Library Coq.Reals.AltSeries
Require Import Rbase.
Require Import Rfunctions.
Require Import Rseries.
Require Import SeqProp.
Require Import PartSum.
Require Import Max.
Open Local Scope R_scope.
Definition tg_alt (Un:nat -> R) (i:nat) : R := (-1) ^ i * Un i.
Definition positivity_seq (Un:nat -> R) : Prop := forall n:nat, 0 <= Un n.
Lemma CV_ALT_step0 :
forall Un:nat -> R,
Un_decreasing Un ->
Un_growing (fun N:nat => sum_f_R0 (tg_alt Un) (S (2 * N))).
Proof.
intros; unfold Un_growing in |- *; intro.
cut ((2 * S n)%nat = S (S (2 * n))).
intro; rewrite H0.
do 4 rewrite tech5; repeat rewrite Rplus_assoc; apply Rplus_le_compat_l.
pattern (tg_alt Un (S (2 * n))) at 1 in |- *; rewrite <- Rplus_0_r.
apply Rplus_le_compat_l.
unfold tg_alt in |- *; rewrite <- H0; rewrite pow_1_odd; rewrite pow_1_even;
rewrite Rmult_1_l.
apply Rplus_le_reg_l with (Un (S (2 * S n))).
rewrite Rplus_0_r;
replace (Un (S (2 * S n)) + (Un (2 * S n)%nat + -1 * Un (S (2 * S n)))) with
(Un (2 * S n)%nat); [ idtac | ring ].
apply H.
cut (forall n:nat, S n = (n + 1)%nat); [ intro | intro; ring ].
rewrite (H0 n); rewrite (H0 (S (2 * n))); rewrite (H0 (2 * n)%nat); ring.
Qed.
Lemma CV_ALT_step1 :
forall Un:nat -> R,
Un_decreasing Un ->
Un_decreasing (fun N:nat => sum_f_R0 (tg_alt Un) (2 * N)).
Proof.
intros; unfold Un_decreasing in |- *; intro.
cut ((2 * S n)%nat = S (S (2 * n))).
intro; rewrite H0; do 2 rewrite tech5; repeat rewrite Rplus_assoc.
pattern (sum_f_R0 (tg_alt Un) (2 * n)) at 2 in |- *; rewrite <- Rplus_0_r.
apply Rplus_le_compat_l.
unfold tg_alt in |- *; rewrite <- H0; rewrite pow_1_odd; rewrite pow_1_even;
rewrite Rmult_1_l.
apply Rplus_le_reg_l with (Un (S (2 * n))).
rewrite Rplus_0_r;
replace (Un (S (2 * n)) + (-1 * Un (S (2 * n)) + Un (2 * S n)%nat)) with
(Un (2 * S n)%nat); [ idtac | ring ].
rewrite H0; apply H.
cut (forall n:nat, S n = (n + 1)%nat); [ intro | intro; ring ].
rewrite (H0 n); rewrite (H0 (S (2 * n))); rewrite (H0 (2 * n)%nat); ring.
Qed.
Lemma CV_ALT_step2 :
forall (Un:nat -> R) (N:nat),
Un_decreasing Un ->
positivity_seq Un ->
sum_f_R0 (fun i:nat => tg_alt Un (S i)) (S (2 * N)) <= 0.
Proof.
intros; induction N as [| N HrecN].
simpl in |- *; unfold tg_alt in |- *; simpl in |- *; rewrite Rmult_1_r.
replace (-1 * -1 * Un 2%nat) with (Un 2%nat); [ idtac | ring ].
apply Rplus_le_reg_l with (Un 1%nat); rewrite Rplus_0_r.
replace (Un 1%nat + (-1 * Un 1%nat + Un 2%nat)) with (Un 2%nat);
[ apply H | ring ].
cut (S (2 * S N) = S (S (S (2 * N)))).
intro; rewrite H1; do 2 rewrite tech5.
apply Rle_trans with (sum_f_R0 (fun i:nat => tg_alt Un (S i)) (S (2 * N))).
pattern (sum_f_R0 (fun i:nat => tg_alt Un (S i)) (S (2 * N))) at 2 in |- *;
rewrite <- Rplus_0_r.
rewrite Rplus_assoc; apply Rplus_le_compat_l.
unfold tg_alt in |- *; rewrite <- H1.
rewrite pow_1_odd.
cut (S (S (2 * S N)) = (2 * S (S N))%nat).
intro; rewrite H2; rewrite pow_1_even; rewrite Rmult_1_l; rewrite <- H2.
apply Rplus_le_reg_l with (Un (S (2 * S N))).
rewrite Rplus_0_r;
replace (Un (S (2 * S N)) + (-1 * Un (S (2 * S N)) + Un (S (S (2 * S N)))))
with (Un (S (S (2 * S N)))); [ idtac | ring ].
apply H.
ring.
apply HrecN.
ring.
