1: \documentclass[runningheads]{cl2emult}
2: \usepackage{psfig}
3: \usepackage{psfrag} % replace labels in figures
4: \usepackage{epsf}
5: \usepackage{makeidx} % allows index generation
6: \usepackage{graphicx} % standard LaTeX graphics tool
7: \usepackage{subeqnar} % subnumbers individual equations
8: \usepackage{multicol} % used for the two-column index
9: \usepackage{citesort}
10: \usepackage{amssymb}
11: \usepackage{amsmath}
12: \usepackage{latexsym}
13: \usepackage{cropmark} % cropmarks for pages without
14: \usepackage{lnp} % placeholder for figures
15: \makeindex % used for the subject index
16:
17: \newcommand{\euler}[1]{{\usefont{U}{eur}{m}{n}#1}}
18: \newcommand{\eulerbold}[1]{{\usefont{U}{eur}{b}{n}#1}}
19: \newcommand{\umu}{\mbox{\euler{\char22}}}
20: \newcommand{\umub}{\mbox{\eulerbold{\char22}}}
21: \begin{document}
22:
23: \frontmatter
24:
25: %\input{Book/Title}
26: %\cleardoublepage
27:
28: %\markboth{Preface}{Preface}
29: %\input{Book/Preface}
30: %\clearpage
31:
32: %\markboth{Contents}{Contents}
33: %\tableofcontents
34: \mainmatter
35: \setcounter{page}{203}
36:
37: %\addcontentsline{toc}{part}{Kinetic Theory and Hydrodynamics}
38: %\part*{Kinetic Theory\\[0.3cm] and\\[0.3cm] Hydrodynamics}
39: %\include{Ernst/Ernst} %FINISHED
40: %\include{Zippelius/Zippelius} %FINISHED (check \stackrel)
41: %\include{Brey/Brey} %FINISHED
42: %\include{Goldhirsch/Goldhirsch} %FINISHED
43: %\include{Brilliantov/PP2} %FINISHED
44: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
45: % \begin{document}
46:
47: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
48: %\include{Jenkins/Jenkins} %FINISHED
49: %\include{Henrique/Henrique} %FINISHED
50:
51: %\addcontentsline{toc}{part}{Collisions and One-dimensional Models}
52: %\part*{Collisions\\[0.3cm] and\\[0.3cm] One-dimensional Models}
53: %\include{Bridges/Bridges} %FINISHED
54: %\include{Thornton/Thornton} %FINISHED
55: %\include{Ramirez/Ramirez}
56: %\include{Poeschel/PoeschelBrilliantov} %FINISHED
57: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
58: %\begin{document}
59:
60: \title*{{\small In: T. P\"oschel and S. Luding (eds.), {\em Granular Gases}, Lecture Notes in Physics Vol. 564, Springer (Berlin, 2000), p. 203}\vspace*{0.5cm}\\ Chains of Viscoelastic Spheres}
61: \label{Poeschelpage}
62:
63: \toctitle{Chains of viscoelastic spheres}
64:
65: \titlerunning{Chains of viscoelastic spheres}
66:
67: \author{Thorsten P\"oschel\inst{1}
68: \and Nikolai V.~Brilliantov\inst{2}$^,$\inst{1}
69: }
70:
71: \authorrunning{T.~P\"oschel \and N.\,V.~Brilliantov}
72: \tocauthor{T.~P\"oschel, N.\,V.~Brilliantov}
73:
74: \institute{Humboldt-Universit\"at, Institut f\"ur Physik,
75: Invalidenstr. 110, D-10115 Berlin, Germany. email thorsten@physik.hu-berlin.de, http://summa.physik.hu-berlin.de/$\sim$thorsten \and Moscow State University, Physics Department, Moscow
76: 119899, Russia.\\ email nbrillia@physik.hu-berlin.de}
77:
78: \maketitle % typesets the title of the contribution
79:
80: \begin{abstract}
81: Given a chain of viscoelastic spheres with fixed masses of the first and last particles. We raise the question: How to chose the masses of the other particles of the chain to assure maximal energy transfer? The results are compared with a chain of particles for which a constant coefficient of restitution is assumed. Our simple example shows that the assumption of viscoelastic particle properties has not only important consequences for very large systems (see~\cite{BPhere}) but leads also to qualitative changes in small systems as compared with particles interacting via a constant restitution coefficient.
82: \end{abstract}
83:
84: \section{Introduction}
85: \label{sec:intro}
86: We consider a linear chain of inelastically colliding particles of masses $m_i$, radii $R_i$ ($i=0\dots n$), with initial velocities $v_0=v>0$ and $v_i=0$ ($i=1\dots n$) at initial positions $x_i>x_j$ for $i>j$ with $x_{i+1}-x_i> R_{i+1}+R_i$. The masses of the first and last particles $m_0$ and $m_n$ are given and we address the questions: How have the masses of the particles in between to be chosen to maximize the energy transfer, i.e., to maximize the after-collisional velocity $v_n^{\prime}$ of the last particle. If $n$ is variable, how should $n$ be chosen to maximize $v_n^{\prime}$? Throughout this paper we assume that the initial distance of the particles is large enough to neglect ``multiple collisions'', i.e., only the first impact of each particle influences the final velocity of the $N$th sphere of the chain.
