 Research
 Open Access
 Published:
On certain new refinements of FinslerHadwiger inequalities
Journal of Inequalities and Applications volume 2017, Article number: 80 (2017)
Abstract
Several refinements of the FinslerHadwiger inequality and its reverse in the triangle are discussed. A new one parameter family of FinslerHadwiger inequalities and their reverses are proved. This allows us to obtain new bounds for the sum of the squares of the side lengths of a triangle in terms of other elements in the triangle. Finally, these new bounds are compared to known ones.
Introduction
In a triangle \(\triangle ABC\) the angle measures are referred to as A, B and C and the lengths of the sides BC, CA and AB are denoted by a, b and c, respectively. As is customary, the circumradius, the inradius, the semiperimeter and the area of the triangle are denoted by R, r, s and F, respectively. The deviation of the triangle from being equilateral is measured by the quantity
With the above notation, the celebrated Weitzenböck inequality [1] states that
This inequality was published in 1919, but the authors in [2] traced it back to 1897 where it was proposed as Problem 273 in the Romanian magazine Gazeta Matematică (III(2), p.52) by Ionescu. So they proposed to call it the ‘IonescuWeitzenböck inequality’. At least five distinct proofs of (1.1) could be found in [3] and [4].
The IonescuWeitzenbök inequality was later refined by Finsler and Hadwiger (see [5, 6]), to give birth to the HadwigerFinsler inequality,
and its reverse,
with equality in each of these inequalities occurring if and only if the triangle is equilateral. In [7], pp.170174, 11 proofs of (1.1) and several proofs of (1.2) are presented. In [8] we find the following equivalent form of (1.2):
and another chain of refinements of (1.1), namely
Note that the rightmost inequality in this chain is called ‘the PolyaSzegö inequality’ (see [9], p.161, Problem 17.1 and [10], p.116).
It is proved in [11], Theorem 3.1, that inequalities (1.2) and (1.3) are optimal in the sense that if for some real numbers λ and μ, we have
in any triangle \(\triangle ABC\), then \(\lambda\ge3\) and \(\mu\le1\). However, it is proved in [12, 13] that the least λ for which the upper inequality holds in any nonobtuse triangle is \((2\sqrt{3})(3+2\sqrt{2})\approx1.5617\).
Surprisingly, it was noted in [14] that the IonescuWeitzenböck inequality (1.1) is equivalent to FinslerHadwiger inequality (1.2) by showing that the second follows from the first by applying the first to another triangle.
Results and discussion
In this paper we consider refinements of (1.2) and (1.3) in the following form:
where \(\psi^{}\) and \(\psi^{+}\) are functions defined on \([0,1]\), such that \(\psi^{}(1)=\psi^{+}(1)=1\) and \(\psi^{+}(t)<1<\psi^{}(t)\) for \(0\le t<1\). Recall that by Euler’s inequality we have \(2r\le R\) with equality if and only if \(\triangle ABC\) is equilateral. We seek nontrivial \(\psi^{}\) and \(\psi^{+}\) but as beautiful and simple as possible, clearly this requirement is not too mathematical, but I think no one would recall an ugly complicated inequality even if it is very sharp. I will let the reader be the judge on this statement.
In [15] the authors show that one can take \(\psi^{} =\psi_{0}^{} \) where
This lower bound was later improved in [16] where the authors proved using the GarfunkelBankoff inequality [17, 18] that one can take \(\psi^{} =\psi_{1}^{}\) where
which is stronger than \(\psi_{0}^{}\) and more beautiful! In fact, we will see that this strengthened inequality is equivalent to Kooi’s inequality [19] (see also [20], Inequality 5.7).
But what about the upper bound \(\psi^{+}\)? To the best of the knowledge of the author, it seems that no refinement better than the constant \(\psi^{+}_{0}\equiv1\) is known or has been published. In this note we will prove that \(\psi^{+}=\psi_{1}^{+}\) with \(\psi_{1}^{+}(t)=\sqrt{t}\) is a refinement of the reverse FinslerHadwiger inequality, and we will also provide an alternative proof (different from the one given in [16]) of the lower bound with \(\psi_{1}^{}\).
Therefore, our first result concerning the refinement of FinslerHadwiger inequality and its reverse (Theorem 3.6) will be the following:
Further, in Theorem 3.7, the next alternative ‘reverse FinslerHadwiger inequality’ is proposed,
These two reverses are then compared and it is proved in Corollary 3.9 that we have the following chain of inequalities:
In the final step of this investigation we consider a one parameter family of inequalities, namely
and we prove in Corollary 3.10 that it holds for \(\lambda\ge0\), its reverse holds for \(\lambda\le\frac{17+8\sqrt{6}}{95}\) and that the constants 0 and \(\frac{17+8\sqrt{6}}{95}\) are the best possible.
Theorems and proofs
The main tool in our proofs is the fundamental inequality in the triangle. This inequality has a long history, the reader may consult [21] or [22], Chapter 1, for more information.
Theorem 3.1
The fundamental inequality
Consider three positive numbers s, r and R with \(2r\le R\). Then s, r and R are, respectively, the semiperimeter, the inradius and the circumradius of a triangle if and only if
where
Moreover, the upper equality holds in (3.1) if and only if the triangle is isosceles and the radian measure of its apex is larger or equal to \(\pi/3\). Similarly, the lower equality holds in (3.1) if and only if the triangle is isosceles and the radian measure of its apex is smaller or equal to \(\pi/3\).
We will use also several algebraic inequalities that are gathered in the next two lemmas.
Lemma 3.2
The following inequalities hold.

(i)
For \(t\in[0,1]\) we have
$$ 2+5t\frac{t^{2}}{4}+2(1t)^{3/2}\le\frac{(8+t)^{2}}{4(4t)}. $$(3.4) 
(ii)
For \(t\in[0,1]\) we have
$$ 2+5t\frac{t^{2}}{4}2(1t)^{3/2}\ge\frac{t ( \sqrt{128+16 t+3t^{2}}\sqrt{3} t )^{2}}{16}. $$(3.5) 
(iii)
For \(t\in[0,1]\) we have
$$ 2+5t\frac{t^{2}}{4}2(1t)^{3/2} \ge\frac{ t(8+t)^{2}}{4(2+t)}. $$(3.6)
Moreover, equality holds in any of the above inequalities if and only if \(t\in\{0,1\}\).
Proof
(i) Let
then it is readily seen that
and (3.4) follows.
To prove (ii) we note that, after expanding the square, the proposed inequality is equivalent to \(G(t)\ge0\) for \(t\in[0,1]\) where
Now, for \(t\in[0,1]\) we have
hence
Finally
This concludes the proof of (3.5).
(iii) According to (ii) we only have to prove that
Again, expanding the square we see that the proposed inequality is equivalent to \(H(t)\ge0\) for \(t\in[0,1]\) where
But
and (3.6) follows. □
Remark 3.3
The proof shows that the ‘ugly’ inequality (3.5) is stronger than (3.6).
Proposition 3.4
Let the semiperimeter, the inradius and the circumradius of a triangle be denoted by s, r and R, respectively. Then
Moreover, equality holds in any of the above inequalities if and only if the triangle is equilateral.
Proof
The wellknown Euler inequality states that \(2r\le R\) (with equality if and only if the triangle is equilateral) so we may apply (3.4) from Lemma 3.2 with \(t=2r/R\) and then multiply both sides of the resulting inequality by \(R^{2}\) to obtain
Combining this with Theorem 3.1 we get
or equivalently
Similarly, applying (3.6) from Lemma 3.2 with \(t=2r/R\), multiplying both sides of the resulting inequality by \(R^{2}\), and finally making use of Theorem 3.1 we obtain
which is equivalent to (3.9).
