cmod_def)
hence "cmod (1 - u' * cnj u') = 1 - (cmod u')⇧2" "cmod (1 - v' * cnj v') = 1 - (cmod v')⇧2"
by (auto simp add: cmod_eq_Re cmod_power2 power2_eq_square[symmetric])
thus ?thesis
using nz
by (simp add: "**"(6) "**"(7) norm_divide norm_mult power2_eq_square)
qed
finally
have 3: "?lhs / ?rhs = (cmod d)⇧2 / ((1 - (cmod u')⇧2)*(1 - (cmod v')⇧2))" .
have "cmod k ≠ 1" "u' ≠ 1 / cnj k" "v' ≠ 1 / cnj k" "u ≠ ∞⇩h" "v ≠ ∞⇩h"
using ‹cmod k < 1› ‹u ∈ unit_disc› ‹v ∈ unit_disc› * ‹k ≠ 0› ** ‹kk ≠ 0› nz
by auto
thus ?thesis using assms
using * ** 3
using moebius_pt_blaschke[of k u']
using moebius_pt_blaschke[of k v']
by (simp add: norm_divide)
qed
text ‹To prove the equivalence between the h-distance definition and the distance formula, we shall
employ the without loss of generality principle. Therefore, we must show that the distance formula
is preserved by h-isometries.›
text‹Rotation preserve @{term poincare_distance_formula}.›
lemma rotation_preserve_distance_formula [simp]:
assumes "u ∈ unit_disc" "v ∈ unit_disc"
shows "poincare_distance_formula (to_complex (moebius_pt (moebius_rotation φ) u)) (to_complex (moebius_pt (moebius_rotation φ) v)) =
poincare_distance_formula (to_complex u) (to_complex v)"
using assms
using inf_or_of_complex[of u] inf_or_of_complex[of v]
by (auto simp: norm_mult)
text‹Unit disc fixing Möbius preserve @{term poincare_distance_formula}.›
lemma unit_disc_fix_preserve_distance_formula [simp]:
assumes "unit_disc_fix M" "u ∈ unit_disc" "v ∈ unit_disc"
shows "poincare_distance_formula (to_complex (moebius_pt M u)) (to_complex (moebius_pt M v)) =
poincare_distance_formula (to_complex u) (to_complex v)" (is "?P' u v M")
proof-
have "∀ u ∈ unit_disc. ∀ v ∈ unit_disc. ?P' u v M" (is "?P M")
proof (rule wlog_unit_disc_fix[OF assms(1)])
fix k
assume "cmod k < 1"
hence "of_complex k ∈ unit_disc"
by simp
thus "?P (blaschke k)"
using blaschke_preserve_distance_formula
by simp
next
fix φ
show "?P (moebius_rotation φ)"
using rotation_preserve_distance_formula
by simp
next
fix M1 M2
assume *: "?P M1" and **: "?P M2" and u11: "unit_disc_fix M1" "unit_disc_fix M2"
thus "?P (M1 + M2)"
by (auto simp del: poincare_distance_formula_def)
qed
thus ?thesis
using assms
by simp
qed
text‹The equivalence between the two h-distance representations.›
lemma poincare_distance_formula:
assumes "u ∈ unit_disc" and "v ∈ unit_disc"
shows "poincare_distance u v = poincare_distance_formula (to_complex u) (to_complex v)" (is "?P u v")
proof (rule wlog_x_axis)
fix x
assume *: "is_real x" "0 ≤ Re x" "Re x < 1"
show "?P 0⇩h (of_complex x)" (is "?lhs = ?rhs")
proof-
have "of_complex x ∈ unit_disc" "of_complex x ∈ circline_set x_axis" "cmod x < 1"
using * cmod_eq_Re
by (auto simp add: circline_set_x_axis)
hence "?lhs = ¦ln (Re ((1 - x) / (1 + x)))¦"
using poincare_distance_zero_x_axis[of "of_complex x"]
by simp
moreover
have "?rhs = ¦ln (Re ((1 - x) / (1 + x)))¦"
proof-
let ?x = "1 + 2 * (cmod x)⇧2 / (1 - (cmod x)⇧2)"
have "0 ≤ 2 * (cmod x)⇧2 / (1 - (cmod x)⇧2)"
by (smt ‹cmod x < 1› divide_nonneg_nonneg norm_ge_zero power_le_one zero_le_power2)
hence arcosh_real_gt: "1 ≤ ?x"
by auto
have "?rhs = arcosh ?x"
by simp
also have "... = ln ((1 + (cmod x)⇧2) / (1 - (cmod x)⇧2) + 2 * (cmod x) / (1 - (cmod x)⇧2))"
proof-
have "1 - (cmod x)⇧2 > 0"
using ‹cmod x < 1›
by (smt norm_not_less_zero one_power2 power2_eq_imp_eq power_mono)
hence 1: "?x = (1 + (cmod x)⇧2) / (1 - (cmod x)⇧2)"
by (simp add: field_simps)
have 2: "?x⇧2 - 1 = (4 * (cmod x)⇧2) / (1 - (cmod x)⇧2)⇧2"
using ‹1 - (cmod x)⇧2 > 0›
apply (subst 1)
unfolding power_divide
by (subst divide_diff_eq_iff, simp, simp add: power2_eq_square field_simps)
show ?thesis
using ‹1 - (cmod x)⇧2 > 0›
apply (subst arcosh_real_def[OF arcosh_real_gt])
apply (subst 2)
apply (subst 1)
apply (subst real_sqrt_divide)
apply (subst real_sqrt_mult)
apply simp
done
qed
also have "... = ln (((1 + (cmod x))⇧2) / (1 - (cmod x)⇧2))"
apply (subst add_divide_distrib[symmetric])
apply (simp add: field_simps power2_eq_square)
done
also have "... = ln ((1 + cmod x) / (1 - (cmod x)))"
using ‹cmod x < 1›
using square_diff_square_factored[of 1 "cmod x"]
by (simp add: power2_eq_square)
also have "... = ¦ln (Re ((1 - x) / (1 + x)))¦"
proof-
have *: "Re ((1 - x) / (1 + x)) ≤ 1" "Re ((1 - x) / (1 + x)) > 0"
using ‹is_real x› ‹Re x ≥ 0› ‹Re x < 1›
using complex_is_Real_iff
by auto
hence "¦ln (Re ((1 - x) / (1 + x)))¦ = - ln (Re ((1 - x) / (1 + x)))"
by auto
hence "¦ln (Re ((1 - x) / (1 + x)))¦ = ln (Re ((1 + x) / (1 - x)))"
using ln_div[of 1 "Re ((1 - x)/(1 + x))"] * ‹is_real x›
by (fastforce simp: complex_is_Real_iff)
moreover
have "ln ((1 + cmod x) / (1 - cmod x)) = ln ((1 + Re x) / (1 - Re x))"
using ‹Re x ≥ 0› ‹is_real x›
using cmod_eq_Re by auto
moreover
have "(1 + Re x) / (1 - Re x) = Re ((1 + x) / (1 - x))"
using ‹is_real x› ‹Re x < 1›
by (smt Re_divide_real eq_iff_diff_eq_0 minus_complex.simps one_complex.simps plus_complex.simps)
ultimately
show ?thesis
by simp
qed
finally
show ?thesis
.