87:
88: Recent investigations show that the properties of very large systems of viscoelastic particles differ significantly from those of particles interacting with constant coefficient of restitution~\cite{BPhere,NBTPDIFF,veldistr,TomThor1,NBTPage}. The system considered here may serve as an example of a {\em small} system which properties change qualitatively when the viscoelastic properties of the particles are considered.
89:
90: The coefficient of restitution is defined via
91: %\begin{equation}
92: $\epsilon = \left|\left({v_{i+1}^\prime}-v_{i}^\prime\right)/\left(v_{i+1}-{v_i}\right)\right|$,
93: %\end{equation}
94: which relates the relative velocity of the particles after the collision to the pre-collisional quantity. The elastic collision corresponds to $\epsilon=\mbox{const.}$ For this case, basic mechanics yields:
95: \begin{equation}
96: \label{8}
97: v_1^\prime=\frac{1+\epsilon}{1+\frac{m_1}{m_0}} \, v_0,~~~~v_{n}^\prime= (1+\epsilon)^n \prod_{k=0}^{n-1} \left(1+\frac{m_{k+1}}{m_k} \right)^{-1}\, v_0\,.
98: \end{equation}
99: The final velocity $v_n^\prime$ is maximized by
100: \begin{equation}
101: \label{5}
102: m_k=\left(\frac{m_n}{m_0}\right)^{k/n}m_0 ~,~~\mbox{yielding}~~~v_n^\prime=\left[ \frac{1+\epsilon}{1+\left(\frac{m_n}{m_0}\right)^{1/n}} \right]^{n} \,v_0\,.
103: \end{equation}
104:
105: The optimal mass distribution for the case of a constant coefficient of restitution $\epsilon$ does not depend on the value of $\epsilon$ and is, therefore, the same as the optimal mass distribution in a chain of elastic particles. Figure~\ref{fig:constantMass} (left) shows the optimal mass distribution for different chain lengths $n$. The mass of the first particle is $m_0=1$ and of the last particle $m_n=0.1$.
106: \begin{figure}[htbp]
107: \centerline{\psfig{figure=mv22m.eps,height=4.7cm,angle=270}\psfig{figure=mv22v.eps,height=4.7cm,angle=270}}
108: % \centerline{\psfig{figure=../../PROGRAM/constant/mv22m.eps,width=5.5cm,angle=270}\psfig{figure=../../PROGRAM/constant/mv22v.eps,width=5.5cm,angle=270}}
109: % \centerline{\psfig{figure=/home/t/NEW/PapersInProcess/Transmission2403/PROGRAM/constant/mv22m.eps,width=5.5cm,angle=270}\psfig{figure=/home/t/NEW/PapersInProcess/Transmission2403/PROGRAM/constant/mv22v.eps,width=5.5cm,angle=270}}
110: %\vspace{0.3cm}
111: \caption{{\bf Left:} Optimal mass distribution $m_i$, $i=1\dots n$, for the case of constant $\epsilon$. Each of the lines shows the mass $m_i$ over the index $i$ for a specified chain length $n$. The masses of the first and last particles are fixed at $m_0=1$ and $m_n=0.1$. {\bf Right:} Velocity distributions of the particles in chains of length $n$ with the optimal mass distribution according to (\ref{5}) as a function of the chain length. The dissipative constant is $b\equiv 1-\epsilon=5\cdot 10^{-4}$. The last particle reaches its maximal velocity for chain length $n^*=44$ (bold drawn). The velocity of the first particle of the chain is $v_0=1$.}
112: %\vspace{0.5cm}
113: \label{fig:constantMass}
114: \end{figure}
115:
116: In contrast to the mass distribution the corresponding velocity distributions do depend on the value of the restitution coefficient $\epsilon$. Figure \ref{fig:constantMass} (right) shows the velocity distribution for $b\equiv 1-\epsilon=5\cdot 10^{-4}$. For the case of dissipative collisions the ratio $R_v=v_n^\prime/v_0$ does not monotonously increase with $n$ as for elastic particles ($\epsilon=1$), but rather it has an extremum which shifts to smaller chain lengths with increasing
117: dissipative parameter $b$. The optimal value of $n$, which maximizes $R_v$ reads
118: \begin{equation}
119: %\label{10}
120: n^*=\log \left( m_0/m_n \right)/\log\left( x_0 \right)
121: \label{optnconst}
122: \end{equation}
123: where $x_0$ is the solution of the equation
124: \begin{equation}
125: \label{11}
126: (1+x_0)=(1+\epsilon)x_0^{x_0/(1+x_0)}\,.
127: \end{equation}
128: Correspondingly, the extremal value of the $R_v$ reads
129: \begin{equation}
130: \label{12}
131: R_v^*=\left[ \frac{1+\epsilon}{1+x_0} \right]^{n^*}\,.
