https://doi.org/10.1007/s42558-019-0006-0

ORIGINAL PAPER

Nonlinear contact mechanics for the indentation of hyperelasticcylindrical bodies

AmyM. Dagro1 · K. T. Ramesh2

Received: 21 January 2019 / Accepted: 10 March 2019© Springer Nature Switzerland AG 2019

AbstractThe mechanical properties of biological materials are commonly found through the application of Hertzian theory to force-displacement data obtained through micro-indentation techniques. Due to their soft nature, biological specimens are oftensubjected to large indentations, resulting in a nonlinear deformation behavior that can no longer be accurately describedby Hertzian contact. Useful models for studying the large deformation response of cylindrical specimens under indentationare not readily available, and the morphologies of biological materials are often closer to cylinders than spheres (e.g.,cellular processes, fibrin, collagen fibrils, etc.). In this study, a computational model is used to analyze the large deformationindentation of an incompressible hyperelastic cylinder in order to provide a generalized formulation that can be used toextract mechanical properties from indentation into soft cylindrical bodies. The effects of specimen size and indentationdepth are examined in order to quantify the deformation at which the proposed force-displacement relationship remainsaccurate.

Keywords Indentation · Hyperelastic · Nonlinear contact

1 Introduction

Measuring the mechanical response of biological specimens is an important key to developing new biomedical materials[1], understanding injury or disease progression [2–4], and potentially improving clinical diagnostic technologies [5, 6].However, obtaining useful mechanical properties of most biological specimens (e.g., individual cells) is a difficult task dueto the nonlinearity, anisotropy, and heterogeneity [7–9] that they exhibit.

Within the last century, a common approach to measuring the mechanical properties in cells is through indentationtesting—an experimental technique that gained popularity due to the relative convenience of probing adherent cells grownon flat, conventional substrates [10]. By indenting the cell surface with a simple geometry probe (e.g., sphere, cone, etc.),it is possible to use the relationship between the applied force and the concurrently measured indentation depth, to estimatean effective local modulus of the cell. The prevalence of indentation testing, such as atomic force microscopy (AFM), isalso attributable to the fact that previously used techniques, such as micropipette aspiration, might disturb the attachmentof the cell membrane to the underlying cytoskeleton [10]. AFM was formally introduced in 1986 [11], although similarindentation-like devices appeared before this, such as the “cell-poker” device of the Elson group in the 1970s [12]. By the1990s, Radmacher and others began using AFM on individual cells to create “elasticity maps,” or images that depict elasticproperties across the various regions of the cell with high spatial resolution [13–15].

� Amy M. Dagroamy.m.dagro.civ@mail.mil

1 U.S. Army Research Laboratory, Aberdeen Proving Ground, Aberdeen, MD 21005-5066, USA

2 Department of Mechanical Engineering, Johns Hopkins University, Baltimore, MD 21218-2682, USA

(2019) 1: 7

Published online: 1/ April 2019

Mechanics of Soft Materials

Since AFM now offers the convenience of being commercially available and compatible with traditional cell culturingtechniques, it has become a commonly employed tool in biology and biomechanics. However, the results from indentationtesting are only useful when considerations are taken with regards to subcellular heterogeneities [9, 10, 16]. Furthermore, itis unfortunately common to find reports that incorrectly apply certain contact mechanics formulations to cellular indentationtests. The well-known classical Hertz solution assumes linear elasticity, which should only be used to estimate the elasticmoduli when describing small indentations. However, cells can be subjected to large deformations due to their soft nature,often surpassing the linear elastic regime associated with classic Hertz theory [17–19]. Therefore, the usefulness of theelastic modulus is limited, and it is desirable to obtain hyperelastic parameters that can account for the finite deformations[20, 21] and strain stiffening behavior [22] of biological materials.

In classical Hertz theory [23], aside from assuming an isotropic linear elastic material, the solution is limited to smalldeformations since the shape of the spherical indenter is given by a parabolic function approximated as r2/2Rs , where r isthe radial coordinate and and Rs is the indenter radius [24, 25]. It is possible to model larger indentation depths (up to 0.8times the indenter radius) by using a higher order function to describe the spherical indenter [26]. However, such a solutionwill still be limited by providing only the linear elastic modulus from the experimentally found forces and displacements.

In recent decades, several approaches have been taken to modify the Hertz solution, in order to obtain parameters thatdescribe the material of the specimen as hyperelastic. The first approach involves the derivation of an approximate analyticalexpression, and validation of the expression with a finite element calculation [27, 28]. A second approach is to use afinite element model to investigate the effects of hyperelasticity on indentation and then create correction factors to use inconjunction with the equations of Hertzian theory [29, 30]. Finally, if the sole purpose of a study is to determine the materialproperties in a specific experiment, it is possible to disregard the classical Hertzian solution, and parametrically run a seriesof finite element simulations, varying the material properties for a given constitutive model, until the experimentally foundforce-displacement relationships are obtained [31, 32].

However, the aforementioned solutions are for the case of a rigid sphere indenting a half-space or another spherical body.There are no force-displacement relationships that exist for finding hyperelastic parameters in structures with a cylindricalmorphology. Quantifying these relationships is essential for finding properties of cylindrical cellular structures—neuronalaxons, cell processes, filopodia, and certain types of bacteria, to name a few—that undergo large deformations duringindentation testing. Here, we only consider a spherical indenter, as previous works have found that pyramidal indentersproduce stresses and indent depths that are too large for cells [33, 34].

The ratio of length scales between the indentation depth, indenter radius, and specimen radius is important for selectingthe appropriate indentation formulation. The indentation depth is given by δ, Rs is the radius of a spherical indenter tip,and Rc is the radius of the cell. For indenter tips with a radius much smaller than the radius of the cell (Rs << Rc), onecan approximate that the cell is a half-space, and it is no longer necessary to consider the cylindrical morphology. AlthoughAFM indenter tips with small radii (<100 nm) are readily available, a larger indenter tip might be desirable in applicationswhere it is necessary to probe mechanical properties over a larger area [35–37], to avoid nonlinear effects [38, 39], or createloads over a larger area when studying cellular injury [40] and mechanotransduction [41].

In this paper, we use a computational model to investigate the force-displacement relationships for spherical indentationinto hyperelastic cylindrical bodies (Fig. 1). We provide analytical expressions, with the calculated corrective functions, thatcan be used to obtain hyperelastic material properties from indentation testing on flat and cylindrical bodies (using bothflat and spherical indenters). The force-displacement relationships were found to be dependent on the ratio of indenter tospecimen radius (Rs/Rc), as well as the ratio of indentation to indenter size (δ/Rs). We observe the strains at which theindentation is large enough that the solution deviates from linear elastic theory. Finally, we evaluate the proposed hyperelasticcorrective functions for large indentations in order to quantify the extent of error due to interactions with the substrate.