qed
ultimately
show ?thesis
by simp
qed
next
fix M u v
assume *: "unit_disc_fix M" "u ∈ unit_disc" "v ∈ unit_disc"
assume "?P (moebius_pt M u) (moebius_pt M v)"
thus "?P u v"
using *(1-3)
by (simp del: poincare_distance_formula_def)
next
show "u ∈ unit_disc" "v ∈ unit_disc"
by fact+
qed
text‹Some additional properties proved easily using the distance formula.›
text ‹@{term poincare_distance} is symmetric.›
lemma poincare_distance_sym:
assumes "u ∈ unit_disc" and "v ∈ unit_disc"
shows "poincare_distance u v = poincare_distance v u"
using assms
using poincare_distance_formula[OF assms(1) assms(2)]
using poincare_distance_formula[OF assms(2) assms(1)]
by (simp add: mult.commute norm_minus_commute)
lemma poincare_distance_formula'_ge_1:
assumes "u ∈ unit_disc" and "v ∈ unit_disc"
shows "1 ≤ poincare_distance_formula' (to_complex u) (to_complex v)"
using unit_disc_cmod_square_lt_1[OF assms(1)] unit_disc_cmod_square_lt_1[OF assms(2)]
by auto
text‹@{term poincare_distance} is non-negative.›
lemma poincare_distance_ge0:
assumes "u ∈ unit_disc" and "v ∈ unit_disc"
shows "poincare_distance u v ≥ 0"
using poincare_distance_formula'_ge_1 assms by (simp add: poincare_distance_formula)
lemma cosh_dist:
assumes "u ∈ unit_disc" and "v ∈ unit_disc"
shows "cosh_dist u v = poincare_distance_formula' (to_complex u) (to_complex v)"
using poincare_distance_formula[OF assms] poincare_distance_formula'_ge_1[OF assms]
by simp
text‹@{term poincare_distance} is zero only if the two points are equal.›
lemma poincare_distance_eq_0_iff:
assumes "u ∈ unit_disc" and "v ∈ unit_disc"
shows "poincare_distance u v = 0 ⟷ u = v"
using assms
apply auto
using poincare_distance_formula'_ge_1[OF assms]
using unit_disc_cmod_square_lt_1[OF assms(1)] unit_disc_cmod_square_lt_1[OF assms(2)]
apply (simp add: poincare_distance_formula)
by (simp add: unit_disc_to_complex_inj)
text‹Conjugate preserve @{term poincare_distance_formula}.›
lemma conjugate_preserve_poincare_distance [simp]:
assumes "u ∈ unit_disc" and "v ∈ unit_disc"
shows "poincare_distance (conjugate u) (conjugate v) = poincare_distance u v"
proof-
obtain u' v' where *: "u = of_complex u'" "v = of_complex v'"
using assms inf_or_of_complex[of u] inf_or_of_complex[of v]
by auto
have **: "conjugate u ∈ unit_disc" "conjugate v ∈ unit_disc"
using * assms
by auto
show ?thesis
using *
using poincare_distance_formula[OF assms]
using poincare_distance_formula[OF **]
by (metis complex_cnj_diff complex_mod_cnj conjugate_of_complex poincare_distance_def poincare_distance_formula'_def poincare_distance_formula_def to_complex_of_complex)
qed
subsection‹Existence and uniqueness of points with a given distance›
lemma ex_x_axis_poincare_distance_negative':
fixes d :: real
assumes "d ≥ 0"
shows "let z = (1 - exp d) / (1 + exp d)
in is_real z ∧ Re z ≤ 0 ∧ Re z > -1 ∧
of_complex z ∈ unit_disc ∧ of_complex z ∈ circline_set x_axis ∧
poincare_distance 0⇩h (of_complex z) = d"
proof-
have "exp d ≥ 1"
using assms
using one_le_exp_iff[of d, symmetric]
by blast
hence "1 + exp d ≠ 0"
by linarith
let ?z = "(1 - exp d) / (1 + exp d)"
have "?z ≤ 0"
using ‹exp d ≥ 1›
by (simp add: divide_nonpos_nonneg)
moreover
have "?z > -1"
using exp_gt_zero[of d]
by (smt divide_less_eq_1_neg nonzero_minus_divide_right)
moreover
hence "abs ?z < 1"
using ‹?z ≤ 0›
by simp
hence "cmod ?z < 1"
by (metis norm_of_real)
hence "of_complex ?z ∈ unit_disc"
by simp
moreover
have "of_complex ?z ∈ circline_set x_axis"
unfolding circline_set_x_axis
by simp
moreover
have "(1 - ?z) / (1 + ?z) = exp d"
proof-
have "1 + ?z = 2 / (1 + exp d)"
using ‹1 + exp d ≠ 0›
by (subst add_divide_eq_iff, auto)
moreover
have "1 - ?z = 2 * exp d / (1 + exp d)"
using ‹1 + exp d ≠ 0›
by (subst diff_divide_eq_iff, auto)
ultimately
show ?thesis
using ‹1 + exp d ≠ 0›
