Abstract
Based on the matrix unfolding technique of a tensor, three easily checkable sufficient conditions for the M-positive definiteness of fourth-order partially symmetric tensors are given. Numerical examples show that the proposed results are efficient.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
The equilibrium equations [1, 2]
are of great importance in the theory of elasticity [3], where \(u_{i}(X) (i =1, 2, 3)\) is the displacement field (X is the coordinate of a material point in the reference configuration), \(c_{i_{1}i_{2}i_{3}i_{4}}\) is the component of elastic modulus tensor \({\mathcal {C}} = (c_{i_{1}i_{2}i_{3}i_{4}})\in {\mathbb {R}}^{3\times 3\times 3\times 3}\) and has the following property:
and Eq. (1.1) is strongly elliptic if and only if
holds for all unit vector \(x\in {\mathbb {R}}^{3}\) and \(y\in {\mathbb {R}}^{3}\). For common usage, (1.2) is called the strong ellipticity condition. In the past several decades, considerable effort has been made to seek the sufficient or necessary criteria for the strong ellipticity condition such as literatures [4,5,6,7,8,9]. However, easily verifiable criteria are few because \({\mathcal {C}}xyxy\) in (1.2) can be equivalently written as \({\mathcal {A}}xyxy\), where \({\mathcal {A}}=(a_{i_{1}i_{2}i_{3}i_{4}})\in {\mathbb {R}}^{3\times 3\times 3\times 3}\) is a partially symmetric tensor, that is, \(a_{i_{1}i_{2}i_{3}i_{4}}=a_{i_{3}i_{2}i_{1}i_{4}}=a_{i_{1}i_{4}i_{3}i_{2}}=a_{i_{3}i_{4}i_{1}i_{2}}\) and \(a_{i_{1}i_{2}i_{3}i_{4}}=\frac{1}{4}(c_{i_{1}i_{2}i_{3}i_{4}}+c_{i_{3}i_{2}i_{1}i_{4}}+c_{i_{1}i_{4}i_{3}i_{2}}+c_{i_{3}i_{4}i_{1}i_{2}})\). In 2009, Qi et al. [10] presented that the strong ellipticity condition holds if and only if the partially symmetric tensor \({\mathcal {A}}\) is M-positive definite which is defined as follows. Without loss of generality, in this paper, we consider a more general partially symmetric tensor, that is, \(\mathcal {A}=(a_{i_{1}i_{2}i_{3}i_{4}})\in {\mathbb {R}}^{m\times n\times m\times n}\) with
Definition 1.1
[10] A partially symmetric tensor \({\mathcal {A}}=(a_{i_{1}i_{2}i_{3}i_{4}})\in {\mathbb {R}}^{m\times n\times m\times n}\) is called an M-positive definite tensor, if
holds for any unit vectors \(x=(x_{i})\in {\mathbb {R}}^{m}\) and \(y=(y_{i})\in {\mathbb {R}}^{n}\).
Qi et al. [10] proved that a fourth-order real partially symmetric tensor is M-positive definite if and only if its smallest M-eigenvalue is positive, whereas the computation for the M-eigenvalues of a partially symmetric tensor is difficult. To derive checkable criteria for the M-positive definiteness of partially symmetric tensors, we use another method to seek these criteria in this paper. Specifically, using the matrix unfolding technique of a tensor, we give three easily verifiable sufficient conditions for the M-positive definiteness of fourth-order partially symmetric tensors. Numerical examples show that the proposed results are efficient.
2 Main Results
We first give one checkable criterion to identify the M-positive definiteness of fourth-order real partially symmetric tensors.
Theorem 2.1
Let \({\mathcal {A}}=(a_{i_{1}i_{2}i_{3}i_{4}})\in {\mathbb {R}}^{m\times n\times m\times n}\) be a partially symmetric tensor with positive diagonal entries \(\{a_{ijij}\}_{i,j=1}^{m,n}\). Then \({\mathcal {A}}\) is M-positive definite, if
where
Proof
Suppose that \({\mathcal {A}}\) is not M-positive definite, by Definition 1.1, there exist at least one \(x\in {\mathbb {R}}^{m}\backslash \{0\}\) and one \(y\in {\mathbb {R}}^{n}\backslash \{0\}\) such that
To derive contradiction, we unfold the tensor \({\mathcal {A}}=(a_{i_{1}i_{2} i_{3}i_{4}})\in {\mathbb {R}}^{m\times n\times m\times n}\) into a matrix \(A=(a_{kh})\in {\mathbb {R}}^{mn\times mn}\) by the following bijective mapping:
where
and we unfold the nonzero matrix \(xy^{{\text {T}}}=(x_{i}y_{j})\in {\mathbb {R}}^{m\times n}\) into a vector \(w=(w_{k})\in {\mathbb {R}}^{mn}\backslash \{0\}\), where \(w_{k}=x_{i_{1}}y_{i_{2}}\). Obviously, the unfolding matrix A is symmetric since the partial symmetry of \({\mathcal {A}}\), and \({\mathcal {A}}xyxy\) can be equivalently rewritten as \(w^{{\text {T}}}Aw\), i.e. \({\mathcal {A}}xyxy=w^{{\text {T}}}Aw\). From (2.2), we know that \(0\ge w^{{\text {T}}}Aw\), which means that the unfolding matrix A is not positive definite. By the properties of positive definite matrices [11], we know that A is not a strictly diagonal dominant matrix, i.e. there is at least one index \(k'=i_{1}'+(i_{2}'-1)m\) such that
Note that (2.4) is equivalent to
which contradicts with (2.1). Hence, \({\mathcal {A}}\) is M-positive definite. \(\square \)
Remark 2.2
From the proof of Theorem 2.1, we conclude that the positive definiteness of the unfolding matrix A implies the M-positive definiteness of the partially symmetric tensor \({\mathcal {A}}\). Hence, using other properties of the positive definite matrices, we can give other criteria for the M-positive definiteness of fourth-order real partially symmetric tensors.