Qed.
A more general inequality
Lemma CV_ALT_step3 :
forall (Un:nat -> R) (N:nat),
Un_decreasing Un ->
positivity_seq Un -> sum_f_R0 (fun i:nat => tg_alt Un (S i)) N <= 0.
Proof.
intros; induction N as [| N HrecN].
simpl in |- *; unfold tg_alt in |- *; simpl in |- *; rewrite Rmult_1_r.
apply Rplus_le_reg_l with (Un 1%nat).
rewrite Rplus_0_r; replace (Un 1%nat + -1 * Un 1%nat) with 0;
[ apply H0 | ring ].
assert (H1 := even_odd_cor N).
elim H1; intros.
elim H2; intro.
rewrite H3; apply CV_ALT_step2; assumption.
rewrite H3; rewrite tech5.
apply Rle_trans with (sum_f_R0 (fun i:nat => tg_alt Un (S i)) (S (2 * x))).
pattern (sum_f_R0 (fun i:nat => tg_alt Un (S i)) (S (2 * x))) at 2 in |- *;
rewrite <- Rplus_0_r.
apply Rplus_le_compat_l.
unfold tg_alt in |- *; simpl in |- *.
replace (x + (x + 0))%nat with (2 * x)%nat; [ idtac | ring ].
rewrite pow_1_even.
replace (-1 * (-1 * (-1 * 1)) * Un (S (S (S (2 * x))))) with
(- Un (S (S (S (2 * x))))); [ idtac | ring ].
apply Rplus_le_reg_l with (Un (S (S (S (2 * x))))).
rewrite Rplus_0_r; rewrite Rplus_opp_r.
apply H0.
apply CV_ALT_step2; assumption.
Qed.
Lemma CV_ALT_step4 :
forall Un:nat -> R,
Un_decreasing Un ->
positivity_seq Un ->
has_ub (fun N:nat => sum_f_R0 (tg_alt Un) (S (2 * N))).
Proof.
intros; unfold has_ub in |- *; unfold bound in |- *.
exists (Un 0%nat).
unfold is_upper_bound in |- *; intros; elim H1; intros.
rewrite H2; rewrite decomp_sum.
replace (tg_alt Un 0) with (Un 0%nat).
pattern (Un 0%nat) at 2 in |- *; rewrite <- Rplus_0_r.
apply Rplus_le_compat_l.
apply CV_ALT_step3; assumption.
unfold tg_alt in |- *; simpl in |- *; ring.
apply lt_O_Sn.
Qed.
This lemma gives an interesting result about alternated series
Lemma CV_ALT :
forall Un:nat -> R,
Un_decreasing Un ->
positivity_seq Un ->
Un_cv Un 0 ->
sigT (fun l:R => Un_cv (fun N:nat => sum_f_R0 (tg_alt Un) N) l).
Proof.
intros.
assert (H2 := CV_ALT_step0 _ H).
assert (H3 := CV_ALT_step4 _ H H0).
assert (X := growing_cv _ H2 H3).
elim X; intros.
apply existT with x.
unfold Un_cv in |- *; unfold R_dist in |- *; unfold Un_cv in H1;
unfold R_dist in H1; unfold Un_cv in p; unfold R_dist in p.
intros; cut (0 < eps / 2);
[ intro
| unfold Rdiv in |- *; apply Rmult_lt_0_compat;
[ assumption | apply Rinv_0_lt_compat; prove_sup0 ] ].
elim (H1 (eps / 2) H5); intros N2 H6.
elim (p (eps / 2) H5); intros N1 H7.
set (N := max (S (2 * N1)) N2).
exists N; intros.
assert (H9 := even_odd_cor n).
elim H9; intros P H10.
cut (N1 <= P)%nat.
intro; elim H10; intro.
replace (sum_f_R0 (tg_alt Un) n - x) with
(sum_f_R0 (tg_alt Un) (S n) - x + - tg_alt Un (S n)).
