1 s2.0 S0022509616302484 Main

Journal of the Mechanics and Physics of Solids 98 (2017) 12–48
Contents lists available at ScienceDirect
Journal of the Mechanics and Physics of Solids

journal homepage: www.elsevier.com/locate/jmps
Nonlinear mechanics of surface growth for cylindrical and

spherical elastic bodies$
Fabio Sozio a,b, Arash Yavari b,c,n
a
Department of Mechanical Engineering, Sapienza University of Rome, Rome, Italy
b
School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA
c
The George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA
a r t i c l e in f o abstract
Article history: In this paper we formulate the initial-boundary value problems of accreting cylindrical
Received 14 April 2016 and spherical nonlinear elastic solids in a geometric framework. It is assumed that the
Received in revised form body grows as a result of addition of new (stress-free or pre-stressed) material on part of
25 August 2016
its boundary. We construct Riemannian material manifolds for a growing body with
Accepted 25 August 2016
Available online 31 August 2016
metrics explicitly depending on the history of applied external loads and deformation
during accretion and the growth velocity. We numerically solve the governing equilibrium
Keywords: equations in the case of neo-Hookean solids and compare the accretion and residual
Surface growth stresses with those calculated using the linear mechanics of surface growth.
Accretion mechanics
& 2016 Elsevier Ltd. All rights reserved.
Geometric mechanics
Nonlinear elasticity
Residual stress
1. Introduction
Mechanics of growing bodies has attracted many researchers in recent years partly because of the emerging applications
in biomechanics. Nonetheless, structures that are built by a process of accretion are common in engineering, e.g., built up of
a concrete dam in successive layers, metal solidification (Schwerdtfeger et al., 1998), electrolytic deposition, layer-by-layer
gluing of composites, wound rolls (Altmann, 1968; Yagoda, 1980), thermal and laser-based 3D printing, etc. As examples of
surface growth in Nature one can mention volcanic and sedimentary rock formation, ice and snow cover build-up (Brown
et al., 1972), formation of planetary objects (Brown and Goodman, 1963), equilibrium shape configurations for rubble piles,
the growth of biological tissues (Skalak et al., 1997), etc.
In a growing body mass is no longer conserved and the body may be residually stressed. There are two types of growth:
bulk growth and surface growth (or accretion). In bulk growth, in the language of continuum mechanics, material points are
preserved; only the mass density and the natural (stress-free) configuration of the body evolve. Because of its similarity with
finite plasticity the idea of the decomposition of the deformation gradient into elastic and growth parts was borrowed from
plasticity (see Sadik and Yavari, 2016a for a historical perspective). There are many theoretical and computational works in
the literature of (bulk) growth mechanics, see Takamizawa and Matsuda (1990), Takamizawa (1991), Rodriguez et al. (1994),
Epstein and Maugin (2000), Garikipati et al. (2004), Ben Amar and Goriely (2005), Klarbring et al. (2007), Yavari (2010),
Sadik et al. (2016) and references therein. In surface growth, instead, new material points are added to or removed from the
☆
Dedicated to the memory of Professor William S. Klug (1976–2016).
n
Corresponding author at: School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA.
E-mail address: arash.yavari@ce.gatech.edu (A. Yavari).
http://dx.doi.org/10.1016/j.jmps.2016.08.012
0022-5096/& 2016 Elsevier Ltd. All rights reserved.
F. Sozio, A. Yavari / J. Mech. Phys. Solids 98 (2017) 12–48 13
boundary of the body, meaning that the set of material points evolves in time. Moreover, the relaxed (natural) configuration
of the body explicitly depends not only on the surface growth characteristics (accretion flux, accretion velocity, etc.), but also
on the history of loading and deformation during surface growth. In the geometric language the material manifold is time
dependent and has a metric that depends on both the surface growth characteristics and the loading and deformation
during the accretion process. We should emphasize that the mechanics of surface growth is much less developed compared
to that of bulk growth perhaps because of its complexity.
The theoretical studies of the mechanics of surface growth go back to the 1940s and the work of Southwell (1941) who
analyzed thick-walled cylinders manufactured by wire winding of an initial elastic tube (see Naumov, 1994 for a more
detailed historical accounts of the mechanics of surface growth). Brown and Goodman (1963) studied the problem of a
growing planet subject to self-gravity. The analysis was carried out in the framework of the linear theory of elasticity and the
growing planet was treated as a hollow sphere made of an isotropic elastic and compressible material. Their approach
consisted in considering an incremental elastic problem at each stage of surface growth in which the contribution of self-
gravity was taken into account by replacing the body force with a traction boundary condition representing the infinitesimal
weight of the added layer. The authors assumed that the infinitesimal contribution of stress, which can also be seen as the
time rate of the Cauchy stress σ̇ , 1 is compatible, and then they integrated it to obtain the total Cauchy stress as:
t
σ (x, t ) = ∫τ (x) σ ̇ (x, s ) ds + σ (x, τ (x)),
(1.1)
where τ (x ) is the time of addition of the particle occupying the position x (since they work in the setting of linear elasticity it
is possible to identify each particle with its spatial position). The second term is the stress tensor on the accretion surface at
the point x at its time of addition τ (x ). In the case of an accreting planet resulting from the addition of stress-free material,
this term vanishes. They observed that stresses induced by surface growth do not satisfy the Beltrami–Michell compatibility
equations. This means that removing gravitational forces the body would be residually stressed. In general, the final state of
stress in an accreted body depends on the way in which new material was added and the state of stress (and loading) during
accretion.
Kadish et al. (2005) extended the work of Brown and Goodman (1963) to a rotating solid sphere. Their approach consists
of a decomposition of the stress tensor into a compatible time-dependent part and an incompatible time-independent part
that represents the residual stresses that would remain if the external forces were removed after completion of accretion.
This is obtained by a time integration of the Cauchy stress rate σ̇ (that is assumed to be compatible) leading to the following
additive decomposition of the total stress:
σ (x, t ) = σ c (x, t ) + σ o (x), (1.2)
where the first term is a time-dependent compatible stress field, and the second term is time independent and represents
that part of the residual stress that does not depend on the time of unloading. Since the time-independent part does not
satisfy the Beltrami–Michell compatibility equations, the problem is underdetermined. However, all the unknowns can be
calculated by imposing some additional boundary conditions on the accretion surface that depend on the nature of the
accretion process. For example, when the accreted material is stress-free at the time of attachment, all the components of
the stress tensor – and not just the traction vector – are zero. They obtained a closed-form solution for this problem and
used it to investigate the failure mechanisms in accreted planets.
Solidification is another phenomenon in which surface growth is relevant. Schwerdtfeger et al. (1998) presented a linear
model of solidification in a round continuous casting mold (assuming a state of generalized plane strain with cylindrical
symmetry) in order to calculate the residual stresses and investigate the formation of an air gap between the solidified
metal and the mold. They explicitly highlighted the fact that in bodies that are subject to surface growth, e.g, solidifying
bodies, there is no stress-free reference state for the body.2 All the equations were written in a rate form and the solution
was integrated numerically using the known initial conditions at the moment of solidification, following therefore a method
that is basically analogous to the one used by Brown and Goodman (1963).
Most of the existing works on surface growth that solve specific examples are restricted to small strains. Metlov (1985)
presented a finite deformation theory of viscoelastic aging solids subject to accretion and discussed a couple of applications.
The reason for considering aging and creep is that actual accretion processes usually take a considerable amount of time and
hence these phenomena become important. In his description each particle is marked by a triplet (τ *, υ1, υ2), where τ * is a
time of attachment and (υ1, υ2) are coordinates on the growth surface. Using convected coordinates, he constructed for each
point a deformation gradient as a composition of the prehistory of deformation and the deformation between the time of
addition and the current time. There are many other works on the mechanics of surface growth (some of which use a
1
Note that here σ̇ is the rate of the Cauchy stress in an accretion process. Surface growth does not need to be parametrized by time. However, if it is
parametrized by time, then σ̇ is a time rate. One needs to be careful with time rates in nonlinear elasticity as not every time rate is objective (frame-
indifferent). Such issues do not arise in linear elasticity. In particular, linearizing any of the objective rates of the Cauchy stress with respect to a stress-free
configuration one simply obtains the partial derivative of the Cauchy stress with respect to time.
2
This is true, however, as we will show in this paper, one can find a stress-free reference configuration that is a Riemannian manifold with a metric
that depends on the accretion process. The material manifold, in general, is not flat and hence cannot be realized in the Euclidean space. In other words, the
material manifold is incompatible with the geometry of the Euclidean ambient space.
14 F. Sozio, A. Yavari / J. Mech. Phys. Solids 98 (2017) 12–48
geometric framework) by Russian researchers (Arutyunyan et al., 1990; Manzhirov, 1995; Lychev and Manzhirov, 2013a;
Manzhirov, 2014; Lychev and Manzhirov, 2013a,b; Lychev, 2011). Most of the recent works are formal. An exception is the
work of Drozdov (1998a,b) who considers three configurations for an accreting body: reference, natural, and actual. These
are local configurations with the natural configuration being stress free. His formulation is motivated by the multiplicative
decomposition of the deformation gradient into elastic and plastic parts in finite plasticity. For nonlinear viscoelastic solids
he models continuous accretion as a limit of discrete accretions and solves a few interesting specific examples.
In the biomechanics literature, Skalak et al. (1997) discussed the kinematics of surface growth for biological applications,
e.g., growth of horns and seashells. They used the concept of generating cells, the material points defining the accretion
surface that generate new material along an arbitrary curve that is not necessarily normal to the accretion surface. Ong and
O'Reilly (2004) discussed surface growth in the context of rigid body dynamics. They first formulated the kinematics of a
deformable body undergoing surface growth using convected coordinates, although with a restrictive global decomposition
of motion into elastic deformation and surface growth. Then they assumed that the body is rigid and obtained some
modified kinematics and balance laws for accreting rigid bodies. As an example, they solved the problem of a cylinder
rolling on an inclined plane while it is accumulating mass. Hodge and Papadopoulos (2010) proposed a theory of surface
growth that they considered to be a continuum theory. They freeze the motion of a body at a time τ and postulate that going
from τ to τ + t , where t is a small enough time increment, motion can be globally decomposed into an elastic deformation
resulting in an intermediate configuration followed by a surface growth motion. In this theory they have to worry about
extensions of different fields into the accreted region. The main issue with this theory is that it is, unfortunately, not a
continuum theory; it is more or less an incremental theory with some questionable postulates.
In bulk growth of a body, one is given a fixed set of material points. Bulk growth changes the local stress-free config-
uration of the body at any point. This can be modeled by a Riemannian metric that depends on the growth distribution
(Yavari, 2010). In surface growth the situation is different; here, the set of material points is not fixed. In an accretion
process, new material points may be added to or some existing material points may be removed from the boundary of the
body in its current configuration. In this paper, instead of trying to formulate a general nonlinear theory of surface growth,
we solve a few particular examples with cylindrical and spherical symmetries in the setting of nonlinear elasticity. We
believe that this will pave the road for developing a general nonlinear theory of surface growth that we will discuss in a
future communication. We consider axisymmetric accretion of cylindrical and spherical bodies, i.e. we start from a cylinder
or sphere that at any instant of time is a cylinder or a sphere throughout the process of accretion. Our model problem is a
hollow cylinder or a hollow spherical ball that is under a time-dependent internal pressure. The loaded body undergoes an
accretion process in which stress-free (or pre-stressed) particles are continuously attached to the body. Unloading the
accreted body at any instant of time the body is residually stressed, in general. In this paper, we are interested in deriving
the governing equations of a nonlinear elastic body going through this special type of surface growth. In our geometric
framework, material manifold is time-dependent and has a Riemannian metric that depends on the surface growth char-
acteristics and the loading and deformation history during accretion.
This paper is organized as follows. In Section 2, we tersely review some elements of Riemannian geometry and the
geometric theory of anelasticity and discuss some aspects of nonlinear surface growth mechanics in a geometric framework.
In Section 3 we analyze, as a prelude to continuous growth, a thick hollow cylinder that is deformed under an internal
pressure and while deformed is accreted by a single layer of stress-free material. We find the material metric and show that
it depends on the external loads during the accretion process and the thickness of the added layer. We then formulate the
continuous analog of this problem in Section 4. We consider a thick hollow cylinder under a time-dependent internal
pressure while undergoing a symmetric and continuous accretion on its outer boundary where the added material is as-
sumed to be stress-free (or pre-stressed) at the time of attachment to the body (outer accretion). We next assume that the
growth surface is the inner boundary of the cylinder (inner accretion), i.e. material is continuously added to the body on its
inner boundary (the added material is under a state of hydrostatic pressure or suction at the time of attachment). We
linearize the nonlinear governing equations and show that the accretion stresses are identical to those calculated using the
linear surface growth theory of Brown and Goodman (1963). In Section 5 we study the outer accretion problem for a hollow
thick spherical ball. In Section 6, we numerically solve the governing equations of accreting cylindrical bars and spherical
balls in the case of an incompressible neo-Hookean solid. In particular, we calculate the residual stresses and demonstrate
the anelastic nature of accretion. Conclusions are given in Section 7.
2. Geometric anelasticity and the mechanics of growth
In this section, we tersely review some elements of Riemannian geometry and the geometric theory of nonlinear elas-
ticity and anelasticity. For more detailed discussions, see Marsden and Hughes (1983), Yavari (2010), and Yavari and Goriely
(2012a).
Riemannian geometry: For a smooth n-manifold 4 , the tangent space to 4 at a point p ∈ 4 is denoted Tp 4 . Suppose 5
is another n-manifold and ψ : 4 → 5 is a smooth and invertible map. A smooth vector field W on 4 assigns a vector Wp to
every p ∈ 4 and p↦Wp ∈ Tp 4 varies smoothly. If W is a vector field on 4 , then ψ W = Tψ ·W○ψ −1 is a vector field on ψ (4)
*
called the push-forward of W by ψ. Similarly, if w is a vector field on ψ (4) ⊂ 5 , then ψ *w = T (ψ −1)·w○ψ is a vector field on
4 that is called the pull-back of w by ψ. Let us denote F = Tψ . In the local charts {XA} and {x a} for 4 and 5 , respectively, F (a
two-point tensor) has the following representation (when ψ is a deformation mapping, F is the so-called deformation
gradient of nonlinear elasticity):
∂ ∂ψ a
F = F aA ⊗ dx a, F aA = ,
∂XA ∂XA (2.1)
where { } is a basis for T 4 and {dx } is a basis for T *

∂
∂XA
p
a
ψ (p ) ψ (4), the co-tangent space, i.e. the dual space of Tψ (p) ψ (4), or the
space of co-vectors (1-forms). The push-forward and pull-back of vectors have the following coordinate representations:
A
(ψ W)a = F aAW A, (ψ *w) A = (F−1) a w a. (2.2)
*
A type ( )− tensor at p ∈ 4 is a bilinear map T: T 4 × T 4 → . In a local coordinate chart {X } for 4
0
2 p p
A
T (U, W) = TAB U AW B, (2.3)
where U, W ∈ Tp 4 .
(4, G) is a Riemannian manifold if 4 is a smooth manifold equipped with an inner product G p on the tangent space Tp 4
that varies smoothly in the sense that if U and W are vector fields on 4 , then
p↦G p (Up, Wp)≕⟪Up, Wp⟫G ,

p (2.4)
is a smooth function. Obviously, G is a symmetric ( )− tensor field on 4 .

0
2
Suppose (4, G) and (5, g) are Riemannian manifolds and ψ : 4 → 5 is a diffeomorphism (smooth map with smooth
inverse). Push-forward of the metric G is a metric on ψ (4) ⊂ 5 , which is denoted by ψ G and is defined as
*
(ψ G) ψ (p) ( u ψ (p) , wψ (p) )≔G p ( (ψ *u)p , (ψ *w)p ). (2.5)
*
In components
A B
( ψ* G)ab = ( F−1) a ( F−1) b GAB. (2.6)
Similarly, pull-back of the metric g is a metric in ψ −1 (5) ⊂ 4 , which is denoted by ψ *g and is defined as
( ψ *g)p (Up, Wp)≔g ψ (p) ( ( ψ* U)ψ (p) , ( ψ* W)ψ (p) ). (2.7)
In components
(ψ *g)AB = F aAF bB gab . (2.8)
If g = ψ G , or equivalently, G = ψ *g, ψ is called an isometry and the Riemannian manifolds (4, G) and (5, g) are isometric.
*
Note that an isometry preserves distances.
Kinematics: In nonlinear elasticity a body B is identified with a Riemannian manifold ( ) , G). A configuration of ) is a
smooth embedding φ : ) → : , where ( :, g) is the ambient space manifold with a (static) background metric g. In most
applications the ambient space is assumed to be the Euclidean space. We denote by ∇G and ∇g the Levi-Civita connections
associated with the Riemannian manifolds ( ) , G) and ( :, g), respectively. The so-called deformation gradient F is the de-
rivative map of φt defined as
F (X , t ) = dφt (X ): TX ) → Tφt (X ) :. (2.9)
The adjoint of F is defined by
F T (X , t ): Tφt (X ) : → TX ), ⟪FV, v⟫g = ⟪V, F Tv⟫G , ∀ V ∈ TX ), v ∈ Tφt (X ) :, (2.10)
where ⟪, ⟫g and ⟪, ⟫G are the inner products induced by g and G, respectively. The right Cauchy–Green deformation tensor is
defined as C (X , t ) = FT (X , t ) F (X , t ) : TX ) → TX ) . In the coordinate charts {XA} and {x a} for ) and : , respectively, C is written as
follows: C AB = GAMF aMF bB gab . The Jacobian of the motion J relates the material and spatial Riemannian volume elements
dV (X , G) and dv (x, g) by dv¼JdV and is given as
det g
J= det F .
det G (2.11)
In classical nonlinear elasticity, a motion is a time-dependent map φt : ) → : , from a fixed material manifold ( ) , G) into an
ambient space manifold ( :, g). In the presence of surface growth, ) is no longer fixed as material points are added to or
removed from the boundary of the body in its current configuration. For an accreting body φt : )t → : , where ( )t , G) is a
time-dependent material manifold. The metric G explicitly depends on the history of loading and deformation during ac-
cretion as we will see shortly in the examples of cylindrical and spherical growing bodies.
In an accretion process new material is added to or existing material is removed from the growth surface ωt, which is
part of the boundary of the body in its deformed configuration, i.e. ωt ⊂ φt (∂)t ). The growth surface in the material manifold
is denoted by Ωt. For any X ∈ )t we define a time of attachment τ (X ) ≤ t . For X ∈ )0 (the initial body), τ (X ) = 0. Note that
Ωt = {X ∈ )t : τ (X ) = t} . (2.12)
Constitutive equations: In this paper we restrict our calculations to incompressible isotropic hyperelastic solids. The left
A B
Cauchy–Green deformation tensor B ♯ = φ* (g ♯) has components B AB = (F−1) a (F−1) b g ab , where gab are components of g ♯ . The
spatial analogs of C♭ and B ♯ are
A B
c♭ = φ (G), ( ) a( F−1)
cab = F−1 GAB , and b ♯ = φ (G ♯), bab = F aAF bBG AB . (2.13)
* b *
b♯ is called the Finger deformation tensor. Note that C and b have the same principal invariants, which we denote by I1, I2,
and I3 (Ogden, 1984). For an isotropic solid the energy function W depends only on I1, I2, and I3. It is seen that, for example,
I1 = tr b = F aAF bBGABgab explicitly depends on the material metric. For an incompressible (I3 = 1) and isotropic hyperelastic
solid with an energy function W = W (I1, I2 ), the Cauchy stress has the following classic representation (Simo and Marsden,
1984):
⎛ ∂W ⎞ ♯ ∂W ♯ ∂W −1
σ = ⎜ −p + 2I2 ⎟g + 2 b −2 b ,
⎝ ∂I2 ⎠ ∂I1 ∂I2 (2.14)
where p is the Lagrange multiplier associated with the internal incompressibility constraint J¼ 1. It should be emphasized
that the Cauchy stress explicitly depends on the material metric G. It should also be emphasized that we assume that the
residually stressed body is isotropic in its stress-free material manifold. In the current configuration the body may not be
isotropic. We assume that a growing body is initially stress-free (the initial body is assumed stress free) and isotropic and
that the added material is isotropic in its stress-free configuration.
Equilibrium equations: As surface growth is a slow process, one can ignore the inertial effects. The local balance of linear
momentum in terms of the Cauchy stress tensor reads
div σ + ρf = 0, (2.15)
where f is the body force per unit deformed mass and div denotes the spatial divergence operator, which in components
reads
ab
( div σ )a = σ ab|b = ∂∂σxb + σ acγ bcb + σ cbγ acb,
(2.16)
where γ abc is the Christoffel symbol of the Levi-Civita connection ∇g in the local chart {x a}, defined as ∇∂gb ∂c = γ abc ∂a (similarly,
for the material manifold ∇∂GB ∂C = Γ ABC ∂A ).
The physical significance of a material manifold: In classical nonlinear elasticity one starts with a reference configuration
that is assumed to be stress free. Motion is then a time-dependent mapping from the reference configuration into an
Euclidean ambient space. In anelasticity (in the sense of Eckart, 1948), the body is residually stressed and hence there exists
no global stress-free reference configuration, in general, that can be realized in the Euclidean ambient space. One may
however define local relaxed configurations using a multiplicative decomposition of deformation gradient. A global stress-
free configuration cannot be isometrically embedded in the Euclidean ambient space, in general, due to its non-Euclidean
geometry. In other words, a global stress-free configuration of the body is incompatible with the geometry of the Euclidean
ambient space. This idea was introduced in the mechanics of defects independently by Kondo (1955a,b) and Bilby et al.
(1955). In a geometric formulation of anelasticity the material manifold has a geometry that depends on the source of
anelasticity. In the case of point and line defects the geometry depends on the (area or volume) density of defects (Yavari
and Goriely, 2012a,b, 2013a, 2014). In thermoelasticity, the material metric depends on both the temperature distribution
and the thermal properties of the solid, e.g., (temperature-dependent) coefficient of thermal expansion (Ozakin and Yavari,
2010; Sadik and Yavari, 2016b). In bulk growth material metric is explicitly time-dependent and is a function of the mass
flux through the balance of mass (Yavari, 2010; Sadik et al., 2016). For inclusions (or inhomogeneities with eigenstrains)
material metric explicitly depends on the distribution of (finite) eigenstrains (Yavari and Goriely, 2013b, 2015; Golgoon et al.,
2016). After solving the problem of surface growth of an infinitely long cylindrical bar (or a hollow spherical ball) we will see
that, as expected, the material manifold is time-dependent and has a metric that is determined by the history of loading and
deformation during surface growth. Interestingly, neither the material manifold nor the material metric is unique. We will
see that there are infinitely many isometric Riemannian material manifolds that are all equivalent in the sense that stress
(and particularly residual stress) does not depend on the particular choice of a material manifold from this equivalence class.
In the material manifold the growing body is stress-free by construction. Having a material manifold a nonlinear surface
growth problem is transformed into a classical nonlinear elasticity problem as long as the non-trivial geometry of the
material manifold is taken into account properly. We will demonstrate this idea for accreting cylindrical and spherical
bodies in this paper.
Fig. 1. Discrete accretion: (a) an annulus (annular cylinder) of inner and outer radii R1 and R2, respectively, in its stress-free configuration in the Euclidean
ambient space. (b) The deformed annulus under a specified internal pressure pi. (c) The deformed annulus with a layer of stress-free material of thickness Δ
attached to its outer boundary. (d) The residually stressed accreted body after removal of the external forces (here the specified internal pressure pi).
3. Attachment of a stress-free layer to a deformed infinitely long thick hollow cylinder
In this section we formulate the boundary-value problem of the attachment of a single stress-free layer on the outer
boundary of a deformed hollow cylinder. The motivation for studying this problem is to gain some insight on the effect of a
single accretion layer on the stress-free material manifold of the body. The calculations of this section can be viewed as a
prelude to the discussion of continuous surface growth in Section 4.
We consider a stress-free and infinitely long hollow cylindrical bar with inner and outer radii R1 and R2, respectively,
subject to an internal pressure pi, which is specified (see Fig. 1(a)). In the deformed configuration (Fig. 1(b)) the inner and
outer radii are r1 and r2, respectively. While keeping the internal pressure a layer of stress-free material (for the sake of
simplicity we assume the same material3) of thickness Δ is added to the outer boundary of the body through surface growth
(accretion) (see Fig. 1(c)). If the internal pressure is removed the accreted body would be residually stressed (see Fig. 1(d)).
Note that this problem is very similar to the classical shrink-fit problem (Antman and Shvartsman, 1994) and can be solved
using standard methods of nonlinear elasticity. Here we present an alternative approach that will be useful in a systematic
analysis of continuous surface growth that will be discussed in the next section. The idea is to find a Riemannian material
manifold in which the accreted body is stress free. Note that such a material manifold may be incompatible with the
geometry of the Euclidean ambient space. This idea is schematically described in Fig. 2. Only the configurations shown in
Figs. 2(b) and (c) are physically observable. Fig. 2(b) shows the loaded body with a layer of stress-free material added to it
and Fig. 2(c) depicts the residually stressed accreted body after it is unloaded. In Fig. 2(a) we show the initial stress-free
body and the stress-free layer that will be attached to it after the initial body is deformed. Note that the stress-free con-
figurations in the Euclidean ambient space are detached, i.e. a global stress-free configuration is incompatible with the
geometry of the Euclidean ambient space and this is why the accreted body will be residually stressed after the removal of
external forces (here the internal pressure pi). Next we will construct a connected Riemannian manifold in which the
accreted body is stress-free by construction (Fig. 2(d)).
3.1. Construction of a material manifold
As the cylindrical bar is infinitely long we can model it as a 2D annulus with polar coordinates (R, Θ) and (r, θ ) in the
reference and current configurations, respectively. The ambient space is equipped with the flat Euclidean metric that in
polar coordinates reads4
⎛1 0⎞
g=⎜ ⎟.
⎝ 0 r2⎠ (3.1)
3
By the “same material” we mean that the undeformed hollow cylinder and the stress-free added layer have the same energy functions.
4
This simply means that the square of the distance between two points with coordinates (r, θ ) and (r + dr , θ + dθ ) is ds2 = dr 2 + r 2dθ 2 .
Fig. 2. Discrete accretion: (a) stress-free configurations of the initial body and the accreted layer in the Euclidean ambient space : . (b) Loaded body with a
layer of stress-free material added to its outer boundary. (c) Residually stressed accreted body after removal of the external forces. (d) Riemannian material
manifold (), G) of the accreted body. Note that the configurations (a)–(c) lie in the Euclidean ambient space (:, g) . In the incompatible Riemannian material
manifold there are no residual stresses.
The (disconnected) Euclidean stress-free configuration shown in Fig. 2(a) is the union of two sets, i.e. )0 = ) −0 ∪ ) +0 ,
where
) −0 = { (R, Θ): R ≤ R ≤ R , 0 ≤ Θ < 2π},

1 2
+
)0 = { (R, Θ): r 2 ≤ R ≤ r2 + Δ, 0 ≤ Θ < 2π . } (3.2)
Here, we know that
− + ⎛1 0 ⎞
G0 = G0 = ⎜ ⎟.
⎝ 0 R2 ⎠ (3.3)
Let us consider the following maps. The deformation of the initial hollow cylinder loaded by an internal pressure pi is
denoted by ψ0−: ) −0 → : . The trivial map (identity map) of the stress-free accreted layer into itself is denoted by ψ0+: ) +0 → : .
The connected material manifold is denoted by ) = ) − ∪ ) + and is shown in Fig. 2(d) where
)− = { (R, Θ): R1 ≤ R ≤ R2, 0 ≤ Θ < 2π}, )+ = { (R, Θ): R2 ≤ R ≤ R2 + Δ′, 0 ≤ Θ < 2π}. (3.4)
Δ′ > 0 has not been specified; any positive value is admissible. A convenient choice is Δ′ = Δ. Deformation mapping from the
material manifold into the loaded configuration (Fig. 2(b)) is denoted by ψ : ) → : . We define ψ − = ψ |)− and ψ + = ψ |)+. These
maps must satisfy the following conditions:
ψ − () −) = ψ0− () −0), ψ + () +) = ψ0+ () +0 ). (3.5)
They have the coordinate representations

ψ0 (R, Θ) = (r0 (R), Θ), ψ (R, Θ) = (r (R), Θ). (3.6)
Note that ψ − (R, Θ) = (r (R ) , Θ) can be calculated using the governing equations of nonlinear elasticity as will be explained in
the following and for ψ + (R, Θ) = (r (R ) , Θ) any differentiable r(R) that satisfies r (R2 ) = r2 and r (R2 + Δ′) = r2 + Δ is admissible.
The material manifold () , G) is isometric to the disconnected Euclidean stress-free configuration, i.e. () −0 , G−0) and () −, G−) are
isometric and the isometry map is η = (ψ0−)−1○ψ −: ) − → ) −0 . Similarly, () +0 , G+0 ) and () +, G+) are isometric. However, knowing
that the accreted layer in the deformed configuration is stress-free, equivalently, () +, G+) is isometric to (ψ0+ () +0 ) , g). In
summary we have
(i) () −, G−) and () −0 , G−0) are isometric and hence

G− = η*G−0 = ((ψ0−)−1○ψ −)*G−o = (ψ −)*○ (ψo−)* G−0 . (3.7)
(ii) () +, G+) and (ψ0+ () +0 ) , g) are isometric and thus

G+ = (ψ +)*g. (3.8)
Therefore, we have
⎛ η′ 2 (R) 0 ⎞
R1 ≤ R ≤ R2: G− = ⎜⎜ ⎟⎟,
⎝ 0 η (R)⎠
2
⎛ r′ 2 (R) 0 ⎞
R2 ≤ R ≤ R2 + Δ′: G+ = ⎜⎜ ⎟⎟,
⎝ 0 r (R)⎠
2
(3.9)
where Δ′ is such that r (R2 + Δ′) = r2 + Δ. Choosing Δ′ = Δ, 5 any differentiable function η (R ) that satisfies η (R1) = R1 and
η (R2 ) = R2 is admissible. We will show that the stress distribution is independent of the choice of an admissible η. In
particular, for the sake of simplicity we choose η (R ) = R , i.e. r (R ) = r0 (R ), meaning that ) − is trivially given by ) −0 .
Calculation of the map ψ − (R, Θ) = ψ0− (R, Θ ) = (r0 (R ) , Θ ) = (r (R ) , Θ ) and the outer radius of the deformed initial body r2: Next
we calculate the configuration r(R) and the corresponding stress distribution in the region R1 ≤ R ≤ R2 when the internal
pressure pi is being applied. Let us assume that the initial annulus and the accreted layer are both made of the same
incompressible isotropic solid (in their stress-free states) with energy function W = W (I1, I2 ). The material metric is
G = diag {1, R2}, and the deformation gradient is F = diag {r′(R ) , 1}. Therefore, the incompressibility condition reads
detg r (R) r′(R)

J= detF = = 1,
detG R (3.10)
from which, denoting r2 = r (R2 ), one obtains
r 2 (R) = R2 − R22 + r22. (3.11)
Using (2.13), the Finger tensor reads

⎛ R2 ⎞
⎜ 2 0⎟
r (R)
b♯ = ⎜ ⎟.
⎜ 1⎟
⎜ 0 ⎟
⎝ R2 ⎠ (3.12)
The principal invariants of b are

r 2 (R) R2
I1 = 2
+ 2 , I2 = 1.
R r (R) (3.13)
Note that (b−1)ab = c ab = g amg bncmn and thus

⎛ r 2 (R) ⎞
⎜ 2 0 ⎟
R
b−1 = ⎜ ⎟.
⎜ R2 ⎟
⎜ 0 ⎟
⎝ r 4 (R) ⎠ (3.14)
∂W ∂W
( )
The Cauchy stress is written as σ = −p + I2 β g ♯ + α b♯ − β b−1, where α = 2 ∂I
1
and β = 2 ∂I ,
2
and in polar coordinates it reads
⎛ R2 ⎛ r 2 (R) ⎞ ⎞
⎜ − p (R) + α (R) + ⎜ 1 − ⎟ β (R) 0 ⎟
⎜ 2
r (R) ⎝ R ⎠2
⎟
σ (R) = ⎜ ⎟.
⎜ p ( R ) α (R ) ⎛ R 2 ⎞ β (R)
⎟
⎜ 0 − + + ⎜ 1 − ⎟ ⎟
⎝ r 2 (R) R2 ⎝ r 2 (R) ⎠ r 2 (R) ⎠ (3.15)
The only non-trivial equilibrium equation is the radial equilibrium equation that is written as
∂σ rr 1
+ σ rr − rσ θθ = 0.
∂r r (3.16)
This is simplified to read
dσ rr ⎡ R 4 ⎤ α (R) + β (R)
= ⎢1 − 4 ⎥ .
dR ⎣ r (R) ⎦ R (3.17)
We know that σ rr (R1) = − pi . Thus, integrating the above ODE from R1 to R, one obtains
5
We will work with a similar choice in the continuous accretion case in the next section.
R α (x) + β (x) ⎡ x4 ⎤
σ rr (R) = − pi + ∫R 1 x
⎢ 1 − 4 ⎥ dx.
⎣ r (x) ⎦ (3.18)
We know that σ rr (R2 ) = 0, and therefore

R2 α (x) + β (x) ⎡ x4 ⎤
∫R 1 x
⎢ 1 − 4 ⎥ dx = pi ,
⎣ r (x) ⎦ (3.19)
which determines the unknown r2 in (3.11).

Calculation of the map ψ + (R, Θ): For the outer layer ) + for which R2 ≤ R ≤ R2 + Δ′ any differentiable map ψ +: ) + → : that
satisfies
ψ + (R2, Θ) = (r2, Θ), ψ + (R2 + Δ′, Θ) = (r2 + Δ, Θ), Θ ∈ [0, 2π ), (3.20)
is admissible. Choosing Δ′ = Δ, the simplest choice for ψ is

þ
ψ + (R , Θ) = (R − R2 + r2, Θ), i.e.
r (R) = R − R2 + r2, R2 ≤ R ≤ R2 + Δ. (3.21)
Therefore, we have the following material metric tensors in the two layers:
− ⎛1 0 ⎞
R1 ≤ R ≤ R2: G = ⎜ ⎟,
⎝ 0 R2 ⎠
+ ⎛1 0 ⎞
R2 ≤ R ≤ R2 + Δ: G = ⎜ ⎟.
⎝ 0 (R − R2 + r2 ) ⎠
2
(3.22)
Notice that G−ΘΘ (R2 )

= ≠ = R22 r22 G+ΘΘ (R2 ). This discontinuity of the material metric will induce residual stresses in the un-
loaded configuration of Fig. 2(c).
3.2. Calculation of residual stresses for an incompressible isotropic solid
We now calculate the residual stress distribution when pi is removed. In this configuration the boundary between the
initial body and the accreted (secondary) body is a circle (infinite cylinder) of unknown radius r̃2 . We denote the deformation
from the reference configuration into the residually stressed configuration by (r˜ , θ˜ ) = (r˜ , Θ) and hence r˜2 = r˜ (R2 ), see Fig. 2.
Note that r˜2 ≠ r2, in general. We have the following boundary and continuity conditions: σ˜ rr (R1) = σ˜ rr (R2 + Δ′) = 0,
σ˜ rr (R2−) = σ˜ rr (R2+), and r˜ (R2−) = r˜ (R2+).
For R1 ≤ R ≤ R2, G = diag {1, R2} and g = diag {1, r˜ 2 (R )}. Therefore, incompressibility gives r˜ 2 (R ) = R2 + r˜22 − R22, with r̃2 to be
calculated. The Cauchy stress has the form given in (3.15) with r̃ instead of r. Therefore, the radial component of the Cauchy
stress reads (note that pi ¼0)
R α (x) + β (x) ⎡ x4 ⎤
σ˜ rr (R) = ∫R 1 x
⎢ 1 − 4 ⎥ dx.
⎣ r (x) ⎦
˜ (3.23)
In particular
R2 α (x) + β (x) ⎡ x4 ⎤
σ˜ rr (R2−) = ∫R 1 x
⎢ 1 − 4 ⎥ dx.
⎣ r˜ (x) ⎦ (3.24)
For R2 ≤ R ≤ R2 + Δ′, G = diag {r′ 2 (R ) , r 2 (R )} and g = diag {1, r˜ 2}. Therefore, incompressibility implies that
r˜ 2 (R ) = r 2 (R ) − r22 + r˜22. The principal invariants of b̃ in this region read
r˜2 (R) r 2 (R)
I˜1 = 2 + 2 , I˜2 = 1.
r (R) r˜ (R) (3.25)
Hence, the Cauchy stress in this region has the following representation:
⎛ r 2 (R) ⎛ r˜2 (R) ⎞ ⎞
⎜ − p (R) + α (R) + ⎜ 1 − 2 ⎟ β (R) 0 ⎟
⎜ r
˜ 2 (R) ⎝ r (R ) ⎠ ⎟
σ˜(R) = ⎜ ⎟.
⎜ p (R) α (R) ⎛ r 2 (R) ⎞ β (R) ⎟
⎜ 0 − + + ⎜ 1 − ⎟ ⎟
⎝ r˜2 (R) r 2 (R) ⎝ r˜2 (R) ⎠ r˜2 (R) ⎠ (3.26)
Using the radial equilibrium equation, one obtains

R α (x) + β (x) ⎡ r 4 (x) ⎤ r (R ) α (ξ ) + β (ξ ) ⎡ ξ4 ⎤
σ˜ rr (R) = σ˜ rr (R2−) + ∫R r (x)
⎢ 1 − 4 ⎥ r′(x) dx = σ˜ 2 +
⎣ r˜ (x) ⎦
∫r ξ
⎢1 − 2
⎣ (ξ − r2 + r2 ) ⎦
2 ˜ 2 2
⎥ dξ ,
2 2 (3.27)
where σ˜ 2 = σ˜ rr (R2−). We know that σ˜ rr (R2 + Δ′) = 0 and hence

R2 α (x) + β (x) ⎡ x4 ⎤ r (R 2 + Δ ′) α (ξ ) + β (ξ ) ⎡ ξ4 ⎤
∫R x
⎢1 − 2
⎣ (x − R2 + r˜2 ) ⎦
2 2 2
⎥ dx + ∫r ξ
⎢1 − 2
⎣ (ξ − r2 + r˜2 ) ⎦
2 2 2
⎥ dξ = 0,
1 2 (3.28)
which is the equation that determines the unknown constant r̃2.
Remark 3.1. Note that the residual stress distribution is independent of the choice of ψ + (R, Θ) = (r (R ) , Θ). One should keep
in mind that to be more precise the material manifold for r2 ≤ r ≤ r2 + Δ is identified by the inverse of the map r(R). Hence,
choosing another map r^ the corresponding ring in the material manifold has radius R^ = r^ (r ). For this choice, Eq. (3.27)
−1
becomes
^
r^ (R ) ⎡ ⎤
^ α (ξ ) + β (ξ ) ⎢ ξ4 ⎥ dξ .
σ˜ rr (R ) = σ˜ 2 + ∫r ξ ⎢⎣
1− 2
(ξ − r2 + r˜2 ) ⎥⎦
2 2 2
2 (3.29)
Noting that r^ (R^ ) = r (R ) one observes that σ˜ rr (R^ ) = σ˜ rr (R ), as expected.
Remark 3.2. If the added material is different from the material of the initial body (note that energy functions are defined
with respect to their stress-free configurations) the calculations do not change much. Suppose the initial body has an energy
function W (i) = W (i) (I1, I2 ) and the accreted body has the energy function W (ii) = W (ii) (I1, I2 ). The material metric would not be
affected. However, in this case α and β functions are defined as
∂W (i) ∂W (i)
R1 ≤ R ≤ R2: α=2 , β=2 ,
∂I1 ∂I2
∂W (ii) ∂W (ii)
R2 ≤ R ≤ R2 + Δ: α=2 , β=2 .
∂I1 ∂I2 (3.30)
4. Continuous surface growth of an infinitely long thick hollow cylinder
In this section we formulate the initial-boundary value problem of symmetric accretion of a thick hollow cylinder made
of an incompressible isotropic hyperelastic solid. We consider two cases: outer and inner accretions. In the outer accretion
problem stress-free (or pre-stressed) material is continuously added to the traction-free outer boundary of the deformed
cylinder. In the inner accretion problem, material is continuously added to the inner boundary of the deformed body. We
will see that the governing equations include a nonlinear partial integral equation that will be solved numerically for a neo-
Hookean solid. We will linearize the governing equations of the nonlinear surface growth and will show that the resulting
stresses are identical to those calculated using the linear theory of Brown and Goodman (1963).
4.1. Outer accretion
Let us assume that the initial stress-free body is an infinitely long solid cylinder with inner radius R1 and outer radius R0.
At time t, under an internal pressure pi(t), the outer radius in the current configuration is r2 (t ). We assume that accretion
velocity is normal to the boundary with magnitude ug(t). This means that in the infinitesimal time interval [t , t + dt ] a stress-
free ring of thickness ug (t ) dt is added to the deformed body. The metric of the Euclidean ambient space in polar coordinates
(r, θ ) reads
⎛1 0⎞
g=⎜ ⎟.
⎝ 0 r2⎠ (4.1)
In the material manifold ) we use the polar coordinates (R, Θ).
4.1.1. Kinematics and material metric

We assume motions of the form (r , θ ) = (r (R, t ) , Θ), and hence the deformation gradient reads
⎛ r′(R, t ) 0⎞
F=⎜ ⎟,
⎝ 0 1⎠ (4.2)
∂r (R, t )
where r′(R, t ) = ∂R . We assign to each layer at radius R in the material manifold a time of accretion τ (R ). This map is
assumed to be invertible, meaning that there is no ablation during the surface growth process. We indicate its inverse with
s = τ −1, which assigns to each time the radius of the accreted ring. We assume that τ (R0 ) = 0. In this problem the growth
surfaces in the material manifold and the current configuration are defined as
Ωt = { (s (t ), Θ): 0 ≤ Θ < 2π}, ωt = { (r (s (t ), t ), θ ): 0 ≤ θ < 2π}. (4.3)
Note that
d ∂r ∂r
r (s (t ), t ) = (s (t ), t ) s ̇ (t ) + (s (t ), t ) = r′(s (t ), t ) Ug (t ) + V r (s (t ), t ),
dt ∂R ∂t (4.4)
∂r
where V r = ∂t
is the radial component of the material velocity on the growth surface and Ug (t ) = s ̇ (t ). We observe that in the
absence of accretion, the spatial velocity of the material points lying on the outer boundary is simply V r (s (t ) , t ). This means
that
ug (t ) = r′(s (t ), t ) Ug (t ). (4.5)
Note that we are assuming that growth velocity is normal to the boundary both in the reference and the current config-
urations. For t = τ (R ), (4.5) can be rewritten as
r′(R, τ (R)) = τ′(R) ug (τ (R)). (4.6)
Similar to our discussion of the discrete accretion problem, for the sake of simplicity of calculations, we assume that in the
reference configuration in the time interval [t , t + dt ] the set s (t ) < R ≤ s (t ) + ug (t ) dt is added to the material manifold. This
choice corresponds to setting Ug (t ) = ug (t ). Therefore, from (4.5) we conclude that
r′(s (t ), t ) = 1 or r′(R, τ (R)) = 1. (4.7)
Also, the choice Ug (t ) = ug (t ) implies that
t
s (t ) = ∫0 ug (η) dη + R 0.
(4.8)
To simplify the calculations, we assume that the spatial growth velocity is constant, i.e. ug (t ) = u0 > 0. Hence
R − R0
s (t ) = R 0 + u0 t or τ (R) = .
u0 (4.9)
In this case, (4.7) reads
⎛ R − R0 ⎞
r′(R 0 + u0 t , t ) = 1 or r′ ⎜ R, ⎟ = 1.
⎝ u0 ⎠ (4.10)
Note that any non-negative function of t whose set of zeros has no limit points would be an acceptable choice for Ug(t) and
will have a corresponding consistent material metric.
For R1 ≤ R ≤ R0 , we know that the material metric has the following representation in polar coordinates:
⎛1 0 ⎞
G=⎜ ⎟.
⎝ 0 R2 ⎠ (4.11)
For R0 ≤ R ≤ s (t ), the accreted ring at any instant of time t is stress-free and hence the material metric at R = s (t ) is the pull
back of the metric of the Euclidean ambient space metric, i.e.
G (s (t )) = φt* g (r (s (t ), t )), or equivalently G (R) = φτ*(R) g (r (R, τ (R))). (4.12)
In components, this reads (cf. (2.8))
GAB (s (t )) = GAB (R) = F aA (R, τ (R)) F bB (R, τ (R)) gab (r (R, τ (R))). (4.13)
Thus, we have
⎛ ⎞ ⎛ ⎞ ⎛1 0 ⎞
⎜ r′ 2 (R, τ (R)) 0 ⎟ ⎜1 0 ⎟ ⎜ ⎟
G=⎜ ⎟=⎜ ⎟=⎜ ⎛ ⎞
R − R 0 ⎟,
r 2 (R, τ (R))⎟ ⎜ 0 r 2 (R, τ (R))⎟ ⎜ 0 r ⎜ R, ⎟⎟
2
⎜ 0
⎝ ⎠ ⎝ ⎠ ⎝ ⎝ u 0 ⎠⎠ (4.14)
where we have used (4.9) and (4.10). Therefore, the material manifold is an evolving Riemannian manifold ()t , G), where
)t = { (R, Θ): 0 ≤ Θ < 2π , R1 ≤ R ≤ s (t ) = R0 + u0 t}, (4.15)
and
⎛ ⎞
⎜1 0 ⎟
R1 ≤ R ≤ R 0: G=⎜ ⎟,
⎜ 0 R2 ⎟
⎝ ⎠
⎛1 0 ⎞
⎜ ⎟
R 0 ≤ R 0 + u 0 t: G=⎜ ⎛ R − R 0 ⎞⎟.
⎜ 0 r ⎜ R, ⎟⎟
2
⎝ ⎝ u 0 ⎠⎠ (4.16)
The incompressibility condition reads
detg
J= detF = 1.
detG (4.17)
For R1 ≤ R ≤ R0 this implies that r (R, t ) r′(R, t ) = R , and therefore
r 2 (R, t ) = r02 (t ) + R2 − R 02, R1 ≤ R ≤ R 0, (4.18)
where r0 (t )≔r (R0, t ). For R0 ≤ R ≤ R0 + u0 t , incompressibility implies that

⎛ R − R0 ⎞
r (R, t ) r′(R, t ) = r ⎜ R, ⎟,
⎝ u0 ⎠ (4.19)
which can be integrated for example between R0 and R to give

R ⎛ x −R 0 ⎞
r 2 (R, t ) = r 2 (R 0, t ) + ∫R 0
2r ⎜ x ,
⎝ u0 ⎠
⎟ dx, R 0 ≤ R ≤ R 0 + u0 t .
(4.20)
Note that the first term on the right-hand side of the above equation only depends on t, while the second term depends only
on R, i.e. incompressibility dictates that r 2 (R, t ) has an additive decomposition into an R-dependent and a t-dependent
function. Defining r¯ (R )≔r R, ( R −R0
u0 ) we can write
R
r 2 (R, t ) = r02 (t ) + ∫R 0
2r¯ (x) dx, R 0 ≤ R ≤ R 0 + u0 t .
(4.21)
Remark 4.1. In the case of constant growth velocity ug (t ) = u0 , as we mentioned earlier Ug (t ) = u0 is not the only choice. One
can choose, for example, Ug (t ) = U0 for any constant U0 (U0 > 0). Note that the material manifold explicitly depends on the
constant U0. In this case, the material manifold and metric have the following representations:
)t = { (R, Θ): 0 ≤ Θ < 2π , R1 ≤ R ≤ s (t ) = R0 + U0 t}, (4.22)
and
⎛1 0 ⎞
R1 ≤ R ≤ R 0: G=⎜ ⎟,
⎝ 0 R2 ⎠
⎛ ⎛ ⎞2 ⎞
⎜ ⎜ u0 ⎟ 0 ⎟
⎜ ⎝ U0 ⎠ ⎟
R 0 ≤ R ≤ R 0 + U0 t: G=⎜ ⎟.
⎜ 0 ⎛ ⎞
R − R0 ⎟
⎜ r 2 ⎜ R, ⎟⎟
⎝ ⎝ U0 ⎠⎠ (4.23)
Note that, similar to what was shown in the case of discrete accretion, the accretion stresses will be independent of this
choice. In other words, the material manifold is not unique. However, the material manifolds given by Eqs. (4.22) and (4.23)
are equivalent in the sense that using any of them the same stresses are calculated. This will be shown in Remark 4.3.
4.1.2. Calculation of stresses for an incompressible isotropic accreting cylinder

We next calculate the stress field during the accretion process. For R1 ≤ R ≤ R0 , Finger tensor, b♯ , and b−1 have the same
forms as given in (3.12) and (3.14), while principal stretches have forms identical to those in (3.13). Using equilibrium
equations, the radial Cauchy stress has the following expression:
R α (x, t ) + β (x, t ) ⎡ x4 ⎤
σ rr (R, t ) = − pi (t ) + ∫R1 x
⎢1 − 4
⎣
⎥ dx.
r (x, t ) ⎦ (4.24)
For R0 ≤ R ≤ R0 + u0 t , the Finger tensor, b♯ , and b−1 read

⎛ r¯2 (R) ⎞ ⎛ r 2 (R, t ) ⎞

⎜ 2 0 ⎟ ⎜ 2 0 ⎟
⎜ r (R, t ) ⎟ ⎜ r¯ (R) ⎟
b♯ = ⎜ , b−1 = ⎜ .
1 ⎟ r¯2 (R) ⎟
⎜ 0 ⎟ ⎜ 0 ⎟
⎝ r¯2 (R) ⎠ ⎝ r 4 (R, t ) ⎠ (4.25)
The principal invariants of b read

r 2 (R, t ) r¯2 (R)
I1 = + 2 , I2 = 1.
r¯2 (R) r (R, t ) (4.26)
The Cauchy stress in this region has the following representation:

⎛ r¯2 (R) ⎛ r 2 (R, t ) ⎞ ⎞
⎜ − p (R, t ) + α (R, t ) + ⎜ 1 − 2 ⎟ β (R, t ) 0 ⎟
⎜ 2
r (R, t ) ⎝ r (R) ⎠
¯ ⎟
σ=⎜
⎛ 2 (R) ⎞ β (R, t )
⎟.
⎜ p ( R , t ) α (R , t ) r
¯ ⎟
⎜ 0 − + + ⎜ 1 − ⎟ ⎟
⎝ r 2 (R, t ) r¯2 (R) ⎝ r 2 (R, t ) ⎠ r 2 (R, t ) ⎠ (4.27)
Using the radial equilibrium equation, the radial Cauchy stress is calculated as
R α (x, t ) + β (x, t ) ⎡ r¯ 4 (x) ⎤
σ rr (R, t ) = σ rr (R 0, t ) + ∫R 0 ¯r (x)
⎢1 − 4
⎣
⎥ dx
r (x, t ) ⎦
R0 α (x, t ) + β (x, t ) ⎡ x4 ⎤ R α (x, t ) + β (x, t ) ⎡ r¯ 4 (x) ⎤
= − pi (t ) + ∫R 1 x
⎢1 − 4
⎣
⎥ dx +
r (x, t ) ⎦
∫R 0 r¯ (x)
⎢1 − 4
⎣
⎥ dx.
r (x, t ) ⎦ (4.28)
6
We know that σ rr (s (t ) , t ) = σ θθ (s (t ) , t ) = 0. In particular, for the radial component this gives us
R0 α (x, t ) + β (x, t ) ⎡ x4 ⎤ R 0 + u0 t α (x, t ) + β (x, t ) ⎡ r¯ 4 (x) ⎤
∫R 1 x
⎢1 − 4
⎣
⎥ dx +
r (x, t ) ⎦
∫R 0 r¯ (x)
⎢1 − 4
⎣
⎥ dx = pi (t ).
r (x, t ) ⎦ (4.29)
Here the unknowns are r0 (t ) and r¯ (R ) with the initial conditions r0 (0) = r¯ (R0 ) = R0 . Note that (σ̄ θθ is the physical component
corresponding to σ θθ .)
⎡ r 2 (R, t ) r¯2 (R) ⎤
σ¯ θθ (R, t ) = r 2 (R, t ) σ θθ (R, t ) = σ rr (R, t ) + ⎢ 2 − 2 ⎥ [α (R, t ) + β (R, t )].
⎣ r¯ (R) r (R, t ) ⎦ (4.30)
Therefore
⎡ r 2 (s (t ), t ) r¯2 (s (t )) ⎤
σ¯ θθ (s (t ), t ) = σ rr (s (t ), t ) + ⎢ 2 − 2 ⎥ [α (s (t ), t ) + β (s (t ), t )] = 0,
⎣ r¯ (s (t )) r (s (t ), t ) ⎦ (4.31)
as by definition r¯ (s (t )) = r (s (t ) , t ), i.e. the above condition is trivially satisfied.

In the case of a homogeneous neo-Hookean solid, α (R, t ) = μ and β (R, t ) = 0. In this case, Eq. (4.29) is simplified to read
R0 1⎡ x4 ⎤ R 0 + u0 t 1 ⎡ r¯ 4 (x) ⎤ p (t )
∫R ⎢1 − 4 ⎥ dx + ∫R ⎢1 − 4 ⎥ dx = i .
1 x⎣ r (x, t ) ⎦ 0 r¯ (x) ⎣ r (x, t ) ⎦ μ (4.32)
In summary, we have the following problem in the triangular region (R, t ) ∈ { 2| R 0 ≤ R ≤ R 0 + u0 t , t ∈ + }:
⎧ ⎛ R − R0 ⎞
⎪ r (R, t ) r′(R, t ) = r ⎜ R, ⎟,
⎪ ⎝ u0 ⎠
⎪
⎨ R0 1 ⎡ x4 ⎤ R 0 + u0 t 1 ⎡ r¯ 4 (x) ⎤ p (t )
⎪ ∫ ⎢1 − 2 ⎥ dx + ∫R ⎢1 − 4 ⎥ dx = i ,
⎪ R1 x ⎣ (r (R 0, t ) + x − R 0 ) ⎦ ¯r (x) ⎣ r (x, t ) ⎦
2 2 2
0 μ
⎪ r (R , 0) = R ,
⎩ 0 0 (4.33)
where the first equation is a partial differential equation, the second is a nonlinear integral equation, and the third is the
initial condition. Note that the condition r (R0, 0) = R0 implies that for R1 ≤ R ≤ R0 , r (R, 0) = R . Furthermore, we assume that
the internal pressure satisfies pi (0) = 0.
Remark 4.2. Suppose that the accreted body is inhomogeneous and is made of a material possibly different from that of the
initial body. This would not affect the material metric. However, for the accreted body (R0 ≤ R ≤ s (t )), the energy function is
^
inhomogeneous, i.e. W = W (R, I1, I2 ) = W (τ (R ) , I1, I2 ). This means that the functions α and β in the interval R0 ≤ R ≤ s (t )
6
It is assumed that the material added to the boundary in the current configuration is stress-free and hence at r2 (t ) , stress tensor, and not just the
traction vector, vanishes.
should be modified as follows:

∂W (R, I1, I2 ) ∂W (R, I1, I2 )
α (R, t ) = 2 , β=2 .
∂I1 ∂I2 (4.34)
Remark 4.3. We now prove that the stress distribution is independent of the choice of U0 > 0 in (4.23). We denote the radial
coordinate in the new material manifold by Ř . For R1 ≤ R ≤ R0 , Rˇ = R . We know that in the two material manifolds the time of
attachment of a layer of stress-free material should be the same, i.e. τˇ (Rˇ ) = τ (R ). Note that R = R0 + u0 τ (R ). Similarly,
Rˇ = R0 + U0 τˇ (Rˇ ) = R0 + U0 τ (R ). Thus, for R0 ≤ R ≤ R0 + u0 t
⎛ U ⎞ U
Rˇ (R) = R 0 ⎜ 1 − 0 ⎟ + 0 R.
⎝ u0 ⎠ u0 (4.35)
Let us denote the radial component of the deformation mapping with respect to the new material manifold by rˇ (Rˇ , t ). We
also know that r (R, t ) = rˇ (Rˇ , t ). Thus
rˇ¯ (Rˇ ) = rˇ (Rˇ , τˇ (Rˇ )) = rˇ (Rˇ , τ (R)) = r (R, τ (R)) = r¯ (R). (4.36)
It is straightforward to show that with respect to the new material manifold (4.23), b♯ ,
I1, and I2 will remain unchanged b−1,
and hence the stress distribution will not change. In other words, the Cauchy stress is independent of the choice U0 > 0.
Remark 4.4. Suppose the accreted ring (cylinder) is not stress-free. This means that if the ring added at time τ (R ) is allowed
to relax, its radius in the relaxed configuration will be χ (R ) ≠ r¯ (R ). We still choose Ug ¼ug, i.e. (4.10) is still valid. What
changes now is that the material manifold inherits its metric from the relaxed (or natural) configuration, and not from the
deformed configuration. Denoting by dΔ the thickness of this ring in its local relaxed configuration, we can express in-
compressibility as
2πχ (R) dΔ = 2πr¯ (R) u0 dt , (4.37)
where u0 dt is the thickness of the added material in the deformed configuration. Thus, dΔ/u0 dt = r¯ (R ) /χ (R ). In the material
manifold the thickness is u0 dt as well because of our choice Ug (t ) = ug (t ). Therefore, the deformation gradient from the
material manifold to the local relaxed configuration is
⎛ r¯ (R) ⎞
⎜ 0⎟
Fχ (R) = ⎜ χ (R) ⎟.
⎜ ⎟
⎝ 0 1⎠ (4.38)
We also know that in the local relaxed configuration the metric gχ is Euclidean, and it is written as
⎛1 0 ⎞
g χ (R) = ⎜ ⎟.
⎝ 0 χ (R)⎠
2
(4.39)
Now the material metric is G = χ * gχ , i.e.

⎛ r¯2 (R) ⎞
⎜ 2 0 ⎟
G (R) = ⎜ χ (R) ⎟.
⎜ ⎟
⎝ 0 χ 2 (R)⎠ (4.40)
Note that for χ (R ) = r¯ (R ) we recover the stress-free case. Note also that detG = r¯ 2 (R ) is independent of χ (R ), and hence the
incompressibility equation remains unchanged. The kinematics and stress analysis in the initial body (R1 ≤ R ≤ R0 ) remains
unchanged as well. For R0 ≤ R ≤ R0 + u0 t , the Finger tensor, b♯ , and b−1 read
⎛ χ 2 (R) ⎞ ⎛ r 2 (R, t ) ⎞
⎜ 2 0 ⎟ ⎜ 2 0 ⎟
⎜ r (R, t ) ⎟ ⎜ χ (R ) ⎟
b ♯(R) = ⎜ , b−1(R) = ⎜ .
1 ⎟ χ 2 (R) ⎟
⎜ 0 ⎟ ⎜ 0 ⎟
⎝ χ 2 (R) ⎠ ⎝ r 4 (R, t ) ⎠ (4.41)
The principal invariants of b read
r 2 (R, t ) χ 2 (R)
I1 = 2
+ 2 , I2 = 1.
χ (R) r (R, t ) (4.42)
The Cauchy stress would have the following representation:

⎛ χ 2 (R) ⎛ r 2 (R, t ) ⎞ ⎞
⎜ − p (R, t ) + α (R, t ) + ⎜ 1 − 2 ⎟ β (R, t ) 0 ⎟
⎜ 2
r (R, t ) ⎝ χ (R) ⎠ ⎟
σ=⎜ ⎟.
⎜ p (R, t ) α (R, t ) ⎛ 2 ⎞
χ (R) β (R, t ) ⎟
⎜ 0 − 2 + 2 + ⎜1 − 2 ⎟ 2 ⎟
⎝ r (R, t ) χ (R) ⎝ r (R, t ) ⎠ r (R, t ) ⎠ (4.43)
The radial Cauchy stress is written as

R α (x, t ) + β (x, t ) ⎡ χ 4 (x) ⎤
σ rr (R, t ) = σ rr (R 0, t ) + ∫R 0 χ 2 (x)
r¯ (x) ⎢ 1 − 4
⎣
⎥ dx
r (x, t ) ⎦
R0 α (x, t ) + β (x, t ) ⎡ x4 ⎤ R α (x, t ) + β (x, t ) ⎡ χ 4 (x) ⎤
= − pi (t ) + ∫R1 x
⎢1 − 4
⎣
⎥ dx +
r (x, t ) ⎦
∫ R0 2
χ (x)
r¯ (x) ⎢ 1 − 4
⎣
⎥ dx.
r (x, t ) ⎦ (4.44)
If we assume that the outer boundary is traction-free at all times we have the boundary condition σ rr (s (t ) , t ) = 0. Therefore
R0 α (x, t ) + β (x, t ) ⎡ x4 ⎤ R α (x, t ) + β (x, t ) ⎡ χ 4 (x) ⎤
∫R 1 x
⎢1 − 4
⎣
⎥ dx +
r (x, t ) ⎦
∫R 0 χ 2 (x)
r¯ (x) ⎢ 1 − 4
⎣
⎥ dx = pi (t ).
r (x, t ) ⎦ (4.45)
Note that in the case of no internal pressure (pi(t) = 0 for each t) the identity map r(R, t) = R is not a solution of (4.45). This
means that, as expected, the addition of prestressed material will, in general, create residual deformations and stresses even
in the absence of external loads during accretion. When the added layer is pre-stressed it can be shown that
⎡ r 2 (R, t ) χ 2 (R) ⎤
σ¯ θθ (R, t ) = σ rr (R, t ) + ⎢ 2 − 2 ⎥ [α (R, t ) + β (R, t )].
⎣ χ (R) r (R, t ) ⎦ (4.46)
Thus
⎡ r¯2 (s (t )) χ 2 (s (t )) ⎤
σ¯ θθ (s (t ), t ) = ⎢ 2 − 2 ⎥ [α (s (t ), t ) + β (s (t ), t )].
⎣ χ (s (t )) r¯ (s (t )) ⎦ (4.47)
It is seen that unlike the previous case, σ̄ θθ (s (t ) , t ) ≠ 0, in general.
4.1.3. Calculation of residual stresses for an incompressible isotropic accreting cylinder

Suppose at some time t = ta , the accretion process stops. For any t > ta , if one removes the internal pressure the accreted
body would be residually stressed. The residual stress distribution depends on the accretion characteristics (here the
constant growth velocity u0) and the history of the applied loads in the time interval t ∈ [0, ta ], i.e. the history of the internal
pressure {pi (t ) , t ∈ [0, ta ]}. Material metric of the additively manufactured cylinder has the following representation:
⎛1 0 ⎞
R1 ≤ R ≤ R 0: G=⎜ ⎟,
⎝ 0 R2 ⎠
⎛1 0 ⎞
R 0 ≤ R ≤ R a: G=⎜ ⎟.
⎝ 0 r¯2 (R)⎠ (4.48)
Note that the function r (R, t ) and hence r¯ (R ) = r R, ( R −R0

u0 ) have already been calculated. The motion from the material
manifold to the residually stressed configuration (under no applied loads) is denoted by φ̃ : ) → : , where in polar co-
ordinates we have φ˜(R, Θ ) = (r˜ , θ˜ ) = (r˜ (R ) , Θ ), see Fig. 3. Using the incompressibility constraint one obtains
r˜2 (R) = r˜2 (R 0 ) + R2 − R 02, R1 ≤ R ≤ R 0,

R
r˜2 (R) = r˜2 (R 0 ) + ∫R 0
2r¯2 (x) dx, R 0 ≤ R ≤ R a,
(4.49)
where r˜ (R0 ) is an unknown that will be determined after enforcing the boundary and continuity conditions. For the de-
formation mapping φ̃ we have
r˜2 (R) R2
I1 = + 2 , I2 = 1, R1 ≤ R ≤ R 0,
R2 r˜ (R)
r˜2 (R) r¯2 (R)
I1 = 2 + 2 , I2 = 1, R 0 ≤ R ≤ R a.
r¯ (R) r˜ (R) (4.50)
In the absence of applied loads the boundary conditions read σ˜ rr (R1) = σ˜ rr (Ra ) = 0. One then has the following distribution of
radial Cauchy stress (the continuity of traction σ˜ rr (R0−) = σ˜ rr (R0+) has been enforced):
Fig. 3. Continuous outer accretion: (a) The material manifold (), G) . The outer radius at time t is s(t). At a later time t + dt the outer radius is changed to
s (t ) + Ug (t ) dt . (b) The deformed annulus under a specified internal pressure pi(t) with a layer of stress-free material of thickness ug (t ) dt attached to its outer
boundary during the time interval [t , t + dt ]. (c) The residually stressed accreted body after removal of external forces (here the internal pressure pi(t)).
α˜(x) + β˜ (x) ⎡
R x4 ⎤
R1 ≤ R ≤ R 0: σ˜ rr (R) = ∫R1 x
⎢ 1 − 4 ⎥ dx,
⎣ r˜ (x) ⎦
R0 α ˜(x) + β˜ (x) ⎡ x4 ⎤ R α˜(x) + β˜ (x) ⎡ r¯ 4 (x) ⎤
R 0 ≤ R ≤ R a: σ˜ rr (R) = ∫R1 x
⎢ 1 − 4 ⎥ dx +
⎣ r˜ (x) ⎦
∫R 0 r¯ (x)
⎢ 1 − 4 ⎥ dx.
⎣ r˜ (x) ⎦ (4.51)
Similarly, for the physical circumferential Cauchy stress component we have

⎡ ⎤ ⎡ r˜2 (R) R2 ⎤
R1 ≤ R ≤ R 0: σ˜¯ θθ (R) = σ˜ rr (R) + ⎢ α˜(R) + β˜ (R) ⎥ ⎢ 2 − 2 ⎥,
⎣ ⎦⎣ R ˜r (R) ⎦
⎡ ⎤ ⎡ r˜2 (R) r¯2 (R) ⎤
R 0 ≤ R ≤ R a: σ˜¯ θθ (R) = σ˜ rr (R) + ⎢ α˜(R) + β˜ (R) ⎥ ⎢ 2 − 2 ⎥.
⎣ ⎦ ⎣ r¯ (R) r˜ (R) ⎦ (4.52)
Note that, in general, r˜ (Ra ) ≠ r¯ (Ra ), and hence σ̃¯ θθ (R a) ≠ 0. The unknown constant r˜ (R0 ) is determined using the following
condition:
R0 α˜(x) + β˜ (x) ⎡ x4 ⎤ Ra α˜(x) + β˜ (x) ⎡ r¯ 4 (x) ⎤
σ˜ rr (R a ) = ∫R
1 x
⎢ 1 − 4 ⎥ dx +
⎣ r (x) ⎦
˜
∫R 0 ¯r (x)
⎢ 1 − 4 ⎥ dx = 0.
⎣ r˜ (x) ⎦ (4.53)
Remark 4.5. For a neo-Hookean solid, in the initial body (R1 ≤ R ≤ R0 ), dσ̃ rr (R ) /dR has the same sign as r˜ (R0 ) − R0 , i.e. the
radial component of the Cauchy stress is either strictly increasing or strictly decreasing. Similarly, dσ̃¯ θθ (R ) /R has the same
sign as −(r˜ (R0 ) − R0 ). In the secondary body (R0 ≤ R ≤ Ra ) the critical points of σ̃ rr (R ) are the zeros of r˜ (R ) − r¯ (R ) = 0.
4.1.4. Accretion stresses in the linearized theory

In this section we calculate the accretion stresses when strains are small. We do this using two approaches. In the first
method, we linearize the present nonlinear theory with respect to a trivial stress-free configuration. In the second approach
we follow Brown and Goodman (1963). We will see that the two methods lead to identical solutions as expected.
Linearization of the nonlinear governing equations: In nonlinear elasticity one can linearize the kinematics and the gov-
erning equations with respect to any motion. Usually, linearization is done with respect to a stress-free configuration.
However, this is not always the case; the so-called small-on-large theory of Green et al. (1952) is linearization about a
finitely deformed and stressed configuration. In geometric elasticity, in order to linearize one needs a reference motion φ∘
and a one-parameter family of motions φ such that φ = φ∘ (Marsden and Hughes, 1983; Yavari and Ozakin, 2008).
ϵ ϵ=0
Let us consider a one-parameter family of motions φϵ that are assumed to be axisymmetric, i.e. φϵ (R, Θ ) = (rϵ (R, t ) , Θ, t ).
Here we linearize about the stress-free configuration φ∘ (R, Θ) = (R, Θ), i.e. rϵ= 0 (R, t ) = R . Note that this reference motion
corresponds to deformation of a hollow cylindrical bar under no external forces going through accretion by adding stress-
free layers to its outer boundary. The variation field is defined as
d
δφt (R, Θ) = φϵ (R, Θ, t ).
dϵ ϵ= 0 (4.54)
7
This is what in linear elasticity is called the displacement field. For the special variations we consider here only the radial
displacement component is non-zero and is defined as
d
δr (R, t ) = rϵ (R, t ).
dϵ ϵ= 0 (4.55)
This implies that δr0 (t ) = δr (R0, t ) and δr¯ (R ) = δr R, ( R −R0

u0 ).
2 2 2 2
Kinematics: For R1 ≤ R ≤ R0 , for the perturbed motions we have rϵ= 0 (R , t ) = rϵ (R 0, t ) + R − R2 . Taking derivative with
respect to ϵ on both sides and evaluating at ϵ = 0 one obtains

R0
δr (R, t ) = δr0 (t ),
R (4.56)
where δr0 (t ) = δr (R0, t ). Knowing that r (R, 0) = R , one concludes that δr (R, 0) = 0 (note that for t¼0, R1 ≤ R ≤ R0 ). For
R 0 ≤ R ≤ s (t )
R
rϵ2 (R, t ) = rϵ2 (R 0, t ) + ∫R 0
2r¯ϵ (x) dx.
(4.57)
Again taking derivative with respect to ϵ on both sides and evaluating at ϵ = 0 one obtains
R0 1 R
δr (R, t ) =
R
δr0 (t ) +
R
∫R 0
δr¯ (x) dx.
(4.58)
Evaluating (4.58) at t = τ (R ), one gets
1 R R0
R
∫R 0
δr¯ (x) dx = δr¯ (R) −
R
δr0 (τ (R)).
(4.59)
Substituting the above relation back into (4.58) one has
R0 ⎡ ⎤
δr (R, t ) − δr¯ (R) = ⎢ δr0 (t ) − δr0 (τ (R)) ⎥⎦.
R ⎣ (4.60)
Accretion stresses: Let us start with a neo-Hookean solid and linearize the surface growth-induced stresses.8 We consider
a one-parameter family of applied internal pressures (pi )ϵ (t ) and assume that (pi )ϵ= 0 (t ) = 0. With an abuse of notation, we
denote the linearized pressure field by pi(t) (instead of δpi (t )). For R1 ≤ R ≤ R0 , for the perturbed motions from (4.24) we have
R μ⎡ x4 ⎤
σϵrr (R, t ) = − (pi )ϵ (t ) + ∫R ⎢1 − 4 ⎥ dx.
1 x⎣ rϵ (x, t ) ⎦ (4.61)
Taking derivative with respect to ϵ on both sides and evaluating at ϵ = 0 one obtains
R δr (R, t )
δσ rr (R, t ) = − pi (t ) + 4μ ∫R 1 x2
dx.
(4.62)
Using (4.56) we have
⎛ 1 1⎞
δσ rr (R, t ) = − pi (t ) + 2μR 0 ⎜ 2 − 2 ⎟ δr0 (t ).
⎝ R1 R ⎠ (4.63)
For R0 ≤ R ≤ R0 + u0 t , from (4.28) one has
7
However, in linear accretion mechanics there is a subtlety in defining the displacement field. The displacement field for the new material points is
defined with respect to their positions at the time of attachment. For surface growth of a cylindrical bar, this means that U (R, Θ, t ) = δφt (R, Θ) − δφτ (R) (R, Θ) .
8
In this particular problem, for an arbitrary isotropic incompressible solid instead of μ one would have the following constant:
⎛ ∂W ∂W ⎞
2⎜ + ⎟ .
⎝ ∂I1 ∂I2 ⎠ I1= 2, I 2= 1
R μ ⎡ r¯ 4 (x) ⎤
σϵrr (R, t ) = σϵrr (R 0, t ) + ∫R ⎢ 1 − 4ϵ ⎥ dx.
0 r¯ϵ (x) ⎣ rϵ (x, t ) ⎦ (4.64)
Therefore
⎛ 1 1⎞ R δr (x, t ) − δr¯ (x)

δσ rr (R, t ) = − pi (t ) + 2μR 0 ⎜ 2 − 2 ⎟ δr0 (t ) + 4μ
⎝ R1 R0 ⎠
∫R x2
dx,
0 (4.65)
and using (4.60) we obtain
⎛ 1 1⎞ R δr0 (τ (x))
δσ rr (R, t ) = − pi (t ) + 2μR 0 ⎜ 2 − 2 ⎟ δr0 (t ) − 4μR 0
⎝ R1 R ⎠
∫R x3
dx.
0 (4.66)
We know that σϵrr (s (t ) , t ) = 0 and hence δσ rr (s (t ) , t ) = 0. Therefore
⎛ 1 1 ⎞ s (t ) δr0 (τ (x)) p (t )
⎜ 2 − 2 ⎟ δr0 (t ) − 2 ∫R dx = i .
⎝ R1 s (t ) ⎠ 0 x3 2μ R 0 (4.67)
Taking derivative with respect to t of both sides one reduces the above integral equation to the following simple ODE:
R12 s2 (t ) pi̇ (t )
δr0̇ (t ) = .
2μR 0 s2 (t ) − R12 (4.68)
Integrating this ODE from 0 to t and noting that δṙ (R0, 0) = 0, one obtains
R12 t s2 (η) pi̇ (η)

δr0 (t ) =
2μ R 0
∫0 s2 (η) − R12
dη .
(4.69)
Therefore, substituting into (4.60), we have
R12 t s2 (η) pi̇ (η)

δr (R, t ) − δr¯ (R) =
2μ R
∫τ (R) s2 (η) − R12
dη .
(4.70)
As was mentioned earlier, the left-hand side of the above relation is what is defined to be the radial displacement in linear
surface growth mechanics. Note that once the difference δr (R, t ) − δr¯ (R ) is known, one can calculate δr (R, t ) by solving Eq.
(4.59) for δr¯ (R ).
Taking time derivative of both sides of (4.66) we obtain
⎛ ⎞
̇ (R, t ) = − ṗ (t ) + 2μR 0 ⎜ 1 − 1 ⎟ δr0̇ (t ).
δσ rr i 2
⎝ R1 R2
⎠ (4.71)
Integrating the above relation from τ (R ) to t and using the condition δσ rr (R, τ (R )) = 0, we have
⎛ R2 ⎞ t s2 (η) pi̇ (η)

δσ rr (R, t ) = − pi (t ) + pi (τ (R)) + ⎜ 1 − 12 ⎟ ∫τ (R) dη .
⎝ R ⎠ s2 (η) − R12 (4.72)
Linearizing (4.30) we get (σ¯ θθ = r 2σ θθ is the physical circumferential component)
4μ ⎡ 2R 2 t s2 (η) pi̇ (η)

δσ¯ θθ (R, t ) = δσ¯ rr (R, t ) + ⎣ δr (R, t ) − δr¯ (R) ⎤⎦ = δσ¯ rr (R, t ) + 21 ∫τ (R) dη ,
R R s2 (η) − R12 (4.73)
which leads to
⎛ R2 ⎞ t s2 (η) pi̇ (η)

δσ¯ θθ (R, t ) = − pi (t ) + pi (τ (R)) + ⎜ 1 + 12 ⎟ ∫τ (R) dη .
⎝ R ⎠ s2 (η) − R12 (4.74)
Calculation of accretion stresses using the linear theory of Brown and Goodman (1963): Next we solve the symmetric
accretion problem of a hollow cylindrical bar in the setting of linear elasticity. We follow Brown and Goodman (1963), but
instead of using the external radius s (t ) = R0 + u0 t as the independent variable we use time. We indicate with UR (R, t ) the
radial displacement, with pi(t) the internal pressure, with σ (R, t ) the Cauchy stress tensor, and with p (R, t ) the pressure field.
All these quantities are considered in the framework of linearized elasticity.
In the initial body, i.e, for R1 ≤ R ≤ R0 , where strains are compatible, the linear problem consists of the following gov-
erning equations:
⎧ ∂U R U R (R , t )
⎪ (R , t ) + =0 (incompressibility condition) ,
⎪ ∂R R
⎪ R R
⎨ σ RR (R, t ) = 2μ ∂U (R, t ) − p (R, t ) , σ ΘΘ (R, t ) = 2μ U (R, t ) − p (R, t ) (constitutive relations) ,
⎪ ∂R R3 R2
⎪ ∂σ RR RR
σ (R , t )
⎪ (R , t ) + − Rσ ΘΘ (R, t ) = 0 (radial equilibrium equation) .
⎩ ∂R R (4.75)
The boundary and continuity conditions are σ RR (R1, t ) = − pi (t ) and σ RR (R0−, t ) = σ RR (R0+, t ). From the incompressibility
condition we can write the radial displacement as
R0
U R (R, t ) = U0R (t ) ,
R (4.76)
where U0R (t ) = UR (R0, t ) is an unknown function to be determined. Therefore, the equilibrium equation is simplified to read
∂σ RR (R, t ) R U R (t )
= 4μ 0 03 ,
∂R R (4.77)
which, taking into account the traction boundary condition at R = R1, leads to
R R 0 U0R (t ) ⎛ 1 1⎞
σ RR (R, t ) = − pi (t ) + ∫R 4μ 3
dx = − pi (t ) + 2μR 0 ⎜ 2 − 2 ⎟ U0R (t ).
1 x ⎝ R1 R ⎠ (4.78)
From (4.75)2 we know that ¯ σ ΘΘ (R, t) = σ RR + 4μUR/R , which gives us

⎛ 1 1⎞
σ¯ ΘΘ (R, t ) = − pi (t ) + 2μR 0 ⎜ 2 + 2 ⎟ U R (R 0, t ).
⎝ R1 R ⎠ (4.79)
Now one can write σ RR (R 0, t ), which is denoted by σ0 (t ), as

⎛ 1 1⎞
σ 0 (t ) = σ RR (R 0, t ) = − pi (t ) + 2μR 0 ⎜ 2 + 2 ⎟ U0R (t ).
⎝ R1 R0 ⎠ (4.80)
For R0 ≤ R ≤ R0 + u0 t = s (t ), following Brown and Goodman (1963), the incremental problem for the growing in-
compressible cylinder at time t is written as:
⎧ ̇R ̇R
⎪ ∂U (R, t ) + U (R, t ) = 0 (incompressibility condition) ,
⎪ ∂R R
⎪ ̇R ̇R
⎨ σ ̇ RR (R, t ) = 2μ ∂U (R, t ) − ṗ (R, t ) , σ ̇ΘΘ (R, t ) = 2μ U (R, t ) − ṗ (R, t ) (constitutive relations) ,
⎪ ∂R R3 R2
⎪ RR RR (R , t )
⎪ ∂σ ̇ σ ̇
(R , t ) + − Rσ ̇ΘΘ (R, t ) = 0 (radial equilibrium equation) .
⎩ ∂R R (4.81)
The boundary conditions for this incremental problem are ̇ ̇ (t ) and

0 , t ) = σ0 σ RR (R σ̇ RR (s (t ) , t ) = 0 , where σ̇0 (t ) is the rate of the
radial stress in (4.80) at the interface with the initial body, i.e.
⎛ 1 1⎞ R
σ 0̇ (t ) = − pi̇ (t ) + 2μR 0 ⎜ 2 − 2 ⎟ U0̇ (t ).
⎝ R1 R0 ⎠ (4.82)
Therefore, from Eqs. (4.76), (4.78) and (4.79) one obtains

R R R
U̇ (R, t ) = 0 U̇0 (t ),
R (4.83)
⎛ 1 1⎞ R ⎛ 1 1⎞ R
σ ̇RR (R, t ) = σ 0̇ (t ) + 2μR 0 ⎜ 2 − 2 ⎟ U0̇ (t ) = − pi̇ (t ) + 2μR 0 ⎜ 2 − 2 ⎟ U0̇ (t ),
⎝ R0 R ⎠ ⎝ R1 R ⎠ (4.84)
⎛ 1 1 ⎞ ̇R ⎛ 1 1⎞ R
̇ (R, t ) = σ 0̇ (t ) + 2μR 0 ⎜ 2 +
σ¯ ΘΘ ⎟ U0 (t ) = − pi̇ (t ) + 2μR 0 ⎜ 2 + 2 ⎟ U0̇ (t ).
⎝ R0 R2 ⎠ ⎝ R1 R ⎠ (4.85)
The function U0R (t ) can now be obtained using the condition σ̇ RR (s (t ) , t ) = 0, i.e.
R R12 s2 (t ) pi̇ (t )
U0̇ (t ) = .
2μR 0 s2 (t ) − R12 (4.86)
Note that
R ̇ (R, t ), ̇ (R, t ).
U̇0 (t ) = δṙ (R 0, t ), σ ̇RR (R, t ) = δσ rr σ¯ ΘΘ
̇ (R, t ) = δσ θθ (4.87)
Moreover, the total displacement and stress fields are given by

t t
R
U R (R, t ) = U R (R, τ (R)) + ∫τ (R) U̇ (R, η) dη, σ (R, t ) = σ (R, τ (R)) + ∫τ (R) σ ̇ (R, η) dη,
(4.88)
where σ RR (R,
τ (R )) = 0, σ ΘΘ (R,
τ (R )) = 0 , and U R (R ,
τ (R )) = 0. Therefore, as expected, the two approaches lead to identical
solutions.
Residual stresses: Integrating Eqs. (4.70), (4.72) and (4.74) by parts, one obtains
R12 s2 (t ) pi (t ) R 2 R p (τ (R)) u R4 t s (η) pi (η)

U R (R, t ) = − 21 2 i + 0 1 ∫ dη ,
2μR s2 (t ) − R12 R − R1 2μ μR τ (R) (s2 (η) − R12 )2
R 2 s2 (t ) − R2 ⎛ R2 ⎞ t s (η) pi (η)
σ RR (R, t ) = − 12 2
R s (t ) − R1 2
pi (t ) + 2u0 R12 ⎜ 1 − 12 ⎟
⎝
∫
R ⎠ τ (R) (s2 (η) − R12 )2
dη ,
R12 s2 (t ) + R2 2R 2 ⎛ R2 ⎞ t s (η) pi (η)

σ¯ ΘΘ (R, t ) = p (t ) − 2 1 2 pi (τ (R)) + 2u0 R12 ⎜ 1 + 12 ⎟
R2 s2 (t ) − R12 i R − R1 ⎝ R ⎠
∫τ (R) (s2 (η) − R12 )2
dη .
(4.89)
If at time ta we stop surface growth and remove the internal applied pressure, we would observe residual stresses in both
the initial and the secondary bodies. In the linear theory of accretion these stresses can be calculated using superposition.
For the cylinder of internal radius R1 and external radius s (ta ) = R0 + u0 ta loaded by an internal pressure −pi (ta ) (unloading)
and traction-free at the external radius one has
R 1 R12 s2 (ta ) pi (ta )

Uunloading (R, ta ) = − ,
2 R s2 (ta ) − R12 μ (4.90)
RR R12 s2 (ta ) − R2
σunloading (R, ta ) = p (ta ),
R2 s2 (ta ) − R12 i (4.91)
ΘΘ R12 s2 (ta ) + R2
σ̄unloading (R, ta ) = − p (ta ).
R2 s2 (ta ) − R12 i (4.92)
Adding the above to their corresponding values at the end of loading one obtains the residual displacement and the residual
stresses as:
R12 R pi (τ (R)) u0 R14 ta s (η) pi (η)

R
Ures (R, ta ) = −
R2 − R12 2μ
+
μR
∫R − R 0
(s2 (η) − R12 )2
dη ,
u0 (4.93)
⎛ R 2 ⎞ ta s (η) pi (η)
RR
σres (R, ta ) = 2u0 R12 ⎜⎜ 1 − 12 ⎟⎟ R − R 0 2
∫ dη ,
⎝ R ⎠ u0 (s (η) − R12 )2 (4.94)
R2 ⎛ ⎞ ⎛ R2 ⎞ ta s (η) pi (η)
R − R0 ⎟
ΘΘ
σ¯res (R, ta ) = − 2 2 1 2 pi ⎜⎜ ⎟ + 2u0 R12 ⎜⎜ 1 + 02 ⎟⎟ ∫R − R dη .
R − R1 ⎝ u 0 ⎠ ⎝ R ⎠ u0
0
(s2 (η) − R12 )2 (4.95)
4.2. Inner accretion
In this section we consider an infinitely long thick hollow cylinder under a specified time-dependent internal pressure
pi(t). Initially the inner and outer radii of the cylinder are R1 and R0, respectively. We assume that material is continuously
added to the cylinder on its inner boundary. The outer boundary is assumed to be traction free (see Fig. 4).
4.2.1. Kinematics and material metric

Again, for the sake of simplicity of calculations, we assume a constant growth velocity u1 > 0 and work with the choice
Ug (t ) = ug (t ) = u1. In this problem the time-dependent material manifold is
)t = { (R, Θ): 0 ≤ Θ < 2π , S (t ) = R1 − u1t ≤ R ≤ R0 }. (4.96)

R1
Note that in order to avoid interpenetration of matter we assume that 0 ≤ t < u1
.
The ambient space metric and the deformation gradient have the same forms as those of the outer accretion example.
Again, we assign to each layer at radius R in the material manifold a time of accretion τ (R ), which is assumed to be invertible
with inverse S = τ −1. We assume that τ (R1) = 0. Note that ug (t ) = r′(S (t ) , t ) Ug (t ), where Ug (t ) = S ̇ (t ). Equivalently,
r′(R, τ (R )) = τ′(R ) ug (τ (R )). The choice Ug (t ) = ug (t ) implies that r′(S (t ) , t ) = 1 or equivalently r′(R, τ (R )) = 1. For constant
growth velocity ug (t ) = u1 > 0 one has
Fig. 4. Continuous inner accretion: (a) The material manifold (), G) . The inner radius at time t is S(t). At a later time t + dt the inner radius is changed to
S (t ) − Ug (t ) dt . (b) The deformed annulus under a specified internal pressure pi(t) with a layer of stress-free material of thickness ug (t ) dt attached to it
during the time interval [t , t + dt ]. (c) The residually stressed accreted body after removal of external forces (here the internal pressure pi(t)).
R1 − R
S (t ) = R1 − u1t or τ (R) = .
u1 (4.97)
For R1 ≤ R ≤ R0 , the material metric is flat and has the representation G = diag {1, R2}
in polar coordinates.
The material points attached to the growth surface at time t are in a state of hydrostatic pressure with magnitude pi(t)
(when pi (t ) < 0 this would be a hydrostatic suction). In the time interval [t , t + dt ] a ring of radius r (S (t ) , t ) with thickness
u1dt is attached to the deformed body. Assuming that the body and the attached particles are made of an incompressible
material, if this ring (which is in a state of hydrostatic pressure or suction) is unloaded its radius would remain r (S (t ) , t ), i.e.
r (S (t ) , t ) is the radius of the accreted ring in its stress-free configuration as well. Therefore, for S (t ) ≤ R ≤ R1, we have
G (S (t )) = φt* g (r (S (t ), t )), or equivalently G (R) = φτ*(R) g (r (R, τ (R))). (4.98)
Hence
⎛1 0 ⎞
G=⎜ ⎟.
⎝ 0 r (R, τ (R))⎠
2
(4.99)
For ug (t ) = u1, this is simplified to read

⎛1 0 ⎞
⎜ ⎟
G=⎜ ⎛ R1 − R ⎞⎟.
⎜ 0 r ⎜ R, ⎟⎟
2
⎝ ⎝ u1 ⎠⎠ (4.100)
Therefore, in summary, the material manifold is an evolving Riemannian manifold ()t , G), where
)t = { (R, Θ): 0 ≤ Θ < 2π , S (t ) ≤ R ≤ R0 }, (4.101)
and
⎛1 0 ⎞
⎜ ⎟
S (t ) ≤ R ≤ R1: G = ⎜ ⎛ R1 − R ⎞⎟,
⎜ 0 r ⎜ R, ⎟⎟
2
⎝ ⎝ u1 ⎠⎠
⎛ ⎞
⎜1 0 ⎟
R1 ≤ R ≤ R 0: G=⎜ ⎟.
⎜ 0 R2 ⎟
⎝ ⎠ (4.102)
Let us define r^ (R )≔r R, ( R1 −R

u1 ). Using the incompressibility constraint we obtain
R1
S (t ) ≤ R ≤ R1: r 2 (R, t ) = r 2 (R1, t ) − ∫R 2r^ (x) dx,
R1 ≤ R ≤ R 0: r 2 (R, t ) = r 2 (R1, t ) − R12 + R2. (4.103)
4.2.2. Calculation of stresses for an incompressible isotropic accreting cylinder

The stress calculation is very similar to that of the outer accretion problem. For R1 ≤ R ≤ R0 , the radial Cauchy stress
component has the following expression (the boundary condition σ rr (R0, t ) = 0 has been enforced):
R0 α (x, t ) + β (x, t ) ⎡ x4 ⎤
σ rr (R, t ) = − ∫R x
⎢1 − 4
⎣
⎥ dx.
r (x, t ) ⎦ (4.104)
For S (t ) ≤ R ≤ R1, the radial Cauchy stress component reads

⎡ 4 ⎤
R1 α (x, t ) + β (x, t ) ⎢ r^ (x) ⎥
σ rr (R, t ) = σ rr (R1, t ) − ∫R ^r (x) ⎢⎣
1− 4
r (x, t ) ⎥⎦
dx
⎡ ⎤ ⎡ 4 ⎤
R0 α (x, t ) + β (x, t ) ⎢ x4 ⎥ R1 α (x, t ) + β (x, t ) ⎢ r^ (x) ⎥
= − ∫R 1 x ⎢⎣
1− 4
r (x, t ) ⎥⎦
dx − ∫R r^ (x) ⎢⎣
1− 4
r (x, t ) ⎥⎦
dx.
(4.105)
Note that σ rr (S (t ) , t ) = − pi (t ) and thus

⎡ ⎤ ⎡ 4 ⎤
R0 α (x, t ) + β (x, t ) ⎢ x4 ⎥ R1 α (x, t ) + β (x, t ) ⎢ r^ (x) ⎥
∫R x ⎢⎣
1− 4
r (x, t ) ⎥⎦
dx + ∫S (t) r^ (x) ⎢⎣
1− 4
r (x, t ) ⎥⎦
dx = pi (t ).
1 (4.106)
Note that similar to the outer accretion problem, t ) = − pi (t ) implies that σ rr (S (t ) , σ̄ θθ (S (t ) , t ) = − pi (t ).

In the case of a homogeneous neo-Hookean solid, (4.106) is simplified to read
⎡ ⎤ ⎡ 4 ⎤
R0 1⎢ x4 ⎥ R1 1 ⎢ r^ (x) ⎥ p (t )
∫R 1− 4 dx + ∫S (t) 1− 4 dx = i .
1 x ⎢⎣ r (x, t ) ⎥⎦ r^ (x) ⎢⎣ r (x, t ) ⎥⎦ μ (4.107)
Eq. (4.107) can be expressed in terms of the function r (R, t ) and its spatial derivative r′(R, t ), leading to the following
problem in the triangular region { (R, t ) ∈  | R − u t ≤ R ≤ R , 0 < t < }:
2
1 1 1
R1
u1
⎧ ⎛ R − R⎞
⎪ r (R, t ) r′(R, t ) = r ⎜ R, 1 ⎟,
⎪ ⎝ u1 ⎠
⎪
⎨ R0 1 ⎡ x4 ⎤ R1 1 − r′ 4 (x, t ) p (t )
⎪ ∫ ⎢1 − 2 ⎥ dx + ∫R − u t dx = i ,
⎪ R1 x ⎣ (r (R 1, t ) + x2 − R 2 )2 ⎦
1 1 1 r (x, t ) r′(x, t ) μ
⎪ r (R , 0) = R .
⎩ 1 1 (4.108)
Note that the condition r (R1, 0) = R1 implies that for R1 ≤ R ≤ R0 , r (R, 0) = R . We also assume that pi (0) = 0.
4.2.3. Calculation of residual stresses for an incompressible isotropic accreting cylinder

After the completion of inner accretion at time t = ta < R1/u1, the material metric has the following representation in
polar coordinates:
⎛1 0 ⎞
⎜ ⎟
R a ≤ R ≤ R1: G=⎜ ⎛ ⎞ ,
0 r 2 ⎜ R, R1 − R ⎟⎟
⎜ ⎟
⎝ ⎝ u1 ⎠⎠
⎛ ⎞
⎜1 0 ⎟
R1 ≤ R ≤ R 0: G=⎜ ⎟,
⎜ 0 R2 ⎟
⎝ ⎠ (4.109)
where Ra = R1 − u1ta . Again, the deformation from the material manifold to the residually stressed configuration (under no
applied loads) is denoted by φ̃ : ) → : , where in polar coordinates we have φ˜(R, Θ ) = (r˜ , θ˜ ) = (r˜ (R ) , Θ ), see Fig. 4. Using the
incompressibility constraint one obtains
R1
R a ≤ R ≤ R1: r˜2 (R) = r˜2 (R1) − ∫R 2r^ (x) dx,
R1 ≤ R ≤ R 0: r˜2 (R) = r˜2 (R1) − R12 + R2, (4.110)
where r˜ (R1) is an unknown that will be determined after enforcing the boundary and continuity conditions. For the residual
deformation mapping φ̃ we have

2
r˜2 (R) r^ (R)
I1 = + 2 , I2 = 1, R a ≤ R ≤ R1,
^r 2 (R) ˜r (R)
r˜2 (R) R2
I1 = + 2 , I2 = 1, R1 ≤ R ≤ R 0.
R2 r˜ (R) (4.111)
In the absence of applied loads the boundary conditions read σ˜ rr (Ra ) = σ˜ rr (R0 ) ¼0. One then has the following distribution of
radial Cauchy stress (the continuity of traction σ˜ rr (R1−) = σ˜ rr (R1+) has been enforced):
⎡ ⎤ ⎡ 4 ⎤
R0 α˜(x) + β˜ (x) ⎢ x4 R1 α˜(x) + β˜ (x) ⎢ r^ (x)
R a ≤ R ≤ R1: σ˜ rr (R) = − ∫R 1 − 4 ⎥ dx − ∫R 1 − 4 ⎥ dx,
1 x ⎢⎣ r˜ (x) ⎥⎦ r^ (x) ⎢⎣ r˜ (x) ⎥⎦
⎡ ⎤
R0 α˜(x) + β˜ (x) ⎢ x4
R1 ≤ R ≤ R 0: σ˜ rr (R) = − ∫R 1 − 4 ⎥ dx.
x ⎢⎣ r˜ (x) ⎥⎦ (4.112)
Similarly, for the physical circumferential Cauchy stress component we have

⎡ ⎤ ⎡ ˜2 2 ⎤
r (R) r^ (R) ⎥
R a ≤ R ≤ R1: σ˜¯ θθ (R) = σ˜ rr (R) + ⎢ α˜(R) + β˜ (R) ⎥ ⎢ 2 − 2 ,
⎢⎣ ⎥⎦ ⎢⎣ r^ (R) r˜ (R) ⎥⎦
⎡ ⎤ ⎡ ˜2 ⎤
r (R) R2 ⎥
R1 ≤ R ≤ R 0: σ˜¯ θθ (R) = σ˜ rr (R) + ⎢ α˜(R) + β˜ (R) ⎥ ⎢ 2 − 2 .
⎢⎣ ⎥⎦ ⎢⎣ R r˜ (R) ⎥⎦ (4.113)
The unknown constant r˜ (R0 ) is determined using the condition σ̃ rr (R a) = 0, which is written as
⎡ ⎤ ⎡ 4 ⎤
R0 α˜(x) + β˜ (x) ⎢ x4 R1 α˜(x) + β˜ (x) ⎢ r^ (x)
∫R 1 − 4 ⎥ dx + ∫R 1 − 4 ⎥ dx = 0.
1 x ⎢⎣ r˜ (x) ⎥⎦ a r^ (x) ⎢⎣ r˜ (x) ⎥⎦ (4.114)
4.2.4. Accretion stresses in the linearized theory

Next we calculate the accretion stresses when strains are small. We consider a one-parameter family of motions φϵ that
are assumed to be axisymmetric, i.e. φϵ (R, Θ) = (rϵ (R, t ) , Θ, t ) and linearize about the stress-free configuration
φ∘ (R, Θ) = (R, Θ), i.e. rϵ= 0 (R, t ) = R . For R1 ≤ R ≤ R0 , the perturbed motion reads rϵ2 (R, t ) = rϵ2 (R1, t ) − R12 + R2. Taking derivative
with respect to ϵ, and evaluating at ϵ = 0, we obtain
R1
δr (R, t ) = δr1 (t ),
R (4.115)
where δr1 (t ) = δr (R1, t ). Knowing that r (R, 0) = R , one concludes that δr (R, 0) = 0. Note that for t¼0, R1 ≤ R ≤ R0 . For
S (t ) ≤ R ≤ R1 we have
R1
rϵ2 (R, t ) = rϵ2 (R1, t ) − ∫R 2r^ϵ (x) dx.
(4.116)
Thus
R1 1 R1
δr (R, t ) = δr1 (t ) − ∫R δr^ (x) dx.
R R (4.117)
Evaluating (4.117)1 at t = τ (R ), one obtains

1 R1 R1
− ∫R δr^ (x) dx = δr^ (R) − δr1 (τ (R)).
R R (4.118)
Substituting the above relation back into (4.117) one gets

R ⎡ ⎤
δr (R, t ) − δr^ (R) = 1 ⎢ δr1 (t ) − δr1 (τ (R)) ⎥.
R⎣ ⎦ (4.119)
Let us define
⎛ ∂W ∂W ⎞
μ = 2⎜ + ⎟ .
⎝ ∂I1 ∂I2 ⎠ I1= 2, I2= 1 (4.120)
Similar to the outer accretion problem, we consider a one-parameter family of applied internal pressures (pi )ϵ (t )
((pi )ϵ= 0 (t ) = 0) and with an abuse of notation denote the linearized pressure field by pi(t). Similar to the linearization of
stresses for the outer accretion problem we have the following linearized radial Cauchy stress. For R1 ≤ R ≤ R0 , using (4.115),
we obtain
R0 δr (x, t ) ⎛ 1 1⎞
δσ rr (R, t ) = − 4μ ∫R x2
dx = − 2μR1 ⎜ 2 − 2 ⎟ δr1 (t ).
⎝R R0 ⎠ (4.121)
For S (t ) ≤ R ≤ R1
R0 δr (x, t ) R1 δr (x, t ) − δr^ (x)
δσ rr (R, t ) = − 4μ ∫R 1 x2
dx − 4μ ∫R x2
dx.
(4.122)
Using (4.117), we can simplify the above equation as

⎛ 1 1⎞ R1 δr (x, t ) − δr^ (x)
δσ rr (R, t ) = − 2μR1 ⎜⎜ 2 − 2 ⎟⎟ δr1 (t ) − 4μ ∫R dx.
⎝ R1 R0 ⎠ x2 (4.123)
From (4.119) one obtains

⎛1 1⎞ R1 δr1 (τ (x))
δσ rr (R, t ) = − 2μR1 ⎜ − 2 ⎟ δr1 (t ) + 2
⎝R R0 ⎠
∫R x3
dx.
(4.124)
Imposing δσ rr (S (t ) , t ) = − pi (t ) one obtains

⎛ 1 1⎞ R1 δr1 (τ (x)) p (t )
⎜ 2 − 2 ⎟ δr1 (t ) − 2 ∫S (t) dx = i , t > 0.
⎝ S (t ) R0 ⎠ x3 2μR1 (4.125)
R 2 S 2 (t ) pi̇ (t )
δr1̇ (t ) = 0 2 .
2μR1 R 0 − S 2 (t ) (4.126)
Integrating this ODE from 0 to t and noting that δr1 (0) = 0, one obtains
R 02 t S 2 (η) pi̇ (η)

δr (R1, t ) =
2μR1
∫0 R 02 − S 2 (η)
dη .
(4.127)
Therefore, substituting into (4.119), we have
R2 t S 2 (η) pi̇ (η)

δr (R, t ) − δr^ (R) = 0 ∫τ (R) dη .
2μ R R 02 − S 2 (η) (4.128)
Taking time derivative of both sides of (4.124)1 we obtain

⎛ 2 ⎞ 2
̇ (R, t ) = − ⎜ R 0 − 1⎟ S (t ) pi̇ (t ) .
δσ rr
⎝R 2
⎠ R 02 − S 2 (t ) (4.129)
Integrating the above relation from τ (R ) to t and noting that δσ rr (R, τ (R )) = − pi (τ (R )), we have
⎛ R2 ⎞ t S 2 (η) pi̇ (η)
δσ rr (R, t ) = − pi (τ (R)) − ⎜ 02 − 1⎟ ∫τ (R) dη .
⎝R ⎠ R 02 − S 2 (η) (4.130)
As regards the component δσ̄ θθ (R, t ), linearizing (4.30), valid in the case of inner accretion as well, we obtain
4μ ⎡ 2R 2 t S 2 (η) pi̇ (η)

δσ¯ θθ (R, t ) = δσ¯ rr (R, t ) + ⎣ δr (R, t ) − δr¯ (R) ⎤⎦ = δσ¯ rr (R, t ) + 20 ∫τ (R) dη ,
R R R 02 − S 2 (η) (4.131)
which leads to
⎛ R2 ⎞ t S 2 (η) pi̇ (η)
δσ¯ θθ (R, t ) = − pi (τ (R)) + ⎜ 02 + 1⎟ ∫τ (R) dξ .
⎝R ⎠ R 02 − S 2 (η) (4.132)
Residual stresses: Suppose surface growth stops at time ta and one removes the internal applied pressure. In the linear
theory of accretion residual stresses can be calculated using superposition. For the cylinder of inner radius S(t) and outer
radius R0 loaded by an internal pressure −pi (ta ) (unloading) and traction-free at the external radius one has
R 1 R 02 S 2 (ta ) pi (ta )
2 R R 02 − S 2 (ta ) μ (4.133)
RR S 2 (ta ) R 02 − R2
R2 R 02 − S 2 (ta ) i (4.134)
ΘΘ S 2 (ta ) R 02 + R2
σ̄unloading (R, ta ) = − p (ta ).
R2 R 02 − S 2 (ta ) i (4.135)
Adding the above to their corresponding values at the end of loading one obtains the residual displacement and the residual
stresses as:
R 02 R pi (τ (R)) u R4 ta S (η) pi (η)

R
Ures (R, t ) = − + 1 1 ∫ dη ,
R 02 − R2 2μ μR τ (R) (R 02 − S 2 (η))2
⎛ R2 ⎞ ta S (η) pi (η)
RR
σres (R, t ) = − 2u1R 02 ⎜ 02 − 1⎟
⎝R
∫
⎠ τ (R) (R 02 − S 2 (η))2
dη ,
2R 02 ⎛ R2 ⎞ ta S (η) pi (η)
ΘΘ
σ¯res (R, t ) = −
R 02 −R 2
pi (τ (R)) + 2u1R 02 ⎜ 12 + 1⎟
⎝R ⎠
∫τ (R) (R 02 − S 2 (η))2
dη .
(4.136)
5. Surface growth of a thick hollow solid sphere
In this section we consider continuous accretion of a hollow spherical ball under a time-dependent internal pressure. We
assume that the initial stress-free body is a hollow spherical ball with inner radius R1 and outer radius R0. At time t, under a
specified internal pressure pi(t), the outer radius in the current configuration is r2 (t ). Note that we assume that the pressure
pi(t) is a priori specified and the inner radius of the ball is part of the unknowns to be solved for. Similar to the cylinder
problem in Section 4, we assume that accretion velocity is normal to the boundary with magnitude ug(t) in the current
configuration. For a hollow ball we only consider outer accretion, i.e. material is continuously added on the outer boundary
of the spherical ball. The metric of the Euclidean ambient space in spherical coordinates (r, θ , ϕ) reads
⎛1 0 0 ⎞
⎜ ⎟
g = ⎜ 0 r2 0 ⎟⎟.
⎜
⎝ 0 0 r 2 sin2 θ ⎠ (5.1)
In the material manifold ) we use the spherical coordinates (R, Θ, Φ ).
5.1. Kinematics and material metric
We assume motions of the form (r , θ , ϕ) = (r (R, t ) , Θ, Φ ), and hence the deformation gradient reads
⎛ r′(R, t ) 0 0⎞
⎜ ⎟
F=⎜ 0 1 0 ⎟,
⎜ ⎟
⎝ 0 0 1⎠ (5.2)
∂r (R, t )
where r′(R, t ) = ∂R
. Again we assign to each layer at radius R in the material manifold a time of accretion τ (R ). This map
has the inverse s = τ −1, which assigns to each time the radius of the new spherical layer in the material manifold. It is
assumed that τ (R0 ) = 0 and r2 (t ) = r (s (t ) , t ). Again we choose Ug (t ) = ug (t ) and assume that ug (t ) = u0 > 0. We still have the
relationships r′(s (t ) , t ) = 1 and s (t ) = R0 + u0 t .
For R1 ≤ R ≤ R0 (the initial body), the material metric has the following representation in spherical coordinates:
⎛1 0 0 ⎞
⎜ ⎟
G = ⎜ 0 R2 0 ⎟⎟.
⎜
⎝ 0 0 R2 sin2 Θ ⎠ (5.3)
Assuming an incompressible solid J = r′r 2/R2 = 1, one finds
r 3 (R, t ) = r 3 (R 0, t ) + R3 − R 03, R1 ≤ R ≤ R 0. (5.4)
The accreted shell at any instant of time t is assumed to be stress free and hence G (R ) = φτ*(R) g (r (R, τ (R ))). In components,
this reads GAB (s (t )) = GAB (R ) = F aA (R, τ (R )) F bB (R, τ (R )) gab (r (R, τ (R ))). Therefore, for R0 ≤ R ≤ R0 + u0 t we have
⎛ r′ 2 (R, τ (R)) 0 0 ⎞ ⎛1 0 0 ⎞
⎜ ⎟ ⎜ 2
⎟
G=⎜ 0 2
r (R, τ (R)) 0 ⎟ = ⎜ 0 r (R, τ (R)) 0 ⎟.
⎜ ⎟ ⎜ ⎟
⎝ 0 0 r 2 (R, τ (R)) sin2 Θ ⎠ ⎝ 0 0 r 2 (R, τ (R)) sin2 Θ ⎠ (5.5)
For the special case of ug (t ) = u0 , this is simplified to read

⎛1 0 0 ⎞
⎜ ⎟
⎛
⎜ 0 r 2 ⎜ R, R − R 0
⎞ ⎟
⎟ 0
G=⎜ ⎝ u0 ⎠ ⎟.
⎜ ⎟
⎜ ⎛ R − R0 ⎞ ⎟
⎜0 0 r 2 ⎜ R, ⎟ sin2 Θ ⎟
⎝ ⎝ u0 ⎠ ⎠ (5.6)
Assuming that the accreted body is made of an incompressible solid we have

⎛ R − R0 ⎞
r 2 (R, t ) r′(R, t ) = r 2 ⎜ R, ⎟.
⎝ u0 ⎠ (5.7)
The above equation can be integrated between R0 and R to give

R ⎛ x − R0 ⎞
r 3 (R, t ) = r 3 (R 0, t ) + ∫R 0
3r 2 ⎜ x ,
⎝
⎟ dx,
u0 ⎠
R 0 ≤ R ≤ R 0 + u0 t ,
(5.8)
from which, defining r0 (t ) = r (R0, t ) and r¯ (R ) = r (R, τ (R )), we obtain a decomposition analogous to (4.21):
R
r 3 (R, t ) = r03 (t ) + ∫R 0
3r¯2 (x) dx, R 0 ≤ R ≤ R 0 + u0 t .
(5.9)
5.2. Calculation of stresses for an incompressible isotropic accreting spherical ball
Next we calculate the stress field during the surface growth. For R1 ≤ R ≤ R0 , Finger tensor, b♯ , and b−1 are given by
⎛ 4 ⎞ ⎛ r 4 (R, t ) ⎞
⎜ R 0 0 ⎟ ⎜ 0 0 ⎟
⎜ r 4 (R, t ) ⎟ ⎜ R4 ⎟
⎜ ⎟ ⎜ R2 ⎟
b♯ = ⎜ 1 ⎟, b−1 = ⎜ 0 0 ⎟.
0 0 4
⎜ R2 ⎟ ⎜ r (R, t ) ⎟
⎜ 1 ⎟ ⎜ R2 ⎟
⎜ 0 0 ⎟ ⎜ 0 0 ⎟
⎝ R sin Θ ⎠
2 2
⎝ r 4 (R, t ) sin Θ ⎠
2
(5.10)
The principal stretches read
R4 2r 2 (R, t ) r 4 (R, t ) 2R2

I1 = + , I2 = + 2 .
r 4 (R, t ) R2 R4 r (R, t ) (5.11)
Therefore, the Cauchy stress has the following representation:

⎛ R4 2R2 ⎞
⎜− p + 4 α + 2 β 0 0 ⎟
⎜ r r ⎟
⎜ 1 ⎡ r 2 ⎛ R2 r 4⎞ ⎤ ⎟
σ=⎜ 0 ⎢ −p + 2 α + ⎜ 2 + 4 ⎟ β ⎥
⎝r R ⎠ ⎦
0 ⎟.
⎜ r ⎣
2 R ⎟
⎜ ⎟
⎜ 1 ⎡ r 2 ⎛ R2 r 4 ⎞ ⎤ ⎟
⎜ 0 0 ⎢ − p + α + ⎜ + ⎟ β ⎥ ⎟
⎝ r 2 sin2 Θ ⎣ R2 ⎝ r2 R 4 ⎠ ⎦⎠ (5.12)
The non-trivial radial equilibrium equation is written as

∂σ rr 2
+ σ rr − rσ θθ − (r sin2 θ ) σ ϕϕ = 0.
∂r r (5.13)
This is simplified to read
∂σ rr (R, t ) 2 ⎡ R6 ⎤ ⎡ r 2 (R, t ) ⎤
= ⎢1 − 6 ⎥ ⎢ α (R, t ) + β (R, t ) ⎥.
∂R r (R, t ) ⎣ r (R, t ) ⎦ ⎣ R2 ⎦ (5.14)
We know that σ rr (R1, t ) = − pi (t ), and hence, integrating the above ODE from R1 to R, one obtains
R 2 ⎡ x6 ⎤ ⎡ r 2 (x, t ) ⎤
σ rr (R, t ) = − pi (t ) + ∫R
1
⎢1 − 6
r (x, t ) ⎣
⎥ ⎢ α (x, t ) +
r (x, t ) ⎦ ⎣ x2
β (x, t ) ⎥ dx.
⎦ (5.15)
For R0 ≤ R ≤ R0 + u0 t , the Finger tensor, b♯ , and b−1 read

⎛ r¯ 4 (R) ⎞ ⎛ r 4 (R, t ) ⎞
⎜ 0 0 ⎟ ⎜ 4 0 0 ⎟
⎜ r (R, t )
4
⎟ ⎜ r¯ (R) ⎟
⎜ 1 ⎟ ⎜ r
¯ 2 (R) ⎟
b♯ = ⎜ 0 0 ⎟, b−1 = ⎜ 0 0 ⎟.
⎜ 2
r¯ (R) ⎟ ⎜ 4
r (R, t ) ⎟
⎜ 1 ⎟ ⎜ 2 (R) ⎟
⎜⎜ ⎟⎟ ⎜⎜ r
¯ ⎟⎟
0 0 2 2 0 0
⎝ r¯ (R) sin Θ ⎠ ⎝ r (R, t ) sin Θ ⎠
4 2
(5.16)
The principal invariants of b are written as
r¯ 4 (R) r 2 (R, t ) r 4 (R, t ) r¯2 (R)

I1 = +2 2 , I2 = +2 2 .
r 4 (R, t ) r¯ (R) r¯ 4 (R) r (R, t ) (5.17)
The Cauchy stress in this region has the following representation:
⎛ r¯ 4 r¯2 ⎞
⎜− p + 4α + 2 2β 0 0 ⎟
⎜ r r ⎟
⎜ 1 ⎡ r 2 ⎛ r4 ¯2 ⎞ ⎤
r ⎟
σ=⎜ 0 ⎢ − p + α + ⎜
⎝ r¯ 4
+ ⎟ β
r2 ⎠ ⎦
⎥ 0 ⎟.
⎜ r2 ⎣ r¯2 ⎟
⎜ ⎟
⎜ 1 ⎡ r2 ⎛ r4 r¯2 ⎞ ⎤ ⎟
⎜ 0 0 ⎢ − p + α + ⎜ 4 + ⎟ β ⎥ ⎟
⎝ r 2 sin2 Θ
⎣ r
¯ 2 ⎝ r
¯ r 2⎠
⎦ ⎠ (5.18)
Using the radial equilibrium equation, the radial Cauchy stress component is calculated as
R 2 ⎡ r¯6 (x) ⎤ ⎡ r 2 (x, t ) ⎤
σ rr (R, t ) = σ rr (R 0, t ) + ∫R 0
⎢1 − 6
r (x, t ) ⎣
⎥ ⎢ α (x, t ) + 2
r (x, t ) ⎦ ⎣ r¯ (x)
β (x, t ) ⎥ dx
⎦
R0 2 ⎡ x6 ⎤ ⎡ r 2 (x, t ) ⎤
= − pi (t ) + ∫R 1
⎢1 − 6
r (x, t ) ⎣
⎥ ⎢ α (x, t ) +
r (x, t ) ⎦ ⎣ x2
β (x, t ) ⎥ dx
⎦
R 2 ⎡ r¯6 (x) ⎤ ⎡ r 2 (x, t ) ⎤
+ ∫R 0 r (x, t )
⎢1 − 6
⎣ r (x, t )
⎥ ⎢ α (x, t ) + 2
⎦ ⎣ r (x)
¯
β (x, t ) ⎥ dx.
⎦ (5.19)
We know that σ rr (s (t ) , t) = σ θθ (s (t ) , t ) = 0. In particular, for the radial component, this gives

R0 2 ⎡ x6 ⎤ ⎡ r 2 (x, t ) ⎤
∫R 1
⎢1 − 6
r (x, t ) ⎣
⎥ ⎢ α (x, t ) +
r (x, t ) ⎦ ⎣ x2
β (x, t ) ⎥ dx
⎦
R 0 + u0 t 2 ⎡ 6
r¯ (x) ⎤ ⎡ r 2 (x, t ) ⎤
+ ∫ R0
⎢1 − 6
r (x, t ) ⎣
⎥ ⎢ α (x, t ) + 2
r (x, t ) ⎦ ⎣ r¯ (x)
β (x, t ) ⎥ dx = pi (t ).
⎦ (5.20)
Note that
r 2 (R, t ) ⎡ r¯6 (R) ⎤ ⎡ r 2 (R, t ) ⎤

σ¯ θθ (R, t ) = σ¯ ϕϕ (R, t ) = σ rr (R, t ) + ⎢1 − 6 ⎥ ⎢ α (R, t ) + 2 β (R, t ) ⎥.
r¯ (R) ⎣
2 r (R, t ) ⎦ ⎣ r¯ (R) ⎦ (5.21)
Therefore, σ¯ θθ (s (t ) , t ) = σ¯ ϕϕ (R, t ) = 0 are trivially satisfied. In the case of a homogeneous neo-Hookean solid, α (R, t ) = μ and
β (R, t ) = 0. In this case, (5.20) is simplified to read
R0 2 ⎡ x6 ⎤ R 0 + u0 t 2 ⎡ r¯6 (x) ⎤ p (t )
∫R ⎢1 − 6 ⎥ dx + ∫R ⎢1 − 6 ⎥ dx = i .
1 r (x, t ) ⎣ r (x, t ) ⎦ 0 r (x, t ) ⎣ r (x, t ) ⎦ μ (5.22)
One obtains the following problem in terms of the function r (R , t ) in the triangular region
{(R, t ) ∈ 2|R0 ≤ R ≤ R0 + u0 t , t ∈ + : }
⎧ ⎛ R − R0 ⎞
⎪ r 2 (R, t ) r′(R, t ) = r 2 ⎜ R, ⎟,
⎪ ⎝ u0 ⎠
⎪ ⎡ ⎤
⎪
⎨ R0 1 ⎢ x6 ⎥ R 0 + u0 t 1 − r′ 3 (x, t ) p (t )
⎪ ∫ 1 ⎢
1− 2⎥
dx + ∫R dx = i ,
⎪ 1 r (R 0, t ) + R3 − R 03 ( ) r (x, t ) 2μ
r 3 (R 0, t ) + R3 − R 03 3 ⎢⎣ ⎥⎦
R 3
⎪ ( ) 0
⎪
⎩ r (R 0, 0) = R 0, (5.23)
analogous to the one that was discussed in Section 4 for a cylindrical bar.
5.3. Calculation of residual stresses for an incompressible isotropic accreting spherical ball
Suppose that at some time t = ta , the accretion process stops. The material metric has the following representation in
spherical coordinates:
⎛1 0 0 ⎞
⎜ ⎟
R1 ≤ R ≤ R 0: G = ⎜ 0 R2 0 ⎟⎟,
⎜
⎝ 0 0 R sin Θ ⎠
2 2
⎛1 0 0 ⎞
⎜ ⎟
⎛
⎜ 0 r 2 ⎜ R, R − R 0
⎞ ⎟
⎟ 0
R 0 ≤ R ≤ R a: G=⎜ ⎝ u0 ⎠ ⎟.
⎜ ⎟
⎜ ⎛ R − R0 ⎞ ⎟
⎜0 r ⎜ R,
2 ⎟ sin Θ ⎟
2
⎝ ⎝ u 0 ⎠ ⎠ (5.24)
Note that the function r (R, t ), and hence r R, ( R −R0

u0 ), has already been calculated. The motion from the material manifold to
the residually stressed configuration (under no applied loads) is denoted by φ̃ : ) → : , where in spherical coordinates we
have φ˜(R, Θ ) = (r˜ , θ˜, ϕ˜ ) = (r˜ (R ) , Θ, Φ ). Using the incompressibility constraint one obtains
r˜ 3 (R) = r˜ 3 (R 0 ) + R3 − R 03, R1 ≤ R ≤ R 0,
R ⎛ x − R0 ⎞
r˜ 3 (R) = r˜ 3 (R 0 ) + ∫R 0
3r 2 ⎜ x ,
⎝
⎟ dx,
u0 ⎠
R 0 ≤ R ≤ R a,
(5.25)
where r˜ (R0 ) is an unknown that will be determined after enforcing the boundary and continuity conditions. In the absence
of applied loads the boundary conditions read σ˜ rr (R1) = σ˜ rr (Ra ) ¼0. One then has the following distribution of radial Cauchy
stress (the continuity of traction σ˜ rr (R0−) = σ˜ rr (R0+) has been enforced):
R 2 ⎡ x6 ⎤ ⎡ r˜2 (x) ⎤
R1 ≤ R ≤ R 0: σ˜ rr (R) = ∫R 1 r (x )
˜
⎢ 1 − 6 ⎥ ⎢ α (x) +
⎣ r (x)
˜ ⎦ ⎣ x 2
β (x) ⎥ dx,
⎦
R0 2 ⎡ x ⎤⎡
6 2
r˜ (x) ⎤ R 2 ⎡ r¯6 (x) ⎤ ⎡ r˜2 (x) ⎤
R 0 ≤ R ≤ R a: σ˜ rr (R) = ∫ R1
⎢ 1 − 6 ⎥ ⎢ α (x) +
r˜(x) ⎣ r˜ (x) ⎦ ⎣ x2
β (x) ⎥ dx +
⎦
∫R 0
⎢ 1 − 6 ⎥ ⎢ α (x) + 2 β (x) ⎥ dx.
r˜(x) ⎣ r˜ (x) ⎦ ⎣ r¯ (x) ⎦ (5.26)
Similarly
r˜2 (R) ⎡ R6 ⎤ ⎡ r˜2 (R) ˜ ⎤

R1 ≤ R ≤ R 0: σ˜¯ θθ (R) = σ˜¯ ϕϕ (R) = σ˜ rr (R) + ⎢1 − 6 ⎥ ⎢ α˜(R) + β (R) ⎥,
R2 ⎣ r˜ (R) ⎦ ⎣ R2 ⎦
2
r˜ (R) ⎡ 6
r¯ (R) ⎤ ⎡ r˜ (R) ˜ ⎤
2
R 0 ≤ R ≤ R a: σ˜¯ θθ (R) = σ˜¯ ϕϕ (R) = σ˜ rr (R) + 2 ⎢1 − 6 ⎥ ⎢ α˜(R) + 2 β (R) ⎥.
r¯ (R) ⎣ r (R) ⎦ ⎣
˜ r¯ (R) ⎦ (5.27)
The unknown constant r˜ (R0 ) is determined using the condition σ̃ rr (R a) = 0, or

R0 2 ⎡ x6 ⎤ ⎡ r˜2 (x) ⎤ Ra 2 ⎡ r¯6 (x) ⎤ ⎡ r˜2 (x) ⎤
∫R 1
⎢ 1 − 6 ⎥ ⎢ α (x) +
r˜(x) ⎣ r˜ (x) ⎦ ⎣ x2
β (x) ⎥ dx +
⎦
∫R 0
⎢ 1 − 6 ⎥ ⎢ α (x) + 2 β (x) ⎥ dx = 0.
r˜(x) ⎣ r˜ (x) ⎦ ⎣ r¯ (x) ⎦ (5.28)
5.4. Surface growth stresses in the linearized theory
Linearization of the governing equations and stresses is similar to that of an infinite cylindrical bar. For R1 ≤ R ≤ R0 , for
the perturbed motions we have rϵ3 (R, t ) = rϵ3 (R0, t ) + R3 − R23. Taking derivative with respect to ϵ on both sides and evaluating
at ϵ = 0, one obtains
R 02
δr (R, t ) = δr0 (t ),
R2 (5.29)
with δr0 (t ) = δr (R0, t ). For R0 ≤ R ≤ R0 + u0 t

R
rϵ3 (R, t ) = rϵ3 (R 0, t ) + ∫R 0
3r¯ϵ2 (x) dx.
(5.30)
Again taking derivatives with respect to ϵ on both sides and evaluating at ϵ = 0, one obtains
R 02 1 R
δr (R, t ) =
R2
δr0 (t ) + 2
R
∫R 0
2xδr¯ (x) dx.
(5.31)
Evaluating (5.31) at t = τ (R ), one obtains

1 R R 02
R
∫R 0
2xδr¯ (x) dx = δr¯ (R) −
R2
δr (R 0, τ (R)).
(5.32)
Substituting the above relation back into (5.31), one has
R 02 ⎡
δr (R, t ) − δr¯ (R) = ⎣ δr (R 0, t ) − δr (R 0, τ (R)) ⎤⎦.
R2 (5.33)
For R1 ≤ R ≤ R0 , the linearized radial Cauchy stress reads

R δr (x, t ) ⎛ 1 1⎞
δσ rr (R, t ) = − pi (t ) + 12μ ∫R 2
dx = − pi (t ) + 4μR 02 ⎜ 3 − 3 ⎟ δr0 (t ),
1 x ⎝ R1 R ⎠ (5.34)
where we used (5.29). For R0 ≤ R ≤ s (t ) we have

⎛ 1 1⎞ R δr (x, t ) − δr¯ (x)
δσ rr (R, t ) = − pi (t ) + 4μR 02 ⎜ 3 − 3 ⎟ δr0 (t ) + 12μ ∫R dx.
⎝ R1 R0 ⎠ 0 x2 (5.35)
Using (5.33) we obtain

⎛ 1 1⎞ R δδr0 (τ (x))
δσ rr (R, t ) = − pi (t ) + 4μR 02 ⎜ 3 − 3 ⎟ δr0 (t ) − 3 ∫R dx.
⎝ R1 R ⎠ 0 x4 (5.36)
We know that σϵrr (s (t ) , t ) = 0 and hence δσ rr (s (t ) , t ) = 0. Therefore

⎛ 1 1 ⎞ R 0 + u0 t δδr0 (τ (x)) p (t )
⎜ 3 − 3 ⎟ δr0 (t ) − 3 ∫R dx = i 2 , t > 0.
⎝ 1
R s (t ) ⎠ 0 x4 4μ R 0 (5.37)
R13 s3 (t ) pi̇ (t )
δr0̇ (t ) = .
4μR 0 s3 (t ) − R13 (5.38)
Integrating this ODE from 0 to t and noting that δṙ (R0, 0) = 0, one obtains
R13 t s3 (η) pi̇ (η)

δr0 (t ) =
4μR 02
∫0 s3 (η) − R13
dη .
(5.39)
Therefore, Eq. (5.33) is simplified to read
R13 t s3 (η) pi̇ (η)

δr (R, t ) − δr¯ (R) =
4μR2
∫τR s3 (η) − R13
dη ,
(5.40)
which, after solving (5.32) for δr¯ (R ), allows one to find δr (R, t ). Taking time derivative of both sides of (5.36) we obtain
⎛ ⎞
̇ (R, t ) = − ṗ (t ) + 4μR 2 ⎜ 1 − 1 ⎟ δṙ (R , t ).
δσ rr i 0 3 3 0
⎝ R1 R ⎠ (5.41)
Integrating the above relation from τ (R ) to t and noting that δσ rr (R, τ (R )) = 0, we have
⎛ R3 ⎞ t s3 (η) pi̇ (η)
δσ rr (R, t ) = − pi (t ) + pi (τ (R)) + ⎜ 1 − 13 ⎟ ∫τ (R) dη .
⎝ R ⎠ s3 (η) − R13 (5.42)
Similarly, the other two non-zero components of the Cauchy stress have the following linearized expressions:
⎛ R3 ⎞ t s3 (η) pi̇ (η)
δσ¯ θθ (R, t ) = δσ¯ ϕϕ (R, t ) = − pi (t ) + pi (τ (R)) + ⎜ 1 + 13 ⎟ ∫τ (R) dη .
⎝ 2R ⎠ s3 (η) − R13 (5.43)
Residual stresses: We follow the same approach that was used for the cylinder examples. The unloading fields are
R R13 s3 (ta ) pi (ta )

R2 s3 (ta ) − R13 4μ
RR R13 s3 (ta ) − R3
R3 s3 (ta ) − R13 i
1 3 3
ΘΘ R13 2 s (ta ) + R
σ¯unloading (R, ta ) = − p (ta ).
R3 s3 (ta ) − R13 i (5.44)
Integrating (Eqs. (5.40), 5.42) and (5.43) by parts, and adding (5.44) we obtain the following residual fields:
⎛ R − R0 ⎞
pi ⎜ ⎟
R13 R ⎝ u0 ⎠ 3u0 R16 ta s2 (η) pi (η)
R
Ures (R, ta ) = −
R − R13
3 4μ
+
4μR2
∫R − R 0
(s3 (η) − R13 )2
dη ,
u0 (5.45)
⎛ R3 ⎞ ta s2 (η) pi (η)
RR
σres (R, ta ) = 3u0 R13 ⎜⎜ 1 − 13 ⎟⎟ ∫R − R dη ,
⎝ R ⎠ u0
0
(s3 (η) − R13 )2 (5.46)
R13⎛ ⎞ ⎛ R3 ⎞ ta s2 (η) pi (η)

3 R − R0 ⎟
θθ
δσ¯res ϕϕ
(R, t ) = δσ¯res (R, t ) = − 3
pi ⎜⎜ ⎟ + 3u0 R13 ⎜⎜ 1 + 13 ⎟⎟ ∫
R − R0
dη .
2 R3 − R1 ⎝ u0 ⎠ ⎝ 2R ⎠ u0 (s3 (η) − R13 )2 (5.47)
6. Numerical examples
In this section we present some numerical examples of surface growth for cylindrical and spherical bodies made of neo-
Hookean solids.
6.1. Numerical solution of the governing equations
We introduce an auxiliary problem to (4.33) in which the original cylinder is the result of accretion “from nothing” with
no internal pressure. The auxiliary problems of inner accretion of a cylindrical bar and outer accretion of a spherical ball are
R −R
similar. The domain of the auxiliary problem is the region R1 ≤ R ≤ R0 + u0 t , and has its starting time at t * = − 0u 1 . Now
0
we show that this approach is equivalent to the problem (4.33).
For t * ≤ t ≤ 0 the governing equations (4.33) read
⎧ ⎛ R − R0 ⎞
⎪ r (R, t ) r′(R, t ) = r ⎜ R, ⎟ for R1 ≤ R ≤ R 0 + u0 t , t * ≤ t ≤ 0,
⎪ ⎝ u0 ⎠
⎪
⎨ R 0 + u0 t 1 ⎡ r¯ 4 (x) ⎤
⎪ ∫ r¯ x
⎢1 − 4
⎣ r
⎥ dx = 0 for t * ≤ t ≤ 0,
(x, t ) ⎦
⎪ 0 R ( )
⎪
⎩ r (R1, 0) = R1. (6.1)
It is straightforward to check that r (R, t ) = R is a solution of this problem. This means that at time t¼0 we obtain the stress-
free cylinder of internal radius R1 and external radius R0. For t ≥ 0 the problem becomes
⎧ ⎛ R − R0 ⎞
⎪ r (R, t ) r′(R, t ) = r ⎜ R, ⎟ for R1 ≤ R ≤ R 0 + u0 t ,
⎪ ⎝ u0 ⎠
⎪
⎨ R 0 + u0 t 1 ⎡ r¯ 4 (x) ⎤ p (t )
⎪ ∫ ⎢1 − 4 ⎥ dx = i for t ≥ 0,
⎪ R1 r¯ (x ) ⎣ r (x , t ) ⎦ μ
⎪
⎩ r (R1, 0) = R1, (6.2)
where the integral can be rewritten as
R0 1 ⎡ r¯ 4 (x) ⎤ R 0 + u0 t 1 ⎡ r¯ 4 (x) ⎤
∫R 1
⎢1 − 4
r¯ (x) ⎣
⎥ dx +
r (x, t ) ⎦
∫R 0
⎢1 − 4
r¯ (x) ⎣
⎥ dx.
r (x, t ) ⎦ (6.3)
Note that in the region R1 ≤ R ≤ R0 , since t * ≤

R −R0
u0
≤ 0, we simply have r¯ (R ) = r R, ( R −R0
u0 ) = R. Thus, we obtain the following
two problems:
⎧ r (R, t ) r′(R, t ) = R for R1 ≤ R ≤ R 0,

⎪
⎪ R0 1 ⎡ 4
r¯ (x) ⎤ p (t )
⎨ ∫ ⎢1 − 4 ⎥ dx = i for t ≥ 0,
⎪ R1 r¯ (x) ⎣ r (x, t ) ⎦ μ
⎪
⎩ r (R, 0) = R, (6.4)
and
Fig. 5. Outer surface growth under a monotonically increasing internal pressure pi(t). (a) The inner, interface, and outer radii and their linear counterparts
as functions of time. (b) The radial and circumferential residual stresses and their linear counterparts.
Fig. 6. Inner surface growth under a monotonically increasing internal pressure pi(t). (a) The outer, interface, and inner radii and their linear counterparts
as functions of time. (b) The radial and circumferential residual stresses and their linear counterparts.
⎧ ⎛ R − R0 ⎞
⎪ r (R, t ) r′(R, t ) = r ⎜ R, ⎟ for R 0 ≤ R ≤ R 0 + u0 t ,
⎪ ⎝ u0 ⎠
⎪
⎨ R0 1 ⎡ x4 ⎤ R 0 + u0 t 1 ⎡ r¯ 4 (x) ⎤ p (t )
⎪ ∫ ⎢1 − 4 ∫
⎥ dx + ⎢1 − 4 ⎥ dx = i for t ≥ 0,
⎪ 1 R x ⎣ r (x , t ) ⎦ R 0 r
¯ (x ) ⎣ r (x , t ) ⎦ μ
⎪
⎩ r (R 0, 0) = R 0, (6.5)
which is consistent with (4.18) for the initial body and is identical to (4.33) for the secondary body. Therefore, considering a
single domain R1 ≤ R ≤ R0 + u0 t starting at time t* with pi ¼0 for t * ≤ t ≤ 0 is equivalent to solving the governing equations in
Fig. 7. Surface growth of a spherical ball under a monotonically increasing internal pressure pi(t). (a) The inner, interface, and outer radii and their linear
counterparts as functions of time. (b) The radial and circumferential residual stresses and their linear counterparts.
two different domains starting at time 0. We solve the auxiliary problem in our numerical simulations, as it is numerically
more convenient. For the problem (6.5) we use finite differences, while the nonlinear equations, including (4.53) for the
residually stressed configuration, are solved using the Newton–Raphson method.
6.2. Numerical results
In the following examples we normalize stresses by the shear modulus μ and work with the dimensionless time t /ta and
the dimensionless radial coordinate R/R1. We also assume that R0/R1 = 1.25 and Ra/R1 = 2.25, which means that u0 = R1/ta = 1.
Example 1. Fig. 5(a) shows the time dependency of the radii of the inner and outer boundaries of a cylinder under outer
accretion and also that of the interface between the initial and the secondary bodies in the current configuration, and their
linear counterparts. The cylinder is under the monotonically increasing pressure shown in the figure. It is seen that for
t > 0.5ta the nonlinear effects are quite noticeable. Fig. 5(b) shows the profiles of the radial and circumferential residual
stresses and their linear counterparts. It is seen that the nonlinear circumferential stress is more than twice as much as the
linear one at some points. Note that the linear and nonlinear stresses do not attain their minima and maxima at the same
points. Also, note that as was explained in Remark 4.5, in the initial body, i.e. for 1 ≤ R/R1 ≤ 1.25, σ̃ rr is strictly increasing
while σ̃¯ θθ is strictly decreasing. Note also that the circumferential component of the residual Cauchy stress changes sign
along the thickness of the cylinder. Fig. 6 shows the same quantities for the inner accretion of the cylinder. We observe the
same change of sign of the circumferential component of the Cauchy stress. For the outer accretion of a spherical ball, Fig. 7
shows the same quantities that were discussed for the cylinder examples.
Example 2. In Fig. 8 we compare two outer accreted cylinders that are manufactured under cycles of internal pressure or
suction. It is seen that the nonlinear deformation and residual stress fields in the two cylinders are quite different. Note also
that under internal pressure the nonlinear stresses and radial displacements are larger (in absolute value) than their linear
counterparts, while under an internal suction the linear quantities are larger. However, as expected, the linear displacement
and residual stress fields are the same up to a sign in the two cases.
Example 3. Next, in Fig. 9, we consider two cylinders that have identical dimensions. One (outer accreted cylinder) is built
in an additive manufacturing process starting with an initial tube. The other one (non-accreted cylinder) has no residual
stresses. For the outer accreted cylinder we consider two cases: one additively manufactured under an internal pressure and
one manufactured under an internal suction. The maximum value of pressure or suction is 0.1μ as shown in Fig. 9. Each
cylinder is loaded under an internal pressure or suction of magnitude pe = 0.3μ. The radial stress srr does not seem to change
much with or without surface growth. The difference is evident especially for σ̄ θθ . In Fig. 10 we make the same comparison
for inner accretion, obtaining similar results. These observations indicate that stress distribution under service loads may be
reduced by controlling the residual stress distribution in an additively manufactured cylindrical bar.
Fig. 8. Outer surface growth under internal pressure (a, b) or internal suction (c, d) cycles. (a) and (c) show the inner, interface and outer radii and their
linear counterparts as functions of time. (b) and (d) show the radial and circumferential residual stresses and their linear counterparts.
Example 4. In Fig. 11 the stiffness of an inner accreted cylinder is compared with that of its corresponding non-accreted
cylinder. We consider two cases: accretion under an internal pressure and accretion under an internal suction. The radial
displacement of the inner boundary is calculated with respect to the position that corresponds to pe ¼0, which is Ra for the
non-accreted cylinder and r˜ (Ra ) for the accreted one. Displacements are normalized using R1. Note that when pi and pe have
opposite signs the accreted cylinder is stiffer than the non-accreted one. This observation suggests that the stiffness of an
additively manufactured structure can be controlled by controlling the surface growth process.
Example 5. In Fig. 12 the history of the inner radius of a cylinder during an outer surface growth under two different
loading cycles is shown (R0/R1 = 1.25 and Ra/R1 = 2.25). It is seen that in both the nonlinear and linear solutions the final
radius is different from its initial value. This is due to the anelastic nature of surface growth even for a hyper-elastic solid.
Fig. 9. Comparison between an accreted and a non-accreted cylinder both with inner and outer radii R1 and Ra, respectively, loaded by an internal pressure
or suction pe /μ = 0.3. The outer accreted cylinder was loaded during the accretion process by pi(t). (a) pi (t ) > 0 and pe > 0 , (b) pi (t ) > 0 and pe < 0 , (c)
pi (t ) < 0 and pe > 0 , (d) pi (t ) < 0 and pe < 0 .
7. Conclusions
In this paper we formulated the initial-boundary value problems of surface growth for cylindrical and spherical bodies
made of nonlinear elastic solids. To simplify the calculations we restricted ourselves to incompressible and isotropic ma-
terials. As model problems we considered infinitely long hollow cylindrical bars and hollow spherical balls. While a cy-
lindrical body is under a time-dependent internal pressure (or suction) material is continuously added to either its outer or
inner boundary. For a hollow spherical ball stress-free material is added to its outer boundary. Our goal was to calculate the
time-dependent stress field during accretion and the residual stress field after the completion of accretion. We solved these
problems in the framework of geometric anelasticity. We first constructed a family of Riemannian material manifolds. The
growing body is stress-free at any instant of time in its material manifold. We calculated the material metric and observed
that it depends on the history of loading and deformation during surface growth and the accretion velocity. Having a
material manifold the surface growth problem is transformed into a classical nonlinear elasticity problem with an evolving
Fig. 10. Comparison between an accreted and a non-accreted cylinder both with inner and outer radii Ra and R0, respectively, loaded by an internal
pressure or suction pe /μ = 0.15. The inner accreted cylinder was loaded during the accretion process by pi(t). (a) pi (t ) > 0 and pe > 0 , (b) pi (t ) > 0 and
pe < 0 , (c) pi (t ) < 0 and pe > 0 , (d) pi (t ) < 0 and pe < 0 .
reference configuration. We showed that calculation of accretion stresses for an incompressible solid is reduced to solving a
nonlinear partial integral equation. We then linearized the nonlinear governing equations about a stress-free equilibrium
configuration and showed that the resulting linear partial integral equation gives displacements and stresses that are
identical to those calculated using the linear theory of accretion of Brown and Goodman (1963). The governing equations
were solved numerically in the case of a neo-Hookean solid. We also calculated the residual stresses.
Extending the present theory to non-normal surface growth of arbitrary-shaped bodies will be the subject of a future
communication. Another extension of the present work would be modeling surface growth in thin shells. Recently, Sadik
et al. (2016) formulated a nonlinear shell theory with bulk growth. Kinematics of a shell is described by the first and the
second fundamental forms that evolve in the material manifold as a result of bulk growth. It would be interesting to see how
the fundamental forms of an accreting shell depend on the history of loading and deformation during surface growth. Other
Fig. 11. Loading of an inner-accreted cylinder (R0/R1 = 1.25 and Ra/R1 = 0.75) and its corresponding non-accreted cylinder. The relation between the
pressure pe /μ and the radial displacement of the inner boundary normalized with R1 for accreted and non-accreted cylinders.
Fig. 12. History-dependence (anelasticity) of the internal radius for (a) a compression–tension cycle and (b) a tension–compression cycle during outer
accretion of a cylindrical bar.
interesting problems for future work would be taking into account material anisotropies and compressibility, surface growth
under twist or uni-axial tension/compression, and the effect of pre-stress of the accreting material on the stiffness of an
additively manufactured structure.
The present nonlinear theory of surface growth can potentially be used for optimizing an additive manufacturing process
for achieving best stiffness and smallest stresses under service loads of the structure. Knowing the service loads one may
reduce the stresses by controlling the residual stresses. The factors that affect the residual stress distribution are loading
during surface growth, speed of adding material, and the state of stress of the added material at the time of attachment to the
structure. The nonlinear surface growth theory presented in this paper will be useful in solving such optimization problems.
Acknowledgments
This research was partially supported by AFOSR – Grant no. FA9550-12-1-0290, NSF – Grant no. CMMI 1130856, and ARO
W911NF-16-1-0064. F.S. was supported by Sapienza University of Rome (Scholarship for Specialization Abroad 2015). We
are grateful to M.F. Shojaei for his help with the numerical simulations.
References
Altmann, H.C., 1968. Formulas for computing stresses in center-wound rolls. J. Tech. Assoc. Pulp Pap. Ind. 51, 176–179.
Antman, S.S., Shvartsman, M.M., 1994. The shrink-fit problem for aeolotropic nonlinearly elastic bodies. J. Elast. 37 (2), 157–166.
Arutyunyan, N.K., Naumov, V., Radaev, Y.N., 1990. A mathematical model of a dynamically accreted deformable body. Part 1: Kinematics and measure of
deformation of the growing body. Izv. Akad. Nauk SSSR. Mekh. Tverd. Tela (6), 85–96.
Ben Amar, M., Goriely, A., 2005. Growth and instability in elastic tissues. J. Mech. Phys. Solids 53, 2284–2319.
Bilby, B.A., Bullough, R., Smith, E., 1955. Continuous distributions of dislocations: a new application of the methods of non-Riemannian geometry. Proc. R.
Soc. Lond. A 231 (1185), 263–273.
Brown, C., Goodman, L., 1963. Gravitational stresses in accreted bodies. Proc. R. Soc. Lond. A: Math. Phys. Eng. Sci. 276, 571–576.
Brown, C.B., Evans, R.J., LaChapelle, E.R., 1972. Slab avalanching and the state of stress in fallen snow. J. Geophys. Res. 77, 4570–4580.
Drozdov, A.D., 1998a. Continuous accretion of a composite cylinder. Acta Mech. 128 (1).
Drozdov, A.D., 1998b. Viscoelastic Structures: Mechanics of Growth and Aging. Academic Press, New York.
Eckart, C., 1948. The thermodynamics of irreversible processes. 4. the theory of elasticity and anelasticity. Phys. Rev. 73 (4), 373–382.
Epstein, M., Maugin, G.A., 2000. Thermomechanics of volumetric growth in uniform bodies. Int. J. Plast. 16, 951–978.
Garikipati, K., Arruda, E.M., Grosh, K., Narayanan, H., Calve, S., 2004. A continuum treatment of growth in biological tissue: the coupling of mass transport
and mechanics. J. Mech. Phys. Solids 52 (7), 1595–1625.
Golgoon, A., Sadik, S., Yavari, A., 2016. Circumferentially-symmetric finite eigenstrains in incompressible isotropic nonlinear elastic wedges. Int. J. Non-
Linear Mech. 84, 116–129.
Green, A.E., Rivlin, R.S., Shield, R.T., 1952. General theory of small elastic deformations superposed on finite elastic deformations. Proc. R. Soc. A 211 (1104),
128–154.
Hodge, N., Papadopoulos, P., 2010. A continuum theory of surface growth. Proc. R. Soc. Lond. A: Math. Phys. Eng. Sci. 466, 3135–3152.
Kadish, J., Barber, J., Washabaugh, P., 2005. Stresses in rotating spheres grown by accretion. Int. J. Solids Struct. 42 (20), 5322–5334.
Klarbring, A., Olsson, T., Stalhand, J., 2007. Theory of residual stresses with application to an arterial geometry. Arch. Mech. 59 (4–5), 341–364.
Kondo, K., 1955a. Geometry of elastic deformation and incompatibility. In: Kondo, K. (Ed.), Memoirs of the Unifying Study of the Basic Problems in
Engineering Science by Means of Geometry, Division C, vol. 1. . Gakujutsu Bunken Fukyo-Kai Tokyo, Japan, pp. 5–17.
Kondo, K., 1955b. Non-Riemannian geometry of imperfect crystals from a macroscopic viewpoint. In: Kondo, K. (Ed.), Memoirs of the Unifying Study of the
Basic Problems in Engineering Science by Means of Geometry, Division D-I, vol. 1. . Gakujutsu Bunken Fukyo-Kai, Tokyo, Japan, pp. 6–17.
Lychev, S., 2011. Universal deformations of growing solids. Mech. Solids 46 (6), 863–876.
Lychev, S., Manzhirov, A., 2013a. The mathematical theory of growing bodies. finite deformations. J. Appl. Math. Mech. 77 (4), 421–432.
Lychev, S., Manzhirov, A., 2013b. Reference configurations of growing bodies. Mech. Solids 48 (5), 553–560.
Manzhirov, A., 1995. The general non-inertial initial-boundary value problem for a viscoelastic ageing solid with piecewise-continuous accretion. J. Appl.
Math. Mech. 59 (5), 805–816.
Manzhirov, A.V., 2014. Mechanics of growing solids: new track in mechanical engineering. In ASME 2014 International Mechanical Engineering Congress
and Exposition. American Society of Mechanical Engineers, Montreal, Canada, pp. V009T12A039–V009T12A039.
Marsden, J., Hughes, T., 1983. Mathematical Foundations of Elasticity. Dover, New York.
Metlov, V., 1985. On the accretion of inhomogeneous viscoelastic bodies under finite deformations. J. Appl. Math. Mech. 49 (4), 490–498.
Naumov, V.E., 1994. Mechanics of growing deformable solids: a review. J. Eng. Mech. 120 (2), 207–220.
Ogden, R.W., 1984. Non-linear Elastic Deformations. Dover, New York.
Ong, J.J., O'Reilly, O.M., 2004. On the equations of motion for rigid bodies with surface growth. Int. J. Eng. Sci. 42 (19), 2159–2174.
Ozakin, A., Yavari, A., 2010. A geometric theory of thermal stresses. J. Math. Phys. 51, 032902.
Rodriguez, E.K., Hoger, A., McCulloch, A.D., 1994. Stress-dependent finite growth in soft elastic tissues. J. Biomech. 27 (455–467).
Sadik, S., Yavari, A., 2016a. On the origins of the idea of the multiplicative decomposition of the deformation gradient. Math. Mech. Solids, http://dx.doi.org/
10.1177/1081286515612280.
Sadik, S., Yavari, A., 2016b. Geometric nonlinear thermoelasticity and the time evolution of thermal stresses. Math. Mech. Solids, http://dx.doi.org/10.1177/
1081286515599458.
Sadik, S., Angoshtari, A., Goriely, A., Yavari, A., 2016. A geometric theory of nonlinear morphoelastic shells. J. Nonlinear Sci. 26 (4), 929–978, http://dx.doi.
org/10.1007/s00332-016-9294-9.
Schwerdtfeger, K., Sato, M., Tacke, K.-H., 1998. Stress formation in solidifying bodies. Solidification in a round continuous casting mold. Metall. Mater. Trans.
B 29 (5), 1057–1068.
Simo, J., Marsden, J., 1984. Stress tensors, Riemannian metrics and the alternative descriptions in elasticity. In: Trends and Applications of Pure Mathematics
to Mechanics. Springer, Berlin, pp. 369–383.
Skalak, R., Farrow, D., Hoger, A., 1997. Kinematics of surface growth. J. Math. Biol. 35 (8), 869–907.
Southwell, R., 1941. Introduction to the Theory of Elasticity for Engineers and Physicists. Oxford University Press, Oxford, England.
Takamizawa, K., 1991. Stress-free configuration of a thick-walled cylindrical model of the artery—an application of Riemann geometry to the biomechanics
of soft tissues. J. Appl. Mech. 58, 840–842.
Takamizawa, K., Matsuda, T., 1990. Kinematics for bodies undergoing residual stress and its applications to the left ventricle. J. Appl. Mech. 57, 321–329.
Yagoda, H.P., 1980. Resolution of a core problem in wound rolls. J. Appl. Mech. 47 (4), 847–854.
Yavari, A., 2010. A geometric theory of growth mechanics. J. Nonlinear Sci. 20 (6), 781–830.
Yavari, A., Goriely, A., 2012a. Riemann–Cartan geometry of nonlinear dislocation mechanics. Arch. Ration. Mech. Anal. 205 (1), 59–118.
Yavari, A., Goriely, A., 2012b. Weyl geometry and the nonlinear mechanics of distributed point defects. Proc. R. Soc. A 468, 3902–3922.
Yavari, A., Goriely, A., 2013a. Riemann–Cartan geometry of nonlinear disclination mechanics. Math. Mech. Solids 18 (1), 91–102.
Yavari, A., Goriely, A., 2013b. Nonlinear elastic inclusions in isotropic solids. Proc. R. Soc. A 469, 20130415.
Yavari, A., Goriely, A., 2014. The geometry of discombinations and its applications to semi-inverse problems in anelasticity. Proc. R. Soc. A 470, 20140403.
Yavari, A., Goriely, A., 2015. The twist-fit problem: finite torsional and shear eigenstrains in nonlinear elastic solids. Proc. R. Soc. A 471, 20150596.
Yavari, A., Ozakin, A., 2008. Covariance in linearized elasticity. Z. Angew. Math. Phys. 59 (6), 1081–1110.

1 s2.0 S0022509616302484 Main

Uploaded by

Copyright:

Available Formats

1 s2.0 S0022509616302484 Main

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S0022509616302484 Main

Uploaded by

Copyright:

Available Formats

Journal of the Mechanics and Physics of Solids 98 (2017) 12–48

Contents lists available at ScienceDirect

Journal of the Mechanics and Physics of Solids

Nonlinear mechanics of surface growth for cylindrical and

2. Geometric anelasticity and the mechanics of growth

where { } is a basis for T 4 and {dx } is a basis for T *

T (U, W) = TAB U AW B, (2.3)

p↦G p (Up, Wp)≕⟪Up, Wp⟫G ,

is a smooth function. Obviously, G is a symmetric ( )− tensor field on 4 .

( ψ *g)p (Up, Wp)≔g ψ (p) ( ( ψ* U)ψ (p) , ( ψ* W)ψ (p) ). (2.7)

(ψ *g)AB = F aAF bB gab . (2.8)

The adjoint of F is defined by

F T (X , t ): Tφt (X ) : → TX ), ⟪FV, v⟫g = ⟪V, F Tv⟫G , ∀ V ∈ TX ), v ∈ Tφt (X ) :, (2.10)

3. Attachment of a stress-free layer to a deformed infinitely long thick hollow cylinder

3.1. Construction of a material manifold

) −0 = { (R, Θ): R ≤ R ≤ R , 0 ≤ Θ < 2π},

Here, we know that

ψ − () −) = ψ0− () −0), ψ + () +) = ψ0+ () +0 ). (3.5)

They have the coordinate representations

(i) () −, G−) and () −0 , G−0) are isometric and hence

(ii) () +, G+) and (ψ0+ () +0 ) , g) are isometric and thus

detg r (R) r′(R)

from which, denoting r2 = r (R2 ), one obtains

r 2 (R) = R2 − R22 + r22. (3.11)

Using (2.13), the Finger tensor reads

The principal invariants of b are

Note that (b−1)ab = c ab = g amg bncmn and thus

This is simplified to read

We know that σ rr (R2 ) = 0, and therefore

which determines the unknown r2 in (3.11).

is admissible. Choosing Δ′ = Δ, the simplest choice for ψ is

Notice that G−ΘΘ (R2 )

3.2. Calculation of residual stresses for an incompressible isotropic solid

Using the radial equilibrium equation, one obtains

where σ˜ 2 = σ˜ rr (R2−). We know that σ˜ rr (R2 + Δ′) = 0 and hence

which is the equation that determines the unknown constant r̃2.

Noting that r^ (R^ ) = r (R ) one observes that σ˜ rr (R^ ) = σ˜ rr (R ), as expected.

4. Continuous surface growth of an infinitely long thick hollow cylinder

4.1. Outer accretion

In the material manifold ) we use the polar coordinates (R, Θ).

4.1.1. Kinematics and material metric

Ωt = { (s (t ), Θ): 0 ≤ Θ < 2π}, ωt = { (r (s (t ), t ), θ ): 0 ≤ θ < 2π}. (4.3)

r′(R, τ (R)) = τ′(R) ug (τ (R)). (4.6)

r′(s (t ), t ) = 1 or r′(R, τ (R)) = 1. (4.7)

Also, the choice Ug (t ) = ug (t ) implies that

In this case, (4.7) reads

G (s (t )) = φt* g (r (s (t ), t )), or equivalently G (R) = φτ*(R) g (r (R, τ (R))). (4.12)

In components, this reads (cf. (2.8))

)t = { (R, Θ): 0 ≤ Θ < 2π , R1 ≤ R ≤ s (t ) = R0 + u0 t}, (4.15)

The incompressibility condition reads

For R1 ≤ R ≤ R0 this implies that r (R, t ) r′(R, t ) = R , and therefore

r 2 (R, t ) = r02 (t ) + R2 − R 02, R1 ≤ R ≤ R 0, (4.18)

where r0 (t )≔r (R0, t ). For R0 ≤ R ≤ R0 + u0 t , incompressibility implies that

which can be integrated for example between R0 and R to give

)t = { (R, Θ): 0 ≤ Θ < 2π , R1 ≤ R ≤ s (t ) = R0 + U0 t}, (4.22)

4.1.2. Calculation of stresses for an incompressible isotropic accreting cylinder

For R0 ≤ R ≤ R0 + u0 t , the Finger tensor, b♯ , and b−1 read

⎛ r¯2 (R) ⎞ ⎛ r 2 (R, t ) ⎞

The principal invariants of b read

( ψ g)p (Up, Wp)≔g ψ (p) ( ( ψ U)ψ (p) , ( ψ* W)ψ (p) ). (2.7)