Skip to main content

New lower bounds for the minimum M-eigenvalue of elasticity M-tensors and applications

Abstract

M-eigenvalues of elasticity M-tensors play an important role in nonlinear elasticity and materials. In this paper, we present several new lower bounds for the minimum M-eigenvalue of elasticity M-tensors and propose numerical examples to illustrate the efficiency of the obtained results. As applications, we provide several checkable sufficient conditions for the strong ellipticity and positive definiteness of irreducible elasticity M-tensors.

Introduction

A tensor \(\mathcal{A}=(a_{ijkl})\in \mathbb{E}_{4,n}\) is called a fourth-order real partially symmetric tensor if

$$ a_{ijkl}=a_{jikl}=a_{ijlk}, \quad i,j,l,k \in [n], $$

where \([n]=\{1,2,\ldots ,n\}\). The tensor of elastic moduli for a linearly anisotropic elastic solid is a fourth-order real partially symmetric tensor [1], and the components of such a tensor are considered as the coefficients of the following optimization problem:

$$ \textstyle\begin{cases} \min \quad f({\mathbf{x}},{\mathbf{y}})=\mathcal{A}{\mathbf{x}}{\mathbf{y}}{\mathbf{x}}{\mathbf{y}}= {\sum_{i,j,k,l\in [n]}}a_{ijkl}x_{i}x_{j}y_{k}y_{l}, \\ \text{s.t.}\quad {\mathbf{x}}^{T}{\mathbf{x}}=1, {\mathbf{y}}^{T}{\mathbf{y}}=1, \\ \hphantom{\text{s.t.}\quad } {\mathbf{x}}, {\mathbf{y}}\in \mathbb{R}^{n}. \end{cases} $$
(1.1)

Problem (1.1) has applications in the ordinary ellipticity and strong ellipticity and nonlinear elastic materials analysis [228]. The strong ellipticity condition is stated as \(f({\mathbf{x}},{\mathbf{y}})>0\) for all nonzero vectors \({\mathbf{x}}, {\mathbf{y}}\in \mathbb{R}^{n} \), which guarantees the existence of solutions of basic boundary-value problems of elastostatics and ensures an elastic material to satisfy some mechanical properties [29]. In fact, the KKT condition of (1.1) can be regarded as the following definition of M-eigenvalues.

Definition 1.1

([1])

Let \(\mathcal{A}\in \mathbb{E}_{4,n}\). If there are \(\lambda \in \mathbb{R}\) and \(\mathbf{x}, \mathbf{y}\in \mathbb{R}^{n}\backslash \{\mathbf{0}\} \) such that

$$ \textstyle\begin{cases} \mathcal{A}\mathbf{x} \mathbf{y}^{2}=\lambda \mathbf{x}, \\ \mathcal{A}\mathbf{x}^{2} \mathbf{y}=\lambda \mathbf{y}, \\ \mathbf{x}^{T}\mathbf{x}=1, \\ \mathbf{y}^{T}\mathbf{y}=1, \end{cases} $$
(1.2)

where \((\mathcal{A}\mathbf{x} \mathbf{y}^{2})_{i}={\sum_{j,k,l \in [n]}}a_{ijkl}x_{j}y_{k}y_{l}\), and \((\mathcal{A} \mathbf{x}^{2} \mathbf{y})_{l}={\sum_{i,j,k \in [n]}}a_{ijkl}x_{i}x_{j}y_{k}\), then the scalar λ is called an M-eigenvalue of \(\mathcal{A}\), and x, y are called the corresponding left and right M-eigenvectors of \(\mathcal{A}\), respectively.

Furthermore, Han et al. revealed that the strong ellipticity condition holds if and only if the smallest M-eigenvalue is positive [1]. Recently, Ding et al. [30] investigated a fourth-order structured partially symmetric tensors named elasticity M-tensors, and some sufficient conditions for the strong ellipticity were provided. Since the strong ellipticity condition and M-positive definiteness can be identified by the smallest M-eigenvalue, He et al. [31] proposed some lower bounds for the minimum M-eigenvalue of elasticity M-tensors.

In this paper, we present several new bounds for the minimum M-eigenvalue of elasticity M-tensors. We prove that the bounds are tighter than those proposed in [31]. Numerical examples illustrate the efficiency of the obtained results. As applications, we give some checkable sufficient conditions for the strong ellipticity and positive definiteness of elasticity tensors.

Main results

For an elasticity tensor \(\mathcal{A} \in \mathbb{E}_{4,n}\), its M-spectral radius is denoted by

$$ \rho (\mathcal{A})=\max \bigl\{ \vert \lambda \vert : \lambda \text{ is an M-eigenvalue of }\mathcal{A} \bigr\} . $$

The identity tensor \(\mathcal{I} = (e_{ijkl})\in \mathbb{E}_{4,n} \) is defined by

$$ e_{ijkl}= \textstyle\begin{cases} 0& \text{if $i = j$, $k = l$}, \\ 1& \text{otherwise}. \end{cases} $$

Let \(\alpha _{i} =\max_{l\in [n]}\{a_{iill}\}\), \(\beta _{l} = \max_{i\in [n]}\{a_{iill}\}\), and

$$\begin{aligned}& r_{i}(\mathcal{A})=\sum_{j,k,l\in [n],k\neq l} \vert {a}_{ijkl} \vert ,\qquad \gamma _{i}=\sum _{j\in [n],j\neq i}\max_{l\in [n]}\bigl\{ \vert {a}_{ijll} \vert \bigr\} ,\\& r_{i}^{i}(\mathcal{A})= \sum_{k,l\in [n],k\neq l} \vert {a}_{iikl} \vert ,\qquad R_{i}( \mathcal{A})= r_{i}(\mathcal{A})+\gamma _{i}, \\& c_{l}(\mathcal{A})=\sum_{i,j,k\in [n],i\neq j} \vert {a}_{ijkl} \vert ,\qquad \delta _{l}=\sum _{k\in [n],k\neq l}\max_{i\in [n]} \bigl\{ \vert {a}_{iikl} \vert \bigr\} ,\\& c_{l}^{l}(\mathcal{A})= \sum_{i,j\in [n],i\neq j} \vert {a}_{ijll} \vert ,\qquad C_{l}(\mathcal{A})=c_{l}(\mathcal{A})+\delta _{l}. \end{aligned}$$

To continue, we need the following definitions and technical results.

Definition 2.1

([30])

A tensor \(\mathcal{A}\in \mathbb{E}_{4,n}\) is called an elasticity M-tensor if there exist a nonnegative tensor \(\mathcal{B}\in \mathbb{E}_{4,n}\) and a real number \(s\geq \rho (\mathcal{B})\) such that \(\mathcal{A} = s\mathcal{I}-\mathcal{B}\), where \(\rho (\mathcal{B})\) is the M-spectral radius of \(\mathcal{B}\). Furthermore, if \(s>\rho (\mathcal{B})\), then \(\mathcal{A}\) is called a nonsingular elasticity M-tensor.

Definition 2.2

([32])

A tensor \(\mathcal{A} = (a_{i_{1}i_{2}\dots i_{m}})\) of order m and dimension n is called reducible if there exists a nonempty proper index subset \(J\in \{1,2,\dots ,n\}\subset [n]\) such that

$$ a_{i_{1}i_{2}\dots i_{m}}=0,\quad \forall i_{1}\in J, \forall i_{2} \dots i_{m} \in [n] \backslash J. $$

If \(\mathcal{A}\) is not reducible, then we say that \(\mathcal{A}\) is irreducible.

Theorem 2.1

([31])

Let\(\mathcal{A}=({{a}}_{ijkl})\in \mathbb{E}_{4,n}\)be an irreducible and nonnegative partially symmetric tensor, and let\(\tau (\mathcal{A})\)be the minimal M-eigenvalue of\(\mathcal{A}\). Then\(\tau (\mathcal{A})\geq 0\)is an M-eigenvalue of\(\mathcal{A}\)with positive eigenvectors. Moreover, there exist a nonnegative tensor\(\mathcal{B}\)and a real number\(c\geq \rho (\mathcal{B})\)such that\(\mathcal{A} = c\mathcal{I} -\mathcal{B}\).

Theorem 2.2

([31])

Let\(\mathcal{A}=({{a}}_{ijkl})\in \mathbb{E}_{4,n}\)be an irreducible elasticity M-tensor. Then

$$ \tau (\mathcal{A})\leq \min_{i,l\in [n]}\{a_{iill}\}. $$

Theorem 2.3

([31])

Let\(\mathcal{A}=({{a}}_{ijkl})\in \mathbb{E}_{4,n}\)be an irreducible elasticity M-tensor. Then

$$ \tau (\mathcal{A})\geq \max \Bigl\{ \min_{i\in [n] } \alpha _{i}-R_{i}( \mathcal{A}), \min_{l\in [n]} \beta _{l}-C_{l}(\mathcal{A}) \Bigr\} . $$

Now we are in a position to propose some lower bounds for \(\tau (\mathcal{A})\).

Theorem 2.4

Let\(\mathcal{A}=({{a}}_{ijkl})\in \mathbb{E}_{4,n}\)be an irreducible elasticity M-tensor. Then the minimum M-eigenvalue satisfies

$$ \tau (\mathcal{A})\geq \max \Bigl\{ \min_{i,j\in [n],i\neq j } \bigl\{ \eta _{1}(\mathcal{A})\bigr\} , \min_{k,l\in [n],k\neq l } \bigl\{ \eta _{2}(\mathcal{A})\bigr\} \Bigr\} , $$

where\(\eta _{1}(\mathcal{A})= \frac{\alpha _{i}- r_{i}^{i}(\mathcal{A})+\alpha _{j}-\Delta _{i,j}^{\frac{1}{2}}}{2}\), \(\eta _{2}(\mathcal{A})= \frac{\beta _{k}- c_{k}^{k}(\mathcal{A})+\beta _{l}-\varTheta _{k,l}^{\frac{1}{2}}}{2}\), and

$$\begin{aligned}& \Delta _{i,j}=\bigl(\alpha _{i}- r_{i}^{i}( \mathcal{A})-\alpha _{j}\bigr)^{2}+4\bigl(r_{i}( \mathcal{A})- r_{i}^{i}(\mathcal{A})+\gamma _{i} \bigr)R_{j}(\mathcal{A}), \\& \varTheta _{k,l}=\bigl(\beta _{k}- c_{k}^{k}( \mathcal{A})-\beta _{l}\bigr)^{2}+4\bigl(c_{k}( \mathcal{A})- c_{k}^{k}(\mathcal{A})+\delta _{k} \bigr)C_{l}(\mathcal{A}). \end{aligned}$$

Proof

By Theorem 2.1 suppose that \({\mathbf{x}}=\{x_{i}\}_{i=1}^{n}>0\in \mathbb{R}^{n}\) and \({\mathbf{y}}=\{y_{l}\}_{l=1}^{n}>0\in \mathbb{R}^{n}\) are the corresponding left and right M-eigenvectors, respectively. Let \(x_{p}\geq x_{s} \geq \max_{i\in [n], i\neq p,s}\{x_{i}\}\). From the pth equation of \(\mathcal{A}\mathbf{x} \mathbf{y}^{2}=\tau (\mathcal{A}) \mathbf{x} \) in (1.2) we obtain

$$ \begin{aligned} \tau (\mathcal{A}){x}_{p}&=\sum _{j,k,l\in [n]}{a}_{pjkl}{x}_{j}{y}_{k}{y}_{l} \\ & =\sum_{k,l\in [n],k\neq l}{a}_{ppkl}{x}_{p}{y}_{k}{y}_{l}+ \sum_{j,k,l \in [n],j\neq p,k\neq l}{a}_{pjkl}{x}_{j}{y}_{k}{y}_{l} \\ &\quad {}+\sum_{j,l \in [n],j\neq p}{a}_{pjll}{x}_{j}{y}_{l}^{2}+ \sum_{l\in [n]}{a}_{ppll}{x}_{p}{y}_{l}^{2}, \end{aligned} $$

that is,

$$ \begin{aligned} &\sum_{l\in [n]}{a}_{ppll}{x}_{p}{y}_{l}^{2}- \tau (\mathcal{A}){x}_{p}\\ &\quad = -\sum_{k,l\in [n],k\neq l}{a}_{ppkl}{x}_{p}{y}_{k}{y}_{l}- \sum_{j,k,l \in [n],j\neq p,k\neq l}{a}_{pjkl}{x}_{j}{y}_{k}{y}_{l} -\sum_{j,l \in [n],j\neq p}{a}_{pjll}{x}_{j}{y}_{l}^{2}. \end{aligned} $$

Let \(\alpha _{p} =\min_{l\in [n]}\{a_{ppll}\}\). It follows from Theorem 2.2 that

$$ \begin{aligned} 0&\leq \bigl(\alpha _{p}-\tau (\mathcal{A}) \bigr){x}_{p} \leq \biggl(\sum_{l\in [n]}{a}_{ppll}{x}_{p}{y}_{l}^{2}- \tau (\mathcal{A})\biggr){x}_{p} \\ &\leq \sum_{k,l\in [n],k\neq l} \vert {a}_{ppkl} \vert {x}_{p}+\sum_{j,k,l\in [n],j \neq p,k\neq l} \vert {a}_{pjkl} \vert {x}_{s} +\sum_{j,l\in [n],j\neq p} \vert {a}_{pjll} \vert {x}_{s} \bigl\vert {y}_{l}^{2} \bigr\vert . \end{aligned} $$

Note that

$$ \begin{aligned} \sum_{j,l\in [n],j\neq p} \vert {a}_{pjll} \vert {x}_{s} \bigl\vert {y}_{l}^{2} \bigr\vert &=\sum_{j \in [n],j\neq p} \biggl(\sum_{l\in [n]} \vert {a}_{pjll} \vert \bigl\vert {y}_{l}^{2} \bigr\vert \biggr){x}_{s} \\ &\leq \sum_{j\in [n],j\neq p}\max_{l\in [n]} \vert {a}_{pjll} \vert \biggl(\sum_{l \in [n]} \bigl\vert {y}_{l}^{2} \bigr\vert \biggr){x}_{s} \\ &= \sum_{j\in [n],j\neq p}\max_{l\in [n]}\bigl\{ \vert {a}_{pjll} \vert \bigr\} {x}_{s}. \end{aligned} $$

Furthermore,

$$ \bigl(\alpha _{p}-\tau (\mathcal{A})- r_{p}^{p}(\mathcal{A})\bigr){x}_{p}\leq \bigl(r_{p}( \mathcal{A})- r_{p}^{p}(\mathcal{A})+ \gamma _{p}\bigr){x}_{s} . $$
(2.1)

From the sth equation of \(\mathcal{A}\mathbf{x} \mathbf{y}^{2}=\tau (\mathcal{A}) \mathbf{x} \) in (1.2) we have

$$ \begin{aligned} \tau (\mathcal{A}){x}_{s}&=\sum _{j,k,l\in [n]}{a}_{sjkl}{x}_{j}{y}_{k}{y}_{l} \\ & =\sum_{j,k,l\in [n],k\neq l}{a}_{sjkl}{x}_{j}{y}_{k}{y}_{l} +\sum_{j,l \in [n],j\neq s}{a}_{sjll}{x}_{j}{y}_{l}^{2}+ \sum_{l\in [n]}{a}_{ssll}{y}_{l}^{2}x_{s}. \end{aligned} $$

Let \(\alpha _{s} =\min_{l\in [n]}\{a_{ssll}\}\). It follows from Theorem 2.2 that

$$ \bigl(\alpha _{s}-\tau (\mathcal{A}) \bigr){x}_{s}\leq R_{s}(\mathcal{A}){x}_{p} . $$
(2.2)

Multiplying (2.1) and (2.2), we have

$$ \bigl(\alpha _{p}-\tau (\mathcal{A})- r_{p}^{p}( \mathcal{A})\bigr) \bigl(\alpha _{s}- \tau (\mathcal{A})\bigr)\leq \bigl(r_{p}(\mathcal{A})- r_{p}^{p}(\mathcal{A})+ \gamma _{p}\bigr)R_{s}(\mathcal{A}) , $$

which means that

$$ \tau (\mathcal{A})\geq \frac{\alpha _{p}- r_{p}^{p}(\mathcal{A})+\alpha _{s}-\Delta _{p,s}^{\frac{1}{2}}}{2} , $$
(2.3)

where \(\Delta _{p,s}=(\alpha _{p}- r_{p}^{p}(\mathcal{A})-\alpha _{s})^{2}+4(r_{p}( \mathcal{A})- r_{p}^{p}(\mathcal{A})+\gamma _{p})R_{s}(\mathcal{A})\).

On the other hand, let \(|y_{q}|\geq |y_{t}| \geq \max_{l\in [n], l\neq q,t}|y_{l}|\). From the qth equation of \(\mathcal{A}\mathbf{x}^{2} \mathbf{y}=\tau (\mathcal{A}) \mathbf{y} \) in (1.2) it follows that

$$ \begin{aligned} \tau (\mathcal{A}){y}_{q}&= \sum_{i,j,k\in [n]}{a}_{ijkq}{x}_{i}{x}_{j}{y}_{k} \\ & =\sum_{i,j\in [n],i\neq j}{a}_{ijqq}{x}_{i}{x}_{j}{y}_{q}+ \sum_{i,j,k \in [n],i\neq j,k\neq q}{a}_{ijkq}{x}_{i}{x}_{j}{y}_{k} +\sum_{i,k \in [n],k\neq q}{a}_{iikq}{x}_{i}^{2}{y}_{k}+ \sum_{i\in [n]}{a}_{iiqq}{y}_{q}{x}_{i}^{2}. \end{aligned} $$

Let \(\beta _{q} =\min_{i\in [n]}\{a_{iiqq}\}\). It follows from Theorem 2.2 that

$$ \begin{aligned} 0 &\leq \bigl(\beta _{q}-\tau (\mathcal{A})\bigr){y}_{q}\leq \biggl(\sum _{i\in [n]}{a}_{iiqq}{y}_{q}{x}_{i}^{2}- \tau (\mathcal{A})\biggr){y}_{q} \\ & =-\sum_{i,j\in [n],i\neq j}{a}_{ijqq}{x}_{i}{x}_{j}{y}_{q}- \sum_{i,j,k \in [n],i\neq j,k\neq q}{a}_{ijkq}{x}_{i}{x}_{j}{y}_{k} -\sum_{i,k \in [n],k\neq q}{a}_{iikq}{x}_{i}^{2}{y}_{k} \\ & \leq \sum_{i,j\in [n],i\neq j} \vert {a}_{ijqq} \vert {y}_{q}+\sum_{i,j,k\in [n],i \neq j,k\neq q} \vert {a}_{ijkq} \vert {y}_{t} +\sum_{i,k\in [n],k\neq q} \vert {a}_{iikq} \vert {y}_{t}, \end{aligned} $$

that is,

$$ \bigl(\beta _{q}-\tau (\mathcal{A})- c_{q}^{q}(\mathcal{A})\bigr){y}_{q}\leq \bigl(c_{q}( \mathcal{A})- c_{q}^{q}(\mathcal{A})+ \delta _{q}\bigr){y}_{t}. $$
(2.4)

From the tth equation of \(\mathcal{A}\mathbf{x}^{2} \mathbf{y}=\tau (\mathcal{A}) \mathbf{y} \) in (1.2) we obtain

$$ \begin{aligned} \tau (\mathcal{A}){y}_{t}&= \sum_{i,j,k\in [n]}{a}_{ijkt}{x}_{i}{x}_{j}{y}_{k} \\ & =\sum_{i,j,k\in [n],i\neq j}{a}_{ijkt}{x}_{i}{x}_{j}{y}_{k} +\sum_{i,k \in [n],k\neq t}{a}_{iikt}{x}_{i}^{2}{y}_{k}+ \sum_{i\in [n]}{a}_{iitt}{x}_{i}^{2}y_{t}. \end{aligned} $$

Let \(\beta _{t} =\min_{i\in [n]}\{a_{iitt}\}\). This yields

$$ \bigl(\beta _{t}-\tau (\mathcal{A}) \bigr){y}_{t}\leq C_{t}(\mathcal{A}){y}_{q} .$$
(2.5)

Multiplying (2.4) and (2.5), we have

$$ \bigl(\beta _{q}-\tau (\mathcal{A})- c_{q}^{q}(\mathcal{A})\bigr) \bigl(\beta _{t}- \tau (\mathcal{A})\bigr)\leq \bigl(c_{q}(\mathcal{A})- c_{q}^{q}(\mathcal{A})+ \delta _{q} \bigr)C_{t}(\mathcal{A}), $$
(2.6)

which means that

$$ \tau (\mathcal{A})\geq \frac{\beta _{q}- c_{q}^{q}(\mathcal{A})+\beta _{t}-\varTheta _{q,t}^{\frac{1}{2}}}{2} , $$
(2.7)

where \(\varTheta _{q,t}=(\beta _{q}- c_{p}^{p}(\mathcal{A})-\beta _{t})^{2}+4(c_{p}( \mathcal{A})- c_{p}^{p}(\mathcal{A})+\delta _{q})C_{t}(\mathcal{A})\). Then the conclusion follows. □

Next, we compare the bound in Theorem 2.3 with that in Theorem 2.4 and obtain the following conclusion.

Theorem 2.5

Let\(\mathcal{A}=({{a}}_{ijkl})\in \mathbb{E}_{4,n}\)be an irreducible elasticity M-tensor. Then

$$ \begin{aligned} \tau (\mathcal{A})&\geq \max \Bigl\{ \min_{i,j\in [n],i\neq j } \bigl\{ \eta _{1}(\mathcal{A})\bigr\} , \min_{k,l\in [n],k\neq l } \bigl\{ \eta _{2}(\mathcal{A})\bigr\} \Bigr\} \\ &\geq \max \Bigl\{ \min _{i\in [n] } \bigl\{ \alpha _{i}-R_{i}( \mathcal{A})\bigr\} , \min_{l\in [n]} \bigl\{ \beta _{l}-C_{l}( \mathcal{A})\bigr\} \Bigr\} . \end{aligned} $$

Proof

We first show that \(\min_{i,j\in [n],i\neq j } \{\eta _{1}(\mathcal{A})\}\geq \min_{i\in [n] } \{\alpha _{i}-R_{i}(\mathcal{A})\}\) and divide the argument into two cases.

Case 1. For any \(i,j\in [n]\), \(i\neq j\), if \(\alpha _{i}-R_{i}(\mathcal{A})\leq \alpha _{j}-R_{j}(\mathcal{A})\), then

$$ \alpha _{j}-\alpha _{i}+R_{i}( \mathcal{A})\geq R_{j}(\mathcal{A}) \geq 0. $$
(2.8)

From (2.8) we deduce

$$ \begin{aligned} &\bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})-\alpha _{j} \bigr)^{2}+4\bigl(r_{i}( \mathcal{A})- r_{i}^{i}( \mathcal{A})+\gamma _{i}\bigr)R_{j}(\mathcal{A}) \\ &\quad \leq \bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})- \alpha _{j}\bigr)^{2}+4\bigl(r_{i}( \mathcal{A})- r_{i}^{i}(\mathcal{A})+\gamma _{i}\bigr) \bigl( \alpha _{j}-\alpha _{i}+R_{i}( \mathcal{A})\bigr) \\ &\quad =\bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})-\alpha _{j}\bigr)^{2}+4\bigl(R_{i}( \mathcal{A})- r_{i}^{i}(\mathcal{A})\bigr) \bigl(\alpha _{j}- \alpha _{i}+R_{i}( \mathcal{A})\bigr) \\ &\quad =\bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})-\alpha _{j}\bigr)^{2}+4\bigl(R_{i}( \mathcal{A})- r_{i}^{i}(\mathcal{A})\bigr) \bigl(\alpha _{j}- \alpha _{i}+r_{i}^{i}( \mathcal{A})-r_{i}^{i}( \mathcal{A})+R_{i}(\mathcal{A})\bigr) \\ &\quad =\bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})-\alpha _{j}\bigr)^{2}+4\bigl(R_{i}( \mathcal{A})- r_{i}^{i}(\mathcal{A})\bigr) \bigl(\alpha _{j}- \alpha _{i}+r_{i}^{i}( \mathcal{A})\bigr)+4 \bigl(R_{i}(\mathcal{A})- r_{i}^{i}(\mathcal{A}) \bigr)^{2} \\ &\quad =\bigl(\alpha _{j}-\alpha _{i}- r_{i}^{i}( \mathcal{A})+2R_{i}(\mathcal{A})\bigr)^{2}. \end{aligned} $$

Thus

$$ \begin{aligned} & \frac{1}{2}\Bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})+\alpha _{j}-\sqrt{\bigl( \alpha _{i}- r_{i}^{i}( \mathcal{A})-\alpha _{j}\bigr)^{2}+4\bigl(r_{i}( \mathcal{A})- r_{i}^{i}(\mathcal{A})+\gamma _{i} \bigr)R_{j}(\mathcal{A})}\Bigr) \\ &\quad \geq \frac{1}{2}\bigl(\alpha _{i}- r_{i}^{i}( \mathcal{A})+\alpha _{j}-\bigl( \alpha _{j}-\alpha _{i}- r_{i}^{i}(\mathcal{A})+2R_{i}( \mathcal{A})\bigr)\bigr) \\ &\quad =\alpha _{i}-R_{i}(\mathcal{A}), \end{aligned} $$

which means that

$$ \begin{aligned} &\frac{1}{2}\min_{i,j\in [n],i\neq j }\Bigl\{ \alpha _{i}- r_{i}^{i}( \mathcal{A})+\alpha _{j}-\sqrt{\bigl(\alpha _{i}- r_{i}^{i}( \mathcal{A})- \alpha _{j}\bigr)^{2}+4\bigl(r_{i}( \mathcal{A})- r_{i}^{i}(\mathcal{A})+\gamma _{i} \bigr)R_{j}( \mathcal{A})}\Bigr\} \\ &\quad \geq \min_{i\in [n] }\bigl\{ \alpha _{i}-R_{i}( \mathcal{A})\bigr\} .\end{aligned} $$

Case 2. For any \(i,j\in [n]\), \(i\neq j\), if \(\alpha _{i}-R_{i}(\mathcal{A})\geq \alpha _{j}-R_{j}(\mathcal{A})\), then

$$ \alpha _{i}-r_{i}(\mathcal{A})- \alpha _{j}+R_{j}(\mathcal{A})\geq \gamma _{i}. $$
(2.9)

From (2.9) we have

$$ \begin{aligned} &\bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})-\alpha _{j} \bigr)^{2}+4\bigl(r_{i}( \mathcal{A})- r_{i}^{i}( \mathcal{A})+\gamma _{i}\bigr)R_{j}(\mathcal{A}) \\ &\quad \leq \bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})- \alpha _{j}\bigr)^{2}+4\bigl(r_{i}( \mathcal{A})- r_{i}^{i}(\mathcal{A})+ \alpha _{i}-r_{i}( \mathcal{A})- \alpha _{j}+R_{j}(\mathcal{A}) \bigr)R_{j}(\mathcal{A}) \\ &\quad = \bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})- \alpha _{j}\bigr)^{2}+4\bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})- \alpha _{j}+R_{j}( \mathcal{A})\bigr)R_{j}( \mathcal{A}) \\ &\quad = \bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})- \alpha _{j}+2R_{j}( \mathcal{A})\bigr)^{2}. \end{aligned} $$

Then

$$ \begin{aligned} & \frac{1}{2}\Bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})+\alpha _{j}-\sqrt{\bigl( \alpha _{i}- r_{i}^{i}( \mathcal{A})-\alpha _{j}\bigr)^{2}+4\bigl(r_{i}( \mathcal{A})- r_{i}^{i}(\mathcal{A})+\gamma _{i} \bigr)R_{j}(\mathcal{A})}\Bigr) \\ &\quad \geq \frac{1}{2}(\alpha _{i}- r_{i}^{i}( \mathcal{A})+\alpha _{j}-\bigl( \alpha _{i}- r_{i}^{i}(\mathcal{A})-\alpha _{j}+2R_{j}( \mathcal{A})\bigr) \\ &\quad =\alpha _{j}-R_{j}(\mathcal{A}), \end{aligned} $$

which implies

$$ \begin{aligned} &\frac{1}{2}\min_{i,j\in [n],i\neq j }\Bigl\{ \alpha _{i}- r_{i}^{i}( \mathcal{A})+\alpha _{j}-\sqrt{\bigl(\alpha _{i}- r_{i}^{i}( \mathcal{A})- \alpha _{j}\bigr)^{2}+4\bigl(r_{i}( \mathcal{A})- r_{i}^{i}(\mathcal{A})+\gamma _{i} \bigr)R_{j}( \mathcal{A})}\Bigr\} \\ &\quad \geq \min_{j\in [n] }\bigl\{ \alpha _{j}-R_{j}( \mathcal{A})\bigr\} .\end{aligned} $$

Therefore we obtain \(\min_{i,j\in [n],i\neq j } \{\eta _{1}(\mathcal{A})\}\geq \min_{i\in [n] } \{\alpha _{i}-R_{i}(\mathcal{A})\}\).

Similarly, we have \(\min_{i,j\in [n],i\neq j } \{\eta _{2}(\mathcal{A})\}\geq \min_{l\in [n] } \{\beta _{l}-C_{l}(\mathcal{A})\}\). Thus we deduce

$$ \max \Bigl\{ \min_{i,j\in [n],i\neq j } \bigl\{ \eta _{1}( \mathcal{A})\bigr\} , \min_{k,l\in [n],k\neq l } \bigl\{ \eta _{2}( \mathcal{A})\bigr\} \Bigr\} \geq \max \Bigl\{ \min_{i\in [n] } \bigl\{ \alpha _{i}-R_{i}(\mathcal{A}) \bigr\} , \min _{l\in [n]} \bigl\{ \beta _{l}-C_{l}(\mathcal{A}) \bigr\} \Bigr\} , $$

and the desired result follows. □

In what follows, we propose another lower bound for \(\tau (\mathcal{A})\).

Theorem 2.6

Let\(\mathcal{A}=({{a}}_{ijkl})\in \mathbb{E}_{4,n}\)be an irreducible elasticity M-tensor. Then

$$ \begin{aligned} \tau (\mathcal{A})&\geq \max \Bigl\{ \min_{i,j\in [n],i\neq j } \bigl\{ \theta _{1}(\mathcal{A}),\alpha _{i}- r_{i}^{i}( \mathcal{A}), \alpha _{j}- r_{j}^{j}(\mathcal{A}) \bigr\} ,\\ &\quad \min_{k,l\in [n],k\neq l } \bigl\{ \theta _{2}(\mathcal{A}), \beta _{k}- c_{k}^{k}(\mathcal{A}),\beta _{l}- c_{l}^{l}(\mathcal{A})\bigr\} \Bigr\} , \end{aligned} $$

where

$$\begin{aligned}& \theta _{1}(\mathcal{A})= \frac{(\alpha _{i}- r_{i}^{i}(\mathcal{A}))+(\alpha _{j}-r_{j}^{j}(\mathcal{A}))-\varOmega _{i,j}^{\frac{1}{2}}}{2}, \\& \theta _{2}(\mathcal{A})= \frac{(\beta _{k}- c_{k}^{k}(\mathcal{A}))+(\beta _{l}-c_{l}^{l}(\mathcal{A}))-\varPhi _{k,l}^{\frac{1}{2}}}{2}, \\& \varOmega _{i,j}=\bigl(\alpha _{i}- r_{i}^{i}( \mathcal{A})-\bigl(\alpha _{j}-r_{j}^{j}( \mathcal{A})\bigr)\bigr)^{2}+4\bigl(R_{i}( \mathcal{A})-r_{i}^{i}(\mathcal{A})\bigr) \bigl(R_{j}( \mathcal{A})- r_{j}^{j}(\mathcal{A}) \bigr), \end{aligned}$$

and

$$ \varPhi _{k,l}=\bigl(\beta _{k}- c_{k}^{k}( \mathcal{A})-\bigl(\beta _{l}-c_{l}^{l}( \mathcal{A})\bigr)\bigr)^{2}+4\bigl(C_{k}( \mathcal{A})-c_{k}^{k}(\mathcal{A})\bigr) \bigl(C_{l}( \mathcal{A})- c_{l}^{l}(\mathcal{A}) \bigr). $$

Proof

Let \(\tau (\mathcal{A})\) be the minimal M-eigenvalue of tensor \(\mathcal{A}\). From Theorem 2.1 we suppose that \({\mathbf{x}}=\{x_{i}\}_{i=1}^{n}>0\in \mathbb{R}^{n}\) and \({\mathbf{y}}=\{y_{l}\}_{l=1}^{n}>0\in \mathbb{R}^{n}\) are the corresponding left and right M-eigenvectors, respectively. Let \(x_{p}\geq x_{s} \geq \max_{i\in [n], i\neq p,s}\{x_{i}\}\). From the sth equation of \(\mathcal{A}\mathbf{x} \mathbf{y}^{2}=\tau (\mathcal{A}) \mathbf{x} \) in (1.2) we have

$$ \begin{aligned} \tau (\mathcal{A}){x}_{s}&=\sum _{j,k,l\in [n]}{a}_{sjkl}{x}_{j}{y}_{k}{y}_{l} \\ & =\sum_{k,l\in [n],k\neq l}{a}_{sskl}{x}_{s}{y}_{k}{y}_{l}+ \sum_{j,k,l \in [n],j\neq s,k\neq l}{a}_{sjkl}{x}_{j}{y}_{k}{y}_{l} +\sum_{j,l \in [n],j\neq s}{a}_{sjll}{x}_{j}{y}_{l}^{2}+ \sum_{l\in [n]}{a}_{ssll}{y}_{l}^{2}x_{s}. \end{aligned} $$

Let \(\alpha _{s} =\min_{l\in [n]}\{a_{ssll}\}\). It follows from Theorem 2.2 that

$$ \begin{aligned} 0&\leq \bigl(\alpha _{s}-\tau (\mathcal{A}) \bigr){x}_{s} \leq \biggl(\sum_{l\in [n]}{a}_{ssll}{x}_{s}{y}_{l}^{2}- \tau (\mathcal{A})\biggr){x}_{p} \\ &\leq \sum_{k,l\in [n],k\neq l} \vert {a}_{sskl} \vert {x}_{s}+\sum_{j,k,l\in [n],j \neq s,k\neq l} \vert {a}_{pjkl} \vert {x}_{p} +\sum_{j,l\in [n],j\neq s} \vert {a}_{sjll} \vert {x}_{p} \bigl\vert {y}_{l}^{2} \bigr\vert . \end{aligned} $$

Moreover,

$$ \bigl(\alpha _{s}-\tau (\mathcal{A})- r_{s}^{s}(\mathcal{A})\bigr){x}_{s}\leq \bigl(r_{s}( \mathcal{A})- r_{s}^{s}( \mathcal{A})+\gamma _{s}\bigr){x}_{p} . $$
(2.10)

