roseking's picture
Add lean source and binary files (part 68)
7d9acbd verified
Raw History Blame Contribute Delete
21.7 kB
import OAI.Geometry.HarmonicGrowth.InsertedWord
import OAI.Geometry.HarmonicGrowth.Profiles
namespace OAI
noncomputable section
namespace HarmonicCounterexample.CausalConstruction
open Filter Set
open scoped Topology ContDiff
variable {P E : Type*}
/-- Prefixes contain genuine future zero controls, rather than the base point
of an endpoint chart (whose control word need not vanish). -/
def prefixes (zero : P) (choose : ℕ → (ℕ → P) → P) : ℕ → ℕ → P
| 0 => fun _ => zero
| n+1 => Function.update (prefixes zero choose n) n (choose n (prefixes zero choose n))
def limitChoice (zero : P) (choose : ℕ → (ℕ → P) → P) (j : ℕ) : P :=
choose j (prefixes zero choose j)
lemma prefixes_zero {zero : P} {choose : ℕ → (ℕ → P) → P} {n j : ℕ} (h : n ≤ j) :
prefixes zero choose n j=zero := by
induction n with
| zero => rfl
| succ n ih =>
rw [prefixes,Function.update_of_ne (by omega : j ≠ n)]
exact ih (by omega)
lemma prefixes_stable {zero : P} {choose : ℕ → (ℕ → P) → P} {n j : ℕ} (h : j < n) :
prefixes zero choose n j=limitChoice zero choose j := by
induction n with
| zero => omega
| succ n ih =>
by_cases hj : j=n
· subst j; simp [prefixes,limitChoice]
· rw [prefixes,Function.update_of_ne hj]
exact ih (by omega)
/-- Pure dependent choice in chronological order. The predicate may depend on
all past controls, but never on a fictitious future control. -/
theorem causal_choice (zero : P) (admissible : Set P)
(good : ℕ → (ℕ → P) → P → Prop)
(produce : ∀ j s,(∀ i < j,s i ∈ admissible) → (∀ i,j ≤ i → s i=zero) →
∃ x ∈ admissible,good j s x) :
∃ s : ℕ → P,(∀ j,s j ∈ admissible) ∧
∀ j,good j (fun i => if i < j then s i else zero) (s j) := by
classical
let choose (j : ℕ) (s : ℕ → P) : P :=
if h : (∀ i < j,s i ∈ admissible) ∧ (∀ i,j ≤ i → s i=zero)
then (produce j s h.1 h.2).choose else zero
have hg : ∀ j,limitChoice zero choose j ∈ admissible ∧
good j (prefixes zero choose j) (limitChoice zero choose j) := by
intro j
induction j using Nat.strong_induction_on with
| h j ih =>
have hp : ∀ i < j,prefixes zero choose j i ∈ admissible := by
intro i hi
rw [prefixes_stable hi]
exact (ih i hi).1
have hz : ∀ i,j ≤ i → prefixes zero choose j i=zero := fun i hi => prefixes_zero hi
have he : limitChoice zero choose j=(produce j (prefixes zero choose j) hp hz).choose := by
change (if h : (∀ i < j,prefixes zero choose j i ∈ admissible) ∧
(∀ i,j ≤ i → prefixes zero choose j i=zero)
then (produce j (prefixes zero choose j) h.1 h.2).choose else zero)=_
rw [dite_eq_left ⟨hp,hz⟩]
rw [he]
exact (produce j (prefixes zero choose j) hp hz).choose_spec
refine ⟨limitChoice zero choose,fun j => (hg j).1,?_⟩
intro j
have he : prefixes zero choose j=(fun i => if i < j then limitChoice zero choose i else zero) := by
funext i
split_ifs with hi
· exact prefixes_stable hi
· exact prefixes_zero (by omega)
rw [← he]
exact (hg j).2
variable [NormedAddCommGroup E] [NormedSpace ℝ E]
open Schedule
/-- Vector-valued local finiteness for the fixed polynomial schedule. -/
theorem smooth_escaping_sum {f : ℕ → ℝ → E} (hf : ∀ j,ContDiff ℝ ∞ (f j))
(hz : ∀ j t,t ≤ time j → f j t=0) :
ContDiff ℝ ∞ (fun t => ∑' j,f j t) := by
apply contDiff_iff_contDiffAt.2
intro t
obtain ⟨N,hN⟩ := eventually_atTop.1 (time_atTop.eventually_gt_atTop (t+1))
apply (ContDiffAt.sum (s:=Finset.range N) (fun j _ => (hf j).contDiffAt)).congr_of_eventuallyEq
filter_upwards [eventually_lt_nhds (show t<t+1 by linarith)] with u hu
apply tsum_eq_sum
intro j hj
have hj' : N ≤ j := by simpa only [Finset.mem_range,not_lt] using hj
exact hz j u (by linarith [hN j hj'])
omit [NormedSpace ℝ E] in
/-- At every actual time in a period there is exactly one possible increment.