For instance, based on the fact that double strictly diagonal dominant symmetric matrices [11] with positive diagonal entries are positive definite, the second criterion for the M-positive definiteness of fourth-order real partially symmetric tensors can be easily obtained.
Corollary 2.3
Let \({\mathcal {A}}=(a_{i_{1}i_{2}i_{3}i_{4}})\in {\mathbb {R}}^{m\times n\times m\times n}\) be a partially symmetric tensor with positive diagonal entries \(\{a_{ijij}\}_{i,j=1}^{m,n}\). Then \({\mathcal {A}}\) is M-positive definite, if
holds for all \(i,k\in \langle m\rangle \), \(j,l\in \langle n\rangle \) and \(\varsigma _{ijkl}=0\).
Proof
Assume that \({\mathcal {A}}\) is not M-positive definite, then similar to the proof of Theorem 2.1, we know that the unfolding matrix A is not positive definite. From the properties of positive definite matrices, we confirm that A is not a double strictly diagonal dominant symmetric matrix, i.e. there are at least two different indexes \(j'=i_{1}'+(i_{2}'-1)m\), \(k'=l_{1}'+(l_{2}'-1)m\), where \(i_{1}',l_{1}'\ne k'\in \langle m\rangle \) and \(i_{2}',l_{2}'\in \langle n\rangle \), such that
By (2.3), we know that (2.6) is equivalent to
which contradicts with (2.5). Therefore, \({\mathcal {A}}\) is M-positive definite. \(\square \)
Remark 2.4
Given a partially symmetric tensor \({\mathcal {A}}=(a_{i_{1}i_{2}i_{3}i_{4}})\in {\mathbb {R}}^{m\times n\times m\times n}\), it is not difficult to see that the criterion (2.1) holds implies that the criterion (2.5) must hold, but the converse is not necessarily true.
Besides, according to the position of the eigenvalue inclusion sets [12] of the unfolding matrix in the complex plane, we can judge whether the corresponding partially symmetric tensor \({\mathcal {A}}\) is M-positive definite or not.
Corollary 2.5
Let \({\mathcal {A}}=(a_{i_{1}i_{2}i_{3}i_{4}})\in {\mathbb {R}}^{m\times n\times m\times n}\) be a partially symmetric tensor. Then \({\mathcal {A}}\) is M-positive definite, if
where
and \({\mathbb {C}}^{-}=\{z\in {\mathbb {C}}: {{\mathbf {R}}}{{\mathbf {e}}}(z)\le 0\}\).
Proof
Provided that \({\mathcal {A}}\) is not M-positive definite, by the proof of Theorem 2.1, we know that the unfolding matrix A is not positive definite. According to the eigenvalue properties of positive definite matrices, we confirm that both regions of the Ger\(\mathrm {\check{s}}\)gorin set [12] \(\Gamma (A)\) and the Brauer set [12] \({\mathcal {K}}(A)\) of A cannot be located only in the right half complex plane, i.e.
where
and
Using (2.3) and replacing A with \({\mathcal {A}}\) in (2.8), we have
which contradicts with (2.7). Hence, \({\mathcal {A}}\) is M-positive definite. \(\square \)
Next we use two examples to illustrate that above criteria are efficient.
Example 2.6
Consider the partially symmetric tensor \({\mathcal {A}}_{1} = (a_{i_{1}i_{2}i_{3}i_{4}})\in {\mathbb {R}}^{3\times 3\times 3\times 3}\) with
\(a_{3333}=5\) and \(a_{i_{1}i_{2}i_{3}i_{4}}=0\), otherwise.
By computing, we find that \({\mathcal {A}}_{1}\) satisfies both the condition (2.1) of Theorem 2.1 and the condition (2.5) of Corollary 2.3; therefore, \({\mathcal {A}}_{1}\) is M-positive definite. Additionally, we draw the regions generated by the set \({\mathcal {K}}({\mathcal {A}}_{1})\) and the set \(\Gamma ({\mathcal {A}}_{1})\) in Fig. 1. From Fig. 1, we know that
which means that \({\mathcal {A}}_{1}\) satisfies the condition (2.7) of Corollary 2.5; so, we also conclude that \({\mathcal {A}}_{1}\) is M-positive definite.
In fact, for all \(x,y\in {\mathbb {R}}^{3}\) with \(x^{{\text {T}}}x=1\) and \(y^{{\text {T}}}y=1\), we have
By Definition 1.1, we know that \({\mathcal {A}}_{1}\) is M-positive definite, which also illustrates that the proposed results are efficient.
Example 2.7
Consider the elastic modulus tensor \({\mathcal {C}}=(c_{i_{1}i_{2}i_{3}i_{4}})\in {\mathbb {R}}^{3\times 3\times 3\times 3}\) in equilibrium Eq. (1.1), where \({\mathcal {C}}\) belongs to the rhombic system with nine elasticities [3], that is,
and \(c_{i_{1}i_{2}j_{1}j_{2}}=0\), otherwise.
To identify whether equilibrium Eq. (1.1) is strongly elliptic or not, we transfer \({\mathcal {C}}\) into a partially symmetric tensor \({\mathcal {A}}_{2}=(a_{i_{1}i_{2}i_{3}i_{4}})\in {\mathbb {R}}^{3\times 3\times 3\times 3}\) by taking
By computing, we have
It is obvious that \({\mathcal {A}}_{2}\) does not satisfy condition (2.1) of Theorem 2.1 but satisfies condition (2.5) of Corollary 2.3; hence, \({\mathcal {A}}_{2}\) is M-positive definite, which means that equilibrium Eq. (1.1) is strongly elliptic. Additionally, we draw the regions generated by \({\mathcal {K}}({\mathcal {A}}_{2})\) and \(\Gamma ({\mathcal {A}}_{2})\) in Fig. 2. From Fig. 2, we know that
which means that \({\mathcal {A}}_{2}\) satisfies the condition (2.7) of Corollary 2.5.
Therefore, we conclude that \({\mathcal {A}}_{2}\) is M-positive definite, and then equilibrium Eq. (1.1) is strongly elliptic.
Remark 2.8
Examples 2.6 and 2.7 also illustrate the conclusion of Remark 2.4 is correct.
3 Conclusions
We give three easily verifiable sufficient conditions for the M-positive definiteness of fourth-order real partially symmetric tensors using the matrix unfolding technique of a tensor. Numerical examples show that the proposed results are efficient. Actually, except for the above three sufficient conditions, we can give many other checkable criteria for the M-positive definiteness by other subclasses of H-matrices and eigenvalue inclusion sets of matrices [12, 13].
References
Knowles, J.K., Sternberg, E.: On the ellipticity of the equations of nonlinear elastostatics for a special material. J. Elast. 5(3–4), 341–361 (1975)
Knowles, J.K., Sternberg, E.: On the failure of ellipticity of the equations for finite elastostatic plane strain. Arch. Ration. Mech. Anal. 63(4), 321–336 (1976)
Gurtin, M.E.: The linear theory of elasticity. In: Truesdell, C. (ed.) Handbuch der Physik, vol. VIa/2. Springer, Heidelberg, New York, Berlin (1972)
Aron, M.: On the role of the strong ellipticity condition in nonlinear elasticity. Int. J. Eng. Sci. 21(11), 1359–1367 (1983)
Knowles, J.: On the representation of the elasticity tensor for isotropic materials. J. Elast. 39(2), 175–180 (1995)
Itskov, M.: On the theory of fourth-order tensors and their applications in computational mechanics. Comput. Methods Appl. Mech. Eng. 189(2), 419–438 (2000)
Walton, J.R., Wilber, J.P.: Sufficient conditions for strong ellipticity for a class of anisotropic materials. Int. J. Nonlinear Mech. 38(4), 411–455 (2003)
Chirita, S., Danescu, A., Ciarletta, M.: On the strong ellipticity of the anisotropic linearly elastic materials. J. Elast. 87(1), 1–27 (2007)
Han, D.R., Dai, H.H., Qi, L.Q.: Conditions for strong ellipticity of anisotropic elastic materials. J. Elast. 97(1), 1–13 (2009)
Qi, L.Q., Dai, H.H., Han, D.R.: Conditions for strong ellipticity and M-eigenvalues. Front. Math. China 4(2), 349–364 (2009)
Horn, R.A., Johnson, C.R.: Matrix analysis, 1st edn. Cambridge University Press, Cambridge (2005)
Varga, R.S.: Gers̆gorin and His Circles. Springer, Berlin, Germany (2004)
Cvetković, L.: H-matrix theory vs. eigenvalue localization. Numerical Algorithms 42(3–4), 229–245 (2006)
Acknowledgements
Suhua Li’s work is supported in part by Yunnan University’s Research Innovation Fund for Graduate Students(Grant number 2018Z057); Yunnan Provincial Doctoral Graduate Academic Newcomer Award; China Scholarship Council (Grant number 201807030004).
Yaotang Li’s work is supported by National Natural Science Foundations of China (Grant number 11861077).
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Abbas Salemi.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Li, S., Li, Y. Checkable Criteria for the M-Positive Definiteness of Fourth-Order Partially Symmetric Tensors. Bull. Iran. Math. Soc. 46, 1455–1463 (2020). https://doi.org/10.1007/s41980-019-00335-y
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s41980-019-00335-y