by simp
qed
ultimately
show ?thesis
using poincare_distance_zero_x_axis[of "of_complex ?z"]
using ‹d ≥ 0› ‹exp d ≥ 1›
by simp (simp add: cmod_eq_Re)
qed
lemma ex_x_axis_poincare_distance_negative:
assumes "d ≥ 0"
shows "∃ z. is_real z ∧ Re z ≤ 0 ∧ Re z > -1 ∧
of_complex z ∈ unit_disc ∧ of_complex z ∈ circline_set x_axis ∧
poincare_distance 0⇩h (of_complex z) = d" (is "∃ z. ?P z")
using ex_x_axis_poincare_distance_negative'[OF assms]
unfolding Let_def
by blast
text‹For each real number $d$ there is exactly one point on the positive x-axis such that h-distance
between 0 and that point is $d$.›
lemma unique_x_axis_poincare_distance_negative:
assumes "d ≥ 0"
shows "∃! z. is_real z ∧ Re z ≤ 0 ∧ Re z > -1 ∧
poincare_distance 0⇩h (of_complex z) = d" (is "∃! z. ?P z")
proof-
let ?z = "(1 - exp d) / (1 + exp d)"
have "?P ?z"
using ex_x_axis_poincare_distance_negative'[OF assms]
unfolding Let_def
by blast
moreover
have "∀ z'. ?P z' ⟶ z' = ?z"
proof-
let ?g = "λ x'. ¦ln (Re ((1 - x') / (1 + x')))¦"
let ?A = "{x. is_real x ∧ Re x > -1 ∧ Re x ≤ 0}"
have "inj_on (poincare_distance 0⇩h ∘ of_complex) ?A"
proof (rule comp_inj_on)
show "inj_on of_complex ?A"
using of_complex_inj
unfolding inj_on_def
by blast
next
show "inj_on (poincare_distance 0⇩h) (of_complex ` ?A)" (is "inj_on ?f (of_complex ` ?A)")
proof (subst inj_on_cong)
have *: "of_complex ` ?A =
{z. z ∈ unit_disc ∧ z ∈ circline_set x_axis ∧ Re (to_complex z) ≤ 0}" (is "_ = ?B")
by (auto simp add: cmod_eq_Re circline_set_x_axis)
fix x
assume "x ∈ of_complex ` ?A"
hence "x ∈ ?B"
using *
by simp
thus "poincare_distance 0⇩h x = (?g ∘ to_complex) x"
using poincare_distance_zero_x_axis
by (simp add: Let_def)
next
have *: "to_complex ` of_complex ` ?A = ?A"
by (auto simp add: image_iff)
show "inj_on (?g ∘ to_complex) (of_complex ` ?A)"
proof (rule comp_inj_on)
show "inj_on to_complex (of_complex ` ?A)"
unfolding inj_on_def
by auto
next
have "inj_on ?g ?A"
unfolding inj_on_def
proof(safe)
fix x y
assume hh: "is_real x" "is_real y" "- 1 < Re x" "Re x ≤ 0"
"- 1 < Re y" "Re y ≤ 0" "¦ln (Re ((1 - x) / (1 + x)))¦ = ¦ln (Re ((1 - y) / (1 + y)))¦"
have "is_real ((1 - x)/(1 + x))"
using ‹is_real x› div_reals[of "1-x" "1+x"]
by auto
have "is_real ((1 - y)/(1 + y))"
using ‹is_real y› div_reals[of "1-y" "1+y"]
by auto
have "Re (1 + x) > 0"
using ‹- 1 < Re x› by auto
hence "1 + x ≠ 0"
by force
have "Re (1 - x) ≥ 0"
using ‹Re x ≤ 0› by auto
hence "Re ((1 - x)/(1 + x)) > 0"
using Re_divide_real ‹0 < Re (1 + x)› complex_eq_if_Re_eq hh(1) hh(4) by auto
have "Re(1 - x) ≥ Re ( 1 + x)"
using hh by auto
hence "Re ((1 - x)/(1 + x)) ≥ 1"
using ‹Re (1 + x) > 0› ‹is_real ((1 - x)/(1 + x))›
by (smt Re_divide_real arg_0_iff hh(1) le_divide_eq_1_pos one_complex.simps(2) plus_complex.simps(2))
have "Re (1 + y) > 0"
using ‹- 1 < Re y› by auto
hence "1 + y ≠ 0"
by force
have "Re (1 - y) ≥ 0"
using ‹Re y ≤ 0› by auto
hence "Re ((1 - y)/(1 + y)) > 0"
using Re_divide_real ‹0 < Re (1 + y)› complex_eq_if_Re_eq hh by auto
have "Re(1 - y) ≥ Re ( 1 + y)"
using hh by auto
hence "Re ((1 - y)/(1 + y)) ≥ 1"
using ‹Re (1 + y) > 0› ‹is_real ((1 - y)/(1 + y))›
by (smt Re_divide_real arg_0_iff hh le_divide_eq_1_pos one_complex.simps(2) plus_complex.simps(2))
have "ln (Re ((1 - x) / (1 + x))) = ln (Re ((1 - y) / (1 + y)))"
using ‹Re ((1 - y)/(1 + y)) ≥ 1› ‹Re ((1 - x)/(1 + x)) ≥ 1› hh
by auto
hence "Re ((1 - x) / (1 + x)) = Re ((1 - y) / (1 + y))"
using ‹Re ((1 - y)/(1 + y)) > 0› ‹Re ((1 - x)/(1 + x)) > 0›
by auto
hence "(1 - x) / (1 + x) = (1 - y) / (1 + y)"
using ‹is_real ((1 - y)/(1 + y))› ‹is_real ((1 - x)/(1 + x))›
using complex_eq_if_Re_eq by blast
hence "(1 - x) * (1 + y) = (1 - y) * (1 + x)"
using ‹1 + y ≠ 0› ‹1 + x ≠ 0›
by (simp add:field_simps)
thus "x = y"
by (simp add:field_simps)
qed
