International Journal of Mechanical Sciences 121 (2017) 151–156

Contents lists available at ScienceDirect

International Journal of Mechanical Sciences

journal homepage: www.elsevier.com/locate/ijmecsci

Constitutive boundary conditions and paradoxes in nonlocal elastic MARK

nanobeams
⁎
Giovanni Romano , Raffaele Barretta, Marina Diaco, Francesco Marotti de Sciarra
Department of Structures for Engineering and Architecture, University of Naples Federico II, via Claudio 21, 80125 Naples, Italy

A R T I C L E I N F O A BS T RAC T

Keywords: A debated issue, in applications of ERINGEN's nonlocal model of elasticity to nanobeams, is the paradox
Nonlocal elasticity concerning the solution of simple beam problems, such as the cantilever under end-point loading. In the
Integral and differential constitutive laws adopted nonlocal model, the bending field is expressed as convolution of elastic curvature with a smoothing
Well-posedness kernel. The inversion of the nonlocal elastic law leads to solution of a FREDHOLM integral equation of the first
Nanobeams
kind. It is here shown that this problem admits a unique solution or no solution at all, depending on whether the
bending field fulfils constitutive boundary conditions or not. Paradoxical results found in solving nonlocal
elastostatic problems of simple beams are shown to stem from incompatibility between the constitutive
boundary conditions and equilibrium conditions imposed on the bending field. The conclusion is that existence
of a solution of nonlocal beam elastostatic problems is an exception, the rule being non-existence for problems
of applicative interest. Numerical evaluations reported in the literature hide or shadow this conclusion since
nodal forces expressing the elastic response are not checked against equilibrium under the prescribed data. The
cantilever problem is investigated as case study and analytically solved to exemplify the matter.

1. Introduction differential constitutive formulation and capable to explain differences

A challenging paradox of nonlocal mechanics is commonly con-
sidered to be faced in looking for the bending solution of elastic beams
obeying the elastic integral nonlocal law Eq. (1) according to which the
bending field is got by convolution of elastic curvature with the special
smoothing kernel of Eq. (3), depicted in Fig. 1.
Striking examples are BERNOULLI-EULER nonlocal cantilever nano-
beams under end-point loading which find applications in microelec-
tromechanical systems (MEMS) and nanoelectromechanical systems
(NEMS) as actuators or sensors.
The paradox, first detected in [1] and later claimed in [2], was that
some bending solutions of integral-based nonlocal elastic beams are
found to be identical to the classical (local) solution. This affirmation
has been repeated several times in literature but a fully clarifying
treatment has not yet been contributed.
Recently the issue has been newly drawn to attention by the
discussion in [3] where the relation between integral and differential
formulations of the nonlocal constitutive law is addressed and a
treatment of paradoxical examples is performed by numerical compu-
tations based on the integral formulation.
The contradiction between equilibrium and nonlocal constitutive
conditions is there considered responsible for preventing the use of the
between the results obtained by means of the integral formulation with
those obtained by the differential formulation.
Our approach is more basic.
It is shown that the nonlocal integral elastic law is equivalent to a
problem composed of constitutive differential and boundary condi-
tions. These boundary conditions arise in a natural way in detecting the
GREEN's function of differential problems defined on a bounded domain
and provide an effective test to discriminate whether a bending field is
obtainable by integral convolution or not.
In Proposition 3.1 it will be proved that fulfilment of constitutive
boundary conditions by the bending field is necessary and sufficient
condition in order to assure existence and uniqueness of the solution of
the integral equation defining the corresponding elastic curvature.
At this point a general discussion of the elastostatic problem is
appropriate. Firstly we observe that:
1. The bending field solution of the elastostatic problem has to fulfil
equilibrium with the imposed loading.
2. The elastic curvature has to fulfil kinematic compatibility under the
imposed boundary constraints and has to be associated with a
bending field that meets the constitutive boundary conditions.

⁎
Corresponding author.
E-mail addresses: romano@unina.it (G. Romano), rabarret@unina.it (R. Barretta), diaco@unina.it (M. Diaco), marotti@unina.it (F. Marotti de Sciarra).

http://dx.doi.org/10.1016/j.ijmecsci.2016.10.036
Received 31 July 2016; Received in revised form 9 October 2016; Accepted 26 October 2016
Available online 08 November 2016
0020-7403/ © 2016 Elsevier Ltd. All rights reserved.
G. Romano et al. International Journal of Mechanical Sciences 121 (2017) 151–156

1 ⎛ |x | ⎞
ϕλ (x ):= exp ⎜ − ⎟ .
2L c ⎝ Lc ⎠ (3)
This kernel, which will be referred to as the special kernel, fulfils
symmetry, positivity, impulsivity:
⎧
⎪i ) ϕλ (x − y ) = ϕλ (y − x ) ≥ 0,
⎨ ii ) lim ϕ (x ) = δ (x ),
⎪
⎩ λ →0
λ
(4)
where δ (x ) is the DIRAC distribution, corresponding to a unit impulse at
the origin.

3. Differential formulation and boundary conditions

Fig. 1. Special kernel Eq. (3); λ = 1/3. The integral equation (1) in the unknown curvature field χel and
with the bending field M as data, is known as FREDHOLM equation of the
It follows that a solution of the elastostatic problem will exist only if first kind. In general a solution of this kind of integral equations does
the bending field, univocally detected among the equilibrated ones by not exist and, when it does, uniqueness cannot be assured [5]. However
imposing the conditions of kinematic compatibility to the correspond- the following peculiar result is consequent to the choice of the special
ing elastic curvature field, will also meet the constitutive boundary kernel Eq. (3).
conditions. This verification generally fails in cases of applicative
interest. Proposition 3.1. The constitutive integral equation (1) with the
The consequent interpretation of paradoxical examples is different special kernel equation (3) admits, for any λ > 0 , either a unique
from the one usually adduced in literature. solution or no solution at all, depending on whether or not the
Our analysis reveals in fact that no paradox occurs since in all bending field fulfils the constitutive boundary conditions1
claimed examples the elastostatic problem does not admit solution, and ⎧ M ′ (a ) = 1 M (a ),
⎪ λ Lc λ
it is exactly the presumed existence of a solution that lies at the root of ⎨
all paradoxical results. ⎪ Mλ (b ) = − 1 Mλ (b ).
′
⎩ Lc (5)
As a matter of fact, elastic beam problems formulated according to
ERINGEN's nonlocal integral law, as a rule do not admit solution,
Under fulfilment of the conditions (5), the unique solution is provided
existence being the exception.
by the differential expression:
Ill-posedness of nonlocal elastostatic problems is put into evidence
by general considerations and by a specific discussion of the well- 1 1
·(K ·χel )(x ) = 2 ·Mλ (x ) − Mλ″(x ).
known paradox of nonlocal cantilevers under end-point loading. Lc2 Lc (6)
Mixing of local and nonlocal material behaviours considered in
literature are discussed in Section 6. It is shown that the local elastic Proof.. Let us split Eq. (1) by setting
fraction of the mixture has a beneficial effect and induces well-
posedness. This effect is however abruptly cancelled when the local Mλ (x ) = M1 (x, λ ) + M2 (x, λ ), (7)
fraction vanishes so that a singular behaviour is expected in the limit of
a vanishing local fraction, the inherent ill-posedness of fully nonlocal with
problems being not eliminated.
Other proposed remedies to overcome paradoxical results, such as ⎧ M (x , λ ):= x ϕ (x − y )·(K ·χ )(y ) dy
numerical computations of discretised formulations, hide or shadow ⎪ 1 ∫a λ el
⎪ ⎛y − x⎞
ill-posedness of nonlocal problems, an effect that should be checked by ⎪ = ∫
x 1
exp ⎜ ⎟ ·(K ·χel )(y ) dy,
explicitly verifying equilibrium between nodal forces, expressing the ⎪ a 2L c ⎝ Lc ⎠
elastic response, and the prescribed data. ⎨ b
⎪ M2 (x , λ ):= ∫ ϕλ (x − y )·(K ·χel )(y ) dy
⎪ x
⎪ b 1 ⎛ x − y⎞
⎪ = ∫x 2Lc exp ⎜ ⎟ ·(K ·χel )(y ) dy.
2. Integral formulation ⎩ ⎝ Lc ⎠ (8)

In the wake of the original formulation of nonlocal elasticity

Taking the first derivative we get
contributed by ERINGEN in [4], a nonlocal elastic law for the BERNOULLI-
EULER beam model is usually formulated by expressing the bending field
⎧ M′(x, λ ) = 1 ( 1 , (,K , ·,χ , ), (,x, ), −,M , (,x, ,,λ , )),
M in terms of the curvature field χel by means of the integral ⎪ 1 Lc 2 el 1
⎨
convolution law ⎪ M ′( x , λ ) =
1
(−
1
( K · χ )( x ) + M2 x , λ )),
(
⎩ 2 Lc 2 el (9)
b
Mλ (x ) = (ϕλ ⋆(K ·χel ))(x ) = ∫a ϕλ (x − y )·(K ·χel )(y ) dy,
(1) and hence
with L = b − a > 0 beam length, K = IE local elastic bending stiffness,
1
with IE second moment of the field of EULER elastic moduli E on the Mλ′ (x ) = (M2 (x, λ ) − M1 (x, λ )).
Lc (10)
beam cross section. The smoothing kernel ϕ depends on the nonlocal
parameter λ > 0 .
Introducing the characteristic length A further derivation yields Eq. (6). Evaluating the expression of the first
derivative Eq. (10) at the boundary points and observing that
Lc :=λ ·L , (2)
1
the kernel is usually defined as, see Fig. 1: An apex denotes differentiation along the beam axis.

152
G. Romano et al.
M2 (a , λ ) = Mλ (a ), M1 (a, λ ) = 0,
M1 (b , λ ) = Mλ (b ), M2 (b , λ ) = 0,
we infer the necessity of the constitutive boundary conditions (5).
Sufficiency is deduced from uniqueness of solutions of Eq. (6)
consequent to the fact that the associated homogeneous equation
admits only the trivial solution under the boundary conditions (5). □
Exact evaluations of nonlocal bending fields,2 performed either
according to the integral law equation (1) or according to the
differential condition (6) with the boundary conditions (5), confirm
coincidence with the nonlocal bending fields got by direct evaluation of
the convolution in Eq. (1), for any curvature data and any positive
value of the nonlocal parameter.
4. Green's function
The result in Proposition 3.1 can also be stated by saying that the
special kernel Eq. (3) with λ > 0 is the GREEN's function associated with
the differential problem
Mλ (x ) − Lc2·Mλ″ (x ) = δ (x ),
with the homogeneous boundary conditions (5).
In fact, by linearity, the solution of the differential expression Eq.
(6) is provided by convolution of the GREEN's function with the datum
K ·χel , as expressed by Eq. (1).
By linearity again, the homogeneous constitutive boundary condi-
tions (5) will be fulfilled also by the bending field got by convolution.
The differential condition Eq. (6) has been widely adopted in
treatments of nonlocal elastic beams, often without any mention of
the constitutive boundary conditions.
Necessity of this boundary conditions has been however evidenced
in [6] and addressed in [3,7]. Thus, when referring to the differential
formulation of the nonlocal constitutive law, one should always mean
the differential problem composed of the constitutive differential
condition (6) with the constitutive boundary conditions (5).
At difference with classical problems of potential theory in the
whole 2D or 3D space, where an unbounded domain is considered and
conditions of rapid decrease at infinity are imposed, in most structural
problems bounded domains are considered and hence the evaluation of
GREEN's functions will naturally involve boundary conditions. It is
however to be underlined that these boundary conditions are of purely
constitutive origin and depend on the nonlocal parameter.
They should therefore be carefully taken as distinct from kinematic
or static boundary conditions generated by equilibrium requirements.
In this respect we observe that the treatment in [8, Eq. (4)] leads to a
constitutive theory in which material properties are dependent on
arbitrarily prescribed kinematic conditions.
5. Nonlocal elastostatic problem
In the geometrically linearised BERNOULLI-EULER beam model, the
displacement fields are required to be square integrable together with
the first and second generalised derivatives, so that boundary values of
the displacement fields and of their first derivatives can be properly
considered. This kinematical space is a HILBERT space denoted by
V :=H2 (a, b ).
Formulation of the nonlocal elastostatic problem is completed by
adding to the constitutive law equation (1) the following items
concerning kinematic compatibility and equilibrium.
1. Denoting by 3 ⊂ V the affine manifold of fields fulfilling kinematic
conditions imposed at the beam boundary, we may set
2
Symbolic and numerical computations and graphical rendering have been carried out
by the software Mathematica due to Stephen Wolfram.
International Journal of Mechanical Sciences 121 (2017) 151–156
3 = w + 3 o, (13)
(11) where w ∈ V is a prescribed displacement field and 3 o ⊂ V is the linear
manifold of fields fulfilling the associated homogenous kinematic
boundary conditions. Denoting by χo ∈ H :=H 0 (a, b )3 a prescribed
curvature, kinematic compatibility between the total curvature field
χel + χo ∈ H and the beam displacement field u ∈ 3 , is expressed by
χel + χo = u″. (14)
2. Equilibrium, expressed by the virtual power variational condition
b
∫a M (x ), δv″(x ) dx = f , δv ,
(15)
for all virtual displacement δv ∈ 3 o , requires that bending fields must
belong to the affine manifold
Σ = Mf + Σo, (16)
described by a square integrable equilibrated bending field
Mf ∈ H (a, b ) and by the linear subspace Σo of self-equilibrated bending
fields Mo , characterised by the variational property
b
(12) ∫a 〈Mo (x ), δv″(x )〉dx = 0, ∀ δv ∈ 3 o.
(17)
Kinematic compatibility can be equivalently expressed by the
implicit variational condition of mean square orthogonality between
the curvature
χel + χo − w″ ∈ H (a, b ) (18)
and all self-equilibrated bending moments δMo ∈ Σo :
b
∫a 〈(χel + χo − w″)(x ), δMo (x )〉dx = 0.
(19)
This condition is necessary and sufficient to assure existence of a
displacement field v = u − w ∈ 3 o such that χel + χo − w″ = v″, which
is equivalent to Eq. (14).
5.1. Existence of a solution
The statement in Proposition 3.1 puts into evidence that the
nonlocal elastic law Eq. (1) imposes stringent requirements to bending
fields in the beam since constitutively admissible bending fields must
belong to the linear subspace Ξλ of fields fulfilling the constitutive
boundary conditions (5).
Compatibility between the constitutive boundary conditions (5) and
the equilibrium condition in Eq. (15) is a necessary requirement for
existence of a bending field solution of the elastostatic problem. What
happens is that the affine manifold Σ of equilibrated bending fields and
the linear subspace Ξλ of those fulfilling the constitutive boundary
conditions (5) may well have a void intersection, i.e.
Σ ∩ Ξλ = ∅. (20)
The relation between boundary values of bending and shearing
fields (Mλ, Mλ′), imposed by constitutive boundary conditions (5) for
λ > 0 , is in fact likely to be in contrast with the equilibrium require-
ments imposed by the prescribed kinematical constraints, in all
nonlocal elastic problems of applicative interest.
Existence of a solution to the nonlocal elastostatic problem,
requires a non-void intersection Σ ∩ Ξλ and results to be an exceptional
event as evidenced also by the fact that only one of the two manifolds is
dependent on the nonlocal parameter and that the constitutive
3
The linear Hilbert space of square integrable fields on [a, b] is more usually denoted
by 32 (a, b ) . The notation H 0 (a, b ) is here preferred to avoid similarity with the subspace
3 of admissible displacement fields.
153
G. Romano et al.
boundary conditions involve a ratio Lc between between the boundary
values of bending and shearing fields that is contrasted by the
requirements of constraints of applicative interest.
As a consequence no bending field solution will in general exist.
Emergence of this lack of existence is usually referred to as a paradox in
the literature, because existence of a solution is assumed a priori as a
reliable guess.4
Unfortunately this confidence in existence is belied by the adopted
nonlocal elastic law.
The situation is akin to the one characteristic of limit analysis,
where lack of possible equilibrium is due to limitations imposed on
admissible bending fields usually prescribed in a pointwise manner by
the assignment of an elastic domain in the stress space at each point.
Lack of existence of a bending solution then occurs when the data
(loading, prescribed displacement and distortion fields) are such that at
some material point in the body domain the intersection, between the
elastic domain and the set of local values of fields in the equilibrium
manifold Σ , becomes void.
In the nonlocal model of Eq. (1) the concealed necessary condition
of existence is that some equilibrated bending field should be in the
range of the convolution with an arbitrary square integrable curvature
field. By means of Proposition 3.1, the class of constitutively permis-
sible bending fields is detected to be the linear manifold Ξλ composed of
those fields that fulfil the constitutive boundary conditions (5).
Statically determinate simple beams (such as cantilevers and simply
supported beams) are those in which the linear subspace Σo of self-
equilibrated bending fields Mo is trivial. There, the equilibrated bending
field is uniquely defined and independent of the nonlocal parameter λ .
Consequently, fulfilment of the boundary conditions (5), for any
positive value (or at least for two values) of the nonlocal parameter
λ , implies that constraint reactions must vanish at both ends of the
beam so that only self-equilibrated loading distributions (those whose
resultant and resultant momentum are both vanishing) are compatible
with the constitutive requirement. This kind of constraints and
loadings is however not usual in engineering applications.
6. Local/nonlocal mixture
A two-phases constitutive mixture, defined by a convex combina-
tion of local and nonlocal phases, was introduced in [9,10]. This
mixture model has been recently resorted to in [11,2,7,12,13].
The nonlocal constitutive law is accordingly expressed by
b problem for ξ → 0 , where the full nonlocal law is recovered. It is then
Mλ (x ) = ξ (K ·χel )(x ) + (1 − ξ ) ∫a ϕλ (x − y )·(K ·χel )(y ) dy,
with 0 ≤ ξ ≤ 1 phase parameter.
The fully nonlocal law is recovered by setting ξ = 0 , while the
standard local law corresponds to ξ = 1.
A procedure analogous to the one in the proof of Proposition 3.1
may be carried out to provide differential and boundary conditions
equivalent to Eq. (21). Splitting the integral term in Eq. (21), as was
made in Eq. (8), gives
Mλ (x ) = ξ (K ·χel )(x ) + (1 − ξ )(M1 (x, λ ) + M2 (x, λ )).
Differentiating from Eq. (10) yields the expression
1−ξ
Mλ′ (x ) = ξ (K ·χel )′(x ) + (M2 (x, λ ) − M1 (x, λ )).
Lc
Differentiating again from Eq. (9) we get
4
In [1] it is affirmed that no nonlocal effects will be exhibited by a cantilever
subjected to any combination of concentrated loads, but it should rather be said that
these nonlocal problems do not admit solution.
International Journal of Mechanical Sciences 121 (2017) 151–156
1−ξ
Mλ″ (x ) = ξ (K ·χel )″(x ) + (−(K ·χel )(x ) + M1 (x, λ ) + M2 (x, λ )),
Lc2
(24)
which can be rewritten as
1 1
·(K ·χel )(x ) − ξ·(K ·χel )″(x ) = 2 ·Mλ (x ) − M ″λ (x ).
Lc2 Lc (25)
Evaluating the expression of the first derivative Eq. (23) at the
boundary points and observing that
(1 − ξ ) M2 (a , λ ) = Mλ (a ) − ξ (K ·χel )(a ), M1 (a, λ ) = 0,
(1 − ξ ) M1 (b , λ ) = Mλ (b ) − ξ (K ·χel )(b ), M2 (b, λ ) = 0, (26)
we get the expression of the constitutive boundary conditions
⎧ M ′ (a ) − 1 ξ
⎪ λ Lc
·Mλ (a ) = ξ (K ·χel )′(a ) − Lc
·(K ·χel )(a ),
⎨
⎪ Mλ′ (b ) + 1
·Mλ (b ) = ξ (K ·χel )′(b ) +
ξ
·(K ·χel )(b ).
⎩ Lc Lc (27)
The equilibrium condition (15) is expressed by the requirement
that the bending field must belong to the affine manifold Σ and in beam
problems the linear subspace Σo of self-equilibrated bending fields is of
finite dimension, say n . Any equilibrated bending field in Σ can then be
expressed as sum of a special equilibrated field Mf and a linear
combination of a basis of n self-equilibrated bending fields in Σo .
Any bending-curvature pair {Mλ, χel } solution of the constitutive
integral equation (21) will fulfil the system of differential and boundary
constitutive conditions (25), (27), and vice versa.
Solving the elastostatic problem with the local/nonlocal mixture
requires the evaluation of n + 2 unknown parameters, and precisely n
unknown static parameters, multipliers of a basis of n linearly
independent self-equilibrated bending fields, and 2 constitutive para-
meters emerging from the integration of the second order equation
(25) in the unknown field K ·χel expressed in terms of the bending fields
in Σ . The task is univocally performed by means of the two scalar
constitutive boundary conditions (27) plus the n scalar conditions of
kinematic compatibility, expressed by orthogonality, in the mean
square sense of Eq. (19), between the curvature field and a basis of
self-equilibrated bending fields. The constitutive law is thus fulfilled by
an equilibrated bending field and a kinematically compatible curvature
field. Integrating twice and imposing the kinematic constraints to
eliminate rigid body displacement fields, the displacement solution is
got.
The mixture trick is thus effective in eliminating ill-posedness of the
elastostatic problem for ξ > 0 , but it will tend anyway to an ill-posed
(21) advisable to be aware that unpredictable numerical results can be
found for small values of the phase parameter, since the problem is ill-
conditioned.
The situation is akin to the one occurring in the limit analysis of a
composite bar with a perfectly plastic matrix, having a definite thresh-
old for the traction, and indefinitely elastic reinforcements. The
displacement solution will exist for any positive value of the elasticity
modulus but, as the percentage of elastic reinforcement tends to
vanish, the elongation will tend to infinity and, for a null reinforcement
percentage, the displacement solution will no more exist. A recent
(22) treatment in [13] confirms the outcomes of this observation. The
truncated analytical expressions there exposed as displacement solu-
tions, show in fact a singular behaviour in the nonlocal limit ξ → 0 ,
with kinematic boundary conditions and equilibrium not fulfilled, since
(23) the limit problem does in fact admit no solution.
7. Paradox of cantilever under end-point loading
A case-study in nonlocal elastic beam theory is represented by a
cantilever under end-point loading. The discussion at the end of
Section 5.1 reveals that this statically determinate problem does not
154
G. Romano et al. International Journal of Mechanical Sciences 121 (2017) 151–156

admit solution.
Anyway non-existence of a solution can also be verified by a direct
analysis which puts into evidence singularities of the involved fields
when a wrong procedure is carried out.
The equilibrium differential condition

Mλ″ (x ) = q (x ) = 0, (28)
and the boundary equilibrium conditions, require that the bending field
is expressed by the linear field
Mλ (x ) = F ·(L − x ). (29)
By Eq. (6) we also should have
(K ·χel )(x ) = Mλ (x ). (30)
This conclusion is contradicted by a direct application of the
constitutive integral law equation (1), for any positive value of the
nonlocal parameter λ > 0 .
Indeed the linear field equation (29) would output, by convolution
with the smoothing kernel in Eq. (1), a nonlinear bending field
expressed by
b
Mλ (x ) = ∫a ϕλ (x − y )·(F ·(L − y )) dy

1 ⎛ ⎛x − L⎞ ⎛ x⎞ ⎛ x ⎞⎞
= F ·L ⎜λ exp ⎜ ⎟ − (1 + λ )exp ⎜ − ⎟ + 2 ⎜1 − ⎟ ⎟ .
2 ⎝ ⎝ Lc ⎠ ⎝ Lc ⎠ ⎝ L ⎠⎠ (31)
The usual interpretation of this contradiction, as given for instance
in [3], is that integral and differential formulations of the nonlocal law
equation (1) with the special kernel equation (3) are not equivalent. Fig. 2. Cantilever bending fields Mλ Eq. (31).
The right conclusion is more stringent since both the integral
formulation Eq. (1) and the equivalent differential formulation Eqs. (5)
and (6), lead to a nonlocal elastostatic problem that does not admit
solution.
The evident contrast between the bending fields expressed respec-
tively by Eqs. (29) and (31) is due to incompatibility between the
equilibrium condition and the constitutive assumption.
This incompatibility emerges clearly from Proposition 3.1 by
observing that the equilibrated bending field Eq. (29) does not fulfil
the constitutive boundary conditions (5).
The proper conclusion is that a solution to this nonlocal elastostatic
problem does not exist and that existence will hold only for those
loading distribution in equilibrium with bending fields that are in the
range of the integral law equation (1), a class of loadings that depends
on the nonlocal parameter and is no interest for applications.
Shearing Tλ (x ):=Mλ′ (x ) and loading fields qλ (x ):=Mλ″ (x ) in the
cantilever, associated with the bending field Mλ (x ) emerging from the
integral law Eq. (31), are expressed by

1 ⎛ ⎛x − L⎞ Lc + L ⎛ x⎞ ⎞
Tλ (x ) = F ⎜exp ⎜ ⎟+ exp ⎜ − ⎟ − 2⎟ ,
2 ⎝ ⎝ Lc ⎠ Lc ⎝ Lc ⎠ ⎠ (32)

F ⎛ ⎛x − L⎞ Lc + L ⎛ x ⎞⎞
qλ (x ) = ·⎜exp ⎜ ⎟− ·exp ⎜ − ⎟ ⎟ .
2Lc ⎝ ⎝ Lc ⎠ Lc ⎝ Lc ⎠ ⎠ (33)
The fields in Eqs. (31), (32), (33) are respectively plotted in Figs. 2,
3, 4 for the nonlocal parameter λ in the list of values
{.0001, 0.005, .03, .05, .10, .15, .333, .50, .75, 1.00} in the upper draw-
ings and in the shorter list {.0001, 1/100, 1/10, 1/3} in the lower
drawings, where a scaling of vertical axis has been performed to
Fig. 3. Cantilever shearing fields Mλ′ Eq. (32).
improve evidence of pathological behaviours in shearing and loading
fields at a boundary layer.
3. the shearing field is not constant and quite different from the unit
We may observe that equilibrium is significantly violated by the
value of the end-force,
bending output of Eq. (31). In particular,
4. the elastic response presents a singular behaviour at the clamped
and at the free ends, due to the effect of successive differentiations in
1. the bending field does not vanish at the right free-end of the
Eqs. (32) and (33).
cantilever,
2. the emerging loading distribution is not vanishing,
Numerical solutions of nonlocal elastic beams according to the

155
G. Romano et al.
Fig. 4. Cantilever loading fields Mλ″ Eq. (33).
nonlocal constitutive model of Eq. (1) should therefore accurately be
checked to verify fulfilment of equilibrium. This control could resolve
the contradictory occurrence of finding numerical solutions of dis-
cretised problems intended to be approximations of continuum parents
that do not admit solution.
Also plotting of bending and curvature fields, not reported in most
treatments, could provide effective elements of solution control.
Indeed, even if the behaviour of displacement fields could be deemed
acceptable, due to the smoothing effects of a double integration, the
bending fields and even more the shearing and the loading fields may
clearly reveal pathological behaviours due to non-existence of solution
for the continuum problem.
8. Concluding remarks
The constitutive law of nonlocal elasticity proposed by ERINGEN in
[4], was adapted to unidimensional beam models in [1] and thence
widely adopted, with a multitude of investigations dealing with static
and dynamic behaviour of micro and nano-beams and applications to
MEMS and NEMS.
A proliferation of contributions has spread out in the literature,
notwithstanding various signals were indicating that something basic
was not going in the right way.
To contradictory outcomes of various analyses, enunciated as
paradoxes, several interpretations and adjustments have been pro-
posed, until recently. The prototype of paradoxical results is the
affirmation in [1] that the nonlocal elastic cantilever under end-point
loading shows no nonlocal effect. The recent discussion in [14] also
elaborates along the same lines.
A two-phases constitutive mixture, composed of a convex combina-
tion of local and nonlocal phases was considered in [9,10] and adopted
in [15,11,2,7,12,16].
This mixture of local-nonlocal elasticity improves the overall
stiffness and renders the elastostatic problem well-posed, but only if
156
International Journal of Mechanical Sciences 121 (2017) 151–156
the local component is a strictly positive fraction of the mixture. The
problem will anyway tend to ill-posedness when this fraction tends to
vanish.
The analytical treatment of elastic beams with the nonlocal law
equation (1) exposed in [17] claims to provide exact solutions to
problems admitting no solution, a conclusion readily got by checking
that the conditions in Eq. (5) of Proposition 3.1 are not fulfilled.
Moreover, the exposed “exact” displacement field for a cantilever
with end-point loading, does not respect the clamping condition of null
rotation [18].
The resolution of the paradox proposed in [3] entails the (numer-
ical) solution of two FREDHOLM integral equations of the first kind [3,
Eqs.(31,32)], which are in fact unsolvable since the constitutive
boundary conditions in Eq. (5) are not fulfilled. The discretised model
considered in [12] for the computation of the displacement field by a
numerical evaluation hides essential difficulties since lack of equili-
brium, not explicitly checked, is not evidenced by the plotted deflec-
tions.
Attempts to overcome paradoxical results generated by ill-posed-
ness, are eventually bound to fail if the basic inconsistency of looking
for a solution to problems that do not admit solution is not eliminated.
Statements about paradoxes should properly be clarified by the explicit
affirmation that the problems under discussion do not admit solution.
In conclusion, the nonlocal elastostatic problem consequent to the
adoption of the nonlocal constitutive law proposed in [4], adapted to
simple beams in [1] and subsequently reproduced in literature, is to be
considered as ill-posed.
References
[1] Peddieson J, Buchanan GR, McNitt RP. Application of nonlocal continuum models
to nanotechnology. Int J Eng Sci 2003;41(3–5):305–12.
[2] Challamel N, Wang CM. The small length scale effect for a non-local cantilever
beam: a paradox solved. Nanotechnology 2008;19:345703.
[3] Fernández-Sáez J, Zaera R, Loya JA, Reddy JN. Bending of Euler-Bernoulli beams
using Eringen's integral formulation: a paradox resolved. Int J Eng Sci
2016;99:107–16.
[4] Eringen AC. On differential equations of nonlocal elasticity and solutions of screw
dislocation and surface waves. J Appl Phys 1983;54:4703.
[5] Wazwaz AM. Linear and nonlinear integral equations methods and applications.
Beijing: Higher Education Press and New York: Springer-Verlag; 2011.
[6] Polyanin AD, Manzhirov AV. Handbook of integral equations, 2nd ed.. Boca Raton,
FL: Chapman & Hall/CRC; 2008.
[7] Benvenuti E, Simone A. One-dimensional nonlocal and gradient elasticity: closed-
form solution and size effect. Mech Res Comm 2013;48:46–51.
[8] Challamel N, Zhang Z, Wang CM, Reddy JN, Wang Q, Michelitsch T, Collet B. On
nonconservativeness of Eringen's nonlocal elasticity in beam mechanics: correction
from a discrete-based approach. Arch Appl Mech 2014;84:1275–92.
[9] Eringen AC. Linear theory of nonlocal elasticity and dispersion of plane waves. Int J
Eng Sci 1972;10(5):425–35.
[10] Eringen AC. Theory of nonlocal elasticity and some applications. Res Mech
1987;21:313–42.
[11] Pisano AA, Fuschi P. Closed form solution for a nonlocal elastic bar in tension. Int J
Solids Struct 2003;40:13–23.
[12] Khodabakhshi P, Reddy JN. A unified integro-differential nonlocal model. Int J Eng
Sci 2015;95:60–75.
[13] Wang YB, Zu XW, Dai HH. Exact solutions for the static bending of Euler-Bernoulli
beams using Eringen's two-phase local/nonlocal model. AIP Adv 2016;6:085114.
http://dx.doi.org/10.1063/1.4961695.
[14] Li C, Yao L, Chen W, Li S. Comments on nonlocal effects in nano-cantilever beams.
Int J Eng Sci 2015;87:47–57.
[15] Polizzotto C. Nonlocal elasticity and related variational principles. Int J Solids
Struct 2001;38:7359–80.
[16] Eptaimeros KG, Koutsoumaris CC, Tsamasphyros GJ. Nonlocal integral approach
to the dynamical response of nanobeams. Int J Mech Sci 2016;115–116:68–80.
[17] Tuna M, Kirca M. Exact solution of Eringens nonlocal integral model for bending of
Euler–Bernoulli and Timoshenko beams. Int J Eng Sci 2016;105:80–92.
[18] Romano G, Barretta R. Comment on the paper Exact solution of Eringen’s nonlocal
integral model for bending of Euler-Bernoulli and Timoshenko beams by Meral
Tuna & Mesut Kirca. Int J Eng Sci 2016;109:240–2. http://dx.doi.org/10.1016/
j.ijengsci.2016.09.009.