2Methods

2.1 Elliptical contact onto linear elastic bodies

In the case of a rigid spherical indenter in contact with a flat half-space made of an isotropic linear elastic material, theforce-displacement relation is given by the classical Hertz solution:

FH = 4

3R1/2s

E

1 − ν2δ3/2 (1)

Mech Soft Mater (2019) 1: 77 Page 2 of 18

Fig. 1 a Example of acylindrical body (e.g., axon)subjected to indentation by aspherical indenter. b Experimentis simplified as a sphereindenting a cylindrical body c Afinite element model of sphericalindentation. Note that due tosymmetry, only one-fourth of themodel needed to be simulated.The indenter was modeled as arigid body and displaced alongthe negative x-direction by δ

where FH is the magnitude of the force applied to the indented body (e.g., a cell), Rs is the effective radius of the sphericalindenter, ν is the Poisson’s ratio of the cell, and E is the Young’s modulus of the cell. This relationship is for a sphericalcontact area and can only be used when probing the somal (spherical) compartment of a cell, or a large, flat cell with a radiusof curvature much larger than the indenter.

In the case of the silica bead indenting an axon or cell process, this can be idealized as a rigid sphere (with radius Rs)indenting a cylindrical body (with radius Rc), and thus an elliptical contact area is formed. A schematic of two generalizedellipsoidal shaped bodies and their respective geometries is shown in Fig. 2. To account for the eccentricity in the contactarea, correction factors are used to modify (1). We will also introduce the parameters Rz and Ry which are the effective radii

Fig. 2 Generalized case of twoellipsoidal volumes (“Body S”and “Body C”) and theirrespective geometries. Theequivalent radii of curvature aregiven in Eqs. 3 and 4. In the caseof a spherical indenter (Body S),Rs,y = Rs,z. For a cylindricalspecimen (Body C) with thelongitudinal axis aligned in they-direction, Rc,y → ∞

Mech Soft Mater (2019) 1: 7 Page 3 of 18 7

of curvature in the z-plane and y-plane, respectively. The effective curvature sum R, a quantity of importance in the analysisof contact mechanics, is defined by [42, 43]:

1

R= 1

Rz

+ 1

Ry

(2)

where

1

Rz

= 1

Rc,z

+ 1

Rs,z

(3)

1

Ry

= 1

Rc,y

+ 1

Rs,y

(4)

For a cylinder where Rc,y goes to infinity, the last expression simplifies to Ry = Rs,y . For a spherical indenter, Rs,z = Rs,y .Now, the relation for force-displacement during an elliptical contact for this linear elastic case can be written as:

FLEellip = 2π

3(2R)1/2

E

1 − ν2δ3/2χellip (5)

where χellip is a product which is solely dependent on the ratio between the indenter radius Rs and the cylinder radius Rc

and can be written as:

χellip = kF (−3/2)E1/2 (6)

where k is the eccentricity parameter of the contact area (given as the ratio of semimajor to semiminor axis), F is thecomplete elliptic integral of the first kind, and E is the complete elliptic integral of the second kind. The eccentricityparameter is related to the elliptic integrals of the first and second kind [24, 43]:

k =√2F − E(1 + γ )

E(1 − γ )(7)

where

γ = R(1

Rz

− 1

Ry

) (8)

F =∫ π/2

0[1 − (1 − 1

k2) sin2 φ]−1/2dφ (9)

E =∫ π/2

0[1 − (1 − 1

k2) sin2 φ]1/2dφ (10)

One can numerically evaluate the elliptical equations using iterative procedures or utilizing lookup tables [44]. Here, inorder to provide the most useful result to the reader, we will utilize the algebraic approximations from Brewe and Hamrock[42] to replace the elliptical integrals of F and E with numerically found approximations of F and E . These are given by:

F = 1.5277 + .6023 lnRy

Rz

(11)

E = 1.0003 + 0.5968Rz

Ry

(12)

The approximate eccentricity parameter is now approximated:

k ≈√2F − E(1 + γ )

E(1 − γ )(13)

It follows that the approximate expression for χellip is now given by:

χellip ≈ kF (−3/2)E1/2(14)

Mech Soft Mater (2019) 1: 77 Page 4 of 18

Thus, we can rewrite the relation for force-displacement during an elliptical contact as a product of the classic Hertzianforce relationship (FH ) and a correction factor that depends on the ratio of Rs to Rc, which we will call �LE

ellip:

FLEellip = 2π

3(2R)1/2

E

1 − ν2δ3/2kF−3/2E1/2

(15)

= FH �LEellip (16)

where

�LEellip = πχellip√

2

√RsRc

2Rc+Rs√RS

. (17)

The superscript LE denotes that these expressions are for linear elastic bodies. Note that the force-displacement F − δ

relationship retains the 3/2 power, arising out of linear elasticity.A comparison between the force-displacement relationships of Eqs. 1 and 16 is shown in Fig. 3a. The ratio between the

indenter tip size versus specimen size will effect the ellipticity of the contact area, determining the usefulness of Eq. 16.Figure 3a shows that for sufficiently small indenter tips (Rs/Rc ≤ 0.4), the elliptical contact solution differs from thecircular contact solution (1) by < 6% for extremely large indentations (δ/Rc = 0.9). In other words, if the radius of thecylindrical specimen is at least 2.5 times larger than the indenter tip, the elliptical contact solution approaches the classic

Fig. 3 aNondimensional plot of the force-displacement relationship for different ratios ofRs/Rc. Solid lines show F −δ curves using the ellipticalcontact formulation of Eq. 16, whereas dashed lines show the F − δ relationships with a circular contact area (classical Hertz theory). b Plot ofthe elliptical contact correction factor �LE

ellip (orange) and eccentricity (k, blue) as a function of varying Rs/Rc (and hence, Ry/Rz)

Mech Soft Mater (2019) 1: 7 Page 5 of 18 7

Hertzian solution (analogous to a sphere indenting a half-space). For large values of the size ratio Rs/Rc, the eccentricitywill become large enough that the behavior will approach a line contact formulation (k ≈ 10).

Figure 3b plots the eccentricity parameter k and the elliptical contact correction factor �LEellip as a function of Rs/Rc. Note

that if Ry/Rz=1, then F = E = π/2, k = 1, and there is no need to use �LEellip ( �LE

ellip = 1) since it is a circular contactarea. As Rs/Rc becomes large, k also becomes large, and will eventually approach the limit of a rectangular line contact.For all values of Rs/Rc, �LE

ellip must be ≤ 1 (and hence FLEellip<FH ).

2.2 Elliptical contact on hyperelastic neo-Hookean bodies

The preceding section considered a linearly elastic material subjected to small deformation. In the case of biologicalmaterials subjected to large deformation, a hyperelastic material model is more appropriate. We begin by considering aneo-Hookean material (perhaps the simplest constitutive model within the class of hyperelastic models). The neo-Hookeanmodel for an incompressible material uses a strain energy density of the form:

WNH = C10(I1 − 3); (18)

where C10 is a material property and I1 is the first invariant of C (the right Cauchy-Green deformation tensor). In the smallstrain regime, the material constant C10 can be related to the linear elastic shear modulus μ and linear elastic Young’smodulus through C10 = μ/2 = E

4(1+ν), where ν is the Poisson’s ratio. For an incompressible material, ν = 0.5.

For the case of a neo-Hoookean body subjected to spherical indentation, we propose a solution which takes the form:

FNH = FH �LEellip�NH (19)

= 2π

3(2R)1/2

E

1 − ν2δ3/2χellip�NH (20)

where �NH is a new correction function that must be solved for. Although �LEellip is solely dependent on the ratio between

the indenter radius and specimen radius (Rs/Rc), the correction factor for the neo-Hookean material will have an additionaldependence on the extent of deformation since hyperelastic materials exhibit stiffening at larger strains. Here, the extentof deformation is characterized by the ratio between indentation depth and the indenter radius (δ/Rs). Thus, �NH will bedependent on (δ/Rs) and (Rs/Rc). For simplicity in calculating the material property C10 (explained in later sections), thefollowing sections will obtain expressions for the corrective function NH (the inverse of �NH ):

NH (δ/Rs, Rs/Rc) = 1

�NH

. (21)

2.3 Elliptical contact on hyperelastic Mooney-Rivlin bodies

A Mooney-Rivlin material model is another hyperelastic constitutive model that is considered an extension of the neo-Hookean model since it incorporates a second invariant of the left Cauchy-Green tensor. For an incompressible material, itsstrain energy takes the form:

WMR = C10(I1 − 3) + C01(I2 − 3) (22)

where C10 and C01 are material constants. Generally speaking, expressions which include the dependence on the secondinvariant can model the stress response of rubber-like materials and soft biomaterials more accurately [45]. For smalldeformations, the constants C10 and C01 are related to the elastic modulus by E = 4(1+ν)(C10+C01). For C01 = 0, Eq. 22reduces to the neo-Hookean strain energy. Similar to the correction factor for the neo-Hookean body �NH , an expressionfor a Mooney-Rivlin body subjected to spherical indentation will have a correction factor �MR:

FMR = FH �LEellip�MR (23)

Again, we will be finding solutions to the inverse corrective function: MR = 1�MR

. In this case, the corrective functionwill have an additional dependence on the ratio between the parameters C10 and C01. Therefore, MR will be a function of(δ/Rs), (Rs/Rc), and κ , where κ is a dimensionless parameter defined by:

κ = C10

C10 + C01(24)

Mech Soft Mater (2019) 1: 77 Page 6 of 18

Note that κ = 1 corresponds to a neo-Hookean material.

2.4 Computational model

2.4.1 Geometry considerations

3D finite element simulations of spherical indentation on cylindrical specimens were performed with the commerciallyavailable software Abaqus (Dassault Systemes Simulia Corp.). The indentation was simulated as frictionless contact betweena rigid sphere (with Rs varying between 2 and 100 μm) and a cylindrical specimen (Rc = 5μm) with either linear elasticor hyperelastic material properties. Figure 1c shows the computational model with the applied boundary conditions. Takingadvantage of symmetry, a quartered section of the model, with approximately 200–400k ten-node quadratic tetrahedralelements, was utilized to reduce computational costs. The magnitude of indentation strain (δ/Rc) remained constant acrossall simulations, while a different Rs/Rc value was used for each simulation. A total of six indenter sizes were simulated,with Rs equal to 2, 3, 5, 7, 10, 20, 30, 50, 75, and 100 μm. These correspond to Rs/Rc ratios of 0.4, 0.6, 1, 1.5, 2, 4, 6, 10,15, and 20, respectively. Consequently, the maximum value of δ/Rs will be different across each simulation.

The range of Rs/Rc values was determined by considering the appropriate amount of ellipticity. It was shown in theprevious section that forRs/Rc < 0.4, the ellipticity parameter k → 1, and the indentation can be approximated by a circularcontact formulation, such as those provided by Lin et al. [27]. At the upper limit, a line contact exists for large values of k, orwhen Rs/Rc ≈ 20 (see Fig. 4). This range is also applicable to experimental data since commonly used silica bead indentertips range from 5 to 100 microns while probing cellular components with cylindrical radii on the order of 1–10 microns.

For simulations with larger ellipticity in the contact area (larger indenters), the cylindrical specimens were made to begeometrically longer along the longitudinal direction (y-axis) so that the free boundary normal to the y-axis would not affectthe solution. Across all models, the length of the specimen was made to be approximately twice as long as the major axis ofthe contact area during the maximum indentation depth.

2.4.2 Material parameters

For each of the seven indenter sizes, three different types of material models were implemented for the cylindrical specimen:linear elastic, neo-Hookean, and Mooney-Rivlin. For the linear elastic material model, we used a Young’s modulus ofE = 100 Pa and assumed that the specimen was incompressible (ν = 0.5). Note that the correction factors for �ellip,NH , and MR are dimensionless functions that do not depend on the specific material parameters. For the Mooney-Rivlinmodels, additional simulations with varying κ values (κ=0.1, 0.2, 0.3, ...1.0) were performed for each ratio of Rs/Rc.

Fig. 4 Example of finiteelement simulations whichdemonstrate the effect of Rs/Rc

ratios on the ellipticity of thecontact area. For Rs/Rc = 0.4(top simulation), the contactarea is almost circular, while forRs/Rc = 10, the contact areahas greater eccentricity (k)

Mech Soft Mater (2019) 1: 7 Page 7 of 18 7

2.4.3 Mesh convergence

The mesh size of the cylindrical specimen was biased to be more refined near the area of contact with the indenter. Amesh convergence study was performed for the linear elastic model of the smallest Rs/Rc ratio, since this was the casethat corresponded to the largest indentation strains. To find the necessary minimum element size, several simulations wereperformed with the smallest element size at the center of contact varying between 2.5 × 10−5 mm and 5 × 10−4 mm.Although all simulations (including the coarsest mesh) were able to accurately resolve the theoretical force-displacementcurves at small strains (FLE

ellip of Eq. 16), it was found that a minimum element size of 1 × 10−4 mm was necessary for theconvergence of the total internal energy.

2.4.4 Boundary conditions

The rigid sphere was displaced by δ = 700 μm in the negative x-direction in 100 ms. This loading rate is comparable tosome AFM experiments [27]; however, it should be noted that viscoelastic effects are not considered in these computations.The interface between the indenter and specimen was assumed to be frictionless. The model was given symmetric boundaryconditions along surfaces normal to the positive y- and z-directions and the negative x-direction, while it remained free toexpand in the negative y-direction. A schematic of the fixed translational and rotational degrees of freedom which create thesymmetric boundary conditions is shown in Fig. 1c. In summary, the negative x-normal plane of symmetry was not allowedto move in the x-direction, the positive y-normal plane was not allowed to move in the y-direction, and the z-normal planeof symmetry was not allowed to move in the z-direction.

2.5 Calculation of corrective functions

Since analytical expressions are known for FH and �ellip, our aim is to utilize the results from the following sectionsto obtain expressions for the correction functions NH and MR . After obtaining these correction functions, simplisticalgebraic expressions are provided and can be used to extract hyperelastic material properties from experiments of cylindricalspecimens undergoing spherical indentation.

In order to obtain algebraic expressions for the corrective functions, we obtained the simulated corrective factorsNH (δ/Rs) and MR(δ/Rs) from variousRs/Rc values over a useful range. Hereafter we will use the symbol ( − ) to denotethe values obtained with the finite element model. Each simulation provides a total reaction force (FNH or FMR ), which canbe divided by the analytical expression of FLE

ellip (16) in order to obtain the corrective factors NH (δ/Rs) and MR(δ/Rs),

respectively. From NH (δ/Rs) and MR(δ/Rs) it is possible to deduce the final expressions for NH (δ/Rs, Rs/Rc) andMR(δ/Rs, Rs/Rc, κ).

3 Results

3.1 Linear elastic specimen simulations

Although the aim of this study is to obtain corrective functions to describe the specimen with a nonlinear constitutive model,an analogous “corrective factor” (LE) can be computed which compares the FEM simulation of a linear elastic (LE)specimen with the theoretical curve for FLE

ellip. The analytical solution of FLEellip can be divided by the total reaction force

obtained during a given simulation (FLE) in order to obtain LE such that:

LE = FLEellip

FLE

= FH �LEellip

FLE

(25)

Due to the discretization of surface elements in the finite element model, the simulation will be inherently inaccurate forvery small indentations. In order to find the minimum indentation depths that allow for an accurate FEM solution, we beganby performing simulations with a linear elastic material model for the cylindrical specimen. At the indentation depth wherethe linear elastic simulation matches the elliptical contact theory, the simulation matches the theoretical curve for FLE

ellip, andwe therefore only consider fitting the corrective functions beyond this indentation depth.

Mech Soft Mater (2019) 1: 77 Page 8 of 18

A comparison between the theoretical force-displacement relationship of Eq. 16 for various values of Rs/Rc andsimulations with linear elastic specimens is shown in Fig. 5. In Fig. 5a, the nondimensional forces (for both the simulationand theoretical FLE

ellip) are plotted against δ/Rs . For small values of Rs/Rc (approaching the behavior of spherical contact),the linear elastic theory for elliptical contact (16) predicts the behavior quite accurately, even for fairly large indentation(δ/Rc > 0.1). For larger values of Rs/Rc and larger indentations (δ/Rc > 0.1), the simulations deviate farther from Eq. 16,as shown more clearly in Fig. 5b. This is due to a decrease in accuracy of the algebraic expressions in Eqs. 7, 9, and 10 forlarger values of eccentricity k [42].

For a given Rs/Rc, when LE = 1, the indentation is large enough for the simulation to be considered accurate enoughfor obtaining corrective functions. Figure 5b plots LE over the relative indentation δ/Rs . The divergence of LE from 1 is aresult of two causes: (1) larger indentations creating a deviation from small strain theory and (2) larger ratios of Rs/Rc (andconsequently, larger eccentricity values) impart less accuracy in ellip. The values of δ/Rs at which LE = 1 are smallestfor the largest values of Rs/Rc, since a larger indenter will impart a larger area of contact onto the specimen (therefore, thelimiting mesh size within the contact area becomes less significant).

3.2 Neo-Hookean specimen simulations

Figure 6a compares (16) to the force-displacement response obtained from simulations with neo-Hookean (NH) specimensacross various ratios of Rs/RC . The neo-Hookean contact also shows increased deviation from the theoretical elliptical

Fig. 5 aNondimensional forces (both simulations and theoretical) plotted over nondimensional indentation depth for various values ofRs/Rc witha linear elastic (LE) specimen. Each curve corresponds to a value of Rs/Rc. Finite element simulations (dashed lines) are compared to the linearelastic theory of elliptical contact (solid lines) given in Eq. 16. b The calculated correction factors (LE ) from the simulations of a linear elasticspecimen. LE measures the deviation from the linear elastic theory of elliptical contact, and is plotted over the nondimensional indentation δ/Rs

Mech Soft Mater (2019) 1: 7 Page 9 of 18 7

Fig. 6 a Nondimensional forces (both the simulations of the neo-Hookean specimen and the theoretical curve for FLEellip) plotted over

nondimensional indentation depth for various values of Rs/Rc. Finite element simulations (dashed lines) are compared to the linear elastic theoryof elliptical contact (solid lines) given in Eq. 16. b The calculated correction factors (NH ) from the simulations of a neo-Hookean specimen. Eachcurve corresponds to a value of Rs/Rc. The correction factors calculated from the simulations (circles) are plotted alongside the correspondingfitted power-law functions (solid lines). Only selected values of Rs/Rc are shown for clarity. A complete list of the power-law function coefficientsis provided in Table 1

contact theory (FLEellip) at higher values of Rs/Rc and at higher relative indentations (δ/Rs). For comparison, the elliptical

contact theory curves used a Young’s modulus of E = 4C10(1+ν). These results are in agreement with previous indentationstudies on hyperelastic materials which have shown that the indentation response is stiffer than that predicted by linearelasticity [28]. As expected, FNH is similar to FLE at small indentation, but begins to differ from FLE at larger indentations(Fig. 6b). The simulated corrective factors NH were calculated by:

NH = FLEellip

FNH

(26)

The corrective factors for the neo-Hookean specimen (shown in Fig. 6b) are plotted as a function of indentation andfollow a similar trend to the LE factors. That is, it is apparent that the neo-Hookean correction factors are a function ofδ/Rs . A power-law function is used to fit the correction factor curves such that:

NH (δ/Rs) = A(Rs/Rc)

(δ

Rs

)m(Rs/Rc)

(27)

For a given Rs/Rc, the power-law functions were fitted to NH for values of δ/Rs that were greater than the indentationat which LE<1. For clarity, only selected values of Rs/Rc are shown; however, a complete list of the fitted power-lawfunction coefficients is provided in Table 1.

Mech Soft Mater (2019) 1: 77 Page 10 of 18

Table 1 Summary ofpower-law coefficients, A andm for the corrective functionsof Eq. 28 (neo-Hookean model)for various values of Rs/Rc

Rs/Rc A m

0.4 0.91 −0.04

0.6 0.85 −0.06

1 0.74 −0.09

1.5 0.68 −0.10

2 0.64 −0.10

4 0.56 −0.10

6 0.47 −0.12

10 0.38 −0.14

15 0.32 −0.16

20 0.27 −0.17

The power-law functions take the form of NH = A( δRs

)m

Figure 7 shows that A(Rs/Rc) and m(Rs/Rc) logarithmically decay with increasing the size ratios Rs/Rc. For smallerratios of Rs/Rc, the solution becomes analogous to circular contact with a half-space, and therefore A → 1, m → 0, andNH (δ/Rs) → 1 for all values of Rs/Rc. The corrective function for a neo-Hookean material can now be written in its finalform as:

NH =[A1 ln(

Rs

Rc

) + A2

] (δ

Rs

)m1 ln(RsRc

)+m2

(28)

A summary of the coefficients for the neo-Hookean corrective function is given in Table 2. Note that the function m(Rs/Rc)

remains small and negative over the range of Rs/Rc.

3.3 Mooney-Rivlin specimen simulations

In order to obtain a correction function for the Mooney-Rivlin model (MR), simulations with varying parameter ratios(κ=0,0.1, 0.2, 0.3, ...1.0) were performed for each size ratio of Rs/Rc. Since ten different size ratios were considered, thisresulted in 110 total simulations analyzed for a Mooney-Rivlin specimen. Similar to the previous approach with the linearelastic and neo-Hookean materials, the corrective factors MR are calculated by the relationship:

MR = FLEellip

FMR

(29)

Fig. 7 Logarithmic functions found for curve fitting of NH (solid lines). The logarithmic functions fitted to the simulation results (circles)showed an excellent fit, with R2 values, or coefficients of determination, greater than 0.90

Mech Soft Mater (2019) 1: 7 Page 11 of 18 7

For each value of κ , the calculated value of MR plotted over δ/Rs still follows a power-law behavior; however, incontrast to the neo-Hookean simulations, the functions A and m now have an additional dependence on the parameter ratioκ such that:

MR(δ/Rs, κ) = A(Rs/Rc, κ)

(δ

Rs

)m(Rs/Rc,κ)

(30)

We begin by finding an expression for the function A for the ten simulations (Rs/Rc =0.4, 0.6, 1, 1.5, 2, 4, 6, 10, 15, 20)for each specified value of κ . The simulations show that the function A can be rewritten as:

A(Rs/Rc, κ) = A1(κ) ln

(Rs

Rc

)+ A2(κ) (31)

where A1(κ) and A2(κ) are functions determined after comparing the functional A(Rs/Rc, κ) across all values of κ .Similarly, the equation for the functional expression m, at a given value of κ , is described by:

m(Rs/Rc, κ) = m1(κ) ln

(Rs

Rc

)+ m2(κ) (32)

Likewise, m1(κ) and m2(κ) are functions determined following a comparison of the functional m(Rs/Rc, κ) across allvalues of κ .

As shown in Fig. 8, the parameters A1 and m1 can be approximated as constant values across κ . The functions A2 andm2, however, show some dependence on κ , and can be described with the linear functions:

A2 = .12κ + .64 (33)

m2 = .04κ − .12 (34)

The generalized form for the Mooney-Rivlin corrective function is now given by:

MR =[A1 ln(

Rs

Rc

) + A2(κ)

] (δ

Rs

)m1 ln(RsRc

)+m2(κ)

(35)

A summary of the calculated parameters for the Mooney-Rivlin corrective function is given in Table 2. In theory, thecalculated model parameters for κ = 1 (the parameter C01 = 0) should be identical to the parameters found with the neo-Hookean form. One can see that while the coefficients for A2 and m2 show excellent agreement between the neo-Hookeanand Mooney-Rivlin forms (when κ = 1), the coefficients A1 and m1 are slightly different. This is most likely due to thefact that the linear functions that describe A1 and m1 have a poorer fit to the data (R2 = 0.50 and 0.81, respectively) incomparison to the functions describing A2 and m2 (R2 > 0.99). Note that Eq. 35 can be further simplified by approximatingm1 = 0, although m1 remains included in the assessment for accuracy.

Fig. 8 Plot of MR coefficients as functions of κ found in simulations (circles) and the corresponding fitted linear functions (solid lines). The R2

values, or coefficients of determination, are 0.50, 0.99, 0.81, and 0.99 for A1, A2, m1, and m2, respectively

Mech Soft Mater (2019) 1: 77 Page 12 of 18

Table 2 Summary ofcoefficients for the generalizedcorrective functions of Eq. 28(Neo-Hookean Model) andEq. 35 (Mooney-Rivlin model)

NH coefficient values MR coefficient values

A1 −0.16 −0.14

A2 0.76 0.12κ + 0.64

m1 −0.03 −0.03

m2 −0.07 0.04κ − 0.12

From the corrective function, we see that MR ∝ δ−0.03 ln Rs

Rc+(.04κ−.12). It follows that

FMR ∝ δ3/2δ0.03 ln Rs

Rc−0.04κ+0.12. (36)

Therefore, our nonlinear solution deviates from linear elastic theory (FLEellip ∝ δ3/2) by increasing the indentation power

term. For the range of Rs/Rc examined in this study, the additional indentation power term varies from 0.05 (Rs/Rc = 0.4,κ = 1) to 0.18 (Rs/Rc = 20, κ = 0). Consequently, the force (FMR) is proportional to the indentation within the range ofδ1.55 to δ1.68.

3.4 Comparison of corrective functions

Smaller values of MR (i.e., MR << 1) indicate stronger deviation from the linear elastic elliptical contact theory. Figure 9shows contour plots of the corrective functions for MR against the nondimensional variables of size ratio (Rs/Rc) andindentation ratio (δ/Rs) for the two extreme cases: κ = 0 (Fig. 9a) and κ = 1 (Fig. 9b). For values of small Rs/Rc and δ/Rs ,the corrective function approaches a value of 1 (MR ≈ 1), and linear elasticity theory may be used. For a given set of δ/Rs

and Rs/Rc, a comparison between Fig. 9a and b shows that when κ = 0, MR is smaller (MR deviates farther from 1).Therefore, the material demonstrates an increase in nonlinear behavior. As expected, the corrective function deviates farthestfrom linear elasticity (MR = .025) when κ = 0, Rs/Rc is large, and δ/Rs is large.

One can see the most variation in MR across Rs/Rc. For a given value of Rs/Rc, the corrective function changes slowlywith respect to the extent of indentation δ/Rs . This is explained by the fact that m1 and m2 are small across all values of κ ,and therefore Eq. 35 will demonstrate greater dependence on Rs/Rc.

4 Discussion

4.1 Effects of large deformations and accuracy of corrective functions

In this section, we discuss the accuracy of our corrective functions. For incompressible thin specimens under largedeformation (large values of indentation strain, δ/Rc), the specimen experiences restricted deformation in the direction ofapplied loading, and will therefore experience more expansion along the sides and length of the cylinder. In the previoussection, all simulations for computing MR utilized an indentation strain of 0.14; however, in this section, we extend theanalysis to larger indentation strains (δ/Rc > 0.30) to assess the accuracy of our solution.

Figure 10a shows the force-displacement curves calculated with the proposed corrective functions (35) in comparison toadditional finite element simulations performed at a larger indentation depth. A plot of the corresponding sum-of-squareserror (SSE) is shown in Fig. 10b (SSE values are given by SSE = �(y − y)2, where y are the forces obtained through thenew simulations, and y are the force values calculated using the corrective function of Eq. 35). One can see that for smallervalues of Rs/Rc, the corrective function provides an excellent fit (SSE < 0.25) up until δ/Rc = 0.45. For larger ratios ofRs/Rc, the corrective function only seems suitable for indentation strains less than 0.25, after which the SSE values exceed0.25.

As expected, the theoretical curves from linear elastic theory (“+” curves) show a much larger error. For Rs/Rc = 1, thelinear elastic theory provides a reasonable estimate for δ/Rc < 0.28. For Rs/Rc = 20, the linear elastic theory of ellipticalcontact shows significant errors (SSE > 0.25) for δ/Rc < 0.08.

Mech Soft Mater (2019) 1: 7 Page 13 of 18 7

Fig. 9 a Contour plot of corrective function MR over the nondimensional variables (Rs/Rc and δ/Rs ) for κ = 0.1. b MR over thenondimensional variables (Rs/Rc and δ/Rs ) for κ = 1. Both contour plots show Eq. 35 with the coefficients specified in Table 1

Fig. 10 a Plot of normalized force over the nondimensional variables (δ/Rc) for the case of Rs/Rc = 1 (blue) and Rs/Rc = 20. Simulations per-formed over large indentations (dashed lines) are compared to force-displacement curves calculated with the correction functions (κ = 1) foundin this study (solid lines) and linear elastic theory (“+” lines). The approximation to the indentation of a cylinder by a half-space is also shown forcomparison (circles). b Plot of the sum of squared errors for the correction functions found in this study (solid lines) and compared to linear elas-tic elliptical contact theory (“+” lines)

Mech Soft Mater (2019) 1: 77 Page 14 of 18

4.2 Substrate effects

As mentioned in the preceding section, one way in which substrate interactions affect the solution is by restriction ofdeformation along the direction of loading. Another way in which the substrate interactions affect the solution is bymodifying the boundary conditions between the specimen and substrate. For the ease of creating a symmetric model (Fig. 1c),we chose to use a boundary condition that allows the specimen to slip along the longitudinal direction, while being restrictedin the direction normal to the loading (i.e., “non-adhered” boundary condition). Alternatively, one could also examine theeffects of a “fully adhered” boundary condition, which is expected to to behave differently than the “non-adhered” boundarycondition at large values of δ/Rc. In reality, the boundary condition between a cell and its substrate is not entirely fullyadhered or entirely non-adhered. As described by Mahaffey et al. 2004, cell membranes will often have a higher abundanceof focal adhesion sites around the edges. A combination of full-adhesion and no-adhesion modeling can be used to obtainmaterial properties for different regions of the cell [46].

4.3 Regimes for Rs/Rc

The correction factors provided in this study were found using computational models with Rs/Rc between 0.4 and 20. Thisrange of Rs/Rc was sufficient to encompass the lower bound of elliptic contact (rigid spherical indenter on a deformablehalf-space specimen) and the upper bound of elliptic contact (rigid half-space indenter on a deformable cylinder). The lowerbound of elliptic contact becomes a circular contact area, while the upper bound becomes a rectangular/line contact area.As mentioned previously, for values of Rs/Rc < 0.4, the cylindrical specimen is relatively “flat,” and the conventionalformulations for circular contact areas (spherical indentation on a half-space) can be utilized. Previous works have alsoconsidered the effects of large indentation and substrate interaction for spherical indentation on a half-space. For a sphericalindentation on a flat cellular specimen (Rs/Rc < 0.4) of finite thickness h (h ≥ 0.1Rs), one can use the analyticalexpressions provided by Dimitriadis [47]. However, they assume linearity, and the result is meant for small indentations(δ ≤ 0.1h). For very thin flat specimens (Rs/Rc < 0.4, h ≤ 0.1Rs), an analytical expression is provided by Chadwick [48].

Here, we do not specify a theoretical upper bound for Rs/Rc, although Rs/Rc = 20 is the maximum value used in fittingthe expression of the corrective function. For very large values of Rs/Rc, the behavior is expected to approach the theoreticallimit for the indentation of a half-space on a cylinder. This is also equivalent to the case of two parallel cylinders in contact(although with one cylinder having a radius that tends to infinity). The two parallel cylinders brought into contact will resultin a rectangular contact area. Precise equations for the contact between two cylinders are provided in previous works [24, 44](although a major drawback of these models is that the contact force cannot be explicitly defined as a function of indentationin closed form [49]). The indentation depth for a rigid half-space with a linear elastic cylinder of length L is given by [44]:

δ = F

πLE∗

(1 + ln

πE∗L3

RcylF

), (37)

assuming that the half-space is rigid, 1E∗ = 1−ν2

E. Equation 37 is plotted in Fig. 10a (circles). It is observed that the theoretical

force-displacement curve for the elliptical contact of Rs/Rc = 20 is very close to the upper bound where Rs approachesinfinity (i.e., rigid half-space indenter), indicating that the range of Rs/Rc values we selected for our computational studiesis sufficient for capturing both the lower and upper bounds of ellipticity.

4.4 Limitations and future considerations

This study examined two simple and well-known hyperelastic (neo-Hookean and the Mooney-Rivlin) models, althoughthese are not the only possibilities for hyperelastic strain energy functions. More sophisticated material models, however,might require more material parameters and could result in more complex corrective functions.

For values of Rs/Rc less than 0.4, one can expect MR ≈ 1 for fairly large indentation depths (until δ/Rc = 0.45). Forlarger size ratios (Rs/Rc > 20), the correction factors are precise until values of δ/Rc > 0.25. Errors at larger indentationdepths should be considered, as discussed in the previous section. Large values of δ/Rc could also result in large errors whenapplying the model to experimental data where the observed interaction between the specimen and substrate is fully adhered.

The work presented herein assumes that the cell is a homogeneous continuous solid in order to obtain an overall effectivemodulus. However, cells are heterogeneous in nature—comprising of a lipid membrane surrounding networks of biopolymerfilaments with cytoplasmic fluid. Our formulation applies only to the case of an incompressible material, ignoring the

Mech Soft Mater (2019) 1: 7 Page 15 of 18 7

possible time-dependent behavior of the cytoplasmic cell volume [50]. To expand this work further into applications of cellmechanics over various time scales, one could use a biphasic material model approach to capture the hyperelastic solidnetwork (comprised of the cytoskeleton and organelles) surrounded by cytoplasmic fluid. Furthermore, there is the caveatthat we have not factored in the effects of strain rate into the indentation formulations, which would also be important whenconsidering the effective mechanical properties of cells.

4.5 Potential applications

In order to obtain the hyperelastic material properties from the experimental data, one could use the following procedure:

1. Multiply the experimental data (Fexp) by the corrective function for the neo-Hookean model (FexpNH ).

2. Set FexpNH = 2π3

4C10(1+ν)

1−ν2

√2Rδ3/2kF−3/2E1/2

where,

NH =[A1 ln(

Rs

Rc) + A2

] (δRs

)m1 ln(RsRc

)+m2

ν = 0.5, and the constants are given by: A1 = −0.16, A2 = 0.76, m1 = −0.03, and m2 = −0.07.3. Calculating k, F , and E with the known dimensions of Rs and Rc, and assuming ν = 0.5, use the preceding expression

to obtain C10.

An equivalent result can also be obtained by using MR (setting κ = 1) instead of NH . Once C10 is obtained, one canuse MR and find an appropriate value for κ by numerically solving for the minimization of squared errors between theestimated FMR curves and the experimental results. Therefore, an initial value of C10 should be determined first, otherwise,κ cannot be uniquely determined [51].

The formulations provided in this study can be applied to several cylindrical-shaped specimens that appear in biologicalmaterials. In Fig. 1, we showed an example of spherical indentation on a neuronal axon (cylinders with diameters on theorder of a few microns). In addition to axons, the vast majority of brain cells contain cylindrical-like morphologies in theirprocesses (glial cell processes, neurites, dendrites, etc.).

Examples of other cells that exhibit a stellate-like morphology (i.e., have processes) include osteocytes, mesenchymalstem cells, pericytes, and fibroblasts. Due to the heterogeneity of the mechanical properties of cells, it is important to obtainproperties with respect to specific cellular compartments [9, 52] (and not just measuring the response of the sphericalsomal body with classical Hertz formulations). Aside from probing fibrous cellular components, other uses of this workinclude probing other types of cylindrical-shaped biological specimens such as certain strains of bacteria (i.e., bacilli), aswell as fibrous materials of the extracellular matrix (such as fibrin, collagen, and elastin). With respect to non-biologicalapplications, this work can be applied to measuring hyperelastic properties of cylindrical synthetic polymer fibers and naturalfibers that can be used in fiber-reinforced composites (provided that they can be approximated as incompressible).

Fig. 11 Application of Mooney-Rivlin formulation presented in this study with experimental data on indentation of a collagen fibril (Andriotis etal. [53]). Fitted hyperelastic properties were found to be C10 = 6.8 × 108 Pa, κ = 0.6, C01 = 4.6 × 108 Pa

Mech Soft Mater (2019) 1: 77 Page 16 of 18

An example of the formulation applied to the experimental data of AFM indentation on a single collagen fibril is shownin Fig. 11. The experimental data is taken from Andriotis et al. [53], with Rs = 5 nm and Rc ∼ 50 nm. Using the procedureoutlined above, and performing a minimization of root-mean-square error to a range of C10 and κ , we obtained hyperelasticparameters of C10 = 6.8 × 108 Pa and κ = 0.6. In the linear elastic regime, this corresponds to a Young’s modulus ofE = 6.8 × 109 Pa, which is within the range of values found by Andriotis et al. (E ∼ 6 − 14 GPa). Although we canreasonably approximate the Young’s modulus with our formulation, the usefulness in the formulation relies in its ability toestimate the nonlinear properties as well.

5 Conclusions

Our approach provides a generalized formulation that can be used to extract mechanical properties from indentation into softcylindrical bodies. The generalized corrective function MR of Eq. 35 with the corresponding coefficients of Table 2 canbe applied to force-displacement relationships for cylindrical incompressible hyperelastic materials subjected to sphericalindentation. These corrective functions are valid for a wide range of specimen and indenter sizes: with small values ofRs/Rc

approaching the limit of a flat specimen subjected to spherical indentation and large values ofRs/Rc approaching the limit ofa flat indenter with a cylindrical specimen. For large values of Rs/Rc and large indentations (δ/Rs), the force-displacementbehavior deviates farther from linear elastic elliptical contact theory.

Although the motivation of this study was to provide a theoretical framework for obtaining useful mechanical propertiesof cylindrical cellular bodies subjected to large indentation and deformation, the results can be applied to experimentaldata of non-cellular soft specimens (collagen, fibrin, biomimetic polymer strands, etc.) where a hyperelastic model is alsorequired. Given the availability of indentation experiments over various length scales, the corrective functions provided inthis study are a promising tool for measuring the hyperelastic properties of soft materials in a wide range of applications.

Funding information This work was funded by the DoD SMART Scholarship Program and the US Army Research Lab (Aberdeen ProvingGround, MD), under Cooperative Agreement Number W911NF-12-2-0022.

Disclaimer The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the officialpolicies, either expressed or implied, of the Army Research Laboratory or the US Government. The US Government is authorized to reproduceand distribute reprints for Government purposes notwithstanding any copyright notation herein.

References

1. Lutolf, M.P., Gilbert, P.M., Blau, H.M.: Designing materials to direct stem-cell fate. Nature 462(7272), 433 (2009)2. Ethier, C.R., Simmons, C.A.: Introductory Biomechanics: From Cells to Organisms. Cambridge University Press (2007)3. Suresh, S., Spatz, J., Mills, J., Micoulet, A., Dao, M., Lim, C., Beil, M., Seufferlein, T.: Connections between single-cell biomechanics and

human disease states: gastrointestinal cancer and malaria. Acta biomaterialia 1(1), 15 (2005)4. Moeendarbary, E., Weber, I.P., Sheridan, G.K., Koser, D.E., Soleman, S., Haenzi, B., Bradbury, E.J., Fawcett, J., Franze, K.: The soft

mechanical signature of glial scars in the central nervous system. Nat. Commun. 8, 14787 (2017)5. Suresh, S.: Biomechanics and biophysics of cancer cells. Acta Mater. 55(12), 3989 (2007)6. Darling, E.M., Di Carlo, D.: High-throughput assessment of cellular mechanical properties. Ann. Rev. Biomed. Eng. 17, 35 (2015)7. Zahalak, G., McConnaughey, W., Elson, E.: Determination of cellular mechanical properties by cell poking, with an application to leukocytes.

J. Biomech. Eng. 112(3), 283 (1990)8. Zahalak, G.I., Wagenseil, J.E., Wakatsuki, T., Elson, E.L.: A cell-based constitutive relation for bio-artificial tissues. Biophys. J. 79(5), 2369

(2000)9. Heidemann, S.R., Wirtz, D.: Towards a regional approach to cell mechanics. Trends Cell Biol. 14(4), 160 (2004)

10. Petersen, N.O., McConnaughey, W.B., Elson, E.L.: Dependence of locally measured cellular deformability on position on the cell,temperature, and cytochalasin b. Proc. Natl. Acad. Sci. 79(17), 5327 (1982)

11. Binnig, G., Quate, C.F., Gerber, C.: Atomic force microscope. Phys. Rev. Lett. 56(9), 930 (1986)12. McConnaughey, W.B., Petersen, N.O.: An apparatus for stress-strain measurements on living cells. Rev. Sci. Instrum. 51(5), 575 (1980)13. Radmacher, M., Tillamnn, R., Fritz, M., Gaub, H.: From molecules to cells: imaging soft samples with the atomic force microscope. Science

257(5078), 1900 (1992)14. Hoh, J.H., Schoenenberger, C.A.: Surface morphology and mechanical properties of mdck monolayers by atomic force microscopy. J. Cell

Sci. 107(5), 1105 (1994)15. Vinckier, A., Semenza, G.: Measuring elasticity of biological materials by atomic force microscopy. FEBS Lett. 430(1-2), 12 (1998)16. Azeloglu, E.U., Costa, K.D.: Atomic force microscopy in mechanobiology: measuring microelastic heterogeneity of living cells. Atomic

Force Microscopy in Biomedical Research, pp. 303–329. Springer (2011)17. Lim, C., Zhou, E., Quek, S.: Mechanical models for living cells: a review. J. Biomech. 39(2), 195 (2006)

Mech Soft Mater (2019) 1: 7 Page 17 of 18 7

18. Sen, S., Subramanian, S., Discher, D.E.: Indentation and adhesive probing of a cell membrane with AFM. Biophys. J. 89(5), 3203 (2005)19. Digiuni, S., Berne-Dedieu, A., Martinez-Torres, C., Szecsi, J., Bendahmane, M., Arneodo, A., Argoul, F.: Single cell wall nonlinear mechanics

revealed by a multiscale analysis of AFM forceindentation curves. Biophys. J. 108(9), 2235 (2015)20. Humphrey, J.D.: Continuum biomechanics of soft biological tissues. Proceedings of the Royal Society of London A: Mathematical, Physical

and Engineering Sciences, vol. 459, pp. 3–46. The Royal Society (2003)21. Mow, V.C., Guilak, F., Tran-Son-Tay, R., Hochmuth, R.M.: Cell mechanics and cellular engineering. Springer Science & Business Media (2012)22. Levental, I., Georges, P.C., Janmey, P.A.: Soft biological materials and their impact on cell function. Soft Matter 3(3), 299 (2007)23. Hertz, H.: On the contact of solid elastic bodies and on hardness. J. Math 92, 156 (1881)24. Johnson, K.L.: Contact Mechanics. Cambridge University Press (1987)25. Williams, J.A., Dwyer-Joyce, R.S.: Contact between solid surfaces. Modern Tribol. Handbook 1, 121 (2001)26. Liu, D., Zhang, Z., Sun, L.: Nonlinear elastic load-displacement relation for spherical indentation on rubberlike materials. J. Mater. Res.

25(11), 2197 (2010)27. Lin, D.C., Shreiber, D.I., Dimitriadis, E.K., Horkay, F.: Spherical indentation of soft matter beyond the hertzian regime: numerical and

experimental validation of hyperelastic models. Biomech. Modeling Mechanobiol. 8(5), 345 (2009)28. Giannakopoulos, A., Triantafyllou, A.: Spherical indentation of incompressible rubber-like materials. J. Mech. Phys. Solids 55(6), 1196 (2007)29. Zhang, M.G., Cao, Y.P., Li, G.Y., Feng, X.Q.: Spherical indentation method for determining the constitutive parameters of hyperelastic soft

materials. Biomech. Model. Mechanobiol. 13(1), 1 (2014)30. Ding, Y., Xu, G.K., Wang, G.F.: On the determination of elastic moduli of cells by AFM based indentation. Sci. Rep. 7, 45575 (2017)31. Bernick, K.B., Prevost, T.P., Suresh, S., Socrate, S.: Biomechanics of single cortical neurons. Acta biomaterialia 7(3), 1210 (2011)32. Kang, I., Panneerselvam, D., Panoskaltsis, V.P., Eppell, S.J., Marchant, R.E., Doerschuk, C.M.: Changes in the hyperelastic properties of

endothelial cells induced by tumor necrosis factor-α. Biophys. J. 94(8), 3273 (2008)33. Mahaffy, R., Shih, C., MacKintosh, F., Kas, J.: Scanning probe-based frequency-dependent microrheology of polymer gels and biological

cells. Phys. Rev. Lett. 85(4), 880 (2000)34. Guz, N., Dokukin, M., Kalaparthi, V., Sokolov, I.: If cell mechanics can be described by elastic modulus: study of different models and probes

used in indentation experiments. Biophys. J. 107(3), 564 (2014)35. Sokolov, I., Iyer, S., Woodworth, C.D.: Recovery of elasticity of aged human epithelial cells in vitro. Nanomed.: Nanotechnol. Biol. Med.

2(1), 31 (2006)36. Codan, B., Del Favero, G., Martinelli, V., Long, C., Mestroni, L., Sbaizero, O.: Exploring the elasticity and adhesion behavior of cardiac

fibroblasts by atomic force microscopy indentation. Mater. Sci. Eng. C 40, 427 (2014)37. Berdyyeva, T.K., Woodworth, C.D., Sokolov, I.: Human epithelial cells increase their rigidity with ageing in vitro: direct measurements. Phys.

Med. Biol. 50(1), 81 (2004)38. Park, S., Koch, D., Cardenas, R., Kas, J., Shih, C.K.: Cell motility and local viscoelasticity of fibroblasts. Biophys. J. 89(6), 4330 (2005)39. Sokolov, I., Dokukin, M.E., Guz, N.V.: Method for quantitative measurements of the elastic modulus of biological cells in AFM indentation

experiments. Methods 60(2), 202 (2013)40. Magdesian, M.H., Sanchez, F.S., Lopez, M., Thostrup, P., Durisic, N., Belkaid, W., Liazoghli, D., Grutter, P., Colman, D.R.: Atomic force

microscopy reveals important differences in axonal resistance to injury. Biophys. J. 103(3), 405 (2012)41. Charras, G.T., Horton, M.A.: Single cell mechanotransduction and its modulation analyzed by atomic force microscope indentation. Biophys.

J. 82(6), 2970 (2002)42. Brewe, D.E., Hamrock, B.J.: Elastic compression of spheres and cylinders at point and line contact. J. Lubr. Technol. 99(4), 485 (1977)43. Hamrock, B.J., Brewe, D.: Simplified solution for stresses and deformations. J. Lubricat. Technol. 105(2), 171 (1983)44. Puttock, M., Thwaite, E.: Elastic Compression of Spheres and Cylinders at Point and Line Contact. Commonwealth Scientific and Industrial

Research Organization, Melbourne Australia (1969)45. Horgan, C.O., Smayda, M.G.: The importance of the second strain invariant in the constitutive modeling of elastomers and soft biomaterials.

Mech. Mater. 51, 43 (2012)46. Mahaffy, R., Park, S., Gerde, E., Kas, J., Shih, C.: Quantitative analysis of the viscoelastic properties of thin regions of fibroblasts using

atomic force microscopy. Biophys. J. 86(3), 1777 (2004)47. Dimitriadis, E.K., Horkay, F., Maresca, J., Kachar, B., Chadwick, R.S.: Determination of elastic moduli of thin layers of soft material using

the atomic force microscope. Biophys. J. 82(5), 2798 (2002)48. Chadwick, R.S.: Axisymmetric indentation of a thin incompressible elastic layer. SIAM J. Appl. Math. 62(5), 1520 (2002)49. Pereira, C.M., Ramalho, A.L., Ambrosio, J.A.: A critical overview of internal and external cylinder contact force models. Nonlin. Dyn. 63(4),

681 (2011)50. Moeendarbary, E., Valon, L., Fritzsche, M., Harris, A.R., Moulding, D.A., Thrasher, A.J., Stride, E., Mahadevan, L., Charras, G.T.: The

cytoplasm of living cells behaves as a poroelastic material. Nat. Mater. 12(3), 253 (2013)51. Pan, Y., Zhan, Y., Ji, H., Niu, X., Zhong, Z.: Can hyperelastic material parameters be uniquely determined from indentation experiments?

RSC Adv. 6(85), 81958 (2016)52. Cartagena, A., Raman, A.: Local viscoelastic properties of live cells investigated using dynamic and quasi-static atomic force microscopy

methods. Biophys. J. 106(5), 1033 (2014)53. Andriotis, O.G., Manuyakorn, W., Zekonyte, J., Katsamenis, O.L., Fabri, S., Howarth, P.H., Davies, D.E., Thurner, P.J.: Nanomechanical

assessment of human and murine collagen fibrils via atomic force microscopy cantilever-based nanoindentation. J. Mech. Behav. Biomed.Mater. 39, 9 (2014)

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Mech Soft Mater (2019) 1: 77 Page 18 of 18