Doing the same manipulation with (3.5) we get
and according to Theorem 3.1 we obtain
or equivalently
Finally, squaring and rearranging we get (3.10). □
Remark 3.5
Inequalities (3.8) and (3.9) are not new, see [23], Inequality 2.15, and particularly, (3.8) is Kooi’s inequality [19].
Now we are ready to give an alternative proof of the refinement of FinslerHadwiger inequality given in [16] and to present our refinement of its reverse, as announced in the Introduction.
Theorem 3.6
In a triangle \(\triangle ABC\) let the inradius, the circumradius, the area and the lengths of sides opposite to angles A, B and C be denoted by r, R, F, a, b and c, respectively. If \(Q=(ab)^{2}+(bc)^{2}+(ca)^{2}\) then the following two inequalities hold:
Moreover, equality holds in any of the above inequalities if and only if the triangle is equilateral.
Proof
Using Heron’s formula and the facts that \(F=rs\) and \(abc=4RF\) we obtain
Hence
Moreover, because \(4s^{2}=a^{2}+b^{2}+c^{2}+2(ab+bc+ca)\) we also conclude that
It follows that
So (3.14) follows from (3.8) by multiplying both sides by 4r and using (3.18).
Similarly, from (3.16) and (3.17) we conclude that
So (3.15) follows from (3.10). □
In the next result we provide an alternative ‘reverse FinslerHadwiger inequality’.
Theorem 3.7
Let \(\triangle ABC\) be a triangle, and let the lengths of sides opposite to angles A, B and C be denoted by a, b and c, respectively. If F represents the area of \(\triangle ABC\), and if \(Q=(ab)^{2}+(bc)^{2}+(ca)^{2}\) then the following inequality holds:
with equality if and only if the triangle is equilateral.
Proof
Indeed, this follows from
where we used (3.9) from Proposition 3.4 and (3.17). □
In fact Theorems 3.6 and 3.7 yield two different upper bounds for the sum of squares of the side lengths of a triangle in terms of its area, and they are both candidates to be called the ‘reverse FinslerHadwiger inequality’. A legitimate question is the following: Are these two bounds comparable? Some numerical experimentation shows that the upper bound given in Theorem 3.7 is better than the corresponding upper bound given in Theorem 3.6. This suggests that the inequality
is valid in any triangle. Rearranging this inequality we see that it is equivalent to
Writing this as
leads to the following question what the least constant μ is such that the following inequality holds:
Indeed, this inequality holds clearly for \(\mu\ge1\) because we have \(s\sqrt{1+\frac{R}{r}}\ge4R+r\), and it does hold for \(\mu=1\) according to Proposition 3.4. Testing it with a triangle \(\triangle ABC\) having side lengths \(a=1\), \(b=c=t\) with \(t>1/2\) but near \(1/2\) shows that a necessary condition for such an inequality to hold is \(\mu\ge\lambda=52\sqrt{6}=\frac{1}{5+2\sqrt{6}}\approx0.10102\). This observation is confirmed in the next theorem.
Theorem 3.8
Let \(\triangle ABC\) be a triangle, and let r, R and s represent the inradius, the circumradius and the semiperimeter of \(\triangle ABC\), respectively. Then the following inequality holds:
for all \(\mu\ge\lambda=52\sqrt{6}=\frac{1}{5+2\sqrt{6}}\), with equality if and only if the triangle is equilateral.
Proof
The previous discussion shows that it is enough to prove the inequality for \(\mu=\lambda\). Note that the proposed inequality is equivalent to the fact that s is larger than the positive root of a second degree polynomial, that is,
and according to Theorem 3.1, for a given r and R, the smallest possible value of s is attained when \(\triangle ABC\) is isosceles. Therefore, it is enough to prove the proposed inequality for isosceles triangles.