132: \end{equation}
133:
134: \section{Chains of viscoelastic particles}
135:
136: It has been shown that for colliding viscoelastic spheres the restitution coefficient depends on the masses of the colliding particles and also on their relative velocity $v_{ij}$ \cite{BSHP}. An explicit expression for the coefficient of restitution is given by the series \cite{TomThor1,Rosa} (see also \cite{BPhere})
137: \begin{equation}
138: \epsilon=1-C_1\left(\frac{3A}{2} \right)\alpha^{2/5} v_{ij}^{1/5}\!+
139: \!C_2 \left(\frac{3A}{2} \right)^2\alpha^{4/5} v_{ij} ^{2/5}\! \mp\! \cdots
140: \label{epsilon}
141: \end{equation}
142: with
143: \begin{equation}
144: \alpha= \frac{2~ Y\sqrt{R^{\,\mbox{\footnotesize eff}}}}{
145: 3~ m^{\mbox{\footnotesize eff}}\left( 1-\nu ^2\right) }
146: \label{rhodef}
147: \end{equation}
148: with $Y$ and $\nu$ being the Young modulus and the Poisson ratio, respectively and $R^{\,\mbox{\footnotesize eff}}=R_iR_j/(R_i+R_j)$,
149: $m^{\mbox{\footnotesize eff}}=m_im_j/(m_i+m_j)$. The material constant $A$ describes the dissipative properties of the spheres (for details see~\cite{BSHP}). The
150: constants $C_1=1.15344$ and $ C_2=0.79826 $ were obtained analytically
151: in Ref.~\cite{TomThor1} and then confirmed by numerical simulations.
152:
153: In the following calculation we neglect terms ${\cal O}\left(v^{2/5}\right)$ and higher and assume for simplicity that all particles are of the
154: same radius $R$, but have different masses. We abbreviate
155: \begin{equation}
156: \label{epsviab}
157: \epsilon=1-b\, v^{1/5}\left(m^{\mbox{\footnotesize eff}}\right)^{-2/5}~~~\mbox{with}~~~b=C_1 \left(\frac{3A}{2} \right)\left(\frac{2}{3}\frac{Y \sqrt{R/2}}{1-\nu^2}\right)^\frac25\,.
158: \end{equation}
159: %with
160: %\begin{equation}
161: %b=C_1 \left(\frac{3A}{2} \right)\left(\frac{2}{3}\frac{Y \sqrt{R/2}}{1-\nu^2}\right)^\frac25.
162: %\label{bdef}
163: %\end{equation}
164:
165: Hence, for viscoelastic particles the velocities of the $k+1$-rst particle after colliding with the $k$-th reads
166: \begin{equation}
167: \label{vkk}
168: v^\prime_{k+1}=
169: \frac{2-b\left(\frac{m_{k+1}+m_k}{m_{k+1}m_k} \right)^{2/5} v_k^{1/5} }{1+\frac{m_{k+1}}{m_k}} \, v_k\,.
170: \end{equation}
171: The masses $m_k$, $k=1\dots n-1$ which maximize $v_n^\prime$ can be determined numerically and the results are shown in Fig. \ref{fig:mv22m} for two different values of the dissipative constant $b$.
172:
173: \begin{figure}[htbp]
174: \centerline{\psfig{figure=mv22mEPS.eps,height=4.5cm,angle=270}\hspace{0.5cm}\psfig{figure=mv42m.eps,height=4.5cm,angle=270}}
175: \vspace{0.4cm}
176: % \centerline{\psfig{figure=../../PROGRAM/1_0.1/mv22m.eps,width=5cm,angle=270}\psfig{figure=../../PROGRAM/1_0.1/mv42m.eps,width=5cm,angle=270}}
177: % \centerline{\psfig{figure=/home/t/NEW/PapersInProcess/Transmission2403/PROGRAM/1_0.1/mv22m.eps,width=5cm,angle=270}\psfig{figure=/home/t/NEW/PapersInProcess/Transmission2403/PROGRAM/1_0.1/mv42m.eps,width=5cm,angle=270}}
178: %\vspace{0.3cm}
179: \caption{Optimal mass distribution of collision chains over the chain length $n$. The dissipative constant was $b=5\cdot 10^{-4}$ (left) and $b=2\cdot 10^{-3}$ (right).}
180: %\vspace{0.5cm}
181: \label{fig:mv22m}
182: \end{figure}
183:
184: For small chain length or small $b$, respectively, the optimal mass distribution is very close to that for the elastic chain as shown in Fig.~\ref{fig:constantMass}. Again we find a monotonously decaying function for the masses. For larger chain length $n$ or larger dissipation $b$, however, the mass distribution is a non-monotonous function. The according velocities of the particles in chains of spheres of optimal masses are drawn in Fig~\ref{fig:mv22v}.
185: \begin{figure}[htbp]
186: \centerline{\psfig{figure=mv22vEPS.eps,height=4.5cm,angle=270}\hspace{0.5cm}\psfig{figure=mv42v.eps,height=4.5cm,angle=270}}
187: \vspace{0.4cm}
188: % \centerline{\psfig{figure=../../PROGRAM/1_0.1/mv22v.eps,width=5cm,angle=270} \psfig{figure=../../PROGRAM/1_0.1/mv42v.eps,width=5cm,angle=270}}
189: % \centerline{\psfig{figure=/home/t/NEW/PapersInProcess/Transmission2403/PROGRAM/1_0.1/mv22v.eps,width=5cm,angle=270} \psfig{figure=/home/t/NEW/PapersInProcess/Transmission2403/PROGRAM/1_0.1/mv42v.eps,width=5cm,angle=270}}
190: %\vspace{0.3cm}
191: \caption{The velocities of particles in optimal chains according to Fig.~\ref{fig:mv22m}.}
192: %\vspace{0.5cm}
193: \label{fig:mv22v}
194: \end{figure}
195:
196: \subsection{Optimal mass-distribution}
197:
198: The ``loss'' of energy, i.e., the amount of kinetic energy which is not transferred from the first particle of the chain to the last one may be subdivided into ``inertial'' and ``viscous'' losses. Inertial losses occur due to mismatch of subsequent masses, which causes incomplete transfer of momentum even for elastic collisions if the masses differ. Viscous losses are caused by the dissipative nature of collisions. The inertial
199: loss is, thus, given by the energy which remains in the $i-1$rst particle after the collision with the $i$th:
200: \begin{eqnarray}
201: \label{Ein1}
202: \Delta E_{in}^{(i)} &=& \frac{m_{i-1}}{2}\left(v_{i-1}^\prime\right)^2 = \frac{m_{i-1}}{2}\left(\frac{m_i-m_{i-1}}{m_i+m_{i-1}}\right)^2 v_{i-1}^2\,.
203: \end{eqnarray}
204: We describe the chain in continuum approximation $m(x)$ with $m_i\approx m_{i-1}+\frac{d m(x)}{d x}\cdot 1 $, where we assume that particles are separated on a line by unit distance. Discarding high-order mass gradients within the continuum picture $\Delta E_{in}^{(i)} \to \frac{d E_{in}}{d x} \cdot 1$
205: we write for the ``line-density'' of the inertial losses
206: \begin{equation}
207: \label{Ein2}
208: \frac{d E_{in}}{d x} \approx \frac{\left(\frac{dm(x)}{dx}\right)^2}{8m(x)}v(x)^2\,.
209: \end{equation}
210:
211: Viscous losses may be quantified as the difference of the kinetic energy of a particle after an {\em elastic}
212: collision and that of after a {\em dissipative} collision:
213: \begin{multline}
214: \Delta E_{vis}^{(i)} = \left.\frac{m_{i} v_{i}^{2}}{2}\right|_{\epsilon=1}-\left.\frac{m_{i} v_{i}^2}{2}\right|_{\epsilon = \epsilon\left(v_i\right)}= \\
215: =\frac{m_{i}}{2}\left(\frac{2}{1+\frac{m_{i}}{m_{i-1}}}\right)^2 v_{i-1}^2 - \frac{m_{i}}{2}\left(\frac{1+\epsilon\left(v_{i-1}\right)}{1+\frac{m_{i}}{m_{i-1}}}\right)^2 v_{i-1}^2 \\
216: =\frac{2 m_{i} v_{i-1}^2}{\left( 1+\frac{m_{i}}{m_{i-1}} \right)^2}
217: \left\{1-\left[1-\frac{b}{2}\left(\frac{m_i+m_{i-1}}{m_i m_{i-1}}\right)^{2/5} v_{i-1}^{1/5}\right]^2 \right\}\,.
218: \label{viss}
219: \end{multline}
220: Expanding (\ref{viss}) up to linear order in the dissipative parameter $b$ which is assumed to be small and neglecting products of $b$ and mass gradients (which are supposed to be small too), the continuum transition of Eq.~(\ref{viss}) yields
221: \begin{equation}
222: \frac{d E_{vis}}{d x} \approx \frac{b}{2^{3/5}} m^{3/5} v^{11/5}.
223: \end{equation}
224: Thus, the total energy loss in the entire chain reads
225: \begin{equation}
226: \label{Etot1}
227: E_{tot} = \int\limits_0^n \left[
228: \frac{m_x^2}{8m}v^2 + \frac{b}{2^{3/5}} m^{3/5} v^{11/5} \right] dx= \int\limits_0^n \left[ \frac{m_x^2}{8m^2} + \frac{b}{2^{3/5}}
229: \frac{1}{m^{1/2}}\right] dx\,.
230: \end{equation}
231: with $m_x \equiv dm/dx$. For the second part of Eq. (\ref{Etot1}) we assume in zero-order approximation the ``ideal chain Ansatz'' for the velocity distribution $v(x)$, which refers to the velocity distribution $v(x)$ in an
232: idealized chain, where the kinetic energy completely transforms through the chain, i.e., where $\frac12 m(x)v^2(x)= {\rm const}=\frac12 m_0v_0^2$. With $m_0=1$, $v_0=1$, i.e., $v(x)=1/\sqrt{m(x)}$, the right hand side of Eq. (\ref{Etot1}) follows.
233:
234: The mass distribution which minimizes $E_{tot}$ satisfies the
235: Euler-equation applied to the integrand in (\ref{Etot1}):
236: \begin{equation}
237: \label{eq:Euler}
238: \frac{d}{dx}\frac{2m_x}{8m^2} -
239: \frac{\partial}{\partial m} \left[ \frac{m_x^2}{8m^2} + \frac{b}{2^{3/5}}
240: \frac{1}{m^{1/2}} \right] = 0
241: \end{equation}
242: which leads to an equation for the mass distribution of
243: the optimal chain, written for $y(x) \equiv 1/m(x)$:
244: \begin{equation}
245: \label{eqy}
246: \frac{d^2 y}{dx^2} - \frac{1}{y}\left(\frac{dy}{dx}\right)^2 -
247: 2^{2/5}b y^{3/2} = 0\,.
248: \end{equation}
249: Figure~\ref{fig:mAnalNumA} (lines) shows the numerical solution of (\ref{eqy}), i.e., the optimal mass distribution, for a
250: chain of length $n=40$ for different damping parameters $b$. The points show the results of a discrete numerical optimization of the
251: full chain problem (see Eq.~(\ref{vkk})) applying a steepest descent method to optimize the masses $m_k$ of all particles. For small dissipation $b$ both results agree.
252: \begin{figure}[htbp]
253: %\psdraft
254: \centerline{\psfig{figure=wurf1.eps,height=4.8cm,angle=270}\psfig{figure=wurf.eps,height=4.8cm,angle=270}}
255: % \centerline{\psfig{figure=../../PROGRAM/wurf/40/wurf1.eps,height=4.8cm,angle=270}\psfig{figure=../../PROGRAM/wurf/40/wurf.eps,height=4.8cm,angle=270}}
256: % \centerline{\psfig{figure=/home/t/NEW/PapersInProcess/Transmission2403/PROGRAM/wurf/40/wurf1.eps,width=5.5cm,angle=270}\psfig{figure=/home/t/NEW/PapersInProcess/Transmission2403/PROGRAM/wurf/40/wurf.eps,width=5.5cm,angle=270}}
257: \psfull
258: %\vspace{0.3cm}
259: \caption{{\bf Left:} Mass distribution of the optimal chain of length $n=40$
260: for different values of the dissipative parameter $b$. Lines: numerical solution of Eq. (\ref{eqy}),
261: Points: discrete numerical optimization (from top to bottom: $\bullet:~b=0.128$,
262: $\blacksquare:~b=0.064$, $\blacklozenge:~b=0.032$,
263: $\blacktriangle:~ b=0.016$, $\blacktriangleleft:~b=0.008$,
264: $\blacktriangledown:~b=0.004$, $\blacktriangleright:~b=0.002, $
265: etc.). {\bf Right:} Same data and symbols as left but plotted in larger scale. }
266: %\vspace{0.5cm}
267: \label{fig:mAnalNumA}
268: \end{figure}
269:
270: For larger values of $b$ the solution of Eq.~(\ref{eqy}) deviates from the discrete
271: optimization which is understandable since in our approximation we assumed the gradients of the mass distribution to be small which is violated for larger $b$. While the absolute values of masses deviates from the discrete calculation, Eq.~(\ref{eqy}) still predicts well the position of the maximum of $m(x)$.
272:
273: Figure \ref{fig:vepsonv40} displays the corresponding distribution of
274: velocities for the optimal chains shown in
275: Fig. \ref{fig:mAnalNumA}. According to the maximum in the mass
276: distribution, the velocity distribution reveals for larger $b$ a
277: pronounced minimum.
278: \begin{figure}[htbp]
279: %\psdraft
280: \centerline{\psfig{figure=V40.eps,height=4.1cm,angle=270}\hspace{0.8cm}\psfig{figure=opt.eps,height=4.1cm,angle=0}}
281: % \centerline{\psfig{figure=../../PROGRAM/1_0.1/vmfor100/V40.eps,height=4.2cm,angle=270}\hspace{0.2cm}\psfig{figure=../../PROGRAM/1_0.1/optchain/opt.eps,height=4.2cm,angle=0}}
282: % \centerline{\psfig{figure=/home/t/NEW/PapersInProcess/Transmission2403/PROGRAM/1_0.1/vmfor100/V40.eps,width=5.5cm,angle=270}\psfig{figure=/home/t/NEW/PapersInProcess/Transmission2403/PROGRAM/1_0.1/optchain/opt.eps,width=5.5cm,angle=0}}
283: \psfull
284: \vspace{0.8cm}
285: \caption{{\bf Left:} The velocity distribution along the optimal chain shown in
286: Fig.~\ref{fig:mAnalNumA} Lines from top to bottom: $b=2.5\cdot 10^{-4}$,
287: $5\cdot 10^{-4}$, $0.001$, $0.002$, $0.004$, $0.008$, $0.016$,
288: $0.032$, $0.064$, $0.128$. }
289: \label{fig:vepsonv40}
290: \caption{{\bf Right:} The mass of the heaviest sphere $m^*$ in an optimal chain
291: depends on the dissipative parameter $b$ and on the chain length
292: $n$. In the figure we plotted $\left(m^*\right)^{1/4}$ over $n\sqrt{b}$ for about
293: 3000 different combinations of $b$ and $n$ ($n=2\dots 300$,
294: $b=0.0001\dots 0.256$) including all data presented in
295: Figs. \ref{fig:mv22m}, \ref{fig:mAnalNumA}. Without any adjustable parameters the data
296: from the numerical optimization of chains agrees well with the
297: analytical expression Eq. (\ref{eq:maxmass}).}
298: %\vspace{0.5cm}
299: \label{fig:maxmass}
300: \end{figure}
301:
302: One can give a simple physical explanation of the appearance of a maximum in
303: the mass distribution (and the corresponding minimum in the velocity
304: distribution): As it is seen from Eq. (\ref{epsviab}) the restitution
305: coefficient increases with decreasing impact velocity and increasing
306: masses of colliding particles; this reduces the viscous losses. Thus,
307: slowing down particles, by increasing their masses in the inner part
308: of the chain, leads to decrease of the viscous losses of the energy
309: transfer. The larger the masses in the middle and the smaller their
310: velocities, the less energy is lost due to dissipation. On the other
311: hand, since the masses $m_0$ and $m_n$ are fixed, very large masses in the
312: middle of the chain will cause large mass mismatch of the subsequent
313: masses and, thus, large inertial losses [see Eq. (\ref{Ein1})]. The
314: optimal mass distribution, minimizing the {\em total} losses,
315: compromises (dictated by $b$) between these two opposite tendencies.