Both endpoints have zero increments, so this holds on closed periods. -/
lemma sum_period {f : ℕ → ℝ → E}
(hz0 : ∀ j t,t ≤ time j → f j t=0)
(hz1 : ∀ j t,time (j+1) ≤ t → f j t=0)
{j : ℕ} {t : ℝ} (hlo : time j ≤ t) (hhi : t ≤ time (j+1)) :
(∑' i,f i t)=f j t := by
apply tsum_eq_single j
intro i hi
rcases lt_or_gt_of_ne hi with hij | hij
· exact hz1 i t ((time_strictMono.monotone hij).trans hlo)
· exact hz0 i t (hhi.trans (time_strictMono.monotone hij))
omit [NormedSpace ℝ E] in
lemma sum_initial {f : ℕ → ℝ → E}
(hz0 : ∀ j t,t ≤ time j → f j t=0) {t : ℝ} (ht : t ≤ time 0) :
(∑' i,f i t)=0 := by
have he : (fun i => f i t)=(fun _ => (0:E)) := by
funext j
exact hz0 j t (ht.trans (time_strictMono.monotone (Nat.zero_le j)))
rw [he,tsum_zero]
omit [NormedSpace ℝ E] in
/-- A future insertion cannot change a previous time, regardless of the
infinite tail of controls. This is the causality needed by regular ODEs. -/
lemma sum_causal {f g : ℕ → ℝ → E} {j : ℕ}
(hf : ∀ i t,t ≤ time i → f i t=0)
(hg : ∀ i t,t ≤ time i → g i t=0)
(he : ∀ i < j,f i=g i) {t : ℝ} (ht : t ≤ time j) :
(∑' i,f i t)=(∑' i,g i t) := by
apply tsum_congr
intro i
by_cases hi : i<j
· rw [he i hi]
· have hti : t ≤ time i := ht.trans (time_strictMono.monotone (by omega))
rw [hf i t hti,hg i t hti]
end HarmonicCounterexample.CausalConstruction
end
noncomputable section
namespace HarmonicCounterexample.AngularStream
open Set Schedule Pulses FiniteControl.SmoothWord
open scoped ContDiff Topology
variable {E : Type*} [NormedAddCommGroup E] [NormedSpace ℝ E] {n : ℕ}
/-- The fixed diagonal leg; initially and during every pulse it is round. -/
def base (qstar : ℝ) (j0 : ℕ) : ℝ → ℝ := Pulses.profile (fun _ _ => 0) qstar j0
lemma base_smooth (qstar : ℝ) (j0 : ℕ) : ContDiff ℝ ∞ (base qstar j0) :=
Pulses.profile_smooth (fun _ => contDiff_const) (by simp) qstar j0
lemma base_initial {qstar t : ℝ} {j0 : ℕ} (ht : t ≤ time j0) : base qstar j0 t=1 :=
Pulses.profile_initial (by simp) ht
lemma base_pulse {qstar t : ℝ} {j0 j : ℕ} (hj : j0 ≤ j)
(hlo : time j ≤ t) (hhi : t ≤ time j+pulse j) : base qstar j0 t=1 := by
have h := Pulses.profile_pulse (z:=fun _ _ => 0) (qstar:=qstar) (by simp) (by simp) hj hlo hhi
simpa [base] using h
lemma base_range {qstar : ℝ} (hq : 1 ≤ qstar) (j0 : ℕ) (t : ℝ) :
base qstar j0 t ∈ Icc 1 qstar := by
by_cases ht : t ≤ time j0
· rw [base_initial ht];exact ⟨le_rfl,hq⟩
· obtain ⟨j,hj,hlo,hhi⟩ := GlobalGrowth.grid_bracket time_atTop j0 (le_of_not_ge ht)
have h := Pulses.profile_bounds_on_period (z:=fun _ _ => 0) (M:=0) (by simp) (by simp)
(by simp) hq hj hlo hhi
simpa [base] using h
/-- Actual angular increment of a period, zero on both open round collars. -/