Now, consider an isosceles triangle \(\triangle ABC\). Since the desired inequality is homogeneous we may suppose that the side lengths are \(a=1\) and \(b=c=1/(2x)\) with \(0< x<1\). In this case we have
Let \(f(x)\) be defined for \(x\in(0,1)\) by
That is,
We can arrange this as follows:
where
Now, in g we consider \(u=\sqrt{\frac{1+x}{x}}\) as a new variable so that \(x=\frac{1}{u^{2}1}\) for \(u>\sqrt{2}\). We have
with
Noting that \(P(\sqrt{2})=P(\sqrt{3})=P'(\sqrt{3})=0\) we conclude that P factors as follows:
Thus
Or
which is clearly positive for \(x\in(0,1)\) and vanishes only if \(x=1/2\). On the other hand
and consequently
which is also positive for \(x\in(0,1)\) and vanishes only if \(x=1/2\). Thus \(f(x)\ge0\) for \(x\in(0,1)\) with equality if and only if \(x=1/2\). The theorem is proved. □
Taking \(\mu=3/4\) we obtain the next corollary.
Corollary 3.9
Let \(\triangle ABC\) be a triangle, and let the lengths of sides opposite to angles A, B and C be denoted by a, b and c respectively. If r, R and F represent the inradius, the circumradius and the area of \(\triangle ABC\), and if \(Q=(ab)^{2}+(bc)^{2}+(ca)^{2}\) then the following inequality holds:
with equality if and only if the triangle is equilateral. Consequently
In fact Theorem 3.8 allows us to give the following ‘parametric FinslerHadwiger inequality’.
Corollary 3.10
Let a, b, c, R, r and F be the side lengths of a triangle \(\triangle ABC\), its circumradius, inradius and its area, respectively, and let \(Q=(ab)^{2}+(bc)^{2}+(ca)^{2}\). The following inequality:
holds for \(\lambda\ge0\), and its reverse holds for \(\lambda\le\frac{8\sqrt{6}+17}{95}\approx0.38522\), with equality if and only if the triangle is equilateral.
Proof
Note that (3.21) is equivalent to
So, according to Theorem 3.6 and Corollary 3.9, the proposed inequality (3.21) does hold for \(\lambda\ge0\).
Now, using (3.18) and the fact that \(4s^{2}=3(a^{2}+b^{2}+c^{2})Q\) we see that Theorem 3.7 can be rephrased as follows:
for \(\mu\ge52\sqrt{6}\). Now if we suppose that \(\mu\in[52\sqrt{6},3/4)\) we conclude that
Setting \(\lambda=\frac{1}{4\mu3}\) we get
for \(\lambda\in (\infty,\frac{1}{8\sqrt{6}17} ]\), and the corollary is proved. □
Remark 3.11
This inequality is the best of its kind in the sense that if \(\lambda \in (\frac{8\sqrt{6}+17}{95},0 )\) then there are triangles \(\triangle ABC\) that satisfy (3.21) and others that violate it. Indeed, testing (3.21) with an isosceles triangle with side lengths \(a=1\), \(b=c=t\ge1/2\) for large t shows that the condition \(\lambda\ge0\) is necessary for its validity. Testing its reverse for t near \(1/2\) (but larger than \(1/2\)) shows that the condition \(\lambda\le\frac{8\sqrt{6}+17}{95}\) is necessary for the validity of the reverse.
Remark 3.12
For \(\lambda\le\frac{8\sqrt{6}+17}{95}\) we have two different lower bounds for \(a^{2}+b^{2}+c^{2}\) given by Theorem 3.6 and Corollary 3.10, respectively. Testing the difference with our famous isosceles triangle with side lengths \(a=1\), \(b=c=t\ge1/2\) shows that these two lower bounds are not comparable.
Conclusion
In this work, we considered the problem of refining the FinslerHadwiger inequality and its reverse in the triangle. Several refinements are proposed and compared, and an optimal parametric refinement of this inequality and its reverse is proved.
