Extensions of the natural approach to refinements and generalizations of some trigonometric inequalities

In this paper we propose a new method for sharpening and refinements of some trigonometric inequalities. We apply these ideas to some inequalities of Wilker–Cusa–Huygens type.


Introduction
Inequalities involving trigonometric functions are used in many applications in various fields of mathematics such as difference equations and inequalities [1], theory of stability, theory of approximations, etc. A method called the natural approach, proposed by Mortici in [2], uses the idea of comparing functions to their corresponding Taylor polynomials. This method has been successfully applied to prove and approximate a wide category of trigonometric inequalities [3,4].
In this paper we extend the ideas of the natural approach by comparing and replacing functions with their corresponding power series. In particular, we focus on the results of Mortici in [2] related to Wilker-Cusa-Huygens's inequalities and give generalizations and refinements of the inequalities stated in Theorems 1, 2, 3, 4, 5, and 6 in that paper. They are cited below. Statement 1 ([2], Theorem 1) For every 0 < x < π/2, we have - x 4 15 < cos x -sin x x 3 < -x 4 15 + 23x 6 1890 .

Preliminaries
First, let us recall some of the well-known power series expansions that will be used in our proofs. For x ∈ R, the following power series expansions hold: Also, according to [5], for x ∈ R, we have the following power series expansions: and For x ∈ (0, π 2 ), according to [5], the following series expansions hold: and where B i are Bernoulli's numbers.
Theorem WD ([6], Theorem 2) Suppose that f (x) is a real function on (a, b), and that n is a positive integer such that f (k) , then for all x ∈ (a, b) the following inequality holds: Furthermore, if (-1) n f (n) (x) is decreasing on (a, b), then the reversed inequality of (7) holds.
, then for all x ∈ (a, b) the following inequality holds: Furthermore, if f (n) (x) is decreasing on (a, b), then the reversed inequality of (8) holds.
Let us mention that an interesting application of Theorem WD is given in [7], see also [8].

Main results
We need the following theorem for the proofs of Theorems 1, 2, 3, and 4.

Proposition 1
Let the series f (x) = ∞ k=1 (-1) k+1 A(k)x 2k converge for x ∈ (0, c), c ∈ R + . Suppose that the following statements are true: (i) If c < 1, then the sequence {A(k)} k∈N is a positive decreasing sequence that converges to 0.
Then, for all x ∈ (0, c) and for all n ∈ N and m ∈ N , we have and Proof Suppose that c < 1. Then, for every x ∈ (0, c), the positive sequence {A(k)x 2k } k∈N decreases monotonically and lim k→∞ A(k)x 2k = 0. Thus, assertions (9) and (10) immediately follow from Leibniz's theorem for the alternating series.
Suppose now that c ≥ 1. We have Let us introduce the substitution t = x c in the previous power series and consider the series For the assumption A(k) > c 2 A(k + 1), we have Hence, we conclude that for every t ∈ (0, 1) power series (11) satisfies Leibniz's theorem for the alternating series, and for all n, m ∈ N we have 2n k=1 and Returning the variable x = tc to (12) and (13) gives the assertions of proposition.

Refinements of the inequalities in Statement 1
We propose the following improvement and generalization of Statement 1.

Refinements of the inequalities in Statement 2
We propose the following improvement and generalization of Statement 2.

Refinements of the inequalities in Statement 3
We propose the following improvement and generalization of Statement 3.
As the last inequality holds for k ≥ 2, the assertions of Theorem 2 immediately follow from Proposition 1.

Refinements of the inequalities in Statement 4
We propose the following improvement and generalization of Statement 4.
For c = π/2, we have Also, As the last inequality holds for k ≥ 2, the assertions of Theorem 2 immediately follow from Proposition 1.

Refinements of the inequalities in Statement 5
We prove the following generalization of Statement 5.

Theorem 5
For every x ∈ (0, π 2 ) and m ∈ N , m ≥ 2, the following inequalities hold: where B i are Bernoulli's numbers.
Proof of Theorem 5 Consider the function Based on series expansion (5) and (6), we have Since all coefficients are positive, by applying Theorem WD, we get the inequalities in the statement of the theorem.
Remark Let us notice that Theorem WD allows for the approximation error to be estimated. The difference between the right-hand side and the left-hand side of the double inequality in Theorem 5 can be represented by the following function: The maximum values of R n (x) are reached at π 2 , and their values for n = 3, 4, 5 and 6 are 6.97 × 10 -2 , 2.26 × 10 -2 , 6.95 × 10 -3 , and 2.06 × 10 -3 , respectively.

Refinements of the inequalities in Statement 6
We propose the following generalization of Statement 6.

Theorem 6
For every x ∈ (0, π 2 ) and m ∈ N , m ≥ 3, the following inequality holds: where B i are Bernoulli's numbers.
Proof of Theorem 6 Consider the function Based on the series expansion (1) and (4), we have It is easy to verify that 3|B 2k |(2 2k -2) > 1 for k ≥ 2, and that it is equal to 1 for k = 1, therefore all corresponding coefficients are positive. Now, using Theorem WD, we get the inequalities in the statement of the theorem.
Examples For x ∈ (0, π 2 ) and f (x) = 3 x sin x + cos x, we show the inequalities for m = 3, 4, 5, 6. • For m = 3: On the right-hand side we see the inequality from Statement 6. Remark The difference between the right-hand side and the left-hand side of the double inequality in Theorem 6 can be represented by the following function: The maximum values of R n (x) are reached at π 2 , and their values for n = 3, 4, 5, and 6 are 3.20 × 10 -2 , 7.78 × 10 -3 , 1.95 × 10 -3 , and 4.88 × 10 -4 , respectively.

Conclusion
The idea to compare and replace functions with their corresponding power series to get more accurate approximations was used in [9,10], and [7]. Following the same idea, in this paper we extended the natural approach. We proposed and proved new inequalities which represent refinements and generalizations of the inequalities stated in [2], related to Wilker-Cusa-Huygens's inequalities.
Note that proofs of the new inequalities (14), (17), (19), (20), (21), and (22) for any fixed n, m ∈ N can be obtained by substituting x = sin t for t ∈ [0, π 2 ] and using the methods and algorithms developed in [11] and [12]. However, our approach provides proofs for the approximation of the corresponding function by the inequality of an arbitrary degree.