316: For the case of a constant coefficient of restitution the relative
317: part of the kinetic energy, which is lost due to dissipation does not
318: depend on the impact velocity. This means that only minimization of
319: the inertial losses, caused by mass gradient, may play a role in
320: the optimization of the mass distribution. Thus, only a monotonous mass
321: distribution with minimal mass gradients along the chain may be
322: observed as an optimal one for the case of the constant restitution
323: coefficient.
324:
325: \subsection{The maximum of the optimal mass-distribution}
326:
327: The mass $m^*$ of the heaviest sphere in the optimal mass distribution can be expressed as a function of the chain
328: length $n$ and the dissipative parameter $b$. With the term $y^{-1}\left( dy/dx \right)^2$ discarded, Eq.~(\ref{eqy}) describes formally
329: scattering of a particle of unit mass by the potential
330: \begin{equation}
331: U(y)=-\frac12 d\,b y^{5/2}~~~~~\mbox{with}~~~~~d \equiv \frac{4}{5} 2^{2/5}.
332: \end{equation}
333: Formally changing notations $x \to t$ (``time'') to emphasize the
334: mechanical analogy, we write the equation of motion
335: \begin{equation}
336: \label{y(t)}
337: \ddot{y}=-\frac{dU}{dy}\,,~~~y_0=y(t=0)=\frac{1}{m_0}\,,~~~y_n=y(t=n)=\frac{1}{m_n}\,.
338: \end{equation}
339: Here we consider the case of mass distributions
340: having a maximum; the generalization, however, is
341: straightforward. Hence,
342: \begin{equation}
343: \label{ydot}
344: \frac12 \dot{y}^2+U(y)={\rm const}=U(y^*)
345: \end{equation}
346: where $y^*$ is the turning point in the scattering problem, i.e., the
347: point where the particle's ``velocity'' $\dot{y}$ is zero (this
348: corresponds to the point $m^*$ of the mass distribution in the initial
349: problem). The ``particle'' reaches this point at ``time'' $t^*$, i.e.,
350: \begin{equation}
351: \label{ydot1}
352: \dot{y}^2=d\, b \left(y^{5/2}-y^{*\, 5/2} \right)\,.
353: \end{equation}
354: Solving Eq. (\ref{ydot1}) with respect to $\dot{y}$
355: yields
356: \begin{equation}
357: \label{ydotsolv}
358: \frac{dy}{dt}=
359: \pm \, \sqrt{d\,b} \, y^{*\, 5/4} \sqrt{\left(y/y^* \right)^{5/2}-1}\,.
360: \end{equation}
361: Integration over ``time'' from $t=0$ to $t=n$ in Eq. (\ref{ydotsolv}), therefore, leads to (with correct choice of signs)
362: \begin{equation}
363: \label{integovert}
364: \frac{y^{*\, -\frac54}}{\sqrt{d\,b}}
365: \left[\int_{y^*}^{y_0} \frac{dy}{\sqrt{\left( \frac{y}{y^*}\right)^{\frac52 }-1}}
366: +\int_{y^*}^{y_n} \frac{dy}{\sqrt{\left( \frac{y}{y^*}\right)^{\frac52 }-1}}
367: \right]=n\,.
368: \end{equation}
369: Using the substitute $z=(y^*/y)^{5/2}$, the integrals in Eq. (\ref{integovert}) may be recasted into the form
370: \begin{equation}
371: \label{inttrans}
372: \frac25\, y^* \int\limits_{\left(\frac{y^*}{y_k} \right)^{\frac52}}^{1}
373: z^{-\frac{9}{10}}\left(1-z \right)^{-\frac12}dz = {\cal B} \left(\frac{1}{10}, \frac12 \right) -
374: {\cal B} \left[\frac{1}{10}, \frac12,\, \left(\frac{y^*}{y_k}
375: \right)^{\frac52} \right]
376: \end{equation}
377: with $k=0$ for the first integral in the the left-hand side of
378: Eq. (\ref{integovert}) and with $k=n$ for the second integral.
379: %These may be also written as
380: %\begin{equation}
381: % \label{intviagamma}
382: %\end{equation}
383: ${\cal B}(x,y)$ is the Beta-function and ${\cal B}(x,y,\,a)$ is
384: the incomplete Beta-function (which has an upper limit $a$ instead of
385: $1$ in its integral representation). If we assume the pronounced
386: maximum in the optimal mass-distribution, so that $a \equiv
387: \left(y^*/y_k \right)^{5/2} =\left(m_k/m^* \right)^{5/2}$ is small,
388: one can approximate the incomplete Beta-function as
389: \begin{equation}
390: \label{Betaapprox}
391: {\cal B} \left(\frac{1}{10}, \frac12, \, a \right) \equiv
392: \int_0^a z^{-\frac{9}{10}} \left(1-z \right)^{-\frac12} dz
393: \approx \int_0^a z^{-\frac{9}{10}} = 10\, a\,.
394: \end{equation}
395: With the use of Eqs. (\ref{inttrans})
396: %,(\ref{intviagamma})
397: and
398: (\ref{Betaapprox}), Eq. (\ref{integovert}) reads
399: \begin{equation}
400: \label{nandBeta}
401: \frac{2 y^{*\,-\frac14}}{5\sqrt{d\,b}}
402: \left\{2 {\cal B} \left(\frac{1}{10}, \frac12 \right)-
403: 10\left[ \left(\frac{y^*}{y_0} \right)^{\frac14}+ \left(\frac{y^*}{y_n}
404: \right)^{\frac14}\right]\right\}=n\,.
405: \end{equation}
406: In the original variables, which refer to the mass distribution, we
407: obtain a {\em scaling} relation, connecting the heaviest mass $m^*$,
408: the chain length $n$ and the dissipative parameter $b$:
409: \begin{equation}
410: \label{eq:maxmass}
411: \left(m^*\right)^{1/4} = p\, \sqrt{b} \, n + q
412: \end{equation}
413: with the constants
414: \begin{equation}
415: \label{pq}
416: p \equiv \frac{5\sqrt{d}}{4{\cal B} \left(\frac{1}{10}, \frac12
417: \right)}
418: ~~~~~~~q \equiv \frac{5}{{\cal B}
419: \left(\frac{1}{10}, \frac12 \right)}
420: \left[ m_0^{\frac14}+m_n^{\frac14} \right]\,.
421: \end{equation}
422: So far we considered the solution of the variational Eq. (\ref{eqy})
423: with the term $y^{-1}\left(dy/dx \right)^2$ discarded. For this case
424: the constant $d$, which has been given above reads: $d=\frac45\,
425: 2^{2/5}$. It may be shown, however, that perturbative (thus
426: approximate) account of this omitted term leads to an equation of the
427: same form as Eq. (\ref{ydotsolv}), but with the {\em renormalized}
428: coefficient $d \to \frac95\,d = \frac{36}{25}\,2^{2/5}$; the details
429: are given in~\cite{BPchain}. Using numerical values for ${\cal B} \left(\frac{1}{10},
430: \frac12 \right)$ (see \cite{GradshteinRyzhik}), the renormalized coefficient
431: $d$ and $m_0=1$, $m_n=0.1$ for the first and the last masses of the
432: chain, yields for $p$ and $q$:
433:
434: \begin{equation}
435: \label{pqval}
436: p=0.15217 \qquad q=0.68989
437: \end{equation}
438:
439: In Fig. \ref{fig:maxmass} we compare the analytical relation
440: Eq.~(\ref{eq:maxmass}) with the constants given in Eq.~(\ref{pqval}) with numerical results for $m^*$. The numerical data follow from the numerical optimization of the mass distribution for
441: different chain length and different dissipative constants, including all data given in Figs.~\ref{fig:mv22m}-\ref{fig:vepsonv40}. As one can
442: see from Fig.\ref{fig:maxmass}, the results of the analytical theory
443: and of the numerical optimization agree well, except for large
444: dissipation values. We would like to stress that no fitting
445: parameters have been used.
446:
447: \section{Conclusion}
448:
449: We investigated analytically and numerically the transmission of
450: kinetic energy through one-dimensional chains of inelastically
451: colliding spheres. For constant restitution coefficient, $\epsilon=\mbox{const.}$, the distribution
452: of the masses which leads to optimal energy transfer, is an exponentially decreasing
453: function which is independent on $\epsilon$, i.e., it is the same as for elastic particles with $\epsilon=1$.
454:
455: For viscoelastic particles where $\epsilon$ depends on the impact velocity, the optimal mass
456: distribution is not necessarily a monotonous function, but depending on the chain length $n$ and on the
457: material parameters of the spheres it may reveal a pronounced maximum.
458:
459: We develop a theory which describes the total energy losses along the chain, so that the optimal mass distribution,
460: minimizing the losses, may be obtained as a solution of a variational
461: equation. We derived an expression relating the heaviest mass in the chain to the chain length and the dissipation constant. Having no fitting parameters,
462: it is in good agreement with the numerical data.
463:
464: It has been demonstrated before that for the case of
465: ``thermodynamically-large'' granular systems the impact-velocity
466: dependence of the restitution coefficient leads to qualitatively different behavior
467: as compared to systems with $\epsilon=\mbox{const.}$ (e.g.~\cite{Luding,Luding1,TomThor1,NBTPDIFF,veldistr,NBTPage,BPhere,Spahn,Spahn1,Spahn2}).
468: Our system demonstrates that the velocity dependence of the
469: restitution coefficient leads to qualitative modifications in small and simple systems too.
470: Therefore, in general, we believe that the assumption of a constant
471: coefficient of restitution is an approximation which justification
472: cannot be assumed {\em \'a priori} but has to be checked for each
473: particular application.
474:
475: \begin{thebibliography}{99}
476: \addcontentsline{toc}{section}{References}
477: \bibitem{BPhere} N.~V.~Brilliantov and T. P\"oschel, {\em Granular Gases with Impact-velocity Dependent Restitution Coefficient}, (in this volume, page~100).
478: %\pageref{BPpage}).
479:
480: \bibitem{NBTPDIFF} N.~V.~Brilliantov and T.~P\"{o}schel, Phys. Rev. E {\bf 61}, 1716 (2000).
481:
482: \bibitem{veldistr} N.~V.~Brilliantov and T.~P\"oschel, Phys. Rev. E, {\bf 61}, 5573 (2000).
483:
484: \bibitem{TomThor1}
485: T.~Schwager and T.~P\"{o}schel,
486: Phys.~Rev.~E, {\bf 57}, 650 (1998).
487:
488: \bibitem{NBTPage} N.~V.~Brilliantov and T.~P\"{o}schel, preprint %AGE???
489:
490: \bibitem{BSHP}
491: N.~V.~Brilliantov, F.~Spahn, J.-M.~Hertzsch, and
492: T.~P\"{o}schel, Phys.~Rev.~E, {\bf 53}, 5382 (1996).
493:
494: \bibitem{Rosa}
495: R. Ram\'{\i}rez, T. P\"oschel, N. V. Brilliantov, and T.