References
Weitzenböck, R: Über eine Ungleichung in der Dreiecksgeometrie. Math. Z. 5(12), 137146 (1919)
Stoica, E, Minculete, N, Barbu, C: New aspects of IonescuWeitzenböck’s inequality. Balk. J. Geom. Appl. 21(2), 95101 (2016)
Sándor, J: Geometric Inequalities. Ed. Dacia, Cluj (1988) (in Hungarian)
Sándor, J: On the Lemoine point of a triangle. Lucr. Semin. Did. Mat. 16, 175182 (2000) (in Romanian)
Finsler, P, Hadwiger, H: Relationen im Dreieck. Comment. Math. Helv. 10(1), 316326 (1937)
Hadwiger, H: Jahresber. Dtsch. Math.Ver. 49, 3539 (1939)
Engel, A: ProblemSolving Strategies. Springer, New York (1998)
Sándor, J: On the cotangent inequality of a triangle. Octogon Math. Mag. 12(2A), 738740 (2004)
Pólya, G, Szegö, G: Problems and Theorems in Analysis, vol. II. Grundlehren der Mathematischen Wissenschaften, vol. 20. Springer, New York (1976). Translated from the revised and enlarged Fourth German Edition by CE Billigheimer
Sándor, J: On the geometry of equilateral triangles. Forum Geom. 5, 107117 (2005)
Marinescu, DŞ, Monea, M, Opincariu, M, Stroe, M: Note on HadwigerFinsler’s inequalities. J. Math. Inequal. 6(1), 5764 (2012)
Cipu, A: Optimal reverse FinslerHadwiger inequalities. Gaz. Mat., Ser. A 34, 6168 (2012)
Wu, SH, Chu, YM: Geometric interpretation of Blundon’s inequality and Ciamberlini’s inequality. J. Inequal. Appl. 2014, 381 (2014)
Wu, SH, Zhang, ZH, Xiao, ZG: On Weitzenböck inequality and its generalizations. Res. Group Math. Inequal. Appl. 6(4), Article 14 (2003). http://rgmia.org/v6n4.php
Lupu, C, Pohoaţă, C: Sharpening the HadwigerFinsler inequality. Crux Math. 34(2), 97101 (2008)
Wu, SH, Debnath, L: Parametrized GarfunkelBanko inequality and improved FinslerHadwiger. Appl. Math. Lett. 23, 331336 (2010)
Bankoff, L: Solution to Problem 825. Crux Math. 10(5), 168 (1984)
Garfunkel, J: Problem 825. Crux Math. 9(2), 79 (1983)
Kooi, O: Inequalities for the triangle. Simon Stevin 32, 97101 (1958)
Bottema, O, Djordjević, RZ, Janic, RR, Mitrinović, DS, Vasić, PM: Geometric Inequalities. WoltersNoordhoff, Groningen (1969)
Blundon, WJ: Inequalities associated with the triangle. Can. Math. Bull. 8, 615626 (1965)
Mitrinović, DS, Pečarić, JE, Volenec, V: Recent Advances in Geometric Inequalities. Mathematics and Its Applications Series. Springer, Dordrecht (1989)
Djodjević, R: Some inequalities for triangle: old and new results. In: Rassias, TM, Srivastava, HM (eds.) Analytic and Geometric Inequalities and Applications, pp. 6992. Springer, Dordrech (1999)
Author information
Authors and Affiliations
Corresponding author
Additional information
Competing interests
The author declares that there is no competing interests with any individual or institution.
Author’s contributions
The author declares that this work was carried out by himself. The author read and approved the final manuscript.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Cite this article
Kouba, O. On certain new refinements of FinslerHadwiger inequalities. J Inequal Appl 2017, 80 (2017). https://doi.org/10.1186/s1366001713565
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s1366001713565
MSC
 11B68
Keywords
 triangle inequalities
 FinslerHadwiger inequality