You are on page 1of 11

journal of the mechanical behavior of biomedical materials 122 (2021) 104649

Contents lists available at ScienceDirect

Journal of the Mechanical Behavior of Biomedical Materials


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

Biomechanical and functional comparison of moulded and 3D printed


medical silicones
Alexandra Zühlke a, *, Michael Gasik a, d, Nihal Engin Vrana b, Celine Blandine Muller c,
Julien Barthes c, Yevgen Bilotsky d, Edwin Courtial e, Christophe Marquette e
a
Aalto University Foundation, Espoo, Finland
b
Spartha Medical SAS, Strasbourg, France
c
INSERM, Strasbourg, France
d
Seqvera Ltd. Oy, Helsinki, Finland
e
3dFAB, Lyon, France

A R T I C L E I N F O A B S T R A C T

Keywords: Modern 3D printing of implantable devices provides an important opportunity for the development of person­
Silicones alized implants with good anatomical fit. Nevertheless, 3D printing of silicone has been challenging and the
Mechanical properties recent advances in technology are provided by the systems which can print medical grade silicone via extrusion.
In vitro
However, the potential impacts of the 3D printing process of silicone on its biomechanical properties has not
In vivo
Biocompatibility
been studied in sufficient detail. Therefore, the present study compares 3D printed and moulded silicone
Implants structures for their cytotoxicity, surface roughness, biomechanical properties, and in vivo tissue reaction. The 3D
printing process creates increased nanoscale roughness and noticeably changes microscale topography. Neither
the presence of these features nor the differences in processes were found to result in an increase in cytotoxicity
or tissue reaction for 3D printed structures, exhibiting limited inflammatory reaction and cell viability above the
threshold values. On the contrary, the biomechanical properties have demonstrated significant differences in
static and dynamic conditions, and in thermal expansion. Our results demonstrate that 3D printing can be used
for establishing a better biomechanical microenvironment for the surrounding tissue of the implant particularly
for fragile soft tissue like epithelial mucosa without having any negative effect on the cytotoxicity or in vivo
reaction to silicone. For engineering of the implants, however, one must consider the differences in mechanical
properties to result in correct and personalized geometry and proper physical interaction with tissues.

1. Introduction The stent might move from its position or even come out of the patient
entirely. Moreover, the mucosa of the upper respiratory tract is highly
There is a growing trend to use 3D printing as the manufacturing fragile and any excessive pressure applied to the mucosa by stents would
method for medical devices enabling their personalization (Vrana et al., lead to an adverse immune reaction and necrosis. In addition to this,
2020; Jammalamadaka and Tappa, 2018; Nagarajan et al., 2018; some patients would have unconventional anatomical features due to
Kacarevic et al., 2018; DiMarzio et al., 2020). Medical grade silicone is trauma or surgical resection to which normal stents would apply addi­
widely used in health care (Sittel, 2009; Emre et al., 2005) and it can be tional pressure and risk of damage. This problem can be prevented by
processed using 3D printing (Jammalamadaka and Tappa, 2018; creating custom devices with a better fit, which can be achieved with a
Kacarevic et al., 2018). An example of a medical device requiring a high 3D printing technology. However, beyond the anatomical fit of the
personalization is stents used in the upper respiratory tract, whose stents, their biomechanical fit should also be considered.
function is to keep the respiratory tract open to minimize the harm of Such 3D printed structures are highly desirable matching the
injury caused by trauma or surgery (Sittel, 2009; Debry et al., 2017; biomechanical properties of the tissue (Gomes and Reis, 2004) and
Cooper, 2018). One of the challenges of such stenting is the possibility of closely mimicking in vivo behaviour, assisting the body to rebuild the
migration, for example, due to coughing or excessive physical loading. surrounding damaged tissue (Chung and Burdick, 2008; Wong, 2007).

* Corresponding author.
E-mail address: alexandra.zuhlke@aalto.fi (A. Zühlke).

https://doi.org/10.1016/j.jmbbm.2021.104649
Received 15 October 2020; Received in revised form 9 June 2021; Accepted 12 June 2021
Available online 16 June 2021
1751-6161/© 2021 Published by Elsevier Ltd.
A. Zühlke et al. Journal of the Mechanical Behavior of Biomedical Materials 122 (2021) 104649

One of the most critical issues to address, however, is the mechanical


properties of 3D printed biomaterials, as additive manufacturing tech­
niques might represent an important potential risk to be considered for
personalized implants (Vrana et al., 2020; Gasik, 2017). Mechanical and
biomechanical stimuli play an important role in organisms’ develop­
ment, homeodynamics and homeostasis (Vrana et al., 2020; van der
Meulen and Huiskes, 2002). On the level of a living cell, mechanical
stimuli are ones of the three fundamental pathways (electrical, (bio)
chemical and mechanical) to communicate with the cell and with its
environment.
Fig. 1. View of the 3D printed (left) and moulded (right) silicone samples as 10
One important gap is still a lack of proper understanding of biome­
x 10 x 2 ​ mm blocks.
chanical properties and their meaning in healthcare (Gasik, 2017). All
cells and organisms are mechanosensitive (Ingber, 2006; Orr et al.,
2006), relaying on conversion of mechanical stimuli from their physical effects including granuloma formation, local necrosis and chronic
surroundings or from within the organism to electrochemical or inflammation induced restenosis (Nesek-Adam et al., 2010; Alshammari
biochemical signals, which then regulate physiological responses (Jaa­ and Monnier, 2012; Ashfaque and Annju, 2018). In order to provide the
louk and Lammerding, 2009). Mechanotransduction also has a funda­ necessary information for the 3D printing based silicone implant
mental role in regulating physiological phenomena in other specialized personalization in stenting the biomechanical properties of medical
tissues that are not directly involved in sensory functions (Vrana et al., silicones produced by traditional moulding and via 3D printing have
2020) and can have a critical influence on cell behaviour, even in tissues been studied and compared with dynamic mechanical analysis (DMA)
and organs that do not serve an apparent biomechanical role in the body complemented with biomaterials enhanced simulation testing (BEST)
(Guilak et al., 2014; Kazantseva et al., 2016). data analysis (Gasik and Bilotsky, 2019). In addition, surface roughness
A key issue in the development of engineered implants is the prior­ and cell behaviour in the presence of 3D printed and moulded implants
itization of various biomechanical properties as design parameters, as it were quantified. The results were combined to evaluate how production
will be difficult (if not impossible) to completely match all the material method affects potential properties of implants for a given application
properties of native tissues (Guilak et al., 2014). There is also an un­ and to which extent these differences needed to be considered in the
derstanding that may not be even necessary to match all of the material design and optimization of these medical devices.
properties of native tissues a priori in view of the remodelling potential
of the implanted tissue in vivo – indeed, some degree of ‘mismatch’ is 2. Materials and methods
required to ensure necessary driving forces aiming on the restoring
functionality (Vrana et al., 2020; Gasik et al., 2018). Although me­ 2.1. Materials tested
chanical strain usually exhibits rather complex patterns in vivo, its im­
pacts and related mechanisms can be investigated with simplified in The silicone samples (moulded silicone marked as “SilM” and 3D
vitro models. For safe and widespread application of 3D printed implant printed silicone marked as “Sil3D”) were produced of a medical grade
technologies, the underlying mechanical differences stemming from the silicone (Elkem Silicones, Norway, Silbione® LSR 4350) by a) moulding
production methods should be carefully considered as a part of in aluminium moulds with industry standards and b) using a custom-
personalization protocols. made extrusion printer (Tobeca, 3dFAB, Lyon, France). The samples
Due to vast number of biomaterials and their forms, one usually were ~2 ​ mm thick and 1 x 1 ​ cm in area. For the compression test, the
needs to choose the biomechanical testing methods, and this choice samples were measured as they were, but for the 3-point bending test the
might not be self-evident. There are constrains in existing standards samples were cut in half, resulting in rectangular samples with 2 x 5 x
(easier to make a test, but more difficult to understand the meaning of 10 ​ mm dimensions. Pictures and specifications of the samples, either 3D
the data for clinical applications) and protocols (ad hoc test devices and printed or moulded, used to perform 3-point bending and compression
methods which are not easy to replicate in another laboratory) (Vrana experiments are given in Fig. 1.
et al., 2020). One may consider elastic modulus, whose definition The surface properties of the moulded and 3D-printed samples were
originally fits only linear elastic materials and only for very small de­ characterised with a Veeco Multimode Atomic Force Microscope (AFM)
formations, as has been prescribed by the theory of elasticity for cen­ in tapping mode for nanoscale surface features and roughness. The Ra
turies. The guidelines of National Physical Laboratory (UK) list nine averages are calculated from 25 x 25 ​ μm ROIs using the Nanoscope
methods of measurement and calculation of elastic modulus (Lord and software. For microscale surface features and roughness, a Keyence
Morrell, 2006), all of which naturally lead to different values. Elasticity VHC-5000 digital microscope was used to have 3D reconstruction of
theory does not consider the time-dependency of this property, the rate surface features via z-stacks (up to 700 ​ μm) using a 60x objective.
(speed) of deformation, or its hysteresis due to cyclic load. Almost all
biomaterials and tissues are clearly not elastic ones, it is a significant 2.2. Mechanical testing
simplification trying to artificially reduce experimental data to some
fixed numbers (Vrana et al., 2020). Silicone, a non-linear elastomer, Before performing the mechanical tests, the samples were cleaned as
would thus expect to have different behaviours when it is moulded or follows with 30 ​ min in ultrasound bath in ethanol (70%) and then by
printed based on extrusion; however, to our best knowledge this dif­ rinsing with deionized water and air dried. All tests were done at
ference has not been studied in the literature up to now. 37 ​ ± ​ 1 ◦ C in dry static air. The biomechanical analysis was carried out
In the present work silicone polymers, intended for otorhinolaryn­ using dynamic mechanical analysis (DMA) with apparatus “Artemis”
gology applications, namely for respiratory tract stenting, have been DMA 242 (models C and E; Netzsch Gerätebau GmbH, Germany) under
characterised for their biomechanical, surface (micro and nanoscale) pseudo-static (creep) and frequency scans (0.01–20 ​ Hz) at 20 ​ μm dy­
and protein adsorption properties together with their cytotoxicity. namic amplitude (resolution ±0.5 ​ nm). In the creep test, the material is
Whereas there are about 100 models of stents known, the problem of subjected to a constant static force over a given time and the resulting
tissue granulation, implant stability and operability has not been solved deformation is measured. Dynamic tests foresee stimulation of the
yet (Trabelsi et al., 2011; Noppen et al., 2005). It was previously sample by an oscillatory stress and the resulting deformation is
observed that the biomechanical mismatches between the silicone im­ measured together with its phase shift (also commonly expressed as loss
plants and the surrounding mucosa were inducing serious adverse tangent). In the frequency scan the amplitude of the oscillation is kept

2
A. Zühlke et al. Journal of the Mechanical Behavior of Biomedical Materials 122 (2021) 104649

Fig. 2. An example of a frequency scan test principle, where the amplitude is constant but the frequency (F1 ​ < ​ F2 ​ < ​ F3) varies.

constant but the frequency is varied (Fig. 2). For all combinations, experimental observation of materials properties without assumption of
changes in length, stiffness, loss tangent, dynamic and static stress/ a material mechanistic model, method linearity, and without applica­
strain ratios, viscosity, creep compliance, aggregate modulus, etc. were tion of Fourier transform or using complex numbers (Gasik and Bilotsky,
analysed. 2019). Whereas Fourier transform is a very useful and perfectly working
Tests were performed for compression mode and for 3-point bending tool for linear operations (in linear viscoelasticity), it fails for properties
at 10 ​ mm span. All sample holders are automatically calibrated on as arbitrary functions of time or frequency (Neumark, 1962a). Gener­
empty system stiffness, static, dynamic and rotation tuning, from which alized Fourier transforms such as the Fourier integral operators and
data are subtracted from the measured signal. In addition, in compres­ pseudo-differential operators could be used (Maslov, 1970), however,
sion mode a thermo-mechanical analysis (TMA) mode was applied to the drawback of this approach is in its strict adherence to certain types of
analyse thermal expansion of silicones up to 45 ◦ C. The deformation pseudo-differential equations and requirements of a deep knowledge of
values recorded by DMA (in μm) were converted into true logarithmic such functional techniques (Maslov, 1970; Neumark, 1962a). On the
strain. True strain is the only measure having solid thermodynamic contrary to these approaches, BEST does not employ differential equa­
grounds among used strain forms (Norris, 2008; Lubarda and Chen, tions or series expansions, neither requires material homogeneity (the
2008; Xiao, 1995), although others might be used especially when the properties are evaluated for the whole specimen, i.e. closer to the con­
deformations are small. True strain in this case is calculated for ditions which might be faced upon implantation). The BEST method
quasi-static loading (creep) conditions as does not stipulate that the material has to be compliant with some
⃒ ( )⃒ pre-selected physical model (elastic, viscoelastic, hyperelastic,
⃒ ΔL ⃒⃒
εstat = ⃒⃒In 1 + , (1) neo-Hookean, etc.). Selection of the material model in any combination
H0 ⃒
is a must for any conventional calculations in viscoelastic analysis or in
numerical computer simulations, but this may lead to wrong predictions
and for dynamic loading conditions as (Gasik and Bilotsky, 2019):
when the model assumptions do not hold. In this work BEST analysis was
( )
1 2adyn used to extract true (invariant) properties of elasticity, viscosity and
εdyn = In 1 + , (2)
2 H0 + ΔL − adyn their time and frequency dependences (Gasik and Bilotsky, 2019).

where H0 - initial dimension (height) of the specimen before the test, ΔL - 2.4. Biocompatibility testing
instant value of change of the static height, adyn - amplitude of the dy­
namic displacement. Equation (2) takes into account the change in the Cytotoxicity by indirect contact was analysed with an extraction
specimen static dimensions during the test, incorporating loading his­ vehicle (the supplemented DMEM cell culture medium) in the contact
tory, and accounts for non-symmetry of the harmonic stimulus signal vs. with the material (moulded and 3D printed silicone samples, PET as
origin due to this. positive control, sodium azide NaN3 as negative control) for 24 ​ h in the
incubator (5% CO2, 37 ◦ C). In parallel, the 3T3 cells are cultured in a cell
2.3. Mechanical data analysis culture dish until reaching 70–80% confluence in 24 ​ h. On the second
day, the extraction vehicle was transferred to the cell mat and left in
With the conventional biomechanical readouts obtained directly contact for next 24 ​ h before evaluating possible cell toxicity at the
from the DMA, advanced data analysis was carried out with BEST morphological level (microscopy) and at the level of viability (MTT
(Biomaterials Enhanced Simulation Testing). Briefly, BEST comprises test).
steps of 1) defining the proper physiologically relevant conditions for Cytotoxicity by direct contact was evaluated with the 3T3 cells first
your clinical application, 2) subjecting the specimen to coherent grown in a cell culture dish to reach 70–80% confluence within 24 ​ h. On
biomechanical stimuli, inducing conditions closer to hostile environ­ the second day, the materials to be tested were placed in contact with
ment, 3) assessing changes in properties of the specimen in time, phase the cell mat and left for 24 ​ h before evaluating possible cell toxicity at
and stimulus domains, 4) post-processing experimental readouts with the morphological level (microscopy) or at the level of viability (MTT
idempotent analysis (Gasik and Bilotsky, 2019; Maslov, 1970), incor­ test). For this test, the dimensions of the material should not exceed 10%
porating specimen history and 5) extracting true invariant values, suit­ of the surface of the cell layer. To keep the material in contact with the
able for material behaviour prediction in a form of model-free cell mat, a cell culture insert was used which was placed on the material.
constitutive equation for given biomaterial.
The differences between BEST and other methods are in direct

3
A. Zühlke et al. Journal of the Mechanical Behavior of Biomedical Materials 122 (2021) 104649

(0.05 ​ mg/kg) was injected into the animal subcutaneously to reduce


pain at the end of the implantation and every 12 ​ h for 5 days. Eutha­
nasia was performed after 14 days. Totally 8 mice were implanted for
each condition. The samples were explanted after 2 weeks and charac­
terised by histology together with the measurement of cytokine levels
for determination of systemic effects. Only cytokines with read-outs
above threshold were reported (IL-1β, IL-23, IL-27, MCP-1, TNF-α,
IFN-γ). Histology images were scored by an independent histopatholo­
gist without access to the sample information (blind analysis).

2.6. Statistical analysis

Data of biomechanical measurements, after time convolution pro­


cedure shown below, were fitted with exponential function of memory
values as it is theoretically predicted by idempotent analysis (Gasik and
Bilotsky, 2019). The presence of leverage points was detected with hat
matrix diagonal components, which did not fit Stephen’s rule (these
points were removed). Influence (outlier) points analysis was made by
calculating Cook’s distances, and data points exceeding unity value (if
Fig. 3. Thermal expansion of silicones.
any) were removed. The consistency of regression coefficients was
independently checked by application of Theil-Shen estimator and the
2.5. In vivo experiments goodness of fit significance by Nelson-modified Anderson-Darling test
(Nelson, 1998). The algorithm for this test has been improved, as the
For testing the tissue reaction to moulded and 3D printed silicones in original source code (Nelson, 1998) does not allow negative normalized
vivo, subcutaneous paravertebral implantation of samples of 1 ​ cm2 with (studentized) values in the data.
5 ​ mm thickness was performed in mice for a duration of 2 weeks after Heteroscedasticity of residuals was estimated with RUNS test and the
the obtention of the ethical authorisation. Experiments were performed residuals autocorrelation with Durbin-Watson parameter. Based on the
in UE-1277 Plateforme d’Infectiologie Expérimentale (INRA, Tours, mathematical analysis, obtained biomechanical invariant values were
France). Protocols were submitted to the C2EA–19 ethics committee considered to be BLUE (best linear unbiased estimators).
with approval number DAP-2017031321109445, in accordance with The statistical significance for in vitro assessed using Student’s t-test
the European directive 2010/63/EC for conducting animal experiments. and in vivo studies of the obtained data was assessed using one-way
Anaesthesia of the animals was performed with an intraperitoneal ANOVA with Tukey’s multiple comparisons or Friedman’s test with
injection of ketamine (Imalgène 1000®, 75 ​ mg/kg) and xylazine Dunn’s multiple comparisons (n ​ ≥ ​ 3). The error bars are representative
(Rompun®, 5 ​ mg/kg). Then mice were placed on a heated pad, shaved, of the standard deviation (SD). Differences with values of p ​ ≥ ​ 0.05
cleaned with betadine and buprenorphine (Buprécare®, 0.05 ​ mg/kg) were considered statistically insignificant.
was injected in subcutaneous. Then incision was made in paravertebral
and silicones patches were implanted in subcutaneous. To finish the
operation, muscles and incisions were closed with stitches. Buprécare

Fig. 4. Surface characterization of moulded (SilM) and 3D printed (Sil3D) silicone samples using digital microscope.

4
A. Zühlke et al. Journal of the Mechanical Behavior of Biomedical Materials 122 (2021) 104649

Fig. 5. Surface characterization and roughness determination (Ra) of moulded (SilM) and 3D printed (Sil3D) silicone samples using AFM.

3. Results and discussion

3.1. Thermal expansion of silicones

Thermal expansion of silicone samples was measured under constant


force of 0.3 ​ N and heating rate of 1 ​ K/min. Data of the expansion (in
μm) were normalized to the initial height of the samples and converted
into true logarithmic strain, expressed in Fig. 3 in micro-strains vs. ab­
solute temperature (K).
It was observed that the expansion of 3D printed silicone samples is
not monotonic, showing jumps in expansion at certain temperatures.
This was seen to originate from a non-uniform surface structure of the
3D sample which wavy surface features might collapse or expand upon
loading and heating (for 3D sample, a rapid increase was observed soon
after heating eventually due to shape irregularities, hence it does no
start from zero strain). Nevertheless, polynomial approximation of the
expansion and its derivation has given coefficient of thermal expansion
(CTE) values (μ/K) as CTESilM ​ = ​ 6.22 ​ T - 1661.45; and Fig. 6. Example of the creep compliance (ratio of strain to stress, 1/kPa) of
CTESil3D ​ = ​ 15.96 ​ T - 4832.45 in the range of 20–45 ◦ C (293–318 ​ K). SilM specimens vs. experiment time (sec) on log-log scales. Numbers in legend
For temperature of 37 ◦ C (310 ​ K) as being of a clinical interest, these are applied creep stresses, kPa.
CTE values are 266.75 μ/K and 115.15 μ/K for moulded and 3D printed
specimens respectively. Literature data (Migwi et al., 1994) reported difference might have an implication for the medical device design and
CTE for silicone at 37 ◦ C of 273.5 μ/K - rather close to moulded material processing, when the exact dimensions at physiological range are
despite values (Migwi et al., 1994) were measured with a different important.
method. Lower CTE values for 3D printed material are considered
originating from a non-uniformity of the surface and the structure, as
well as air bubbles, residing inside the material during its 3.2. Creep data
manufacturing.
The probe of the DMA machine is highly sensitive (sub-nm resolu­ Creep tests were carried out at 1–60 ​ kPa pressure in compression
tion), therefore a good surface quality of the sample is required for mode at 37 ◦ C. This pressure range cover most physiological and over-
obtaining consistent data in compression mode. There are differences in physiological conditions, from a very weak contact to extreme
surface topology between the 3D printed and moulded samples. The compression. If an implant exerts on surrounding soft tissues pressure of
surface of 3D printed silicone is “wavy”, with some leftover material 5–10 ​ kPa (37–75 ​ mmHg), this still would allow hemocirculation, but
appearing as semi-random protrusions shown in Fig. 4. These “tails” long-lasting pressures over >20 ​ kPa (>150 ​ mmHg) would ultimately
confused the sensor probe apparently reaching the surface, when in fact lead to tissue necrosis and hypoxia following severe complications
it only had a tiny contact with a very little material. This has caused a leading to the removal of the implant. Higher pressures however might
major uncertainty when trying to measure Sil3D samples mechanical be faced for short times during the implant placement and its setting, so
properties in compression. The digital microscopy composite images of it is reasonable to cover the wider pressure range for materials
the surfaces of moulded and 3D printed surfaces (Fig. 4) clearly show the evaluation.
presence of the “waves” due to the printing process, which also signif­ The creep compliance data, defined as the ratio of true strain to
icantly increases their roughness (Ra) to 160-120 ​ nm vs. 20–30 ​ nm for applied stress are shown in Fig. 6 for applied creep stresses. It is seen that
moulded samples (Fig. 5). creep compliance clearly depends not only on experiment time, but also
Whereas CTE data for 3D printed material are more scattered, it is on the level of applied stress. An interesting experimental observation is
evident that material features coming from the 3D printing process lead that the creep compliance level is not proportional to the applied stress
to at least twice lower CTE than its moulded (non-porous) analogue. This (Fig. 6), indicating that materials behaviour is non-linear.

5
A. Zühlke et al. Journal of the Mechanical Behavior of Biomedical Materials 122 (2021) 104649

propagation (elastic, poroelastic or viscoelastic) (DeAlmeida and Fer­


reira, 2012; Hanyga and Seredynska, 2003), however, such waves are
not present in pseudo-static experiments here.
Experimental memory values (time-invariant) are shown in Fig. 7 for
both materials as function of applied stress. It is seen that higher
compressive stresses results in more short-time memory (the materials
adapt to the stimulus faster), but in low stress range moulded silicone
exhibit more long-time memory (revealing its more viscous nature) with
a steeper dependence from the applied stress. Low memory values for
Sil3D might be also due to presence of structural non-uniformity, uneven
surface, and trapped air bubbles.
With these experimental memory values, data of Fig. 6 and Eq. (3)
the values of invariant modulus and viscosity have been obtained, Fig. 8.
It is seen that moulded material is stiffer and more viscous than 3D
printed one. Both silicones in compression also exhibit substantial
stiffening with increased loading.
Combing all the above for the creep test, the constitutive equation for
calculation of the expected behaviour of silicones as function of stress
Fig. 7. Memory values (α) in creep tests for both silicone materials. Vertical and time from (3) could be presented as:
bars are standard error. ( ) ( t )α(σ)
σ
εcreep t, σ = (4)
Γ(1 + α(α))⋅E0 (α) τ(α)
For this loading case (creep), the theoretical idempotent solution
(Gasik and Bilotsky, 2019; Gunawardena, 1996) of the behaviour of the
where the parameters for α(σ), E0(σ) and characteristic time τ0(σ) are
materials predicts the power-law formal dependence of strain vs. time
shown in Table 1. These functions are obtained from experiment without
for every fixed value of applied stress (Gasik et al., 2018, 2021; Magin,
considering a material model or its linearity, and they can be directly
2010; Nelson, 1998):
used for the prediction of these silicone materials behaviour in creep
∫t conditions.
εcreep (t) 1 dτ tα
= α = , (3)
σstar Γ(α) Cα (t − τ)1− Γ(1 + α)E01− α ηα0
0
3.3. Three-point bending tests
where σstat is the applied creep (static) stress, Cα is the viscostiffness
(quasi-property in units of kPa⋅sα) (Gasik and Bilotsky, 2019), Γ(⋅) - In dynamic bending mode the behaviour of a polymer is different
gamma-function, α – material memory, E0 – intrinsic stiffness, η0 – from static or creep loading. The results taken from the DMA outputs at
intrinsic viscosity. The last two values are invariant in a sense that they scans from 0.1 to 20 ​ Hz under 20 ​ μm deformation amplitude, are
do not depend on the applied stress or time of experiment (Gasik et al., shown in Fig. 9. This deformation amplitude corresponds to the physi­
2018, 2021; Gasik and Bilotsky, 2019; Nelson, 1998). Equation (3) can ological ranges (Vrana et al., 2020; Gasik et al., 2018; van Mow and
be always numerically explicitly computed without need of assumptions Huiskes, 2005), taking into account proper pressure intervals. Ampli­
of linearity of Cα (α, t) and α(t). tudes of complex elastic modulus by DMA seem to be nearly
It is noteworthy that the rhs of (3) in practice is calculated with
convolution integration in time steps, and not by log-log fitting of the Table 1
data (“power law”; (Bagley and Torvik, 1983; Rodríguez et al., 2018; Parameters of the constitutive creep Eq. (4) for silicones. Invariant (steady)
Coussot et al., 2009; Magin, 2010)). The values of memory parameter in viscosity η0(σ) can be assessed as a product E0(σ)⋅τ0(σ) where stress variable (σ)
equation (3) must be non-negative to ensure causality principle (Hanyga is expressed in kPa.
and Seredynska, 2003; Nelson, 1998; Gasik et al., 2021), and in the Function SilM Sil3D
range 0 ​ < ​ α ​ < ​ 1 they present the fading memory (zero means only
Memory value α(σ) 0.2382⋅σ-1.071 0.0179⋅σ-0.318
short-time memory, i.e. ideally elastic behaviour, whereas unity means Stiffness E0(σ), kPa 1391 ​ + ​ 17.12⋅σ 98.21 ​ + ​ 8.845⋅σ
long-time memory i.e. ideally viscous behaviour). For the range Characteristic time τ0(σ), sec 71.92⋅σ-0.154 104.26⋅σ-0.205
1 ​ < ​ α ​ < ​ 2 the memory values are covering inertial effects and waves

Fig. 8. Invariant stiffness (a) and viscosity (b) in compression for two silicone materials as function of applied stress. Vertical bars are standard errors.

6
A. Zühlke et al. Journal of the Mechanical Behavior of Biomedical Materials 122 (2021) 104649

Fig. 9. Amplitude of the complex elastic modulus (a) and the loss tangent (b) as obtained by DMA, without data post-processing.

viscoelasticity models) usually leads to wrong predictions, artefacts or


Table 2
just improper conclusions (Neumark, 1962b). Authors (Ewoldt et al.,
Dynamic parameters of the constitutive equation (5) for silicone samples
2015) have emphasized that presentation of oscillatory stress or strain
(bending mode, 0.1–20 ​ Hz).
response in real Fourier series is not at all equivalent to complex Fourier
Parameter SilM Sil3D transformation when only the first harmonic real and imaginary moduli
Dynamic invariant 8.03 ​ ± ​ 0.15 3.44 ​ ± ​ 0.04 are calculated (which itself automatically imposes linearity trans­
stiffness E0ω, MPa formation requirements).
Dynamic invariant 0.022 ​ ± ​ 0.004 0.0125 ​ ± ​ 0.0015
One may understand the meaning of the invariants relating the value
viscosity, η0ω,
MPa⋅s of (E0ω)1-α(ω) as elastic contribution to the dynamic stress/strain ampli­
Dynamic memory (0.06265 ​ ± ​ 0.005)⋅ω0.16 ±
(0.1054 ​ ± ​ 0.010)⋅ω0.483 ± tudes ratio, and (η0ωω)α(ω) - as viscous contribution to the whole
value α(ω) 0.01 0.005
stiffness:
σ dyn
= E01−ω α(ω) ⋅(η0ω ⋅ω)α(ω) (6)
independent on frequency, but there are substantial differences in loss εdyn
tangent (Fig. 9, b). At low frequencies Sil3D material does not generate
The smaller the values of η0ω and α(ω) are, the less dynamic stress/
stable DMA readouts, possibly due to specimens’ uneven surface and
strain ratio (dynamic stiffness) would depend of frequency. There
instability in the sample holder.
intrinsic stiffness E0ω and intrinsic viscosity η0ω themselves do not
Application of the lhs integral of (3) to harmonic stress input (Gasik
depend on frequency or time (being invariants), and they are not com­
and Bilotsky, 2019; Bagley and Torvik, 1983; Rodríguez et al., 2018;
plex numbers.
Coussot et al., 2009) results in the non-linear equation for dynamic
It is noteworthy that this analysis does not have a simple decompo­
strain as function of time, frequency, and stress (without postulating any
sition of (6) into “elastic” and “viscous” parts contribution, alike in
models of elastic behaviour of the material):
simple ‘spring-dashpot’ models because of general non-linearity of the
viscoelasticity of the system. It is also notable that these elastic (E0ω)1-
⎛ ⎞ ( )
∫t ( ) σdyn sin ωτ + πα2(ω) α(ω)
εdyn ⎝t, ω, σ dyn ⎠ =
1 σ dyn sin ωτ d τ
= 1− α(ω) (5) and viscous (η0ωω)α(ω) contributions cannot be simply related to
Γ(α) Cω (t − τ)1− α E0ω (η0ω ω)α(ω) storage E’ = Re(E*) and loss E” = Im(E*) moduli usually calculated in
0
linear viscoelasticity obtained with the Fourier transform, since the
where σ dyn is the applied dynamic stress amplitude, ω - circular fre­ presence of a material memory (even when not being explicitly
quency, Cω is the dynamic viscous stiffness (quasi-property in units of frequency-dependent) makes the product (6) being non-linear.
kPa⋅sα) (Gasik and Bilotsky, 2019), α – dynamic material memory The benefits of using data as in Tables 1 and 2 comprise simpler,
parameter, E0ω – intrinsic dynamic stiffness, η0ω – intrinsic dynamic faster, and more reliable biomechanical estimation of personalized
viscosity. This equation comprises time-convolution of the specimen implant quality, minimizing the risks of complications. It was reported
loading history without a need to involve complex algebra (i.e. without that already during prolonged intubation and pressure, the tracheal
using complex numbers), assumptions of a material model (Maxwell, mucosa becomes ischemic which increases the likelihood of tracheal
Burger, standard linear solid, Prony series, etc.) or use of local differ­ wall damage and subsequent stenosis (Nesek-Adam et al., 2010). These
entiation (Gasik and Bilotsky, 2019; Gasik et al., 2021; Zühlke et al., pathological changes may eventually lead to fibrosis of circumferential
2021). In the case the stress input in (5) is not harmonic, the integral part lesions (Alshammari and Monnier, 2012), resulting in progressive
will have different appearance, but the parameters will keep their tracheal stenosis, mucosal ulceration and cartilage destruction, with
meanings. The integration limit in (5) covers separate periods with injury occurring even within minutes after ballooning of the cuff and
constant frequency, where the same harmonic stimulus is applied. subsequent fibrotic changes within the following 3–6 weeks (Ashfaque
The material parameters in (5) have the same meaning as for the and Annju, 2018) – despite all cautions, glottic stenosis continues to
creep tests in (4), but their values are now different, Table 2, and can be occur at a high frequency of 6–21% in intensive care units.
used for prediction of the bending stiffness of these silicones in marked This critical compressive pressure was mentioned to be about
frequency range. These values do not depend on time, frequency, or 30–35 ​ mmHg (4–5 ​ kPa) (Nesek-Adam et al., 2010; Alshammari and
applied force, except for the memory value α(ω), which frequency Monnier, 2012), however, the original study (Knowlson and Basset,
dependence brings major non-linearity in the materials behaviour. If 1970) noted that clinically accepted capillary flow pressure of
α(ω) ​ = ​ const, the material behaviour is linear and can be equally well 32 ​ mmHg needed to be complemented with pressure resistance of the
described in frequency domain with a Fourier transform as in conven­ surrounding tissues, which was estimated to span of 30–60 ​ mmHg extra
tional linear viscoelasticity. For the opposite case, application of com­ values, resulting in total of 60–95 ​ mmHg (8–13 ​ kPa). Taking ≈10 ​ kPa
plex math transformation (such as in traditional rheology or of extra stress as an average value, one can estimate with Tables 1 and 2

7
A. Zühlke et al. Journal of the Mechanical Behavior of Biomedical Materials 122 (2021) 104649

Table 3 for analogous moulded part. Due to this deformation, the exerted stress
Expected static and dynamic (1 ​ Hz) properties for silicone samples under might drop faster and therefore the primary stability of the implant
10 ​ kPa stress. might be compromised even if biomechanical risk of stenosis would be
Parameter estimations SilM Sil3D minimized in this case. Dynamic stiffness at 1 ​ Hz (i.e. closer to physi­
Creep memory value α 0.020 0.009
ological pulse frequency) in these conditions for moulded silicone would
Creep stiffness E0, MPa 1.56 0.186 be about 3.8 times higher than for 3D printed one, meaning that the
Characteristic time τ0, sec 50.45 65.03 deflection of moulded part would be smaller and the exerted pressure to
Strain at creep at 10 ​ kPa after 5 hours 0.007 0.054 soft tissues would be respectively higher if the deformation of the
Dynamic stiffness E0ω, MPa 8.03 3.44
implant is constrained by some means.
Dynamic invariant viscosity, η0ω, MPa⋅s 0.022 0.0125
Dynamic memory value αω 0.084 0.256 Pseudo-static (creep) and dynamic (frequency scan) analyses were
Dynamic stiffness, MPa 6.80 1.79 made with DMA, but the readouts were also additionally processed with
BEST (Vrana et al., 2020; Gasik and Bilotsky, 2019), which does not use
complex math, artificial models neither Fourier transforms. This leads to
data how much the static and dynamic deformation of the silicone stent a minimal set of parameters describing the mechanical process in static
would be feasible for a selected design, size, geometry and the status of and dynamics. As known, there are tangible practical and experimental
the patient (Table 3). costs for adding extra variables and parameters to the constitutive
It is seen that expected static (creep) deformation of 3D printed sil­ equation for describing the underlaying processes (Magin, 2010). When
icones after 5 ​ h post-implantation would be about 8 times higher than the method uses less parameters, and when most of the parameters are

Fig. 10. Indirect cytotoxicity results for moulded (SilM) and 3D printed silicone (Sil3D), with PET as a positive control (~100% viability) and sodium azide imbibed
silicone as a negative control (~10% viability). Corresponding brightfield images of the cells after exposure.

Fig. 11. Direct contact cytotoxicity tests for moulded (SilM) and 3D printed silicone (Sil3D) with PET as positive and sodium azide imbibed silicone as negative
control. Corresponding brightfield microscopy images (10x).

8
A. Zühlke et al. Journal of the Mechanical Behavior of Biomedical Materials 122 (2021) 104649

Fig. 12. Paravertebral subcutaneous implantation of moulded and 3D printed silicones. A) The implantation sites, the histological scoring and scoring criteria. B)
Histological sections after 2 weeks of implantation with haemotoxylin and eosin and Masson Trichrome staining.

invariants, it provides simpler, more concise and more consistent pre­ can be extended for the cases dominated by inertial or wave-propagation
diction for the biomaterial system without unnecessary complication of conditions (1 ​ < ​ α ​ < ​ 2), when linear approach becomes invalid (such
the calculations or ersatz-models. The appearance of Eqs. (3)–(6) for the as with negative loss tangent and strain energy density).
case of the testing modalities is close to known fractional viscoelasticity Whereas biomechanical properties and the surface of the materials
(Bagley and Torvik, 1983; Rodríguez et al., 2018; Coussot et al., 2009; show substantial differences, cytotoxicity of 3D-printed and moulded
Magin, 2010). However, here it was shown that the memory values α are samples does not alert about cell toxicity by direct contact since the cell
not constants and therefore traditional methods of fractional calculus viability is respectively 80% and 95% compared to the TCPS control. In
(Bagley and Torvik, 1983; Magin, 2010) or linear viscoelasticity (Neu­ addition, there is no morphological change in the cells which is another
mark, 1962b; Ewoldt et al., 2015) cannot be directly used. Biome­ criterion for confirming that all materials are not cytotoxic (Fig. 10). In
chanical data obtained with new method (Gasik and Bilotsky, 2019) do indirect contact, moulded silicone shows no cytotoxicity since full
not impose restrictions on the linearity of the system or material, and viability (100%) is maintained compared to the control, without cells

Fig. 13. In vivo systemic cytokine levels in mice at D1, D7 and D14 of implantation of moulded and 3D printed silicone samples compared to baseline read-outs.

9
A. Zühlke et al. Journal of the Mechanical Behavior of Biomedical Materials 122 (2021) 104649

morphological change, with no clear effect by cell density populations Author statement
which one could potentially able to determine (Volfson et al., 2008).
There was a slightly higher viability results compared to 3D printed As the corresponding author I confirm on behalf of all the authors,
samples in this configuration but the viability with 3D printed samples that we have read and approved the revisions to this manuscript.
was still above 70% (the cytotoxicity threshold), Fig. 11.
After 2 weeks of implantation, the histology shows good healing
around both moulded and 3D printed silicone. The histological scoring Declaration of competing interest
resulted on a zero score for the 3D printed silicone (good healing, within
the scale of 0–3; where 3 corresponds to chronic inflammation), Fig. 12. The authors declare that they have no known competing financial
The moulded samples, on the other hand had a score of 0.4; but still interests or personal relationships that could have appeared to influence
within good healing index. The quantification of systemic cytokine the work reported in this paper.
levels showed a slight increase in the pro-inflammatory cytokines after
implantation at similar levels for both silicones. In most cases the dif­ Acknowledgments
ferences were insignificant; except on D1 moulded silicone samples
shown a higher TNF-α, MCP-1 and IL-27 levels - all pro-inflammatory This project has received funding from the European Union’s Hori­
markers (Fig. 13). However, over the course of 2 weeks (D14), the zon 2020 research and innovation programme under grant agreement
levels were stable and cannot be considered to induce a systemic high No 760921 (“PANBioRA”; A.Z., M.G., N.E.V., J.B.) and from the “FUI-
level of inflammation where the read-outs <100 ​ pg/ml range for any of FASSIL” project (J.B., C.B.M., N.E.V.)
the cytokines.
References
4. Conclusions
Alshammari, J., Monnier, P., 2012. Airway stenting with the LT-Mold™ for severe glotto-
subglottic stenosis or intractable aspiration: experience in 65 cases. Eur. Arch. Oto-
In this work biomechanical properties of moulded and 3D printed
Rhino-Laryngol. 269, 2531–2538.
silicones in statics (creep) and dynamics have been measured and Ashfaque, A., Annju, T., 2018. Multimodality surgical approach in management of
quantified for the ranges of boundary and loading conditions which are laryngotracheal stenosis. Case Rep. Otolaryngol., 4583726
Bagley, R.L., Torvik, P.J., 1983. A theoretical basis for the application of fractional
physiologically relevant for implants. It was observed that 3D material
calculus to viscoelasticity. J. Rheol. 27, 201–210.
has very distinct differences in structure and surface quality which are Chung, C., Burdick, J.A., 2008. Engineering cartilage tissue. Adv. Drug Deliv. Rev. 60,
immediately reflected in mechanical properties. On the contrary, cyto­ 243–262.
toxicity tests carried out have shown that processing method (moulding Cooper, J.D., 2018. Use of silicone tubes in the management of complex airway
problems. Thorac. Surg. Clin. 28, 441–447.
and 3D printing) does not have an effect on the cytotoxicity of silicones, Coussot, C., Kalyanam, S., Yapp, R., Insana, M.F., 2009. Fractional derivative models for
and the differences in topology do not effectively impact on the cell ultrasonic characterization of polymer and breast tissue viscoelasticity. IEEE Trans.
behaviour. Thus, biomechanical properties might be the decisive factor Ultrason. Ferroelectrics Freq. Contr. 56, 715–726.
DeAlmeida, M.F., Ferreira, L.C.F., 2012. Self-similarity, symmetries and asymptotic
for preference of certain implant types, designs and geometry, depend­ behavior in Morrey spaces for a fractional wave equation. Differ. Integr. Equ. 25,
ing on their intended purpose and patient-specific requirements. 957–976.
Moulded silicone has nearly twice higher thermal expansion coeffi­ Debry, C., Vrana, N.E., Dupret-Bories, A., 2017. Implantation of an artificial larynx after
total laryngectomy. N. Engl. J. Med. 376, 97–98.
cient vs. 3D printed so for a precise fitting this is needed to be considered DiMarzio, N., Eglin, D., Serra, T., Moroni, L., 2020. Bio-fabrication: convergence of 3D
when designing a personalized stent. These differences might also have bioprinting and nanobiomaterials in tissue engineering and regenerative medicine.
implications for implants with a complex shape or fine geometric fea­ Front. Bioeng. Biotechnol. 8, 326.
Emre, U., Zerrin, B., Gurkan, K., Ahmet, K., Haluk, O., 2005. Soft tissue response of the
tures, as uneven thermal expansion might lead to peaks of stress con­
larynx to silicone, Gore-Tex, and irradiated cartilage Implants. Laryngoscope 115,
centration which could become potential weak spots for failures. 1009–1014.
Obtained experimental biomechanical data of moulded and 3D Ewoldt, R.H., Johnston, M.T., Caretta, L.M., 2015. Experimental challenges of shear
rheology: how to avoid bad data. In: Spagnolie, S. (Ed.), Complex Fluids in Biological
printed silicones have demonstrated substantial differences which
Systems. Springer.
needed to be taken into account for design and manufacturing of precise Gasik, M., 2017. Understanding biomaterial-tissue interface quality: combined in vitro
medical implants. Model-free data for invariant stiffness, viscosity and evaluation. Sci. Techn. Adv. Mater. 18, 550–562.
material memory can be used for numerical modelling and optimization Gasik, M., Bilotsky, Y., 2019. In vitro method for measurement and model-free evaluation
of time-invariant biomaterials functions. US Patent, 10379106 B2.
of these silicone devices as they enable much better prediction for set Gasik, M., Zühlke, A., Haaparanta, A.M., Muhonen, V., Laine, K., Bilotsky, Y.,
stress and frequency values. Kellomäki, M., Kiviranta, I., 2018. The importance of controlled mismatch of
biomechanical compliances of implantable scaffolds and native tissue for articular
cartilage regeneration. Front. Bioeng. Biotechnol. 6, 187.
Authors contributions Gasik, M., Ivanov, R., Kazantseva, J., Bilotsky, Y., Hussainova, I., 2021. Biomechanical
features of graphene-augmented inorganic nanofibrous scaffolds and their physical
A.Z. has made the experimental DMA testing, analysed DMA read­ interaction with viruses. Materials 14, 164, 1.
Gomes, M.E., Reis, R.L., 2004. Biodegradable polymers and composites in biomedical
outs and made manuscript figures as well as revise the final version. M. applications: from catgut to tissue engineering. Int. Mater. Rev. 49, 261–273.
G. has composed the manuscript, arranged the text, and checked the Guilak, F., Butler, D.L., Goldstein, S.A., Baijens, F.P.T., 2014. Biomechanics and
numerical outcomes. E.C., C.M. optimized 3D printing process, prepared mechanobiology in functional tissue engineering. J. Biomech. 47, 1933–1940.
Gunawardena, J., 1996. An Introduction to Idempotency. HP Laboratories Bristol,
and produced 3D-printed silicone samples. N.E.V., J.B., C.B.M. carried
Publication HPL-BRIMS-96-24, p. 50.
out AFM analysis and studied cell reactions on the samples. Y.B. has Hanyga, A., Seredynska, M., 2003. Power-law attenuation in acoustic and isotropic
processed the readouts with BEST and wrote the method explanation. anelastic media. Geophys. J. Int. 155, 830–838.
Ingber, D.E., 2006. Cellular mechanotransduction: putting all the pieces together again.
Faseb. J. 20, 811–827.
Disclosures Jaalouk, D.E., Lammerding, J., 2009. Mechanotransduction gone awry. Nat. Rev. Mol.
Cell Biol. 10, 63–73.
Y.B. and M.G. are shareholders of Seqvera Ltd. N.E.V. is a share­ Jammalamadaka, U., Tappa, K., 2018. Recent advances in biomaterials for 3D printing
and tissue engineering. J. Funct. Biomater. 9, 22.
holder of Spartha Medical SAS. These ownerships did not have any Kacarevic, Ž.P., Rider, P.M., Alkildani, S., Retnasingh, S., Smeets, R., Jung, O.,
implication on the data or the interpretation of the results. Ivaniševic, Z., Barbeck, M., 2018. An introduction to 3D bioprinting: possibilities,
challenges and future aspects. Materials 11, 2199.
Kazantseva, J., Ivanov, R., Gasik, M., Neuman, T., Hussainova, I., 2016. Graphene-
augmented nanofiber scaffolds demonstrate new features in cells behaviour. Sci.
Rep. 6, 30150.

10
A. Zühlke et al. Journal of the Mechanical Behavior of Biomedical Materials 122 (2021) 104649

Knowlson, G.T.G., Basset, H.F.M., 1970. The pressures exerted on the trachea by Noppen, M., Stratakos, G., D’Haese, J., Meysman, M., Vicken, W., 2005. Removal of
endotracheal inflatable cuffs. Br. J. Anesth. 42, 834–837. covered self-expandable metallic airway stents in benign disorders. Chest 127,
Lord, J.D., Morrell, R., 2006. Elastic Modulus Measurement: Measurement Good Practice 938–944.
Guide No. 98. NPL Teddington, UK, p. 100. Norris, A., 2008. Eulerian conjugate stress and strain. J. Mech. Mater. Struct. 3, 243–260.
Lubarda, V.A., Chen, M.C., 2008. On the elastic moduli and compliances of transversely Orr, A.W., Helmke, B.P., Blackman, B.R., Schwartz, M.A., 2006. Mechanisms of
isotropic and orthotropic materials. J. Mech. Mater. Struct. 3, 155–170. mechanotransduction. Dev. Cell 10, 11–20.
Magin, R.L., 2010. Fractional calculus models of complex dynamics in biological tissues. Rodríguez, R.F., Salinas-Rodríguez, E., Fujioka, J., 2018. Fractional time fluctuations in
Comput. Math. Appl. 59, 1586–1593. viscoelasticity: a comparative study of correlations and elastic moduli. Entropy 20,
Maslov, V., 1970. The characteristics of pseudo-differential operators and difference 28.
schemes. Actes Congrès Int. Math. 2, 755–769. Sittel, C., 2009. Larynx: implants and stents. Laryngo-Rhino-Otol. 88, S119–S124.
Migwi, C.M., Darby, M.I., Wostenholm, G.H., Yates, B., Duffy, R., Moss, M., 1994. Trabelsi, O., Prez del Palomar, A., Mena Tobar, A., Lopez-Villalobos, J.L., Ginel, A.,
A method of determining the shear modulus and Poisson’s ratio of polymer Doblare, M., 2011. FE simulation of human trachea swallowing movement before
materials. J. Mater. Sci. 29, 3430–3432. and after the implantation of endoprosthesis. Appl. Math. Model. 35, 4902–4912.
Nagarajan, N., Dupret-Bories, A., Karabulut, E., Zorlutuna, P., Vrana, N.E., 2018. van der Meulen, M.C.H., Huiskes, R., 2002. Why mechanobiology? A survey article.
Enabling personalized implant and controllable biosystem development through 3D J. Biomech. 35, 401–414.
printing. Biotechnol. Adv. 36, 521–533. van Mow, C., Huiskes, R., 2005. Basic Orthopedic Biomechanics and Mechanobiology,
Nelson, L.S., 1998. The Anderson-Darling test for normality. J. Qual. Technol. 30, third ed. Lippincott, Williams & Wilkins, USA, p. 720.
298–299. Volfson, D., Cookson, S., Hasty, J., Tsimring, T.S., 2008. Biomechanical ordering of dense
Nesek-Adam, V., Mršić, V., Oberhofer, D., Grizelj-Stojčić, E., Košuta, D., Rašić, Ž., 2010. cell populations. Proc. Natl. Acad. Sci. Unit. States Am. 105, 15346–15351.
Post-intubation long-segment tracheal stenosis of the posterior wall: a case report Biomaterials for organ and tissue regeneration: new technologies and future prospects.
and review of the literature. J. Anesth. 24, 621–625. In: Vrana, N.E., Knopf-Marques, H., Barthes, J. (Eds.), 2020. Woodhead Publ., UK,
Neumark, S., 1962a. Concept of Complex Stiffness Applied to Problems of Oscillations with p. 828.
Viscous and Hysteretic Damping. A.R.C. Reports and Memoranda No. 3269. Ministry of Wong, J.Y., Bronzino, J.D., 2007. Biomaterials. CRC Press/Taylor & Francis, p. 296.
Aviation, London, UK, p. 36. Xiao, H., 1995. Invariant characteristic representations for classical and micropolar
Neumark, S., 1962b. Concept of Complex Stiffness Applied to Problems of Oscillations anisotropic elasticity tensors. J. Elasticity 40, 239–265.
with Viscous and Hysteretic Damping. Aeronaut. Res. Council Rep. 3269. Ministry of Zühlke, A., Gasik, M., Shahramian, K., Närhi, T., Bilotsky, Y., Kangasniemi, I., 2021.
Aviation, London, UK, p. 36. Enhancement of gingival tissue adherence of zirconia implant posts: in vitro study.
Materials 14, 455, 2.

11

You might also like