Communication

Bone Flexoelectricity www.advmat.de

Flexoelectricity in Bones
Fabian Vasquez-Sancho,* Amir Abdollahi, Dragan Damjanovic, and Gustau Catalan*

accumulation) has been observed near

Bones generate electricity under pressure, and this electromechanical
behavior is thought to be essential for bone’s self-repair and remodeling
properties. The origin of this response is attributed to the piezoelectricity of
collagen, which is the main structural protein of bones. In theory, however,
any material can also generate voltages in response to strain gradients,
thanks to the property known as flexoelectricity. In this work, the flexoelec-
tricity of bone and pure bone mineral (hydroxyapatite) are measured and
found to be of the same order of magnitude; the quantitative similarity
suggests that hydroxyapatite flexoelectricity is the main source of bending-
induced polarization in cortical bone. In addition, the measured flexoelectric
coefficients are used to calculate the (flexo)electric fields generated by cracks
in bone mineral. The results indicate that crack-generated flexoelectricity is
theoretically large enough to induce osteocyte apoptosis and thus initiate
the crack-healing process, suggesting a central role of flexoelectricity in bone
repair and remodeling.
All animals—including of course humans—require electricity
to perform functions as basic as muscle contraction or nervous
impulse sensing and transmission. In the case of vertebrates,
bones are also known to generate electricity,[1,2] and this is con-
sidered essential for bone regeneration.[3,4] One way to generate
electricity is through piezoelectricity, which in bones can be pro-
vided by collagen.[2,5] In addition, ionic streaming potentials[6]
also contribute to the electromechanical properties of wet bones.
Intriguingly, however, bone-repair functionality (osteoblast
F. Vasquez-Sancho, Prof. G. Catalan
Institut Català de Nanociencia i Nanotecnologia (ICN2)
CSIC and The Barcelona Institute of Science and Technology (BIST)
Campus UAB, Bellaterra, 08193 Barcelona, Catalonia, Spain
E-mail: fabian.vasquez@icn2.cat; gustau.catalan@icn2.cat
F. Vasquez-Sancho
Centro de Investigación en Ciencia e Ingeniería de Materiales (CICIMA)
Universidad de Costa Rica
San José 11501, Costa Rica
Dr. A. Abdollahi
Laboratori de Càlcul Numèric (LaCàN)
Universitat Politècnica de Catalunya (UPC)
Campus Nord UPC-C2, E-08034 Barcelona, Spain
Prof. D. Damjanovic
Ecole Politechnique Federale de Lausanne (EPFL)
Lausanne CH-1015, Switzerland
Prof. G. Catalan
Institució Catalana de Recerca i Estudis Avançats (ICREA)
Pg. Lluís Companys 23
E-08010, Barcelona, Catalonia, Spain
The ORCID identification number(s) for the author(s) of this article
can be found under https://doi.org/10.1002/adma.201705316.
DOI: 10.1002/adma.201705316
cracks at the surface of pure hydroxyapa-
tite ceramics, where there is neither col-
lagen nor streaming currents.[7] This result
indicates that hydroxyapatite itself can also
generate signals for the repairing cells. The
nature and origin of such signals, how-
ever, is not known, and is one of the most
intriguing and enduring problems in the
field of osteogenesis.[8–10]
One potential explanation is bone min-
eral piezoelectricity. Another possible cul-
prit is flexoelectricity, which is a coupling
between strain gradients and polarization
allowed by symmetry in all materials,
including non-piezoelectric ones.[11,12]
Early studies suggested that hydroxyapa-
tite is centrosymmetric and therefore not
piezoelectric,[13] although more recent
structural refinements[14] suggest that it
might be. Meanwhile, functional measure-
ments are ambiguous. Thin films yield substantial piezoelectric
coefficients,[15] but thin films can easily become polarized by
built-in fields, strain gradients, or defects.[16] Bulk ceramics,
meanwhile, sometimes yield a small piezoelectricity[17] and
sometimes no piezoelectricity at all.[2] These variations prob-
ably reflect differences in sample composition or morphology,
making it difficult to make definite statements about intrinsic
properties. Our own hydroxyapatite ceramic and commer-
cially acquired ceramics from Berkeley Advanced Biomate-
rials, Inc., were measured by a direct load method,[16] yielding
piezoelectric coefficients smaller than 0.001 pC N−1. This is at
least two orders of magnitude smaller than the piezoelectricity
of bone,[7] and is comparable to the residual (defect-induced)
piezoelectricity of SrTiO3, a reference nonpiezoelectric mate-
rial used for comparison (Figure S1, Supporting Information).
Our bone ceramics are therefore not significantly piezoelectric,
a result consistent with the lack of piezoelectricity in decolla-
genized bones.[2] Macroscopic measurements of course do not
rule out the existence of piezoelectricity on a microscopic level:
piezoelectric grains with different orientation can in theory
average out their aggregate contribution; however, piezore-
sponse force microscopy (Figures S2–S4, Supporting Informa-
tion) showed no phase contrast between grains. If we discard
piezoelectricity, however, how does hydroxyapatite direct the
activity of osteoblasts towards damaged regions?[9]
A plausible hypothesis is that bone mineral generates elec-
tromechanical signals due to flexoelectricity, which is a property
of all dielectric (and even semiconductor[18]) materials whereby
they polarize in response to an inhomogeneous deformation
such as bending.[19] The combination of built-in structural
flexibility and mechanical texture at the microscale—the scale
in which cells operate and build—is inherent to biological

Adv. Mater. 2018, 1705316 1705316 (1 of 5) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

Figure 1. Left: Schematic of the experimental setup for measuring flexoelecricity: a DMA is used to induce three-point bending, and a lock-in amplifier
collects the bending-induced current in bone and hydroxyapatite samples. Right: Strain gradients, and thus flexoelectricity, are biggest around small
structural singularities such as fracture fronts (crack tips) in bone mineral.

tissues, and constitutes an optimal environment for flexoelec- Bones and hydroxyapatite presented some variation from
tricity. For example, flexoelectricity has already been identified sample to sample. The dispersion of the flexoelectric coef-
in stereocillia (inner ear microhairs), as an important ingre- ficient for each material is presented as the shadowed area:
dient of mammalian hearing.[20] The highly textured and inho- red for hydroxyapatite and blue for bones. The effective flexo-
mogeneous structure of bones, with radial porosity gradients electric coefficients are between 0.2 and 2.3 nC m−1 for bone,
and curved walls, also lends itself to flexoelectric phenomena. and between 0.7 and 1.6 nC m−1 for hydroxyapatite. Collagen
Already in 1975 Williams and Breger[21] claimed that some elec- increases the mechanical toughness of bones, allowing them to
tromechanical properties of bones could perhaps be explained withstand bigger bending than brittle hydroxyapatite ceramics;
by “gradient polarization” or inhomogeneous piezoelectricity. but, for any given curvature, hydroxyapatite flexoelectricity is
Around the same time, Lakes also performed a theoretical comparable to the flexoelectricity of bones. Hydroxyapatite
analysis of the potential role of gradients in bones, which could flexoelectricity can by itself account for most of the bending-
not be substantiated due to lack of quantitative knowledge of induced polarization of bones without needing to invoke
their flexoelectric coefficients.[22] Later, Fu reported in a confer- collagen piezoelectricity.
ence the existence of bending-induced polarization in bones,[23] The next important question is: considering that bones already
wrongly attributing this flexoelectric-like response to collagen. generate electromechanical voltages from streaming potentials
Though these antecedents are few and scattered, together they and collagen piezoelectricity, what (if any) is the additional ben-
provide tantalizing evidence that there may be an important efit of having a flexoelectric contribution from bone mineral?
role for flexoelectricity in bones. The answer appears to be related to the multiscale functional
In this paper, we have quantified the flexoelectricity of
hydroxyapatite and its participation in the electromechanical
response of bones. The results indicate that most of the elec-
tromechanical response of a bone to bending comes from the
flexoelectricity of bone mineral rather than from collagen.
We have then used our measured flexoelectric coefficient of
hydroxyapatite to calculate the flexoelectricity generated by
cracks in bone mineral (see Figure 1). The calculated inten-
sity exceeds 5 kV m−1 within a perimeter of 40 µm around the
crack tip, and it therefore can provide a powerful electrical
signal from the center of damage to stimulate bone repair.
Fresh bovine femurs were cut in beams oriented parallel
to the bone axis and electroded for measuring flexoelectricity.
The same femurs were also ground to powder, calcined, and
sintered into ceramic pellets (see Experimental Section). We
used a dynamic mechanical analyzer to deliver an oscillatory
bending and a lock-in amplifier to detect the bending-induced
polarization (see Experimental Section). The bending-induced
polarization of bone, natural hydroxyapatite, and commer-
Figure 2. The flexoelectric coefficient is the constant of proportionality
cially acquired synthetic hydroxyapatite are shown in Figure 2.
between strain gradient (bending) and bending-induced polarization. For
The effective flexoelectric coefficients µeff are extracted from greater accuracy, measurements were made for different applied forces
the slopes of the linear fits of the polarization as a function of (which induced different curvatures).The shadowed areas represent the
bending (see Experimental Section). dispersion of the data for bones (blue) and hydroxyapatite (red).

Adv. Mater. 2018, 1705316 1705316 (2 of 5) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

architecture of bones. Strain gradients grow

in inverse proportion to feature size.[19,24] This
means that although at macroscopic scales the
average strain (and thus piezoelectricity) can
dictate the global response, at small scales the
strain gradient, and thus flexoelectricity, can
be much larger and dominate the local electro-
mechanical response.[23] A dramatic manifes-
tation of this principle takes place at the apex
of cracks, which concentrate in a very small
volume (a crack junction is atomically sharp)
the maximum stress that a material can with-
stand before rupture; according to theoretical
calculations, the flexoelectric polarization
near a crack apex can exceed the piezoelectric
polarization for even the best piezoelectric
materials.[25] In the context of bones, micro­
cracks are common flaws formed due to cycli-
cally applied stress, but they usually represent
no risk for the integrity of the bone thanks to
the process of remodeling.[11,26] As our calcu-
lations show, crack-generated flexoelectricity
is capable of triggering the process of damage Figure 3. Calculated flexoelectric field distribution around a microcrack in bone mineral. The
repair and remodeling. dashed line marks the region where the field is theoretically strong enough to be able to induce
osteocyte apoptosis.
The critical intensity factor KC, which
1
in
bones is in the order of 3 MPa m ; [27]
2
this
is the stress concentration at which cracks propagate through of the orientation of the collagen inside the bone; in this case, all the
bone. Using our measured flexoelectric coefficients, we have samples were longitudinal to the long axis of the bone. The samples
calculated the flexoelectric field (Figure 3) around a microcrack were polished up to 0.1 µm grain size disk with an Allied precision
polishing system at low velocity to minimize damage to the samples.
under critical load (see Experimental Section). The flexoelectric
Hydroxyapatite compact disks were commercially obtained from
field is biggest at the crack tip and decays progressively away, Clarkson Chromatography Products, INC., with certified purity greater
being bigger than 103 V m−1 up to a distance 50 µm around than 95%. Also hydroxyapatite was produced from bovine bones
the crack apex. These numbers are significant because pulsed following the procedure of Ooi et al.[31] The commercially acquired
electric fields of 5 kV m−1 are known to induce apoptosis in hydroxyapatite samples had a porosity between 8% and 13%, and
bone cells,[28] osteocyte apoptosis being the first step of bone natural hydroxyapatite samples that were prepared by grounding, firing
regeneration; when dead, osteocytes release chemical triggers and sintering bone had a porosity smaller than 7%. In order to do the
compact disks, the hydroxyapatite was milled and the powder was sieved
that signal the osteoclasts to initiate the repair by cleaning the to 125 µm particle size. Then, the powder was uniaxially pressed into
damaged region, followed by osteoblasts that segregate new pellets of 22.58 mm of diameter with 25 metric tons. Finally the pellets
bone mineral.[10,23] Electric fields also attract screening ions were air sintered at 1360 °C during 4 h. Samples were cut and polished
creating electrochemical gradients that assist osteogenesis,[29] using the same procedure as for the bones. Platinum electrodes were
thus further increasing the velocity of reparation of the dam- deposited using pulsed laser deposition. Special care was taken to
aged region.[30] have low resistance across the electrode (<100 Ω). The quality if the
electrodes were checked by measuring the capacitance and dielectric
Osteoblast tends to attach near by the tip of cracks in pure
[9]
losses, being 0.2 for the latest.
bone mineral, suggesting that osteoblasts do indeed detect a Polarization was induced by a DMA8000 dynamic mechanical
crack tip as the center of damage. Moreover, the apex is itself analyzer (DMA) of Perkin-Elmer and was measured using the method
a movable entity: as the crack is healed, its apex will recede, described by Zubko et al.[32] The DMA was used to apply a periodic
continually pointing to the osteoclasts and osteoblasts the new three-point bending stress at room temperature. This periodic signal
[10] was used as a reference for a lock-in amplifier, model 830 of Stanford
position of the region to repair. Flexoelectricity is strong Research Instruments, while the signal obtained from the electrodes fed
enough to act as the beacon in this process, suggesting a new the measurement channel of the lock-in amplifier, which recorded the
line of inquiry for tissue regeneration where stress-gradient bending-induced displacement currents. The current was converted into
I
engineering may be used as an additional degree of freedom in polarization using P = 2π vA , where v is the frequency of the bending
bone-forming prosthetic designs. force and A is the area of the electrodes. The polarization measured by
the lock-in is related to the effective flexoelectric coefficient µeff by
Experimental Section
∂ε 11 ∂ε 12z
Sample Preparation and Characterization: Freshly cut (less than 48 h P3 = µ 13
eff and 11 = 3 0 (L − a) (1)
from slaughter) bovine femurs were obtained from a butcher's shop and
∂x 3 ∂x 3 L
stored in a physiological solution. Pieces of compact bone (porosity
< 15%) were then cut using a diamond wire at low speed in order to where L is the separation between the standing points of the sample,
avoid damage to the tissue. The samples were cut in consideration a is the half-length of the electrodes, and z0 is the maximum vertical

Adv. Mater. 2018, 1705316 1705316 (3 of 5) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

∂ε 11 ∂ε
E 1 = f 11 + f 12 22 (6)
∂x 1 ∂x 1

∂ε 22 ∂ε
E 2 = f 22 + f 21 11 (7)
∂x 2 ∂x 2

E = E 12 + E 22 (8)

and fij is the flexocoupling tensor. The flexocoupling tensor was

calculated with the effective flexoelectric coefficient μeff and the dielectric
susceptibility χ

µ eff = f eff χ (9)

For this calculation, f11 = f22 = f12 = f21 = feff = 10 V and the shear
component was taken as null. We used 10 V as a general case since we
measured several samples and all the flexocoupling values ranged from
9.3 to 11.8 V. Same happened with the relative dielectric constant, all the
samples showed different values ranging from 10 to 15.
Figure 4. Flexoelectric coefficients were calculated as the slopes of the
linear fits between the curvature and the bending-induced polarization.
Because bones are more flexible than hydroxyapatite, it could withstand
much larger curvatures, but the slope was still almost the same as for
Supporting Information
pure hydroxyapatite. Inset: sketch of the measurement apparatus. Supporting Information is available from the Wiley Online Library or
from the author.
deflection in the middle of the sample. Typical values used in our
measurements were L = 12 mm, a = 2 mm, and z0 = 2 µm.
Measurements were taken after all samples had been dried in an Acknowledgements
oven at a temperature of 90 °C for 7 h. All the results presented in this
paper were obtained by measurements at 13 Hz, which in the range of This research was funded by an ERC Starting grant (ERC 308023).
frequencies that covers the characteristic timescales of the periodic loads All research in ICN2 was supported by the Severo Ochoa Excellence
in vivo.[33] We also measured for some samples at 1.3, 23, 31, 37 and Programme (SEV-2013-0295). F.V.-S. thanks MICITT, CONICIT and
43 Hz to ensure the results. Flexoelectric measurements are not affected Universidad de Costa Rica for support during his PhD. The authors
by variations in the frequency. Although lower frequencies are particularly thank D. Torres for the illustration in Figure 1, R. Nunez for advice and
relevant because it is the characteristic timescale of periodic mechanical F. Belarre for help with sample polishing.
loads such as walking, a compromise between noise (noise scales as f−1)
and relevant frequencies was made to obtain lower dispersion.
Data Analysis: From Equation (1), the effective flexoelectric coefficient
is defined as the relation between the polarization and the stress
Conflict of Interest
gradient. For more accuracy, several strain gradients were applied to The authors declare no conflict of interest.
each sample and the flexoelectric coefficient was extracted from the
slope of the plots of polarization as a function of strain gradient,[19] as
can be seen in Figure 4.
Calculation of Flexoelectric Field: From the equations of strain around Keywords
a crack mode 1[34]
bone remodeling, cracks, flexoelectricity, hydroxyapatite
1+ υ υ
ε ijel = σ ij − 3 σ mδ ij (2) Received: September 15, 2017
Y Y
Revised: October 17, 2017
where Y is the Young’s modulus, ε is the Poisson’s ratio, and σij is the Published online:
stress applied to the crack in each direction

σ 11 =
KI
2π r
θ
( θ 3θ
cos 1 − sin sin
2 2 2 ) (3) [1] E. Fukada, I. Yasuda, J. Phys. Soc. Japan 1957, 12, 1158.
[2] A. A. Marino, R. O. Becker, S. C. Sidney, Calcif. Tissue Res. 1971, 8, 177.

( ) (4)
KI θ θ 3θ [3] C. A. L. Bassett, R. O. Becker, Science 1962, 137, 1063.
σ 22 = cos 1+ sin sin [4] J. A. Spadaro, Bioelectromagnetics 1997, 18, 193.
2π r 2 2 2
[5] E. Fukada, I. Yasuda, Jpn. J. Appl. Phys. 1964, 3, 502B.
[6] D. Gross, W. S. Williamsi, J. Biomech. 1982, 15, 277.
KI θ θ 3θ (5)
τ 12 = cos sin cos [7] Y. Shu, M. J. Baumann, E. D. Case, R. K. Irwin, S. E. Meyer,
2π r 2 2 2
C. S. Pearson, L. R. McCabe, Mater. Sci. Eng., C. 2014, 44, 191.
1
− [8] O. Verborgt, G. J. Gibson, M. B. Schaffler, J. Bone Miner. Res. 2000,
where KI is the intensity factor taken as 3MPam 2. For the calculations
Y = 20 GPa was used as the Young’s modulus in the direction along 15, 60.
the bone and Y = 6 GPa was used for the direction perpendicular to the [9] D. Taylor, J. A. N. G. Hazenberg, T. C. Lee, Nat. Mater. 2007, 6, 263.
bone’s axis, both values obtained from our measurements. Equations [10] L. E. Lanyon, Calcif. Tissue Int. 1993, 53, 102.
were transformed to Cartesian coordinates in order to compute the [11] S. M. Kogan, Sov. Phys. Solid State 1964, 5, 2069.
flexoelectric field [12] E. Bursian, O. I. Zaikovskii, Sov. Phys. Solid State 1968, 10, 1121.

Adv. Mater. 2018, 1705316 1705316 (4 of 5) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com
[13] M. I. Kay, R. A. Young, Nature 1964, 204, 1050.
[14] D. Haverty, S. A. M. Tofail, K. T. Stanton, J. B. McMonagle,
Phys. Rev. B 2005, 71, 94103.
[15] S. B. Lang, S. A. M. Tofail, A. L. Kholkin, M. Wojtaś, M. Gregor,
A. A. Gandhi, Y. Wang, S. Bauer, M. Krause, A. Plecenik, Sci. Rep.
2013, 3, 2215.
[16] A. Biancoli, C. M. Fancher, J. L. Jones, D. Damjanovic, Nat. Mater.
2015, 14, 224.
[17] S. A. M. Tofail, D. Haverty, F. Cox, J. Erhart, P. Hána, V. Ryzhenko,
J. Appl. Phys. 2009, 105, 064103.
[18] J. Narvaez, F. Vasquez-Sancho, G. Catalan, Nature 2016, 538, 219.
[19] P. Zubko, G. Catalan, A. K. Tagantsev, Annu. Rev. Mater. Res. 2013,
43, 387.
[20] K. D. Breneman, W. E. Brownell, R. D. Rabbitt, PLoS One 2009, 4,
e5201.
[21] W. S. Williams, L. Breger, Biomechanics 1975, 8, 407.
[22] R. Lakes, IEEE Trans. Biomed. Eng. 1980, BME-27, 282.
Adv. Mater. 2018, 1705316 1705316 (5 of 5)
www.advmat.de
[23] Fu, John, APS Meeting Abstracts, T18013, 2010.
[24] R. Maranganti, N. D. Sharma, P. Sharma, Phys. Rev. B 2006, 74, 14110.
[25] A. Abdollahi, C. Peco, D. Millán, M. Arroyo, G. Catalan, I. Arias,
Phys. Rev. B 2015, 92, 94101.
[26] T. C. Lee, A. Staines, D. Taylor, J. Anat. 2002, 201, 437.
[27] T. Norman, D. Vashishtht, D. B. Burr, J. Biomech. 1995, 28, 309.
[28] X. Miao, S. Yin, Z. Shao, Y. Zhang, X. Chen, J. Orthop. Surg. Res.
2015, 10, 104.
[29] K. E. Hammerick, A. W. James, Z. Huang, F. B. Prinz,
M. T. Longaker, Tissue Eng., Part A 2010, 16, 917.
[30] A. H. Rajabi, M. Jaffe, T. L. Arinzeh, Acta Biomater. 2015, 24, 12.
[31] C. Y. Ooi, M. Hamdi, S. Ramesh, Ceram. Int. 2007, 33, 1171.
[32] P. Zubko, G. Catalan, A. Buckley, P. Welche, J. Scott, Phys. Rev. Lett.
2007, 99, 167601.
[33] M. B. Schaffler, D. B. Burr, J. Biomech. 1988, 21, 13.
[34] B. Lawn, Fracture of Brittle Solids, Cambridge University Press,
Cambridge 1993.
© 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim