The Elasticity of Nuclear Pasta
M. E. Caplan,1, ∗ A. S. Schneider,2 and C. J. Horowitz3, †
1McGill Space Institute, McGill University, 3600 Rue University, Montreal, QC Canada H3A 2T82TAPIR, Walter Burke Institute for Theoretical Physics,
California Institute of Technology, Pasadena, CA 91125, USA3Nuclear Theory Center, Indiana University, Bloomington, IN 47401, USA
(Dated: July 10, 2018)
The elastic properties of neutron star crusts are relevant for a variety of currently observable ornear-future electromagnetic and gravitational wave phenomena. These phenomena may depend onthe elastic properties of nuclear pasta found in the inner crust. We present large scale classicalmolecular dynamics simulations where we deform nuclear pasta. We simulate idealized samplesof nuclear pasta and describe their breaking mechanism. We also deform nuclear pasta that isarranged into many domains, similar to what is known for the ions in neutron star crusts. Ourresults show that nuclear pasta may be the strongest known material, perhaps with a shear modulusof 1030 ergs/cm3 and breaking strain greater than 0.1.
Introduction. The breaking strain of materials in neu-tron star (NS) crusts are relevant for a variety of elec-tromagnetic and gravitational wave phenomena. Crustbreaking may occur in magnetar outbursts [1], resonantcrust shattering in NS mergers [2], and in the starquakemodel for pulsar glitches [3, 4]. Likewise, the shear mod-ulus of crust matter may affect the oscillation frequencyof magnetar flares [5], the damping of r-modes [6], anddetermine the height and lifetime of mountains on a NS[7, 8].
The crust comprises the outermost kilometer of theNS, where the density goes from near vacuum at the sur-face to nuclear density (1014 g/cm
3) at the base. The
outer crust is a bcc lattice of nuclei embedded in a gasof degenerate electrons, which becomes increasingly neu-tron rich with depth [9, 10]. At the base of the inner crustthe separation between nuclei becomes comparable to nu-clei radii and nucleons rearrange themselves into complexshapes known as nuclear pasta [11, 12]. The pasta layermay be 100-250 m thick, but due to its high density maycontain half the total mass of the crust or more [13]. Itis therefore essential to understand the elastic proper-ties of this material, as it may play an important role incrust breaking and be the dominant source of continuousgravitational waves that aLIGO is now searching for [14].
Past works have studied elastic properties (break-ing strain and shear modulus) of the ion crust withboth molecular dynamics (MD) and analytic methods[4, 9, 10, 15–17]. The ion crust is understood to be a poly-crystalline bcc lattice, i.e., composed of microscopic do-mains with different orientations. Thus, over sufficientlylarge length scales the crust is treated as an isotropicmaterial whose properties are found by angle averagingbcc lattice properties [17]. MD simulations suggest thatdomains lead to an effective angle-averaged shear mod-ulus [4, 16] and that high pressure prevents voids fromforming and fractures from propagating, thus, leading toa relatively high breaking strain ε ≈ 0.1 [16].
Presently, the pasta is less well understood. Unlike
the ion lattice pasta is not a crystal but, rather, a liquidcrystal. Pethick & Potekhin, Ref. [18], were the first tostudy the elasticity of nuclear pasta. They consideredthe energy of deformation Ed of parallel pasta plates tobe
Ed =B
2
[∂u
∂z− 1
2(∇⊥u)2
]2
+K1
2(∇2⊥u)2 , (1)
where B and K1 are elastic constants and u the displace-ment between plates along the z axis [18]. The firstterm is the energy due to the separation of the plates.The (∇⊥u)2 term is due to rotational invariance, as out-of-plane shear is rotationally equivalent to changing theplate spacing. The last term is due to splay deformations(plate curvature).
These analytic techniques are difficult to apply toasymmetric pasta or pasta lacking long range order. Evenshort range disorder such as helicoidal defects (filamentsconnecting the plates) may support shear stresses be-tween plates. These may be difficult to predict analyt-ically, but can be studied using MD simulations. Thesame is true for the effect of domains and their bound-aries in nuclear pasta on crust elasticity.
Watanabe et al. compare the typical thermal energyto the energy of deformation of nuclear pasta and findthat pasta deforms on length scales of order ten timesits spacing [19]. This result is supported by quantummechanical simulations that suggest that the pasta, likethe ion crust, does not have a uniform orientation acrossthe star [20]. Rather, nuclear pasta should be composedof many microscopic domains with distinct orientations.Thus, calculations of the elastic properties of the crustshould consider ‘polycrystalline’ nuclear pasta with manydomains. Currently, such studies are limited to MD sim-ulations since those can simulate large volumes with mul-tiple domains [21] while quantum mechanical simulationsare limited a few thousand nucleons [22].
In this Letter we discuss MD simulations (1) deform-ing nuclear pasta plates, to understand the breaking of
arX
iv:1
807.
0255
7v1
[nu
cl-t
h] 6
Jul
201
8
2
its elementary units, and (2) deforming a large nuclearpasta system with multiple domains, to understand angleaveraged elastic properties of nuclear pasta.
Simulations. Our simulations are performed usingthe Indiana University Molecular Dynamics code v6.3.1(IUMD), which has been used extensively to simulate nu-clear pasta [10, 23, 24]. In our model, two nucleons i andj separated by a distance r interact via the two-bodypotential
Vij(r) = ae−r2/Λ + [b± c]e−r
2/2Λ +eiejre−r/λ . (2)
The parameters a, b, c, and Λ are chosen to reproducethe binding energy of nuclei, pure neutron matter, andsymmetric nuclear matter while λ is the Thomas-Fermiscreening length, fixed at 10 fm for simplicity [23]. Theb+ c (b− c) term sets a weak (strong) attraction betweenlike (alike) nucleons. The electric charges ei and ej pro-duce long range Coulomb repulsion between protons. Allsystems are simulated at temperature T = 1 MeV andnucleon number density n = 0.05 fm−3.
Our first set of runs contain 102 400 nucleons at a pro-ton fraction YP = 0.4, where pasta is expected to form‘lasagna’ plates [10, 24]. This proton fraction is higherthan expected in the inner crust, but is where our modelforms pasta; at more realistic proton fractions the classi-cal model produces a gas of protons and neutrons [25].
We initialize our simulation with a cubic volume (sidelength l0) with periodic boundary conditions and applyvolume preserving deformations. We apply constant ex-tensional strain along axis r and contract along the othertwo, s and t, by adjusting the box boundary positionslightly after each MD timestep, i.e.,
lr(t) = l0(1 + εt) and ls(t) = lt(t) =l0√
1 + εt. (3)
where ε is the strain rate and r, s, and t are permutationsof x, y, and z. Nucleons respond dynamically to thechanging simulation volume and the induced stresses arecalculated using
σαβ =1
V
∑i
−mu(i)α u
(j)β +
∑i<j
(x(i)β − x
(i)α )f
(ij)β
. (4)
Above, V is the simulation volume, m the nucleon mass,
x(i)α (u
(i)α ) the α component of the position (velocity) of
nucleon i and f(ij)β the β component of the force nucleon
i exerts on nucleon j. Repeated (mixed) indices denotetensile (shear) stress.
Although these are purely tensile strains, they canbe transformed into shear strains by having the pastastructures oriented at an angle with respect to the sim-ulation boundaries. Furthermore, nuclear pasta is notnecessarily linearly elastic. Lasagna, for example, hastransverse isotropy implying five independent elastic con-stants. Misalignment of the pasta structures with respect
to the boundary and their time evolution makes it dif-ficult to isolate these constants for all but the simplestcases. For most simulations, we observe a rotated stresstensor which mixes all σ terms.
We strain at rates ε = 1×10−7 and 2×10−7 c/fm until afinal strain εr along the axis of extension is achieved. Thisresults in a compressional strain of εs = εt = 1/ 2
√2 ≈
0.29 in the other two axes. These choices of strain ratedoes not affect our results. Although these deformationsare much greater than those realized in astrophysical sys-tems, they allow us to observe the breaking behavior ofnuclear pasta.
First we deform ‘perfect’ plates which are aligned withthe xy plane of the simulation volume [27]. In the firstcase, Fig. 1a, we strain along the z direction, pullingthe plates apart, while contracting along the plate plane.For εz > 0.05 the plates start to buckle to maintain theirinitial spacing. Tensile stress along z remains constantfrom then on, while the stress in the plates decrease asthe plates buckle. The plates deform continuously with-out breaking up till a strain εz = 1. As expected, nosignificant shear stresses are observed in this run.
In the second case, Fig. 1b, we strain along the xaxis, pulling along the plates and allowing their spacingto contract. Tensile stress along the z direction grows asCoulomb pressure between plates increases. Meanwhile,in-plane stress decreases with the thinning of the plates.At εx ≈ 0.3 (εz ≈ 0.13), the plates break and the stressesabruptly change. The stresses reset and the process con-tinues until another breaking event occurs at εx ≈ 0.8.To our knowledge, this breaking mechanism has not beendescribed elsewhere in the literature and, thus, we de-scribe it here in detail. We observe that (1) for smallstrains the spacing between plates decreases while theplates themselves become thiner. (2) Once the plates arethin enough, thermal fluctuations nucleate holes in them.(3) By displacing nucleons in the plate the holes relievesthe Coulomb pressure between adjacent plates. (4) Ad-jacent plates become convex in the region of the holes.This partially fills the space vacated by the holes and al-lows the plates to increase their spacing locally, furtherreducing the system’s pressure. (5) Convex lobes fromthe adjacent plates connect near the hole, forming shorthelicoidal bridges that connects adjacent plates. Thisallows connected plates to exchange nucleons. This pro-cess occurs in several locations simultaneously, allowingthe plates to thicken and increase their interspacing lo-cally. The bridges then drift through the plates until theyare aligned, as seen in Fig. 1b, suggesting there is longrange attraction between them as predicted by Guven etal. [28]. The filaments become thinner and disconnect asthe plates reach their ideal thickness, resulting again inplanar lasagna sheets but with one fewer plate spanningthe simulation volume. During the first breaking eventwe observe that the shear stresses in the xz and yz planesare briefly nonzero, suggesting that filaments can support
3
t = 0.0 (107 fm/c) t = 0.2 (107 fm/c)
z
0 . 0 0 . 5 1 . 0
0 . 2 6
0 . 2 7
0 . 2 8
0 . 2 9
� (M
eV/fm
3 )
� z0 . 0 0 . 5 1 . 0
- 8
0
8
� x x � y y � z z � x y � x z � y z
� (10
-4 MeV
/fm3 )
� za) Tensile deformations pulling lasagna sheets apart.
t = 0.0 (107 fm/c) t = 0.3 (107 fm/c)
z
x 0 . 0 0 . 5 1 . 0
0 . 2 6
0 . 2 7
0 . 2 8
0 . 2 9
� (M
eV/fm
3 )
� x0 . 0 0 . 5 1 . 0
- 2
0
2
4
� x x � y y � z z � x y � x z � y z
� (10
-4 MeV
/fm3 )
� xb) Tensile deformation pulling lasagna sheets laterally while compressing them
t = 0.0 (107 fm/c) t = 0.2 (107 fm/c)
z
x 0 . 0 0 . 5 1 . 0
0 . 2 6
0 . 2 7
0 . 2 8
0 . 2 9
� (M
eV/fm
3 )
� z0 . 0 0 . 5 1 . 0- 8
- 4
0
4
8
� x x � y y � z z � x y � x z � y z
� (10
-3 MeV
/fm3 )
� zc) Lasagna sheets experiencing both tensile and shear strains
t = 0.0 (107 fm/c) t = 0.1 (107 fm/c)
z
x 0 . 0 0 . 5 1 . 0
0 . 2 6
0 . 2 7
0 . 2 8
0 . 2 9
� (M
eV/fm
3 )
� z0 . 0 0 . 5 1 . 0- 8
- 4
0
4
8
� x x � y y � z z � x y � x z � y z�
(10-3 M
eV/fm
3 )
� zd) Lasagna sheets with defects experiencing both tensile and shear strains, arrows show shear stress due to defects
FIG. 1. Simulations deforming idealized nuclear pasta. Left to right, (1) initial configuration of the pasta, (2) deformed pastaat some later time, (3) tensile stresses vs. extensional strain, and (4) shear stresses vs. extensional strain. Strain εr = lr(t)− l0,see Eq. (4), is proportional to time. Animations avaiable in SM1-4 [26].
shear stresses between plates.
While the helicoidal bridges were short lived in thissimulation, past work has identified stable configurationswhere filaments connecting plates are long lived [29].These may support shear stresses between plates and,perhaps, even act as grain boundaries between domainswhere pasta plates have different orientations.
We also consider simulations of lasagna plates whichare misaligned with the sides of the box allowing us toinduce shear stresses in the material. In one simulation
pasta plates are rotated in the xz plane, Fig. 1c, whilein another plates are similarly oriented but connected bystable and long lived filaments, Fig. 1d. Both systemshave similar energy, suggesting that pasta has multipleavailable phases to it [20].
The system shown in Fig. 1c has three topologicallydisconnected plates due to the periodic boundary, whilein the previous simulations all plates were topologicallydisconnected. These plates are free to adjust their incli-nation in response to the applied strain. They first break
4
0 . 0 0 . 1 0 . 20 . 1 7 6 4
0 . 1 7 6 5
0 . 1 7 6 6
0 . 1 7 6 7
�
(MeV
/fm3 )
� x0 . 0 0 . 1 0 . 2
- 2
0
2
� x x � y y � z z � x y � x z � y z
� (10
-4 MeV
/fm3 )
� x
εx= 0.20εx= 0.00
FIG. 2. Strain multidomain simulation of nuclear pasta.(Top) Tensile, left, and shear stresses, right. (Bottom) Evolu-tion of domains, identified by color, from no strain (εx = 0.0),left, to εx = 0.2, right. Animations in SM5,6 [26].
near εz ≈ 0.68, implying that pasta breaks less easilywhen the plates can adjust their inclination with respectto each other. The shear stress in the xz plane growsapproximately linearly until the break, transforming atensile stress into a shear stress as expected. However,these shear stresses can be eliminated by applying a ro-tation matrix to the stress tensor whose angle is takenfrom the inclination of the plates, approximately recov-ering the stresses seen in Fig. 1b, as expected. Thisdirectly confirms that disconnected plates have near zeroshear modulus between plates, as claimed by Pethick &Potekhin [18].
In the run with inclined plates connected by helicoidaldefects, Fig. 1d, the tensile deformation of the bound-ary produces a shear deformation of the plates due totheir inclination. However, the filaments connecting theplates experience a shear stress as the plates slide, whichis stronger in the xz plane. The filaments break nearεy ≈ 0.2, at which point the evolution proceeds identi-cally to the case discussed in Fig. 1c.
The difference in σxz at small strains seen between Fig.1c and Fig. 1d cannot be eliminated by applying a rota-tion and is a clear indication of a shear modulus producedby the filaments. Their difference gives a maximum shearstress of 3× 10−4 MeV/fm
3; estimating the shear strain
to be 0.1 gives us a shear modulus of 5 × 1030 erg/cm3.
The rotation of the plates partially transforms the shearstress into a tensile stress at small strains, so this shearmodulus may be a factor of two larger.
With the elastic response of individual domains of nu-
clear pasta understood, we now consider a larger sim-ulation with many domains. We present results of asimulation with 3 276 800 nucleons and proton fractionYP = 0.3. These parameters are known to produce par-allel plates with a hexagonal arrangement of holes, the‘waffle’ phase [25]. This proton fraction was chosen fortwo reasons. First, the simulation is less computation-ally expensive and allows us to evolve a larger number ofparticles. Second, the filamentary structure of the wafflephase allows us to study how its multiple domain struc-ture affects pasta elasticity.
To achieve the initial condition for the strain run weevolved the system from a random initial condition forover 15,250,000 MD timesteps. This is the largest sim-ulation of nuclear pasta yet and is discussed in detail inRef. [21]. The system formed seven differently orien-tated domains, see Fig. 2. The plates are misalignedat domain boundaries, but topologically connected byfilaments. The relative volume fractions of domains re-mained approximately constant during this initial evolu-tion, suggesting the domains are long lived.
This simulation was strained at a rate εx = 2 × 10−7
c/fm until εx = 0.25. The stresses are shown in Fig.2. No breaking was observed. Shear and tensile stresseswith components perpendicular to the strain direction xgrow approximately linearly until εx ∼ 0.12 and then flat-ten out. Meanwhile, stresses with x components decreaseslightly before flattening out. This is expected for a lin-ear isotropic material which is misaligned with the strainaxis. Another simulation was strained at εx = 1 × 10−6
c/fm until εx = 0.77 and finds similar results. Whileit is not possible to extract individual elastic constantsfrom this simulation, the stresses grow comparably to ourYp = 40% simulations, suggesting they may be similar.
In Fig. 2 we show our simulation for two differentstrains, εx = 0.0 and 0.2. At their boundaries, domainsare connected by filaments similar to the helicoidal de-fects that connect lasagna plates and considered above.When strained, domains move relative to each other andslide with only limited resistance. At the boundaries nu-cleons flow freely between adjacent domains, rearrangingnuclear matter and allowing domains to shrink or growin response to the applied stress. This prevents the largestresses that result in catastrophic failures from buildingup. As nuclear pasta in the neutron star crust is ex-pected to form domains, the inner crust may not breakfor realistic astrophysical strains.
Discussion. We find, using MD simulations, that theanalytic model of Pethick & Potekhin describes well thequalitative elastic response of idealized lasagna platesto large deformations. Idealized nuclear pasta is strongin our model, possibly with a shear modulus as largeas 1031 erg/cm
3. This is comparable to the strength of
the outer crust extrapolated to the high densities at thecrust-core boundary, which may make nuclear pasta thestrongest material in the known universe.
5
For comparison, a typical nucleus predicted at pastadensities by Haensel & Zdunik is 88Ti (Z=22) with freeneutron fraction Xn = 0.80 [30]. Using the effective shearmodulus of Horowitz & Hughto µeff ∼ 0.11nZ2e2/a (ionnumber density n, a = [3/(4πn)]1/3) we find µ = 1.1 ×1030 [31]. We emphasize that the large shear modulusand breaking strain for our pasta simulations are modeldependent. Our simulations use proton fractions muchlarger than those expected in the inner crust of NSs. Alower proton fraction likely reduces the shear modulus,perhaps to 1030erg/cm
3. This motivates further work on
pasta elasticity, especially the role that superfluidity mayplay, which is beyond the scope of our simulations [32].
We explicitly describe the breaking mechanism, andfind that the breaking strain of idealized nuclear pastain the lasagna and waffle phases is large in our model,perhaps ε ≈ 0.3. This suggests that the ion crust breakssignificantly earlier than the pasta. This may have conse-quences for crust breaking in a variety of systems, such asresonant shattering flares during neutron star mergers ormagnetar outbursts. Additionally, the large strength anddensity of nuclear pasta predicted by this work suggeststhat neutron stars may support large ‘buried’ mountainsin the inner crust. These could be an efficient sourceof continuous gravitational waves, and further motivatessearches by LIGO, Virgo, and other GW detectors.
While we can describe the elastic response of some sim-ple astromaterials in this work, the large variety of shapesthat nuclear pasta adopts and the variety of elastic re-sponses to induced strains observed here suggests that atheory of pasta elasticity may be difficult to formulateanalytically, and motivates future work exploring pastaelasticity with MD, such as isolating the elastic constantsof the various phases of pasta.
Acknowledgements. M.C. is a CITA National Fellow.A.S.S. is supported by the National Science Foundationunder award No. AST-1333520 and CAREER PHY-1151197. C.J.H. supported in part by US Departmentof Energy grants DE-FG02-87ER40365 (Indiana Univer-sity) and DE - SC0018083 (NUCLEI SciDAC-4 Collab-oration). This research was supported in part by LillyEndowment, Inc., through its support for the IndianaUniversity Pervasive Technology Institute. We thank A.Cumming and A. Chugnuov for conversation.
∗ [email protected]† [email protected]
[1] C. Thompson and R. C. Duncan, The AstrophysicalJournal 561, 980 (2001).
[2] D. Tsang, J. S. Read, T. Hinderer, A. L. Piro, andR. Bondarescu, Phys. Rev. Lett. 108, 011102 (2012).
[3] M. Ruderman, Nature (London) 223, 597 (1969).[4] C. A. I. and H. C. J., Monthly Notices of the Royal As-
tronomical Society: Letters 407, L54.
[5] T. E. Strohmayer and A. L. Watts, The AstrophysicalJournal 653, 593 (2006).
[6] B. Haskell, N. Degenaar, and W. C. G. Ho, Monthly No-tices of the Royal Astronomical Society 424, 93 (2012).
[7] A. I. Chugunov and C. J. Horowitz, Monthly Notices ofthe Royal Astronomical Society: Letters 407, L54 (2010).
[8] S. Mahmoodifar and T. Strohmayer, The AstrophysicalJournal 773, 140 (2013).
[9] N. Chamel and P. Haensel, Living Reviews in Relativity11, 10 (2008).
[10] M. E. Caplan and C. J. Horowitz, Rev. Mod. Phys. 89,041002 (2017).
[11] D. G. Ravenhall, C. J. Pethick, and J. R. Wilson, Phys.Rev. Lett. 50, 2066 (1983).
[12] M.-a. Hashimoto, H. Seki, and M. Yamada, Progress ofTheoretical Physics 71, 320 (1984).
[13] W. G. Newton, Nature Physics 9, 396 (2013).[14] B. P. Abbott et al. (LIGO Scientific Collaboration and
Virgo Collaboration), Phys. Rev. D 96, 122006 (2017).[15] M. E. Caplan, A. Cumming, D. K. Berry, C. J. Horowitz,
and R. Mckinven, The Astrophysical Journal 860, 148(2018).
[16] C. J. Horowitz and K. Kadau, Phys. Rev. Lett. 102,191102 (2009).
[17] D. Kobyakov and C. J. Pethick, Monthly Notices of theRoyal Astronomical Society: Letters 449, L110 (2015).
[18] C. Pethick and A. Potekhin, Physics Letters B 427, 7(1998).
[19] G. Watanabe, K. Iida, and K. Sato, Nuclear Physics A676, 455 (2000).
[20] W. Newton, (unpublished).[21] A. S. Schneider, M. E. Caplan, D. K. Berry, and C. J.
Horowitz, (to be published).[22] F. J. Fattoyev, C. J. Horowitz, and B. Schuetrumpf,
Phys. Rev. C 95, 055804 (2017).[23] C. J. Horowitz, M. A. Perez-Garcıa, and J. Piekarewicz,
Phys. Rev. C 69, 045804 (2004).[24] A. S. Schneider, C. J. Horowitz, J. Hughto, and D. K.
Berry, Phys. Rev. C 88, 065807 (2013).[25] A. S. Schneider, D. K. Berry, C. M. Briggs, M. E. Caplan,
and C. J. Horowitz, Phys. Rev. C 90, 055805 (2014).[26] Animations available at: www.physics.mcgill.ca/
~mecaplan/pasta-elasticity/.[27] A. S. Schneider, D. K. Berry, M. E. Caplan, C. J.
Horowitz, and Z. Lin, Phys. Rev. C 93, 065806 (2016).[28] J. Guven, G. Huber, and D. M. Valencia, Phys. Rev.
Lett. 113, 188101 (2014).[29] C. J. Horowitz, D. K. Berry, C. M. Briggs, M. E. Caplan,
A. Cumming, and A. S. Schneider, Phys. Rev. Lett. 114,031102 (2015).
[30] P. Haensel and J. L. Zdunik, Astronomy and Astro-physics 229, 117 (1990).
[31] C. J. Horowitz and J. Hughto, ArXiv e-prints (2008),arXiv:0812.2650.
[32] D. N. Kobyakov and C. J. Pethick, ArXiv e-prints(2018), arXiv:1803.06254 [nucl-th].