thus "inj_on ?g (to_complex ` of_complex ` ?A)"
using *
by simp
qed
qed
qed
thus ?thesis
using ‹?P ?z›
unfolding inj_on_def
by auto
qed
ultimately
show ?thesis
by blast
qed
lemma ex_x_axis_poincare_distance_positive:
assumes "d ≥ 0"
shows "∃ z. is_real z ∧ Re z ≥ 0 ∧ Re z < 1 ∧
of_complex z ∈ unit_disc ∧ of_complex z ∈ circline_set x_axis ∧
poincare_distance 0⇩h (of_complex z) = d" (is "∃ z. is_real z ∧ Re z ≥ 0 ∧ Re z < 1 ∧ ?P z")
proof-
obtain z where *: "is_real z" "Re z ≤ 0" "Re z > -1" "?P z"
using ex_x_axis_poincare_distance_negative[OF assms]
by auto
hence **: "of_complex z ∈ unit_disc" "of_complex z ∈ circline_set x_axis"
by (auto simp add: cmod_eq_Re)
have "is_real (-z) ∧ Re (-z) ≥ 0 ∧ Re (-z) < 1 ∧ ?P (-z)"
using * **
by (simp add: circline_set_x_axis)
thus ?thesis
by blast
qed
lemma unique_x_axis_poincare_distance_positive:
assumes "d ≥ 0"
shows "∃! z. is_real z ∧ Re z ≥ 0 ∧ Re z < 1 ∧
poincare_distance 0⇩h (of_complex z) = d" (is "∃! z. is_real z ∧ Re z ≥ 0 ∧ Re z < 1 ∧ ?P z")
proof-
obtain z where *: "is_real z" "Re z ≤ 0" "Re z > -1" "?P z"
using unique_x_axis_poincare_distance_negative[OF assms]
by auto
hence **: "of_complex z ∈ unit_disc" "of_complex z ∈ circline_set x_axis"
by (auto simp add: cmod_eq_Re circline_set_x_axis)
show ?thesis
proof
show "is_real (-z) ∧ Re (-z) ≥ 0 ∧ Re (-z) < 1 ∧ ?P (-z)"
using * **
by simp
next
fix z'
assume "is_real z' ∧ Re z' ≥ 0 ∧ Re z' < 1 ∧ ?P z'"
hence "is_real (-z') ∧ Re (-z') ≤ 0 ∧ Re (-z') > -1 ∧ ?P (-z')"
by (auto simp add: circline_set_x_axis cmod_eq_Re)
hence "-z' = z"
using unique_x_axis_poincare_distance_negative[OF assms] *
by blast
thus "z' = -z"
by auto
qed
qed
text‹Equal distance implies that segments are isometric - this means that congruence could be
defined either by two segments having the same distance or by requiring existence of an isometry
that maps one segment to the other.›
lemma poincare_distance_eq_ex_moebius:
assumes in_disc: "u ∈ unit_disc" and "v ∈ unit_disc" and "u' ∈ unit_disc" and "v' ∈ unit_disc"
assumes "poincare_distance u v = poincare_distance u' v'"
shows "∃ M. unit_disc_fix M ∧ moebius_pt M u = u' ∧ moebius_pt M v = v'" (is "?P' u v u' v'")
proof (cases "u = v")
case True
thus ?thesis
using assms poincare_distance_eq_0_iff[of u' v']
by (simp add: unit_disc_fix_transitive)
next
case False
have "∀ u' v'. u ≠ v ∧ u' ∈ unit_disc ∧ v' ∈ unit_disc ∧ poincare_distance u v = poincare_distance u' v' ⟶
?P' u' v' u v" (is "?P u v")
proof (rule wlog_positive_x_axis[where P="?P"])
fix x
assume "is_real x" "0 < Re x" "Re x < 1"
hence "of_complex x ∈ unit_disc" "of_complex x ∈ circline_set x_axis"
unfolding circline_set_x_axis
by (auto simp add: cmod_eq_Re)
show "?P 0⇩h (of_complex x)"
proof safe
fix u' v'
assume "0⇩h ≠ of_complex x" and in_disc: "u' ∈ unit_disc" "v' ∈ unit_disc" and
"poincare_distance 0⇩h (of_complex x) = poincare_distance u' v'"
hence "u' ≠ v'" "poincare_distance u' v' > 0"
using poincare_distance_eq_0_iff[of "0⇩h" "of_complex x"] ‹of_complex x ∈ unit_disc›
using poincare_distance_ge0[of "0⇩h" "of_complex x"]
by auto
then obtain M where M: "unit_disc_fix M" "moebius_pt M u' = 0⇩h" "moebius_pt M v' ∈ positive_x_axis"
using ex_unit_disc_fix_to_zero_positive_x_axis[of u' v'] in_disc
by auto
then obtain Mv' where Mv': "moebius_pt M v' = of_complex Mv'"
using inf_or_of_complex[of "moebius_pt M v'"] in_disc unit_disc_fix_iff[of M]
by (metis image_eqI inf_notin_unit_disc)
have "moebius_pt M v' ∈ unit_disc"
using M(1) ‹v' ∈ unit_disc›
by auto
have "Re Mv' > 0" "is_real Mv'" "Re Mv' < 1"
using M Mv' of_complex_inj ‹moebius_pt M v' ∈ unit_disc›
unfolding positive_x_axis_def circline_set_x_axis
using cmod_eq_Re
by auto fastforce
have "poincare_distance 0⇩h (moebius_pt M v') = poincare_distance u' v'"
using M(1)
using in_disc
by (subst M(2)[symmetric], simp)
have "Mv' = x"
using ‹poincare_distance 0⇩h (moebius_pt M v') = poincare_distance u' v'› Mv'
using ‹poincare_distance 0⇩h (of_complex x) = poincare_distance u' v'›
using unique_x_axis_poincare_distance_positive[of "poincare_distance u' v'"]
‹poincare_distance u' v' > 0›
using ‹Re Mv' > 0› ‹Re Mv' < 1› ‹is_real Mv'›
using ‹is_real x› ‹Re x > 0› ‹Re x < 1›
unfolding positive_x_axis_def
by auto
thus "?P' u' v' 0⇩h (of_complex x)"
using M Mv'
by auto
qed
next
show "u ∈ unit_disc" "v ∈ unit_disc" "u ≠ v"
by fact+
next
fix M u v
let ?Mu = "moebius_pt M u" and ?Mv = "moebius_pt M v"
assume 1: "unit_disc_fix M" "u ∈ unit_disc" "v ∈ unit_disc" "u ≠ v"
hence 2: "?Mu ≠ ?Mv" "?Mu ∈ unit_disc" "?Mv ∈ unit_disc"
by auto
assume 3: "?P (moebius_pt M u) (moebius_pt M v)"
show "?P u v"
proof safe
fix u' v'
assume 4: "u' ∈ unit_disc" "v' ∈ unit_disc" "poincare_distance u v = poincare_distance u' v'"
hence "poincare_distance ?Mu ?Mv = poincare_distance u v"
using 1
by simp
then obtain M' where 5: "unit_disc_fix M'" "moebius_pt M' u' = ?Mu" "moebius_pt M' v' = ?Mv"
using 2 3 4
by auto
let ?M = "(-M) + M'"
have "unit_disc_fix ?M ∧ moebius_pt ?M u' = u ∧ moebius_pt ?M v' = v"
using 5 ‹unit_disc_fix M›
using unit_disc_fix_moebius_comp[of "-M" "M'"]
using unit_disc_fix_moebius_inv[of M]
by simp
thus "∃M. unit_disc_fix M ∧ moebius_pt M u' = u ∧ moebius_pt M v' = v"
by blast
qed
qed
then obtain M where "unit_disc_fix M ∧ moebius_pt M u' = u ∧ moebius_pt M v' = v"
using assms ‹u ≠ v›
by blast
hence "unit_disc_fix (-M) ∧ moebius_pt (-M) u = u' ∧ moebius_pt (-M) v = v'"
using unit_disc_fix_moebius_inv[of M]
by auto
thus ?thesis
by blast
qed
lemma unique_midpoint_x_axis:
assumes x: "is_real x" "-1 < Re x" "Re x < 1" and
y: "is_real y" "-1 < Re y" "Re y < 1" and
"x ≠ y"
shows "∃! z. -1 < Re z ∧ Re z < 1 ∧ is_real z ∧ poincare_distance (of_complex z) (of_complex x) = poincare_distance (of_complex z) (of_complex y)" (is "∃! z. ?R z (of_complex x) (of_complex y)")
proof-
let ?x = "of_complex x" and ?y = "of_complex y"
let ?P = "λ x y. ∃! z. ?R z x y"
have "∀ x. -1 < Re x ∧ Re x < 1 ∧ is_real x ∧ of_complex x ≠ ?y ⟶ ?P (of_complex x) ?y" (is "?Q (of_complex y)")
proof (rule wlog_real_zero)
show "?y ∈ unit_disc"
using y
by (simp add: cmod_eq_Re)
next
show "is_real (to_complex ?y)"
using y
by simp
next
show "?Q 0⇩h"
proof (rule allI, rule impI, (erule conjE)+)
fix x
assume x: "-1 < Re x" "Re x < 1" "is_real x"
let ?x = "of_complex x"
assume "?x ≠ 0⇩h"
hence "x ≠ 0"
by auto
hence "Re x ≠ 0"
using x
using complex_neq_0
by auto
have *: "∀ a. -1 < a ∧ a < 1 ⟶
(poincare_distance (of_complex (cor a)) ?x = poincare_distance (of_complex (cor a)) 0⇩h ⟷
(Re x) * a * a - 2 * a + Re x = 0)"
proof (rule allI, rule impI)
fix a :: real
assume "-1 < a ∧ a < 1"
hence "of_complex (cor a) ∈ unit_disc"
by auto
moreover
have "(a - Re x)⇧2 / ((1 - a⇧2) * (1 - (Re x)⇧2)) = a⇧2 / (1 - a⇧2) ⟷
(Re x) * a * a - 2 * a + Re x = 0" (is "?lhs ⟷ ?rhs")
proof-
have "1 - a⇧2 ≠ 0"
using ‹-1 < a ∧ a < 1›
by (metis cancel_comm_monoid_add_class.diff_cancel diff_eq_diff_less less_numeral_extra(4) power2_eq_1_iff right_minus_eq)
hence "?lhs ⟷ (a - Re x)⇧2 / (1 - (Re x)⇧2) = a⇧2"
by (smt divide_cancel_right divide_divide_eq_left mult.commute)
also have "... ⟷ (a - Re x)⇧2 = a⇧2 * (1 - (Re x)⇧2)"
proof-
have "1 - (Re x)⇧2 ≠ 0"
using x
by (smt power2_eq_1_iff)
thus ?thesis
by (simp add: divide_eq_eq)
qed
also have "... ⟷ a⇧2 * (Re x)⇧2 - 2*a*Re x + (Re x)⇧2 = 0"
by (simp add: power2_diff field_simps)
also have "... ⟷ Re x * (a⇧2 * Re x - 2 * a + Re x) = 0"
by (simp add: power2_eq_square field_simps)
also have "... ⟷ ?rhs"
using ‹Re x ≠ 0›
by (simp add: mult.commute mult.left_commute power2_eq_square)
finally
show ?thesis
.
qed
moreover
have "arcosh (1 + 2 * ((a - Re x)⇧2 / ((1 - a⇧2) * (1 - (Re x)⇧2)))) = arcosh (1 + 2 * a⇧2 / (1 - a⇧2)) ⟷ ?lhs"
using ‹-1 < a ∧ a < 1› x mult_left_cancel[of "2::real" "(a - Re x)⇧2 / ((1 - a⇧2) * (1 - (Re x)⇧2))" "a⇧2 / (1 - a⇧2)"]
by (subst arcosh_eq_iff, simp_all add: square_le_1)
ultimately
show "poincare_distance (of_complex (cor a)) (of_complex x) = poincare_distance (of_complex (cor a)) 0⇩h ⟷
(Re x) * a * a - 2 * a + Re x = 0"
using x
by (auto simp add: poincare_distance_formula cmod_eq_Re)
qed
show "?P ?x 0⇩h"
proof
let ?a = "(1 - sqrt(1 - (Re x)⇧2)) / (Re x)"
let ?b = "(1 + sqrt(1 - (Re x)⇧2)) / (Re x)"
have "is_real ?a"
by simp
moreover
have "1 - (Re x)⇧2 > 0"
using x
by (smt power2_eq_1_iff square_le_1)
have "¦?a¦ < 1"
proof (cases "Re x > 0")
case True
have "(1 - Re x)⇧2 < 1 - (Re x)⇧2"
using ‹Re x > 0› x
by (simp add: power2_eq_square field_simps)
hence "1 - Re x < sqrt (1 - (Re x)⇧2)"
using real_less_rsqrt by fastforce
thus ?thesis
using ‹1 - (Re x)⇧2 > 0› ‹Re x > 0›
by simp
next
case False
hence "Re x < 0"
using ‹Re x ≠ 0›
by simp
have "1 + Re x > 0"
using ‹Re x > -1›
by simp
hence "2*Re x + 2*Re x*Re x < 0"
using ‹Re x < 0›
by (metis comm_semiring_class.distrib mult.commute mult_2_right mult_less_0_iff one_add_one zero_less_double_add_iff_zero_less_single_add)
hence "(1 + Re x)⇧2 < 1 - (Re x)⇧2"
by (simp add: power2_eq_square field_simps)
hence "1 + Re x < sqrt (1 - (Re x)⇧2)"
using ‹1 - (Re x)⇧2 > 0›
using real_less_rsqrt by blast
thus ?thesis
using ‹Re x < 0›
by (simp add: field_simps)
qed
hence "-1 < ?a" "?a < 1"
by linarith+
moreover
have "(Re x) * ?a * ?a - 2 * ?a + Re x = 0"
using ‹Re x ≠ 0› ‹1 - (Re x)⇧2 > 0›
by (simp add: field_simps power2_eq_square)
ultimately
show "-1 < Re (cor ?a) ∧ Re (cor ?a) < 1 ∧ is_real ?a ∧ poincare_distance (of_complex ?a) (of_complex x) = poincare_distance (of_complex ?a) 0⇩h"
using *
by auto
fix z
assume **: "- 1 < Re z ∧ Re z < 1 ∧ is_real z ∧
poincare_distance (of_complex z) (of_complex x) = poincare_distance (of_complex z) 0⇩h"
hence "Re x * Re z * Re z - 2 * Re z + Re x = 0"
using *[rule_format, of "Re z"] x
by auto
moreover
have "sqrt (4 - 4 * Re x * Re x) = 2 * sqrt(1 - Re x * Re x)"
proof-
have "sqrt (4 - 4 * Re x * Re x) = sqrt(4 * (1 - Re x * Re x))"
by simp
thus ?thesis
by (simp only: real_sqrt_mult, simp)
qed
moreover
have "(2 - 2 * sqrt (1 - Re x * Re x)) / (2 * Re x) = ?a"
proof-
have "(2 - 2 * sqrt (1 - Re x * Re x)) / (2 * Re x) =
(2 * (1 - sqrt (1 - Re x * Re x))) / (2 * Re x)"
by simp
thus ?thesis
by (subst (asm) mult_divide_mult_cancel_left) (auto simp add: power2_eq_square)
qed
moreover
have "(2 + 2 * sqrt (1 - Re x * Re x)) / (2 * Re x) = ?b"
proof-
have "(2 + 2 * sqrt (1 - Re x * Re x)) / (2 * Re x) =
(2 * (1 + sqrt (1 - Re x * Re x))) / (2 * Re x)"
by simp
thus ?thesis
by (subst (asm) mult_divide_mult_cancel_left) (auto simp add: power2_eq_square)
qed
ultimately
have "Re z = ?a ∨ Re z = ?b"
using discriminant_nonneg[of "Re x" "-2" "Re x" "Re z"] discrim_def[of "Re x" "-2" "Re x"]
using ‹Re x ≠ 0› ‹-1 < Re x› ‹Re x < 1› ‹1 - (Re x)⇧2 > 0›
by (auto simp add:power2_eq_square)
have "¦?b¦ > 1"
proof (cases "Re x > 0")
case True
have "(Re x - 1)⇧2 < 1 - (Re x)⇧2"
using ‹Re x > 0› x
by (simp add: power2_eq_square field_simps)
hence "Re x - 1 < sqrt (1 - (Re x)⇧2)"
using real_less_rsqrt
by simp
thus ?thesis
using ‹1 - (Re x)⇧2 > 0› ‹Re x > 0›
by simp
next
case False
hence "Re x < 0"
using ‹Re x ≠ 0›
by simp
have "1 + Re x > 0"
using ‹Re x > -1›
by simp
hence "2*Re x + 2*Re x*Re x < 0"
using ‹Re x < 0›
by (metis comm_semiring_class.distrib mult.commute mult_2_right mult_less_0_iff one_add_one zero_less_double_add_iff_zero_less_single_add)
hence "1 - (Re x)⇧2 > (- 1 - (Re x))⇧2"
by (simp add: field_simps power2_eq_square)
hence "sqrt (1 - (Re x)⇧2) > -1 - Re x"
using real_less_rsqrt
by simp
thus ?thesis
using ‹Re x < 0›
by (simp add: field_simps)
qed
hence "?b < -1 ∨ ?b > 1"
by auto
hence "Re z = ?a"
using ‹Re z = ?a ∨ Re z = ?b› **
by auto
thus "z = ?a"
using ** complex_of_real_Re
by fastforce
qed
qed
next
fix a u
let ?M = "moebius_pt (blaschke a)"
let ?Mu = "?M u"
assume "u ∈ unit_disc" "is_real a" "cmod a < 1"
assume *: "?Q ?Mu"
show "?Q u"
proof (rule allI, rule impI, (erule conjE)+)
fix x
assume x: "-1 < Re x" "Re x < 1" "is_real x" "of_complex x ≠ u"
let ?Mx = "?M (of_complex x)"
have "of_complex x ∈ unit_disc"
using x cmod_eq_Re
by auto
hence "?Mx ∈ unit_disc"
using ‹is_real a› ‹cmod a < 1› blaschke_unit_disc_fix[of a]
using unit_disc_fix_discI
by blast
hence "?Mx ≠ ∞⇩h"
by auto
moreover
have "of_complex x ∈ circline_set x_axis"
using x
by auto
hence "?Mx ∈ circline_set x_axis"
using blaschke_real_preserve_x_axis[OF ‹is_real a› ‹cmod a < 1›, of "of_complex x"]
by auto
hence "-1 < Re (to_complex ?Mx) ∧ Re (to_complex ?Mx) < 1 ∧ is_real (to_complex ?Mx)"
using ‹?Mx ≠ ∞⇩h› ‹?Mx ∈ unit_disc›
unfolding circline_set_x_axis
by (auto simp add: cmod_eq_Re)
moreover
have "?Mx ≠ ?Mu"
using ‹of_complex x ≠ u›
by simp
ultimately
have "?P ?Mx ?Mu"
using *[rule_format, of "to_complex ?Mx"] ‹?Mx ≠ ∞⇩h›
by simp
then obtain Mz where
"?R Mz ?Mx ?Mu"
by blast
have "of_complex Mz ∈ unit_disc" "of_complex Mz ∈ circline_set x_axis"
using ‹?R Mz ?Mx ?Mu›
using cmod_eq_Re
by auto
let ?Minv = "- (blaschke a)"
let ?z = "moebius_pt ?Minv (of_complex Mz)"
have "?z ∈ unit_disc"
using ‹of_complex Mz ∈ unit_disc› ‹cmod a < 1›
by auto
moreover
have "?z ∈ circline_set x_axis"
using ‹of_complex Mz ∈ circline_set x_axis›
using blaschke_real_preserve_x_axis ‹is_real a› ‹cmod a < 1›
by fastforce
ultimately
have z1: "-1 < Re (to_complex ?z)" "Re (to_complex ?z) < 1" "is_real (to_complex ?z)"
using inf_or_of_complex[of "?z"]
unfolding circline_set_x_axis
by (auto simp add: cmod_eq_Re)
have z2: "poincare_distance ?z (of_complex x) = poincare_distance ?z u"
using ‹?R Mz ?Mx ?Mu› ‹cmod a < 1› ‹?z ∈ unit_disc› ‹of_complex x ∈ unit_disc› ‹u ∈ unit_disc›
by (metis blaschke_preserve_distance_formula blaschke_unit_disc_fix moebius_pt_comp_inv_right poincare_distance_formula uminus_moebius_def unit_disc_fix_discI unit_disc_iff_cmod_lt_1)
show "?P (of_complex x) u"
proof
show "?R (to_complex ?z) (of_complex x) u"
using z1 z2 ‹?z ∈ unit_disc› inf_or_of_complex[of ?z]
by auto
next
fix z'
assume "?R z' (of_complex x) u"
hence "of_complex z' ∈ unit_disc" "of_complex z' ∈ circline_set x_axis"
by (auto simp add: cmod_eq_Re)
let ?Mz' = "?M (of_complex z')"
have "?Mz' ∈ unit_disc" "?Mz' ∈ circline_set x_axis"
using ‹of_complex z' ∈ unit_disc› ‹of_complex z' ∈ circline_set x_axis› ‹cmod a < 1› ‹is_real a›
using blaschke_unit_disc_fix unit_disc_fix_discI
using blaschke_real_preserve_x_axis circline_set_x_axis
by blast+
hence "-1 < Re (to_complex ?Mz')" "Re (to_complex ?Mz') < 1" "is_real (to_complex ?Mz')"
unfolding circline_set_x_axis
by (auto simp add: cmod_eq_Re)
moreover
have "poincare_distance ?Mz' ?Mx = poincare_distance ?Mz' ?Mu"
using ‹?R z' (of_complex x) u›
using ‹cmod a < 1› ‹of_complex x ∈ unit_disc› ‹of_complex z' ∈ unit_disc› ‹u ∈ unit_disc›
by auto
ultimately
have "?R (to_complex ?Mz') ?Mx ?Mu"
using ‹?Mz' ∈ unit_disc› inf_or_of_complex[of ?Mz']
by auto
hence "?Mz' = of_complex Mz"
using ‹?P ?Mx ?Mu› ‹?R Mz ?Mx ?Mu›
by (metis ‹moebius_pt (blaschke a) (of_complex z') ∈ unit_disc› ‹of_complex Mz ∈ unit_disc› to_complex_of_complex unit_disc_to_complex_inj)
thus "z' = to_complex ?z"
by (simp add: moebius_pt_invert)
qed
qed
qed
thus ?thesis
using assms
by (metis to_complex_of_complex)
qed
subsection‹Triangle inequality›
lemma poincare_distance_formula_zero_sum:
assumes "u ∈ unit_disc" and "v ∈ unit_disc"
shows "poincare_distance u 0⇩h + poincare_distance 0⇩h v =
(let u' = cmod (to_complex u); v' = cmod (to_complex v)
in arcosh (((1 + u'⇧2) * (1 + v'⇧2) + 4 * u' * v') / ((1 - u'⇧2) * (1 - v'⇧2))))"
proof-
obtain u' v' where uv: "u' = to_complex u" "v' = to_complex v"
by auto
have uv': "u = of_complex u'" "v = of_complex v'"
using uv assms inf_or_of_complex[of u] inf_or_of_complex[of v]
by auto
let ?u' = "cmod u'" and ?v' = "cmod v'"
have disc: "?u'⇧2 < 1" "?v'⇧2 < 1"
using unit_disc_cmod_square_lt_1[OF ‹u ∈ unit_disc›]
using unit_disc_cmod_square_lt_1[OF ‹v ∈ unit_disc›] uv
by auto
thm arcosh_add
have "arcosh (1 + 2 * ?u'⇧2 / (1 - ?u'⇧2)) + arcosh (1 + 2 * ?v'⇧2 / (1 - ?v'⇧2)) =
arcosh (((1 + ?u'⇧2) * (1 + ?v'⇧2) + 4 * ?u' * ?v') / ((1 - ?u'⇧2) * (1 - ?v'⇧2)))" (is "arcosh ?ll + arcosh ?rr = arcosh ?r")
proof (subst arcosh_add)
show "?ll ≥ 1" "?rr ≥ 1"
using disc
by auto
next
show "arcosh ((1 + 2 * ?u'⇧2 / (1 - ?u'⇧2)) * (1 + 2 * ?v'⇧2 / (1 - ?v'⇧2)) +
sqrt (((1 + 2 * ?u'⇧2 / (1 - ?u'⇧2))⇧2 - 1) * ((1 + 2 * ?v'⇧2 / (1 - ?v'⇧2))⇧2 - 1))) =
arcosh ?r" (is "arcosh ?l = _")
proof-
have "1 + 2 * ?u'⇧2 / (1 - ?u'⇧2) = (1 + ?u'⇧2) / (1 - ?u'⇧2)"
using disc
by (subst add_divide_eq_iff, simp_all)
moreover
have "1 + 2 * ?v'⇧2 / (1 - ?v'⇧2) = (1 + ?v'⇧2) / (1 - ?v'⇧2)"
using disc
by (subst add_divide_eq_iff, simp_all)
moreover
have "sqrt (((1 + 2 * ?u'⇧2 / (1 - ?u'⇧2))⇧2 - 1) * ((1 + 2 * ?v'⇧2 / (1 - ?v'⇧2))⇧2 - 1)) =
(4 * ?u' * ?v') / ((1 - ?u'⇧2) * (1 - ?v'⇧2))" (is "sqrt ?s = ?t")
proof-
have "?s = ?t⇧2"
using disc
apply (subst add_divide_eq_iff, simp)+
apply (subst power_divide)+
apply simp
apply (subst divide_diff_eq_iff, simp)+
apply (simp add: power2_eq_square field_simps)
done
thus ?thesis
using disc
by simp
qed
ultimately
have "?l = ?r"
using disc
by simp (subst add_divide_distrib, simp)
thus ?thesis
by simp
qed
qed
thus ?thesis
using uv' assms
using poincare_distance_formula
by (simp add: Let_def)
qed
lemma poincare_distance_triangle_inequality:
assumes "u ∈ unit_disc" and "v ∈ unit_disc" and "w ∈ unit_disc"
shows "poincare_distance u v + poincare_distance v w ≥ poincare_distance u w" (is "?P' u v w")
proof-
have "∀ w. w ∈ unit_disc ⟶ ?P' u v w" (is "?P v u")
proof (rule wlog_x_axis[where P="?P"])
fix x
assume "is_real x" "0 ≤ Re x" "Re x < 1"
hence "of_complex x ∈ unit_disc"
by (simp add: cmod_eq_Re)
show "?P 0⇩h (of_complex x)"
proof safe
fix w
assume "w ∈ unit_disc"
then obtain w' where w: "w = of_complex w'"
using inf_or_of_complex[of w]
by auto
let ?x = "cmod x" and ?w = "cmod w'" and ?xw = "cmod (x - w')"
have disc: "?x⇧2 < 1" "?w⇧2 < 1"
using unit_disc_cmod_square_lt_1[OF ‹of_complex x ∈ unit_disc›]
using unit_disc_cmod_square_lt_1[OF ‹w ∈ unit_disc›] w
by auto
have "poincare_distance (of_complex x) 0⇩h + poincare_distance 0⇩h w =
arcosh (((1 + ?x⇧2) * (1 + ?w⇧2) + 4 * ?x * ?w) / ((1 - ?x⇧2) * (1 - ?w⇧2)))" (is "_ = arcosh ?r1")
using poincare_distance_formula_zero_sum[OF ‹of_complex x ∈ unit_disc› ‹w ∈ unit_disc›] w
by (simp add: Let_def)
moreover
have "poincare_distance (of_complex x) (of_complex w') =
arcosh (((1 - ?x⇧2) * (1 - ?w⇧2) + 2 * ?xw⇧2) / ((1 - ?x⇧2) * (1 - ?w⇧2)))" (is "_ = arcosh ?r2")
using disc
using poincare_distance_formula[OF ‹of_complex x ∈ unit_disc› ‹w ∈ unit_disc›] w
by (subst add_divide_distrib) simp
moreover
have *: "(1 - ?x⇧2) * (1 - ?w⇧2) + 2 * ?xw⇧2 ≤ (1 + ?x⇧2) * (1 + ?w⇧2) + 4 * ?x * ?w"
proof-
have "(cmod (x - w'))⇧2 ≤ (cmod x + cmod w')⇧2"
using norm_triangle_ineq4[of x w']
by (simp add: power_mono)
thus ?thesis
by (simp add: field_simps power2_sum)
qed
have "arcosh ?r1 ≥ arcosh ?r2"
proof (subst arcosh_mono)
show "?r1 ≥ 1"
using disc
by (smt "*" le_divide_eq_1_pos mult_pos_pos zero_le_power2)
next
show "?r2 ≥ 1"
using disc
by simp
next
show "?r1 ≥ ?r2"
using disc
using *
by (subst divide_right_mono, simp_all)
qed
ultimately
show "poincare_distance (of_complex x) w ≤ poincare_distance (of_complex x) 0⇩h + poincare_distance 0⇩h w"
using ‹of_complex x ∈ unit_disc› ‹w ∈ unit_disc› w
using poincare_distance_formula
by simp
qed
next
show "v ∈ unit_disc" "u ∈ unit_disc"
by fact+
next
fix M u v
assume *: "unit_disc_fix M" "u ∈ unit_disc" "v ∈ unit_disc"
assume **: "?P (moebius_pt M u) (moebius_pt M v)"
show "?P u v"
proof safe
fix w
assume "w ∈ unit_disc"
thus "?P' v u w"
by (metis "*" "**" unit_disc_fix_discI unit_disc_fix_preserve_poincare_distance)
qed
qed
thus ?thesis
using assms
by auto
qed
end