apply Rle_lt_trans with
(Rabs (sum_f_R0 (tg_alt Un) (S n) - x) + Rabs (- tg_alt Un (S n))).
apply Rabs_triang.
rewrite (double_var eps); apply Rplus_lt_compat.
rewrite H12; apply H7; assumption.
rewrite Rabs_Ropp; unfold tg_alt in |- *; rewrite Rabs_mult;
rewrite pow_1_abs; rewrite Rmult_1_l; unfold Rminus in H6;
rewrite Ropp_0 in H6; rewrite <- (Rplus_0_r (Un (S n)));
apply H6.
unfold ge in |- *; apply le_trans with n.
apply le_trans with N; [ unfold N in |- *; apply le_max_r | assumption ].
apply le_n_Sn.
rewrite tech5; ring.
rewrite H12; apply Rlt_trans with (eps / 2).
apply H7; assumption.
unfold Rdiv in |- *; apply Rmult_lt_reg_l with 2.
prove_sup0.
rewrite (Rmult_comm 2); rewrite Rmult_assoc; rewrite <- Rinv_l_sym;
[ rewrite Rmult_1_r | discrR ].
rewrite double.
pattern eps at 1 in |- *; rewrite <- (Rplus_0_r eps); apply Rplus_lt_compat_l;
assumption.
elim H10; intro; apply le_double.
rewrite <- H11; apply le_trans with N.
unfold N in |- *; apply le_trans with (S (2 * N1));
[ apply le_n_Sn | apply le_max_l ].
assumption.
apply lt_n_Sm_le.
rewrite <- H11.
apply lt_le_trans with N.
unfold N in |- *; apply lt_le_trans with (S (2 * N1)).
apply lt_n_Sn.
apply le_max_l.
assumption.
Qed.
Theorem alternated_series :
forall Un:nat -> R,
Un_decreasing Un ->
Un_cv Un 0 ->
sigT (fun l:R => Un_cv (fun N:nat => sum_f_R0 (tg_alt Un) N) l).
Proof.
intros; apply CV_ALT.
assumption.
unfold positivity_seq in |- *; apply decreasing_ineq; assumption.
assumption.
Qed.
Theorem alternated_series_ineq :
forall (Un:nat -> R) (l:R) (N:nat),
Un_decreasing Un ->
Un_cv Un 0 ->
Un_cv (fun N:nat => sum_f_R0 (tg_alt Un) N) l ->
sum_f_R0 (tg_alt Un) (S (2 * N)) <= l <= sum_f_R0 (tg_alt Un) (2 * N).
Proof.
intros.
cut (Un_cv (fun N:nat => sum_f_R0 (tg_alt Un) (2 * N)) l).
cut (Un_cv (fun N:nat => sum_f_R0 (tg_alt Un) (S (2 * N))) l).
intros; split.
apply (growing_ineq (fun N:nat => sum_f_R0 (tg_alt Un) (S (2 * N)))).
apply CV_ALT_step0; assumption.
assumption.
apply (decreasing_ineq (fun N:nat => sum_f_R0 (tg_alt Un) (2 * N))).
apply CV_ALT_step1; assumption.
assumption.
unfold Un_cv in |- *; unfold R_dist in |- *; unfold Un_cv in H1;
unfold R_dist in H1; intros.
elim (H1 eps H2); intros.
exists x; intros.
apply H3.
unfold ge in |- *; apply le_trans with (2 * n)%nat.
apply le_trans with n.
assumption.
assert (H5 := mult_O_le n 2).
elim H5; intro.
cut (0%nat <> 2%nat);
[ intro; elim H7; symmetry in |- *; assumption | discriminate ].
assumption.
apply le_n_Sn.
unfold Un_cv in |- *; unfold R_dist in |- *; unfold Un_cv in H1;
unfold R_dist in H1; intros.
elim (H1 eps H2); intros.
exists x; intros.
apply H3.
unfold ge in |- *; apply le_trans with n.
assumption.
assert (H5 := mult_O_le n 2).
elim H5; intro.
cut (0%nat <> 2%nat);
[ intro; elim H7; symmetry in |- *; assumption | discriminate ].
assumption.
Qed.
Definition PI_tg (n:nat) := / INR (2 * n + 1).
Lemma PI_tg_pos : forall n:nat, 0 <= PI_tg n.
Proof.
intro; unfold PI_tg in |- *; left; apply Rinv_0_lt_compat; apply lt_INR_0;
replace (2 * n + 1)%nat with (S (2 * n)); [ apply lt_O_Sn | ring ].
Qed.
Lemma PI_tg_decreasing : Un_decreasing PI_tg.
Proof.
unfold PI_tg, Un_decreasing in |- *; intro.
apply Rmult_le_reg_l with (INR (2 * n + 1)).
apply lt_INR_0.
replace (2 * n + 1)%nat with (S (2 * n)); [ apply lt_O_Sn | ring ].
rewrite <- Rinv_r_sym.
apply Rmult_le_reg_l with (INR (2 * S n + 1)).
apply lt_INR_0.
replace (2 * S n + 1)%nat with (S (2 * S n)); [ apply lt_O_Sn | ring ].
rewrite (Rmult_comm (INR (2 * S n + 1))); rewrite Rmult_assoc;
rewrite <- Rinv_l_sym.
do 2 rewrite Rmult_1_r; apply le_INR.
replace (2 * S n + 1)%nat with (S (S (2 * n + 1))).
apply le_trans with (S (2 * n + 1)); apply le_n_Sn.
ring.
apply not_O_INR; discriminate.
apply not_O_INR; replace (2 * n + 1)%nat with (S (2 * n));
[ discriminate | ring ].
Qed.
Lemma PI_tg_cv : Un_cv PI_tg 0.
Proof.
unfold Un_cv in |- *; unfold R_dist in |- *; intros.
cut (0 < 2 * eps);
[ intro | apply Rmult_lt_0_compat; [ prove_sup0 | assumption ] ].
assert (H1 := archimed (/ (2 * eps))).
cut (0 <= up (/ (2 * eps)))%Z.
intro; assert (H3 := IZN (up (/ (2 * eps))) H2).
elim H3; intros N H4.
cut (0 < N)%nat.
intro; exists N; intros.
cut (0 < n)%nat.
intro; unfold Rminus in |- *; rewrite Ropp_0; rewrite Rplus_0_r;
rewrite Rabs_right.
unfold PI_tg in |- *; apply Rlt_trans with (/ INR (2 * n)).
apply Rmult_lt_reg_l with (INR (2 * n)).
apply lt_INR_0.
replace (2 * n)%nat with (n + n)%nat; [ idtac | ring ].
apply lt_le_trans with n.
assumption.
apply le_plus_l.
rewrite <- Rinv_r_sym.
apply Rmult_lt_reg_l with (INR (2 * n + 1)).
apply lt_INR_0.
replace (2 * n + 1)%nat with (S (2 * n)); [ apply lt_O_Sn | ring ].
rewrite (Rmult_comm (INR (2 * n + 1))).
rewrite Rmult_assoc; rewrite <- Rinv_l_sym.
do 2 rewrite Rmult_1_r; apply lt_INR.
replace (2 * n + 1)%nat with (S (2 * n)); [ apply lt_n_Sn | ring ].
apply not_O_INR; replace (2 * n + 1)%nat with (S (2 * n));
[ discriminate | ring ].
replace n with (S (pred n)).
apply not_O_INR; discriminate.
symmetry in |- *; apply S_pred with 0%nat.
assumption.
apply Rle_lt_trans with (/ INR (2 * N)).
apply Rmult_le_reg_l with (INR (2 * N)).
rewrite mult_INR; apply Rmult_lt_0_compat;
[ simpl in |- *; prove_sup0 | apply lt_INR_0; assumption ].
rewrite <- Rinv_r_sym.
apply Rmult_le_reg_l with (INR (2 * n)).
rewrite mult_INR; apply Rmult_lt_0_compat;
[ simpl in |- *; prove_sup0 | apply lt_INR_0; assumption ].
rewrite (Rmult_comm (INR (2 * n))); rewrite Rmult_assoc;
rewrite <- Rinv_l_sym.
do 2 rewrite Rmult_1_r; apply le_INR.
apply (fun m n p:nat => mult_le_compat_l p n m); assumption.
replace n with (S (pred n)).
apply not_O_INR; discriminate.
symmetry in |- *; apply S_pred with 0%nat.
assumption.
replace N with (S (pred N)).
apply not_O_INR; discriminate.
symmetry in |- *; apply S_pred with 0%nat.
assumption.
rewrite mult_INR.
rewrite Rinv_mult_distr.
replace (INR 2) with 2; [ idtac | reflexivity ].
apply Rmult_lt_reg_l with 2.
prove_sup0.
rewrite <- Rmult_assoc; rewrite <- Rinv_r_sym; [ idtac | discrR ].
rewrite Rmult_1_l; apply Rmult_lt_reg_l with (INR N).
apply lt_INR_0; assumption.
rewrite <- Rinv_r_sym.
apply Rmult_lt_reg_l with (/ (2 * eps)).
apply Rinv_0_lt_compat; assumption.
rewrite Rmult_1_r;
replace (/ (2 * eps) * (INR N * (2 * eps))) with
(INR N * (2 * eps * / (2 * eps))); [ idtac | ring ].
rewrite <- Rinv_r_sym.
rewrite Rmult_1_r; replace (INR N) with (IZR (Z_of_nat N)).
rewrite <- H4.
elim H1; intros; assumption.
symmetry in |- *; apply INR_IZR_INZ.
apply prod_neq_R0;
[ discrR | red in |- *; intro; rewrite H8 in H; elim (Rlt_irrefl _ H) ].
apply not_O_INR.
red in |- *; intro; rewrite H8 in H5; elim (lt_irrefl _ H5).
replace (INR 2) with 2; [ discrR | reflexivity ].
apply not_O_INR.
red in |- *; intro; rewrite H8 in H5; elim (lt_irrefl _ H5).
apply Rle_ge; apply PI_tg_pos.
apply lt_le_trans with N; assumption.
elim H1; intros H5 _.
assert (H6 := lt_eq_lt_dec 0 N).
elim H6; intro.
elim a; intro.
assumption.
rewrite <- b in H4.
rewrite H4 in H5.
simpl in H5.
cut (0 < / (2 * eps)); [ intro | apply Rinv_0_lt_compat; assumption ].
elim (Rlt_irrefl _ (Rlt_trans _ _ _ H7 H5)).
elim (lt_n_O _ b).
apply le_IZR.
simpl in |- *.
left; apply Rlt_trans with (/ (2 * eps)).
apply Rinv_0_lt_compat; assumption.
elim H1; intros; assumption.
Qed.
Lemma exist_PI :
sigT (fun l:R => Un_cv (fun N:nat => sum_f_R0 (tg_alt PI_tg) N) l).
Proof.
apply alternated_series.
apply PI_tg_decreasing.
apply PI_tg_cv.
Qed.
Now, PI is defined
Definition PI : R := 4 * match exist_PI with
| existT a b => a
end.
We can get an approximation of PI with the following inequality
Lemma PI_ineq :
forall N:nat,
sum_f_R0 (tg_alt PI_tg) (S (2 * N)) <= PI / 4 <=
sum_f_R0 (tg_alt PI_tg) (2 * N).
Proof.
intro; apply alternated_series_ineq.
apply PI_tg_decreasing.
apply PI_tg_cv.
unfold PI in |- *; case exist_PI; intro.
replace (4 * x / 4) with x.
trivial.
unfold Rdiv in |- *; rewrite (Rmult_comm 4); rewrite Rmult_assoc;
rewrite <- Rinv_r_sym; [ rewrite Rmult_1_r; reflexivity | discrR ].
Qed.
Lemma PI_RGT_0 : 0 < PI.
Proof.
assert (H := PI_ineq 0).
apply Rmult_lt_reg_l with (/ 4).
apply Rinv_0_lt_compat; prove_sup0.
rewrite Rmult_0_r; rewrite Rmult_comm.
elim H; clear H; intros H _.
unfold Rdiv in H;
apply Rlt_le_trans with (sum_f_R0 (tg_alt PI_tg) (S (2 * 0))).
simpl in |- *; unfold tg_alt in |- *; simpl in |- *; rewrite Rmult_1_l;
rewrite Rmult_1_r; apply Rplus_lt_reg_r with (PI_tg 1).
rewrite Rplus_0_r;
replace (PI_tg 1 + (PI_tg 0 + -1 * PI_tg 1)) with (PI_tg 0);
[ unfold PI_tg in |- * | ring ].
simpl in |- *; apply Rinv_lt_contravar.
rewrite Rmult_1_l; replace (2 + 1) with 3; [ prove_sup0 | ring ].
rewrite Rplus_comm; pattern 1 at 1 in |- *; rewrite <- Rplus_0_r;
apply Rplus_lt_compat_l; prove_sup0.
assumption.
Qed.