def increment (r : ℝ) (b : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ) (j : ℕ) (t : ℝ) : E :=
nonlinearBergerWord r b D (s j) (pulse j) ((t-time j)/pulse j)-b
lemma increment_before (r : ℝ) (b : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ)
{j : ℕ} {t : ℝ} (ht : t ≤ time j) : increment r b D s j t=0 := by
have hu : (t-time j)/pulse j ≤ 0 := div_nonpos_of_nonpos_of_nonneg (sub_nonpos.2 ht) (pulse_pos j).le
rw [increment,nonlinearBergerWord_round_before r b D (s j) (pulse j) _
(lt_of_le_of_lt (mul_nonpos_of_nonneg_of_nonpos (duration_pos n).le hu) (by norm_num)),sub_self]
lemma increment_after (r : ℝ) (b : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ)
{j : ℕ} {t : ℝ} (ht : time j+pulse j ≤ t) : increment r b D s j t=0 := by
have hu : 1 ≤ (t-time j)/pulse j := (le_div_iff₀ (pulse_pos j)).2 (by linarith)
have hu' : duration n ≤ duration n*((t-time j)/pulse j) := by
simpa using mul_le_mul_of_nonneg_left hu (duration_pos n).le
rw [increment,nonlinearBergerWord_round_after r b D (s j) (pulse j) _
(lt_of_lt_of_le (by simp [duration]) hu'),sub_self]
lemma increment_zero (r : ℝ) (b : E) (D : Fin n → E) {s : ℕ → Fin n → ℝ}
{j : ℕ} (hj : s j=0) (t : ℝ) : increment r b D s j t=0 := by
have he : packetStretch (s j) (pulse j) ((t-time j)/pulse j)=(fun _ => 1) := by
funext i;simp [packetStretch,hj]
simp only [increment,nonlinearBergerWord,he,nonlinearPacket_round,sub_self]
def activeIncrement (r : ℝ) (b : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ) (j0 j : ℕ) : ℝ → E :=
if j0 ≤ j then increment r b D s j else 0
lemma active_before (r : ℝ) (b : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ) (j0 j : ℕ)
(t : ℝ) (ht : t ≤ time j) : activeIncrement r b D s j0 j t=0 := by
dsimp [activeIncrement];split_ifs
· exact increment_before r b D s ht
· rfl
lemma active_after (r : ℝ) (b : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ) (j0 j : ℕ)
(t : ℝ) (ht : time (j+1) ≤ t) : activeIncrement r b D s j0 j t=0 := by
dsimp [activeIncrement];split_ifs
· apply increment_after r b D s
rw [time_succ] at ht
linarith [pulse_pos j,dwell_pos j]
· rfl
/-- Simultaneously smooth infinite angular coefficient. No angular selector is
used in its definition, so discontinuous labels cannot spoil regularity. -/
def angular (r : ℝ) (b D0 : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ)
(qstar : ℝ) (j0 : ℕ) (t : ℝ) : E :=
PulseTaylor.angular r b D0 (base qstar j0 t)+∑' j,activeIncrement r b D s j0 j t
lemma angular_smooth (r : ℝ) (b D0 : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ)
{qstar : ℝ} (hq : 1 ≤ qstar) (j0 : ℕ)
(hpos : ∀ j,j0 ≤ j → ∀ t i,packetStretch (s j) (pulse j) t i ≠ 0) :
ContDiff ℝ ∞ (angular r b D0 D s qstar j0) := by
have hb : ContDiff ℝ ∞ (fun t => PulseTaylor.angular r b D0 (base qstar j0 t)) := by
have h1 := (base_smooth qstar j0).rpow_const_of_ne (p:=r) (fun t =>
(ne_of_gt (lt_of_lt_of_le zero_lt_one (base_range hq j0 t).1)))
have h2 := (base_smooth qstar j0).rpow_const_of_ne (p:=r-1) (fun t =>
(ne_of_gt (lt_of_lt_of_le zero_lt_one (base_range hq j0 t).1)))
exact (h1.smul contDiff_const).add ((h1.sub h2).smul contDiff_const)
apply hb.add (CausalConstruction.smooth_escaping_sum ?_ (active_before r b D s j0))
intro j
dsimp [activeIncrement]
split_ifs with hj
· exact ((nonlinearBergerWord_smooth r b D (s j) (pulse j) (fun i t => hpos j hj t i)).comp
((contDiff_id.sub contDiff_const).div_const _)).sub contDiff_const
· exact contDiff_const
lemma angular_initial (r : ℝ) (b D0 : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ)
{qstar t : ℝ} {j0 : ℕ} (ht : t ≤ time j0) : angular r b D0 D s qstar j0 t=b := by
have he : (fun j => activeIncrement r b D s j0 j t)=(fun _ => (0:E)) := by
funext j
dsimp [activeIncrement]
split_ifs with hj
· exact increment_before r b D s (ht.trans (time_strictMono.monotone hj))
· rfl
simp only [angular,base_initial ht,PulseTaylor.angular_one,he,tsum_zero,add_zero]
lemma angular_period (r : ℝ) (b D0 : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ)
{qstar t : ℝ} {j0 j : ℕ} (hj : j0 ≤ j) (hlo : time j ≤ t) (hhi : t ≤ time (j+1)) :
angular r b D0 D s qstar j0 t=
PulseTaylor.angular r b D0 (base qstar j0 t)+increment r b D s j t := by
rw [angular,CausalConstruction.sum_period (active_before r b D s j0)
(active_after r b D s j0) hlo hhi]
simp only [activeIncrement,ite_eq_left hj]
lemma angular_pulse (r : ℝ) (b D0 : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ)
{qstar t : ℝ} {j0 j : ℕ} (hj : j0 ≤ j) (hlo : time j ≤ t) (hhi : t ≤ time j+pulse j) :
angular r b D0 D s qstar j0 t=nonlinearBergerWord r b D (s j) (pulse j) ((t-time j)/pulse j) := by
rw [angular_period r b D0 D s hj hlo (by rw [time_succ];linarith [pulse_pos j,dwell_pos j]),
base_pulse hj hlo hhi,PulseTaylor.angular_one,increment]
abel
lemma angular_leg (r : ℝ) (b D0 : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ)
{qstar t : ℝ} {j0 j : ℕ} (hj : j0 ≤ j) (hlo : time j+pulse j ≤ t) (hhi : t ≤ time (j+1)) :
angular r b D0 D s qstar j0 t=PulseTaylor.angular r b D0 (base qstar j0 t) := by
rw [angular_period r b D0 D s hj (by linarith [pulse_pos j]) hhi,
increment_after r b D s hlo,add_zero]
lemma angular_causal (r : ℝ) (b D0 : E) (D : Fin n → E) {s s' : ℕ → Fin n → ℝ}
(qstar : ℝ) (j0 : ℕ) {j : ℕ} (he : ∀ i < j,s i=s' i) {t : ℝ} (ht : t ≤ time j) :
angular r b D0 D s qstar j0 t=angular r b D0 D s' qstar j0 t := by
unfold angular
congr 1
apply CausalConstruction.sum_causal (active_before r b D s j0) (active_before r b D s' j0) ?_ ht
intro i hi
funext u
dsimp [activeIncrement]
split_ifs <;> simp [increment,he i hi]
end HarmonicCounterexample.AngularStream
end
noncomputable section
namespace HarmonicCounterexample.AngularStream
open Set Filter Schedule Pulses FiniteControl.SmoothWord
open scoped ContDiff Topology BigOperators
variable {E : Type*} [NormedAddCommGroup E] [NormedSpace ℝ E] {n : ℕ}
lemma active_summable (r : ℝ) (b : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ)
(j0 : ℕ) (t : ℝ) : Summable (fun j => activeIncrement r b D s j0 j t) := by
obtain ⟨N,hN⟩ := eventually_atTop.1 (time_atTop.eventually_ge_atTop t)
apply summable_of_ne_finset_zero (s:=Finset.range N)
intro j hj
apply active_before r b D s j0 j t
exact hN j (by simpa only [Finset.mem_range,not_lt] using hj)
/-- Updating a previously zero pulse is exactly the actual smooth insertion,
not merely equality of an ideal packet. -/
lemma angular_update (r : ℝ) (b D0 : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ)
(x : Fin n → ℝ) (qstar : ℝ) {j0 j : ℕ} (hj : j0 ≤ j) (hz : s j=0) (t : ℝ) :
angular r b D0 D (Function.update s j x) qstar j0 t=
angular r b D0 D s qstar j0 t+
(nonlinearBergerWord r b D x (pulse j) ((t-time j)/pulse j)-b) := by
classical
have he : (fun i => if i=j then 0 else activeIncrement r b D (Function.update s j x) j0 i t)=
(fun i => if i=j then 0 else activeIncrement r b D s j0 i t) := by
funext i
split_ifs with hi
· rfl
· dsimp [activeIncrement];split_ifs <;> simp [increment,Function.update_of_ne hi]
have hz' : activeIncrement r b D s j0 j t=0 := by
dsimp [activeIncrement];rw [ite_eq_left hj];exact increment_zero r b D hz t
rw [angular,angular,(active_summable r b D (Function.update s j x) j0 t).tsum_eq_add_tsum_ite j,
(active_summable r b D s j0 t).tsum_eq_add_tsum_ite j,he,hz',zero_add]
simp only [activeIncrement,ite_eq_left hj,increment,Function.update_self]
abel
lemma coefficient_update (r : ℝ) (b D0 : E) (D : Fin n → E) (s : ℕ → Fin n → ℝ)
(x : Fin n → ℝ) (qstar : ℝ) (α : ℝ → ℝ) {j0 j : ℕ} (hj : j0 ≤ j) (hz : s j=0) :
(fun t => α t • angular r b D0 D (Function.update s j x) qstar j0 t)=
insertWord (fun t => α t • angular r b D0 D s qstar j0 t) α r b D x (time j) (pulse j) := by
funext t
rw [angular_update r b D0 D s x qstar hj hz t,smul_add]
rfl
def direction (D0 : E) (D : Fin n → E) : Option (Fin n) → E
| none => D0
| some i => D i
lemma angular_representation [NeZero n] (r : ℝ) (b D0 : E) (D : Fin n → E)
(s : ℕ → Fin n → ℝ) {qstar : ℝ} (hq : 1 ≤ qstar) (j0 : ℕ)
(hpos : ∀ j,j0 ≤ j → ∀ t i,packetStretch (s j) (pulse j) t i ∈ Icc (1/2:ℝ) (3/2))
(t : ℝ) : ∃ (d : Option (Fin n)), ∃ q ∈ Icc (1/2:ℝ) (qstar+1/2),
angular r b D0 D s qstar j0 t=PulseTaylor.angular r b (direction D0 D d) q := by
by_cases ht : t ≤ time j0
· refine ⟨none,1,⟨by norm_num,by linarith⟩,?_⟩
rw [angular_initial r b D0 D s ht,PulseTaylor.angular_one]
· obtain ⟨j,hj,hlo,hhi⟩ := GlobalGrowth.grid_bracket time_atTop j0 (le_of_not_ge ht)
by_cases htp : t ≤ time j+pulse j
· obtain ⟨d,hd,_⟩ := exists_unit_round_gap_selector n
let u := (t-time j)/pulse j
refine ⟨some (d u),packetStretch (s j) (pulse j) u (d u),?_,?_⟩
· have hh := hpos j hj u (d u)
exact ⟨hh.1,hh.2.trans (by linarith)⟩
· rw [angular_pulse r b D0 D s hj hlo htp,
nonlinearBergerWord_selected r b D (s j) (pulse j) u (d u) (fun i hi => (hd u i hi).1),
packetStretch_selected (s j) (pulse j) u (d u) (fun i hi => (hd u i hi).1)]
rfl
· refine ⟨none,base qstar j0 t,?_,?_⟩
· have hh := base_range hq j0 t
exact ⟨(by norm_num : (1/2:ℝ) ≤ 1).trans hh.1,hh.2.trans (by linarith)⟩
· exact angular_leg r b D0 D s hj (le_of_not_ge htp) hhi
/-- An actual all-time coefficient bound chosen before the entire sequence of
controls and the starting index. Uniformity is compactness of one fixed angular
family, not a bound on a posteriori chosen histories. -/
theorem angular_uniform_bound [NeZero n] (r : ℝ) (b D0 : E) (D : Fin n → E)
{qstar : ℝ} (hq : 1 ≤ qstar) :
∃ C : ℝ,0 ≤ C ∧ ∀ (s : ℕ → Fin n → ℝ) (j0 : ℕ),
(∀ j,j0 ≤ j → ∀ t i,packetStretch (s j) (pulse j) t i ∈ Icc (1/2:ℝ) (3/2)) →
∀ t,‖angular r b D0 D s qstar j0 t‖ ≤ C := by
classical
have hc (d : Option (Fin n)) : ContinuousOn (PulseTaylor.angular r b (direction D0 D d))
(Icc (1/2:ℝ) (qstar+1/2)) := by
apply ContinuousOn.add
· exact (continuousOn_id.rpow_const (fun t ht => Or.inl (by change t ≠ 0;linarith [ht.1]))).smul continuousOn_const
· exact ((continuousOn_id.rpow_const (fun t ht => Or.inl (by change t ≠ 0;linarith [ht.1]))).sub
(continuousOn_id.rpow_const (fun t ht => Or.inl (by change t ≠ 0;linarith [ht.1])))).smul continuousOn_const
choose C hC using fun d => isCompact_Icc.exists_bound_of_continuousOn (hc d)
let M := ∑ d : Option (Fin n),max (C d) 0
have hM : 0 ≤ M := Finset.sum_nonneg (fun d _ => le_max_right _ _)
refine ⟨M,hM,fun s j0 hpos t => ?_⟩
obtain ⟨d,q,hq',he⟩ := angular_representation r b D0 D s hq j0 hpos t
rw [he]
exact (hC d q hq').trans ((le_max_left _ _).trans
(Finset.single_le_sum (fun i _ => le_max_right (C i) 0) (Finset.mem_univ d)))
end HarmonicCounterexample.AngularStream
end
noncomputable section
namespace HarmonicCounterexample.AngularStream
open Set Schedule Pulses FiniteControl.SmoothWord
open scoped ContDiff InnerProductSpace
variable {E : Type*} [NormedAddCommGroup E] [InnerProductSpace ℝ E] {n : ℕ} [NeZero n]
local instance : NormedAddCommGroup (E →L[ℝ] E) := ContinuousLinearMap.toNormedAddCommGroup
local instance : NormedSpace ℝ (E →L[ℝ] E) := ContinuousLinearMap.toNormedSpace
lemma angular_nonnegative (r B : ℝ) (D0 : E →L[ℝ] E) (D : Fin n → E →L[ℝ] E)
(s : ℕ → Fin n → ℝ) {qstar : ℝ} (hq : 1 ≤ qstar) (j0 : ℕ)
(hpos : ∀ j,j0 ≤ j → ∀ t i,packetStretch (s j) (pulse j) t i ∈ Icc (1/2:ℝ) (3/2))
(hD : ∀ d x,inner ℝ x (direction D0 D d x) ≤ 0)
(hBD : ∀ d x,-(B*‖x‖^2) ≤ inner ℝ x (direction D0 D d x))
(t : ℝ) (x : E) :
0 ≤ inner ℝ x (angular r (B • (1:E →L[ℝ] E)) D0 D s qstar j0 t x) := by
obtain ⟨d,q,hq',he⟩ := angular_representation r (B • (1:E →L[ℝ] E)) D0 D s hq j0 hpos t
rw [he]
exact berger_angular_nonneg r B q (direction D0 D d) (by linarith [hq'.1]) (hD d) (hBD d) x
lemma angular_symmetric (r B : ℝ) (D0 : E →L[ℝ] E) (D : Fin n → E →L[ℝ] E)
(s : ℕ → Fin n → ℝ) {qstar : ℝ} (hq : 1 ≤ qstar) (j0 : ℕ)
(hpos : ∀ j,j0 ≤ j → ∀ t i,packetStretch (s j) (pulse j) t i ∈ Icc (1/2:ℝ) (3/2))
(hD : ∀ d x y,inner ℝ (direction D0 D d x) y=inner ℝ x (direction D0 D d y))
(t : ℝ) (x y : E) :
inner ℝ (angular r (B • (1:E →L[ℝ] E)) D0 D s qstar j0 t x) y=
inner ℝ x (angular r (B • (1:E →L[ℝ] E)) D0 D s qstar j0 t y) := by
obtain ⟨d,q,_,he⟩ := angular_representation r (B • (1:E →L[ℝ] E)) D0 D s hq j0 hpos t
rw [he]
simp only [PulseTaylor.angular,add_apply,smul_apply,
one_apply_eq_self,inner_add_left,inner_add_right,real_inner_smul_left,
inner_smul_right,hD d x y]
omit [NeZero n] in
lemma angular_zero_pulse (r : ℝ) (b D0 : E →L[ℝ] E) (D : Fin n → E →L[ℝ] E)
(s : ℕ → Fin n → ℝ) {qstar t : ℝ} {j0 j : ℕ} (hj : j0 ≤ j) (hz : s j=0)
(hlo : time j ≤ t) (hhi : t ≤ time j+pulse j) : angular r b D0 D s qstar j0 t=b := by
rw [angular_period r b D0 D s hj hlo (by rw [time_succ];linarith [pulse_pos j,dwell_pos j]),
base_pulse hj hlo hhi,PulseTaylor.angular_one,increment_zero r b D hz t,add_zero]
omit [NeZero n] in
/-- The preceding round rest is round even for the first active period, whose
entire previous history is round. This uniformity is essential for recursion. -/
lemma angular_rest (r : ℝ) (b D0 : E →L[ℝ] E) (D : Fin n → E →L[ℝ] E)
(s : ℕ → Fin n → ℝ) {qstar t : ℝ} {j0 j : ℕ}
(hlo : time j+3*pulse j+dwell j ≤ t) (hhi : t ≤ time (j+1)) :
angular r b D0 D s qstar j0 t=b := by
by_cases hj : j0 ≤ j
· rw [angular_leg r b D0 D s hj (by linarith [pulse_pos j,dwell_pos j]) hhi]
have hb : base qstar j0 t=1 := Pulses.profile_rest (by simp) (by simp) hj hlo hhi
rw [hb,PulseTaylor.angular_one]
· apply angular_initial r b D0 D s
exact hhi.trans (time_strictMono.monotone (by omega))
end HarmonicCounterexample.AngularStream
end
end OAI