When \(\alpha _{s}- r_{s}^{s}(\mathcal{A})>\tau (\mathcal{A})\) or \(\alpha _{p}- r_{p}^{p}(\mathcal{A})>\tau (\mathcal{A})\), multiplying (2.1) and (2.10), we have

$$ \bigl(\alpha _{p}-\tau (\mathcal{A})- r_{p}^{p}(\mathcal{A})\bigr) \bigl(\alpha _{s}- \tau (\mathcal{A})- r_{s}^{s}(\mathcal{A})\bigr) \leq \bigl(r_{p}(\mathcal{A})- r_{p}^{p}( \mathcal{A})+ \gamma _{p}\bigr) \bigl(r_{s}(\mathcal{A})- r_{s}^{s}(\mathcal{A})+ \gamma _{s}\bigr) , $$
(2.11)

that is,

$$ \tau (\mathcal{A})\geq \frac{(\alpha _{p}- r_{p}^{p}(\mathcal{A}))+(\alpha _{s}-r_{s}^{s}(\mathcal{A}))-\varOmega _{p,s}^{\frac{1}{2}}}{2}, $$
(2.12)

where \(\varOmega _{p,s}=(\alpha _{p}- r_{p}^{p}(\mathcal{A})-(\alpha _{s}-r_{s}^{s}( \mathcal{A})))^{2}+4(R_{p}(\mathcal{A})-r_{p}^{p}(\mathcal{A}))(R_{s}( \mathcal{A})- r_{s}^{s}(\mathcal{A}))\).

On the other hand, let \(|y_{q}|\geq |y_{t}| \geq \max_{l\in [n], l\neq q,t}|y_{l}|\). From the tth equation of \(\mathcal{A}\mathbf{x}^{2} \mathbf{y}=\tau (\mathcal{A}) \mathbf{y} \) in (1.2) we obtain

$$ \begin{aligned} \tau (\mathcal{A}){y}_{t}&= \sum_{j,k,l\in [n]}{a}_{ijkt}{x}_{i}{x}_{j}{y}_{k} \\ & =\sum_{i,j\in [n],i\neq j}{a}_{ijtt}{x}_{i}{x}_{j}{y}_{t}+ \sum_{i,j,k \in [n],i\neq j,k\neq t}{a}_{ijkt}{x}_{i}{x}_{j}{y}_{k} +\sum_{i,k \in [n],k\neq t}{a}_{iikt}{x}_{i}^{2}{y}_{k}+ \sum_{i\in [n]}{a}_{iitt}{x}_{i}^{2}y_{t}. \end{aligned} $$

Let \(\beta _{t} =\min_{i\in [n]}\{a_{iitt}\}\). It follows from Theorem 2.2 that

$$ \begin{aligned} 0& \leq \bigl(\beta _{t}- \tau (\mathcal{A})\bigr){y}_{t}\leq \biggl(\sum _{i\in [n]}{a}_{iitt}{y}_{t}{x}_{i}^{2}- \tau (\mathcal{A})\biggr){y}_{t} \\ & =-\sum_{i,j\in [n],i\neq j}{a}_{ijtt}{x}_{i}{x}_{j}{y}_{t}- \sum_{i,j,k \in [n],i\neq j,k\neq t}{a}_{ijkt}{x}_{i}{x}_{j}{y}_{k} -\sum_{i,k \in [n],k\neq t}{a}_{iikt}{x}_{i}^{2}{y}_{k} \\ & \leq \sum_{i,j\in [n],i\neq j} \vert {a}_{ijtt} \vert {y}_{t}+\sum_{i,j,k\in [n],i \neq j,k\neq t} \vert {a}_{ijkt} \vert {y}_{q} +\sum_{i,k\in [n],k\neq t} \vert {a}_{iikt} \vert {x}_{i}^{2}{y}_{q}, \end{aligned} $$

that is,

$$ \bigl(\beta _{t}-\tau (\mathcal{A})- c_{t}^{t}(\mathcal{A})\bigr){y}_{t}\leq \bigl(c_{t}( \mathcal{A})- c_{t}^{t}( \mathcal{A})+\delta _{t}\bigr){y}_{q}. $$
(2.13)

When \(\beta _{t}- c_{t}^{t}(\mathcal{A})>\tau (\mathcal{A})\) or \(\beta _{q}- c_{q}^{q}(\mathcal{A})>\tau (\mathcal{A})\), multiplying (2.6) and (2.13), we have

$$ \bigl(\beta _{q}-\tau (\mathcal{A})- c_{q}^{q}(\mathcal{A})\bigr) \bigl(\beta _{t}- \tau (\mathcal{A})- c_{t}^{t}(\mathcal{A})\bigr) \leq \bigl(c_{q}(\mathcal{A})- c_{q}^{q}( \mathcal{A})+ \gamma _{q}\bigr) \bigl(c_{t}(\mathcal{A})- c_{t}^{t}(\mathcal{A})+ \gamma _{t}\bigr) , $$
(2.14)

which means that

$$ \tau (\mathcal{A})\geq \frac{(\beta _{q}- c_{q}^{q}(\mathcal{A}))+(\beta _{t}-c_{t}^{t}(\mathcal{A}))-\varPhi _{q,t}^{\frac{1}{2}}}{2}, $$
(2.15)

where \(\varPhi _{q,t}=(\beta _{q}- c_{q}^{q}(\mathcal{A})-(\beta _{t}-c_{t}^{t}( \mathcal{A})))^{2}+4(C_{q}(\mathcal{A})-c_{q}^{q}(\mathcal{A}))(C_{t}( \mathcal{A})- c_{t}^{t}(\mathcal{A}))\). □

Next, we compare the bound in Theorem 2.3 with that in Theorem 2.6 and obtain the following result.

Theorem 2.7

Let\(\mathcal{A}=({{a}}_{ijkl})\in \mathbb{E}_{4,n}\)be an irreducible elasticity M-tensor. Then

$$ \begin{aligned} \tau (\mathcal{A})&\geq \max \Bigl\{ \min _{i,j\in [n],i\neq j } \bigl\{ \theta _{1}(\mathcal{A}),\alpha _{i}- r_{i}^{i}(\mathcal{A}), \alpha _{j}- r_{j}^{j}(\mathcal{A})\bigr\} ,\\ &\quad \min _{k,l\in [n],k\neq l } \bigl\{ \theta _{2}(\mathcal{A}),\beta _{k}- c_{k}^{k}(\mathcal{A}),\beta _{l}- c_{l}^{l}(\mathcal{A})\bigr\} \Bigr\} \\ &\geq \max \Bigl\{ \min_{i\in [n] } \bigl\{ \alpha _{i}-R_{i}( \mathcal{A})\bigr\} , \min_{l\in [n]} \bigl\{ \beta _{l}-C_{l}( \mathcal{A})\bigr\} \Bigr\} .\end{aligned} $$

Proof

We will show \(\min_{i,j\in [n],i\neq j } \{\theta _{1}(\mathcal{A}), \alpha _{i}- r_{i}^{i}(\mathcal{A}), \alpha _{j}- r_{j}^{j}( \mathcal{A})\}\geq \min_{i\in [n] } \{\alpha _{i}-R_{i}( \mathcal{A})\}\) and divide the argument into two cases.

Case 1. For any \(i,j\in [n]\), \(i\neq j\), if \(\alpha _{i}-R_{i}(\mathcal{A})\leq \alpha _{j}-R_{j}(\mathcal{A})\), then from (2.8) we have

$$ \begin{aligned} &\bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})-\bigl(\alpha _{j}-r_{j}^{j}( \mathcal{A})\bigr)\bigr)^{2}+4\bigl(R_{i}( \mathcal{A})-r_{i}^{i}(\mathcal{A})\bigr) \bigl(R_{j}( \mathcal{A})- r_{j}^{j}(\mathcal{A}) \bigr) \\ &\quad \leq \bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})- \bigl(\alpha _{j}-r_{j}^{j}( \mathcal{A})\bigr) \bigr)^{2}\\ &\qquad {}+4\bigl(R_{i}(\mathcal{A})-r_{i}^{i}( \mathcal{A})\bigr) \bigl( \alpha _{j}-\alpha _{i}+r_{i}^{i}( \mathcal{A})- r_{j}^{j}(\mathcal{A})+R_{i}( \mathcal{A})-r_{i}^{i}(\mathcal{A})\bigr) \\ &\quad =\bigl( \alpha _{j}-\alpha _{i}+r_{i}^{i}( \mathcal{A})- r_{j}^{j}( \mathcal{A})+2R_{i}( \mathcal{A})-2r_{i}^{i}(\mathcal{A})\bigr)^{2}. \end{aligned} $$

Since

$$ \begin{aligned} &\alpha _{j}-\alpha _{i}+r_{i}^{i}(\mathcal{A})- r_{j}^{j}( \mathcal{A})+2R_{i}(\mathcal{A})-2r_{i}^{i}( \mathcal{A}) \\ &\quad =\alpha _{j}-\alpha _{i}+R_{i}( \mathcal{A})-R_{j}(\mathcal{A})+R_{j}( \mathcal{A})- r_{j}^{j}(\mathcal{A})+R_{i}( \mathcal{A})-r_{i}^{i}( \mathcal{A})\geq 0, \end{aligned} $$

we have

$$ \begin{aligned} \theta _{1}(\mathcal{A})&= \frac{(\alpha _{i}- r_{i}^{i}(\mathcal{A}))+(\alpha _{j}-r_{j}^{j}(\mathcal{A}))-\varOmega _{i,j}^{\frac{1}{2}}}{2} \\ &\geq \frac{1}{2}\bigl(\alpha _{i}- r_{i}^{i}( \mathcal{A})\bigr)+\bigl(\alpha _{j}-r_{j}^{j}( \mathcal{A})-\bigl( \alpha _{j}-\alpha _{i}+r_{i}^{i}( \mathcal{A})- r_{j}^{j}( \mathcal{A})+2R_{i}( \mathcal{A})-2r_{i}^{i}(\mathcal{A})\bigr)\bigr) \\ &=\alpha _{i}-R_{i}(\mathcal{A}), \end{aligned} $$

which means that

$$ \min_{i,j\in [n],i\neq j } \bigl\{ \theta _{1}(\mathcal{A}), \alpha _{i}- r_{i}^{i}(\mathcal{A}), \alpha _{j}- r_{j}^{j}( \mathcal{A})\bigr\} \geq \min _{i\in [n] } \bigl\{ \alpha _{i}-R_{i}( \mathcal{A})\bigr\} . $$

Case 2. For any \(i,j\in [n]\), \(i\neq j\), if \(\alpha _{i}-R_{i}(\mathcal{A})\geq \alpha _{j}-R_{j}(\mathcal{A})\), then

$$ \alpha _{i}- \alpha _{j}+R_{j}( \mathcal{A})\geq R_{i}(\mathcal{A}). $$
(2.16)

From (2.16) we have

$$ \begin{aligned} &\bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})-\bigl(\alpha _{j}-r_{j}^{j}( \mathcal{A})\bigr)\bigr)^{2}+4\bigl(R_{i}( \mathcal{A})-r_{i}^{i}(\mathcal{A})\bigr) \bigl(R_{j}( \mathcal{A})- r_{j}^{j}(\mathcal{A}) \bigr) \\ &\quad \leq \bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})- \bigl(\alpha _{j}-r_{j}^{j}( \mathcal{A})\bigr) \bigr)^{2}+4\bigl(\alpha _{i}- \alpha _{j}+R_{j}( \mathcal{A})-r_{i}^{i}( \mathcal{A})\bigr) \bigl(R_{j}(\mathcal{A})- r_{j}^{j}(\mathcal{A}) \bigr) \\ &\quad = \bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})- \bigl(\alpha _{j}-r_{j}^{j}( \mathcal{A})\bigr) \bigr)^{2}\\ &\qquad {}+4\bigl(\alpha _{i}- \alpha _{j}+r_{j}^{j}( \mathcal{A})-r_{i}^{i}( \mathcal{A})+R_{j}( \mathcal{A})-r_{j}^{j}(\mathcal{A})\bigr) \bigl(R_{j}( \mathcal{A})- r_{j}^{j}(\mathcal{A}) \bigr) \\ &\quad =\bigl(\alpha _{i}- r_{i}^{i}(\mathcal{A})-\bigl( \alpha _{j}-r_{j}^{j}( \mathcal{A})\bigr) +2R_{j}(\mathcal{A})-2r_{j}^{j}(\mathcal{A}) \bigr)^{2}. \end{aligned} $$

Since

$$ \begin{aligned} &\alpha _{i}- r_{i}^{i}(\mathcal{A})-\bigl(\alpha _{j}-r_{j}^{j}( \mathcal{A})\bigr) +2R_{j}(\mathcal{A})-2r_{j}^{j}( \mathcal{A}) \\ &\quad =\alpha _{i}-\alpha _{j}+R_{j}( \mathcal{A})-R_{i}(\mathcal{A})+R_{i}( \mathcal{A})- r_{i}^{i}(\mathcal{A})+R_{j}(\mathcal{A})-r_{j}^{j}( \mathcal{A})\geq 0, \end{aligned} $$

we have

$$ \begin{aligned} \theta _{1}(\mathcal{A})&= \frac{(\alpha _{i}- r_{i}^{i}(\mathcal{A}))+(\alpha _{j}-r_{j}^{j}(\mathcal{A}))-\varOmega _{i,j}^{\frac{1}{2}}}{2} \\ &\geq \frac{1}{2}\bigl(\alpha _{i}- r_{i}^{i}( \mathcal{A})\bigr)+\bigl(\alpha _{j}-r_{j}^{j}( \mathcal{A})-\bigl( \alpha _{i}- r_{i}^{i}( \mathcal{A})-\bigl(\alpha _{j}-r_{j}^{j}( \mathcal{A})\bigr) +2R_{j}(\mathcal{A})-2r_{j}^{j}( \mathcal{A})\bigr)\bigr) \\ &=\alpha _{j}-R_{j}(\mathcal{A}), \end{aligned} $$

which means that

$$ \min_{i,j\in [n],i\neq j } \bigl\{ \theta _{1}(\mathcal{A}), \alpha _{i}- r_{i}^{i}(\mathcal{A}), \alpha _{j}- r_{j}^{j}( \mathcal{A})\bigr\} \geq \min _{i\in [n] } \bigl\{ \alpha _{i}-R_{i}( \mathcal{A})\bigr\} . $$

Similarly, we have \(\min_{k,l\in [n],k\neq l } \{\theta _{2}(\mathcal{A}), \beta _{k}- c_{k}^{k}(\mathcal{A}),\beta _{l}- c_{l}^{l}(\mathcal{A}) \}\geq \min_{l\in [n] } \{\beta _{l}-C_{l}(\mathcal{A})\}\). Thus we deduce

$$ \begin{aligned} \tau (\mathcal{A})&\geq \max \Bigl\{ \min _{i,j\in [n],i\neq j } \bigl\{ \theta _{1}(\mathcal{A}),\alpha _{i}- r_{i}^{i}(\mathcal{A}), \alpha _{j}- r_{j}^{j}(\mathcal{A})\bigr\} ,\\ &\quad \min _{k,l\in [n],k\neq l } \bigl\{ \theta _{2}(\mathcal{A}),\beta _{k}- c_{k}^{k}(\mathcal{A}),\beta _{l}- c_{l}^{l}(\mathcal{A})\bigr\} \Bigr\} \\ &\geq \max \Bigl\{ \min_{i\in [n] } \bigl\{ \alpha _{i}-R_{i}( \mathcal{A})\bigr\} , \min_{l\in [n]} \bigl\{ \beta _{l}-C_{l}( \mathcal{A})\bigr\} \Bigr\} ,\end{aligned} $$

and the desired result follows. □

The following example shows the superiority of the conclusions obtained in Theorems 2.4 and 2.6.

Example 2.1

([30])

Let \(\mathcal{A}=({{a}}_{ijkl})\in \mathbb{E}_{4,2}\) be an elasticity M-tensor defined by

$$ a_{ijkl}= \textstyle\begin{cases} a_{1111}=13,\qquad a_{2222}=12,\qquad a_{1122}=2,\qquad a_{2211}=2, \\ a_{1112}=a_{1121}=-2,\qquad a_{2212}=a_{2221}=-1, \\ a_{2111}=a_{1211}=-2,\qquad a_{1222}=a_{2122}=-1, \\ a_{1212}=a_{2112}=a_{1221}=a_{2121}=-4. \end{cases} $$

From the matrices

$$\begin{aligned}& \mathcal{A}(\cdot ,f_{1})= \begin{bmatrix} a_{1111} & a_{1211} \\ a_{2111} & a_{2211}\end{bmatrix} = \begin{bmatrix} 13 & -2 \\ -2 & 2 \end{bmatrix} ,\qquad \mathcal{A}(\cdot ,f_{2})= \begin{bmatrix} a_{1122} & a_{1222} \\ a_{2122} & a_{2222}\end{bmatrix} = \begin{bmatrix} 2 & -1 \\ -4 & 12 \end{bmatrix} , \\& \mathcal{A}(g_{1},\cdot )= \begin{bmatrix} a_{1111} & a_{1112} \\ a_{1121} & a_{1122}\end{bmatrix} = \begin{bmatrix} 13 & -2 \\ -2 & 2 \end{bmatrix} ,\qquad \mathcal{A}(g_{2}, \cdot ,)= \begin{bmatrix} a_{2211} & a_{2212} \\ a_{2221} & a_{2222}\end{bmatrix} = \begin{bmatrix} 2 & -1 \\ -1 & 12 \end{bmatrix} , \end{aligned}$$

we know that \(\mathcal{A}\) is irreducible. By simple computation, \(\mathcal{A}\) has six M-eigenvalues: 13.4163, 12.1118, 11.2036, 6.1778, 0.2442, and 0.1964. The minimal M-eigenvalue of \(\mathcal{A}\) is 0.1964. Furthermore, we obtain

$$\begin{aligned}& \alpha _{1} =13,\qquad \alpha _{2} =12,\qquad \beta _{1} =13,\qquad \beta _{2} =12, \\& r_{1}(\mathcal{A})=12,\qquad r_{2}(\mathcal{A})=10,\qquad c_{1}(\mathcal{A})=12,\qquad c_{2}(\mathcal{A})=10, \\& \gamma _{1}=2,\qquad \gamma _{2}=2,\qquad \delta _{1}=2,\qquad \delta _{2}=2, \\& r_{1}^{1}(\mathcal{A})=4,\qquad r_{2}^{2}( \mathcal{A})=2,\qquad c_{1}^{1}( \mathcal{A})=4,\qquad c_{2}^{2}(\mathcal{A})=2, \\& R_{1}(\mathcal{A})=14,\qquad R_{2}(\mathcal{A})=12,\qquad C_{1}(\mathcal{A})=14,\qquad C_{2}(\mathcal{A})=12. \end{aligned}$$

From Theorems 3.1 and 3.2 in [31] we have \(\tau (\mathcal{A})\geq -1\) and \(\tau (\mathcal{A})\geq -0.8655\), respectively. By Theorem 2.4 we have \(\tau (\mathcal{A})\geq -0.5567\), and by Theorem 2.6 we have \(\tau (\mathcal{A})\geq -0.5125\). Their comparison is drawn in Fig. 1, which reveals that our bounds are tighter than those of [31].

Figure 1
figure1

Comparison of the bound estimations of Theorems 3.1 and 3.2 in [31], Theorem 2.4, and Theorem 2.6

Strong ellipticity and positive definiteness

In this section, based on the results in Theorems 2.4 and 2.6, we present some sufficient conditions for the strong ellipticity and positive definiteness.

Theorem 3.1

Let\(\mathcal{A}=({{a}}_{ijkl})\in \mathbb{E}_{4,n}\)be an irreducible elasticity M-tensor. If

$$ \max \Bigl\{ \min_{i,j\in [n],i\neq j } \bigl\{ \eta _{1}( \mathcal{A})\bigr\} , \min_{k,l\in [n],k\neq l } \bigl\{ \eta _{2}( \mathcal{A})\bigr\} \Bigr\} >0, $$

then\(\mathcal{A}\)is positive definite, and the strong ellipticity condition holds.

Proof

From Theorem 2.4 we have

$$ \tau (\mathcal{A})\geq \max \Bigl\{ \min_{i,j\in [n],i\neq j } \bigl\{ \eta _{1}(\mathcal{A})\bigr\} , \min_{k,l\in [n],k\neq l } \bigl\{ \eta _{2}(\mathcal{A})\bigr\} \Bigr\} >0. $$

Hence \(\mathcal{A}\) is positive definite, and the strong ellipticity condition holds. □

Theorem 3.2

Let\(\mathcal{A}=({{a}}_{ijkl})\in \mathbb{E}_{4,n}\)be an irreducible elasticity M-tensor. If

$$ \max \Bigl\{ \min_{i,j\in [n],i\neq j } \bigl\{ \theta _{1}( \mathcal{A}), \alpha _{i}- r_{i}^{i}(\mathcal{A}), \alpha _{j}- r_{j}^{j}( \mathcal{A})\bigr\} , \min _{k,l\in [n],k\neq l } \bigl\{ \theta _{2}( \mathcal{A}),\beta _{k}- c_{k}^{k}(\mathcal{A}),\beta _{l}- c_{l}^{l}( \mathcal{A})\bigr\} \Bigr\} >0, $$

then\(\mathcal{A}\)is positive definite, and the strong ellipticity condition holds.

Proof

From Theorem 2.6 we have

$$ \begin{aligned} \tau (\mathcal{A})&\geq \max \Bigl\{ \min_{i,j\in [n],i\neq j } \bigl\{ \theta _{1}(\mathcal{A}),\alpha _{i}- r_{i}^{i}( \mathcal{A}), \alpha _{j}- r_{j}^{j}(\mathcal{A}) \bigr\} ,\\ &\quad \min_{k,l\in [n],k\neq l } \bigl\{ \theta _{2}(\mathcal{A}), \beta _{k}- c_{k}^{k}(\mathcal{A}),\beta _{l}- c_{l}^{l}(\mathcal{A})\bigr\} \Bigr\} \\ &>0. \end{aligned} $$

Hence \(\mathcal{A}\) is positive definite, and the strong ellipticity condition holds. □

The following example reveals that Theorems 3.1 and 3.2 can identify the positive definiteness of elasticity M-tensors.

Example 3.1

Let \(\mathcal{A}=({{a}}_{ijkl})\in \mathbb{E}_{4,2}\) be an elasticity M-tensor such that

$$ a_{ijkl}= \textstyle\begin{cases} a_{1111}=7,\qquad a_{2222}=8,\qquad a_{1122}=7,\qquad a_{2211}=8, \\ a_{1112}=a_{1121}=-1,\qquad a_{2212}=a_{2221}=-1, \\ a_{2111}=a_{1211}=-1,\qquad a_{1222}=a_{2122}=-1, \\ a_{1212}=a_{2112}=a_{1221}=a_{2121}=-0.5. \end{cases} $$

By a direct computation we have

$$\begin{aligned}& \mathcal{A}(\cdot ,f_{1})= \begin{bmatrix} 7 & -1 \\ -1 & 8 \end{bmatrix} ,\qquad \mathcal{A}(\cdot ,f_{2})= \begin{bmatrix} 7 & -1 \\ -1 & 8 \end{bmatrix} , \\& \mathcal{A}(g_{1},\cdot )= \begin{bmatrix} 7 & -1 \\ -1 & 7 \end{bmatrix} ,\qquad \mathcal{A}(g_{2},\cdot ,)= \begin{bmatrix} 8 & -1 \\ -1 & 8 \end{bmatrix} . \end{aligned}$$

Then \(\mathcal{A}\) is irreducible. Furthermore, by simple computation we obtain

$$\begin{aligned}& \alpha _{1} =7, \qquad \alpha _{2} =8,\qquad \beta _{1} =7,\qquad \beta _{2} =7, \\& r_{1}(\mathcal{A})=3,\qquad r_{2}(\mathcal{A})=3,\qquad c_{1}(\mathcal{A})=3,\qquad c_{2}(\mathcal{A})=3, \\& \gamma _{1}=1,\qquad \gamma _{2}=1,\qquad \delta _{1}=1,\qquad \delta _{2}=1, \\& r_{1}^{1}(\mathcal{A})=2,\qquad r_{2}^{2}( \mathcal{A})=2,\qquad c_{1}^{1}( \mathcal{A})=2,\qquad c_{2}^{2}(\mathcal{A})=2, \\& R_{1}(\mathcal{A})=4,\qquad R_{2}(\mathcal{A})=4,\qquad C_{1}(\mathcal{A})=4,\qquad C_{2}(\mathcal{A})=4. \end{aligned}$$

From Theorem 3.1 we have

$$ \tau (\mathcal{A})\geq \max \Bigl\{ \min_{i,j\in [n],i\neq j } \bigl\{ \eta _{1}(\mathcal{A})\bigr\} , \min_{k,l\in [n],k\neq l } \bigl\{ \eta _{2}(\mathcal{A})\bigr\} \Bigr\} =3.2984>0. $$

From Theorem 3.2 we have

$$ \begin{aligned} \tau (\mathcal{A})&\geq \max \Bigl\{ \min_{i,j\in [n],i\neq j } \bigl\{ \theta _{1}(\mathcal{A}),\alpha _{i}- r_{i}^{i}( \mathcal{A}), \alpha _{j}- r_{j}^{j}(\mathcal{A}) \bigr\} ,\\ &\quad \min_{k,l\in [n],k\neq l } \bigl\{ \theta _{2}(\mathcal{A}), \beta _{k}- c_{k}^{k}(\mathcal{A}),\beta _{l}- c_{l}^{l}(\mathcal{A})\bigr\} \Bigr\} \\ &=3.4384>0. \end{aligned} $$

Thus from Theorems 3.1 and 3.2 we obtain that \(\mathcal{A}\) is positive definite.

Conclusion

In this paper, we present some bounds for the minimum M-eigenvalue of elasticity M-tensors, which are tighter than some existing results. We propose numerical examples that illustrate the efficiency of the obtained results. As applications, we provide some checkable sufficient conditions for the strong ellipticity and positive definiteness.

References

  1. 1.

    Han, D., Dai, H., Qi, L.: Conditions for strong ellipticity of anisotropic elastic materials. J. Elast. 97, 1–13 (2009)

    MathSciNet  Article  Google Scholar 

  2. 2.

    Knowles, J.K., Sternberg, E.: On the ellipticity of the equations of non-linear elastostatics for a special material. J. Elast. 5, 341–361 (1975)

    Article  Google Scholar 

  3. 3.

    Knowles, J.K., Sternberg, E.: On the failure of ellipticity of the equations for finite elastostatic plane strain. Arch. Ration. Mech. Anal. 63, 321–336 (1977)

    MathSciNet  Article  Google Scholar 

  4. 4.

    Walton, J.R., Wilber, J.P.: Sufficient conditions for strong ellipticity for a class of anisotropic materials. Int. J. Non-Linear Mech. 38, 411–455 (2003)

    MathSciNet  MATH  Google Scholar 

  5. 5.

    Chiritǎ, S., Danescu, A., Ciarletta, M.: On the strong ellipticity of the anisotropic linearly elastic materials. J. Elast. 87, 1–27 (2007)

    MathSciNet  Article  Google Scholar 

  6. 6.

    Padovani, C.: Strong ellipticity of transversely isotropic elasticity tensors. Meccanica 37, 515–525 (2002)

    MathSciNet  Article  Google Scholar 

  7. 7.

    Che, H., Li, M.: The conjugate gradient method for split variational inclusion and constrained convex minimization problems. Appl. Math. Comput. 290, 426–438 (2016)

    MathSciNet  MATH  Google Scholar 

  8. 8.

    Che, H., Chen, H., Li, M.: A new simultaneous iterative method with a parameter for solving the extended split equality problem and the extended split equality fixed point problem. Numer. Algorithms 79(4), 1231–1256 (2018)

    MathSciNet  Article  Google Scholar 

  9. 9.

    Che, H., Chen, H., Wang, Y.: On the M-eigenvalue estimation of fourth-order partially symmetric tensors. J. Ind. Manag. Optim. 16(1), 309–324 (2020)

    MathSciNet  MATH  Google Scholar 

  10. 10.

    Che, H., Chen, H., Wang, Y.: M-positive semi-definiteness and M-positive definiteness of fourth-order partially symmetric Cauchy tensors. J. Inequal. Appl. 2019(1), 32 (2019)

    MathSciNet  Article  Google Scholar 

  11. 11.

    Li, S., Li, C., Li, Y.: M-eigenvalue inclusion intervals for a fourth-order partially symmetric tensor. J. Comput. Appl. Math. 356, 391–401 (2019)

    MathSciNet  Article  Google Scholar 

  12. 12.

    Chen, H., Huang, Z., Qi, L.: Copositivity detection of tensors: theory and algorithm. J. Optim. Theory Appl. 174, 746–761 (2017)

    MathSciNet  Article  Google Scholar 

  13. 13.

    Chen, H., Chen, Y., Li, G., Qi, L.: A semi-definite program approach for computing the maximum eigenvalue of a class of structured tensors and its applications in hypergraphs and copositivity test. Numer. Linear Algebra Appl. 25, e2125 (2018)

    Article  Google Scholar 

  14. 12.

    Chen, H., Huang, Z., Qi, L.: Copositive tensor detection and its applications in physics and hypergraphs. Comput. Optim. Appl. 69, 133–158 (2018)

    MathSciNet  Article  Google Scholar 

  15. 15.

    Chen, H., Wang, Y.: On computing minimal H-eigenvalue of sign-structured tensors. Front. Math. China 12, 1289–1302 (2017)

    MathSciNet  Article  Google Scholar 

  16. 16.

    Chen, H., Qi, L., Song, Y.: Column sufficient tensors and tensor complementarity problems. Front. Math. China 13(2), 255–276 (2018)

    MathSciNet  Article  Google Scholar 

  17. 17.

    Wang, Y., Caccetta, L., Zhou, G.: Convergence analysis of a block improvement method for polynomial optimization over unit spheres. Numer. Linear Algebra Appl. 22, 1059–1076 (2015)

    MathSciNet  Article  Google Scholar 

  18. 18.

    Wang, Y., Zhang, K., Sun, H.: Criteria for strong H-tensors. Front. Math. China 11, 577–592 (2016)

    MathSciNet  Article  Google Scholar 

  19. 19.

    Zhou, G., Wang, G., Qi, L., Alqahtani, M.: A fast algorithm for the spectral radii of weakly reducible nonnegative tensors. Numer. Linear Algebra Appl. 25(2), e2134 (2018)

    MathSciNet  Article  Google Scholar 

  20. 20.

    Zhang, K., Wang, Y.: An H-tensor based iterative scheme for identifying the positive definiteness of multivariate homogeneous forms. J. Comput. Appl. Math. 305, 1–10 (2016)

    MathSciNet  Article  Google Scholar 

  21. 21.

    Wang, X., Chen, H., Wang, Y.: Solution structures of tensor complementarity problem. Front. Math. China 13, 935–945 (2018)

    MathSciNet  Article  Google Scholar 

  22. 22.

    Wang, Y., Qi, L., Zhang, X.: A practical method for computing the largest M-eigenvalue of a fourth-order partially symmetric tensor. Numer. Linear Algebra Appl. 16, 589–601 (2009)

    MathSciNet  Article  Google Scholar 

  23. 23.

    Che, H., Chen, H., Wang, Y.: C-eigenvalue inclusion theorems for piezoelectric-type tensors. Appl. Math. Lett. 89, 41–49 (2019)

    MathSciNet  Article  Google Scholar 

  24. 24.

    Wang, C., Chen, H., Wang, Y., Zhou, G.: On copositiveness identification of partially symmetric rectangular tensors. J. Comput. Appl. Math. 372, 112678 (2020)

    MathSciNet  Article  Google Scholar 

  25. 25.

    Zhang, K., Chen, H., Zhao, P.: A potential reduction method for tensor complementarity problems. J. Ind. Manag. Optim. 15(2), 429–443 (2019)

    MathSciNet  MATH  Google Scholar 

  26. 26.

    Chen, H., Qi, L., Lous, C., Zhou, G.: Birkhoff–von Neumann theorem and decomposition for doubly stochastic tensors. Linear Algebra Appl. 583, 119–133 (2019)

    MathSciNet  Article  Google Scholar 

  27. 27.

    Chen, H., Wang, Y., High-order copositive tensors and its applications. J. Appl. Anal. Comput. 8(6), 1863–1885 (2018)

    MathSciNet  Google Scholar 

  28. 28.

    Wang, W., Chen, H., Wang, Y.: A new C-eigenvalue interval for piezoelectric-type tensors. Appl. Math. Lett. 100, 106035 (2020)

    MathSciNet  Article  Google Scholar 

  29. 29.

    Gurtin, M.E.: The linear theory of elasticity. In: Linear Theories of Elasticity and Thermoelasticity. Springer, Berlin (1973)

    Google Scholar 

  30. 30.

    Ding, W., Liu, J., Qi, L., Yan, H.: Elasticity M-tensors and the strong ellipticity condition. Appl. Math. Comput. 373, 124982 (2020)

    MathSciNet  MATH  Google Scholar 

  31. 31.

    He, J., Li, C., Wei, Y.: M-eigenvalue intervals and checkable sufficient conditions for the strong ellipticity. Appl. Math. Lett. 102, 106137 (2020)

    MathSciNet  Article  Google Scholar 

  32. 32.

    Friedland, S., Gaubert, A., Han, L.: Perron–Frobenius theorems for nonnegative multilinear forms and extensions. Linear Algebra Appl. 438, 738–749 (2013)

    MathSciNet  Article  Google Scholar 

Download references

Acknowledgements

We thank the editor and the anonymous referee for their constructive comments and suggestions, which greatly improved this paper.

Availability of data and materials

The data used to support the findings of this study are available from the corresponding author upon request.

Funding

This work was supported by the Natural Science Foundation of China (11401438, 11671228, 11601261, 11571120), Shandong Provincial Natural Science Foundation (ZR2019MA022), Project of Shandong Province Higher Educational Science and Technology Program (Grant No. J14LI52), and China Postdoctoral Science Foundation (Grant No. 2017M622163).

Author information

Affiliations

Authors

Contributions

All authors contributed equally to the manuscript. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Haibin Chen.

Ethics declarations

Competing interests

The authors declare that they have no competing interests.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Che, H., Chen, H., Xu, N. et al. New lower bounds for the minimum M-eigenvalue of elasticity M-tensors and applications. J Inequal Appl 2020, 188 (2020). https://doi.org/10.1186/s13660-020-02454-1

Download citation

MSC

  • 15A06
  • 74B20
  • 47J25

Keywords

  • M-eigenvalue
  • Elasticity M-tensors
  • Strong ellipticity
  • M-positive definiteness