496: Schwager, Phys.~Rev.~E. {\bf 60}, 4465 (1999).
497:
498: \bibitem{BPchain} N.~V.~Brilliantov and T.~P\"{o}schel, Phys. Rev. E (Nov. 2000, in press), cond-mat/9906138.
499:
500: \bibitem{GradshteinRyzhik}
501: I.~S.~Gradshtein and I.~M.~Ryzhik, {\em Table of Integrals, Series
502: and Products}, 5-th ed., edited by A.~Jeffrey (Academic Press,
503: London, 1965).
504: %Note, that it may be convenient to exploit the relation, ${\cal B}(x,y)=\Gamma(x)\Gamma(y)/\Gamma(x+y)$, which allows to use data for Gamma-function.
505:
506: \bibitem{Luding} S.~Luding, E.~Cl\'ement, A.~Blumen, J.~Rajchenbach,
507: J.~Duran, Phys. Rev. E {\bf 50}, 4113 (1994).
508:
509: \bibitem{Luding1} S.~Luding, E.~Cl\'ement, J.~Rajchenbach, and J.~Duran, Europhys. Lett. {\bf 36}, 247 (1996).
510:
511: %\bibitem{bdef}
512: %If we require the density $\rho$ (and not radii) to be constant for all spheres
513: %Eq. (\ref{epsviab}) reads for particles $i$ and $j$ \begin{equation}
514: %\epsilon=1-2^{-\frac15} b\! \left(\!\frac{4}{\pi\rho}\right)^\frac1{15}\!\!
515: %\left(\frac{m_i^\frac13
516: %m_j^\frac13}{m_i^\frac13+m_j^\frac13}\right)^\frac15\!\!
517: %\left(\frac{m_i+m_j}{m_i m_j}\right)^\frac15 v_{ij}^\frac15\,.\nonumber
518: %\end{equation} Numerical calculations show that the assumption of
519: %equal particle density does not change the results qualitatively,
520: %however, the analysis will be much more complicated if not impossible.
521:
522: \bibitem{Spahn} H.~Salo, J.~Lukkari, and J.~Hanninen, Earth, Moon, and
523: Planets {\bf 43}, 33 (1988).
524:
525: \bibitem{Spahn1} J.~O.~Petzschmann, U.~Schwarz, F.~Spahn, C.~Grebogi, and J.~Kurths, Phys. Rev. Lett. {\bf 82}, 4819 (1999).
526:
527: \bibitem{Spahn2} F.~Spahn, U.~Schwarz, and J.~Kurths, Phys. Rev. Lett. {\bf 78}, 1596 (1997).
528:
529: \end{thebibliography}
530:
531: %INDEX%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
532: %\clearpage
533: %\addcontentsline{toc}{section}{Index}
534: %\flushbottom
535: %\printindex
536: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
537:
538: %\end{document}
539:
540:
541: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
542: %\addtocontents{toc}{\vfill\newpage}
543:
544: %\addcontentsline{toc}{part}{Vibrated Granular Media}
545: %\addtocontents{toc}{\vspace*{-0.07cm}}
546: %\part*{Vibrated Granular Media}
547: %\include{Huntley/Huntley}
548: %\addtocontents{toc}{\vspace*{-0.07cm}}
549: %\include{Clement/Clement} %FINISHED
550: %\addtocontents{toc}{\vspace*{-0.07cm}}
551: %\include{Falcon/Falcon} %FINISHED
552: %\addtocontents{toc}{\vspace*{-0.07cm}}
553: %\include{Farkas/Farkas} %FINISHED
554: %\addtocontents{toc}{\vspace*{-0.07cm}}
555: %\include{Goldshtein/goldshtein} %FINISHED
556: %\addtocontents{toc}{\vspace*{-0.07cm}}
557:
558: %\addcontentsline{toc}{part}{Granular Astrophysics}
559: %\part*{Granular Astrophysics}
560: %\addtocontents{toc}{\vspace*{-0.07cm}}
561: %\include{Brahic/Brahic}
562: %\addtocontents{toc}{\vspace*{-0.07cm}}
563: %\include{Salo/salo} %FINISHED (confirm e-mail)
564: %\addtocontents{toc}{\vspace*{-0.07cm}}
565: %\include{Hanninen/Hanninen} %FINISHED
566: %\addtocontents{toc}{\vspace*{-0.07cm}}
567: %\include{Spahn/Spahn2} %FINISHED
568: %\addtocontents{toc}{\vspace*{-0.07cm}}
569:
570: %\addcontentsline{toc}{part}{Towards Dense Granular Systems}
571: %\addtocontents{toc}{\vspace*{-0.07cm}}
572: %\part*{Towards Dense Granular Systems}
573: %\include{Luding/Luding} %FINISHED
574: %\addtocontents{toc}{\vspace*{-0.07cm}}
575: %\include{Urbach/Urbach} %FINISHED
576: %\addtocontents{toc}{\vspace*{-0.07cm}}
577: %\include{Hong/hong_new} %FINISHED
578: %\addtocontents{toc}{\vspace*{-0.07cm}}
579: %\include{Herrmann/honnef99} %FINISHED
580:
581: %\include{Book/author_index}
582:
583: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
584: %\clearpage
585: %\addcontentsline{toc}{chapter}{Index}
586: %\flushbottom
587: %\printindex
588: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
589:
590: \end{document}
591:
592: