1: \documentclass{article}

2: \usepackage{epsfig}

3: 

4: \renewcommand{\baselinestretch}{1.33}

5: \newcommand{\be}{\begin{equation}}

6: \newcommand{\ee}{\end{equation}}

7: 

8:    \begin{document}

9:    \title{Dynamics of point Josephson junctions in a microstrip line}

10:    \author{  J.G. Caputo and L. Loukitch  \\

11:    {\normalsize \it Laboratoire de Math\'ematiques, INSA de Rouen} \\

12:    {\normalsize \it B.P. 8, 76131 Mont-Saint-Aignan cedex, France }

13:    \date{\today}}

14: 

15: \maketitle

16: \begin{abstract}

17: We model the dynamics of point Josephson junctions in a 1D microstrip

18: line using a wave equation with delta distributed sine nonlinearities.

19: The model is suitable for both low T$_c$ and high T$_c$ systems (0 and

20: $\pi$ junctions).

21: For a single junction in the line, we found two limiting behaviors: the

22: ohmic mode where the junction acts as a pure resistor which stops

23: waves and separates the cavity and the junction mode where the wave

24: is homogeneous throughout the strip. This classification allows to bound the

25: IV curves of the system. Two junctions in a strip give generally ohmic

26: modes and combined junction/ohmic modes and yield information about the

27: behavior with an array with many junctions. Finally we use this analysis

28: to understand the many junction case for 0 and $\pi$ junctions and the

29: effect of an external magnetic field.

30: \end{abstract}

31: 

32: \section{Introduction}

33: 

34: Type I Superconductors are characterized by the phase of a complex

35: number, the macroscopic order parameter. Josephson \cite{josephson} 

36: showed that when 

37: two such superconductors are very close, electrons can tunnel from one

38: to the other according to the relations

39: \be\label{joseph}

40: V = {\Phi_0 \over 2 \pi}{d \phi \over dt}~~, ~~ I = s J_c 

41: \sin(\phi)~,\ee

42: where $V$ and $I$ are respectively the voltage and current across 

43: the barrier, $\phi$ is the difference between the phases in 

44: the two superconductors, $s$ is the contact surface, $J_c$ is the 

45: critical current density and

46: $\Phi_0$ is the flux quantum. 

47: The two Josephson relations together with Maxwell's equations 

48: imply the modulation of DC current by 

49: an external magnetic field in the static regime (SQUIDs) and the 

50: conversion of AC current into microwave radiation. 

51: 

52: 

53: The ratio of the electromagnetic flux to

54: the flux quantum gives the Josephson length, the typical distance of 

55: variation of the phase $\lambda_J = { \Phi_0 \over  L J_c}$.

56: Depending whether their length $l$ is larger or not than $\lambda_J$, 

57: Josephson junctions exhibit significantly different behaviors. 

58: Small junctions for which $l < \lambda_J$ are used as SQUIDS to measure

59: magnetic fields in the static regime or as signal mixers where the 

60: steep IV characteristic at the gap is used to downshift an 

61: unknown high frequency signal by combining it with a known local

62: oscillator signal. 

63: In this latter context Salez et al \cite{Salez} suggested to 

64: use several unequally spaced junctions

65: , instead of just one, to get a better frequency response

66: over a large frequency band. Long Josephson junctions 

67: for which $l >> \lambda_J$ are described by the 1D

68: sine-Gordon equation

69: \be\label{sg1d}

70: \phi_{tt} - \phi_{xx} + \sin (\phi)=0 ,\ee

71: which admit the kink solutions varying from 0 to $2 \pi$. 

72: Long junctions have a characteristic dependence

73: of the maximum critical current in the presence of magnetic field 

74: because of the screening of the applied field by the supercurrent.

75: In the dynamical regime long junctions are used as microwave

76: generators but provide small output power and poor impedance

77: matching to the transmission line. To improve on this, series arrays

78: of junctions have been used, with success since they have been 

79: shown to synchronize

80: in the presence of external forcing \cite{bcsl99} \cite{vbsl02} giving

81: rise to an $N^2$ power output for $N$ junctions. Such behavior

82: has been understood both from the quantum point of view \cite{as01}

83: and the Resistive Shunted Junction (RSJ) equivalent circuit model 

84: \cite{fp02}. The cavity surrounding the junctions

85: stores radiation which can increase the output power of

86: the device. 

87: 

88: The mathematical description of such a system is a wave equation for

89: the cavity coupled to the 2D version of the sine-Gordon equation 

90: (\ref{sg1d}) via interface conditions.  

91: In \cite{bcf02,bc02} one of the authors 

92: studied this coupling for a long junction with a small surrounding passive

93: region. When the cavity is present 

94: only at the junction tip and the miss-match at the interface is

95: not too large, the kink crosses it unaffected and adapts its 

96: speed. When the passive region surrounds the junction, it sets the

97: kink speed. This ballistic kink motion will be absent for small

98: junctions so in the present work we consider the

99: coupling of a large cavity to one or a few small junctions. We want to

100: understand how junctions can talk to each other via the surrounding

101: passive region. Two approximations are assumed, we consider 

102: the 1D limit where transverse modes of the microstrip do not play a role

103: and neglect the phase variation inside the junctions, in other words

104: we neglect the junction length. This is valid for an array of short

105: junctions whose length remains smaller than the wave length of the

106: radiation.

107: 

108: This approach can be applied to high Tc superconductors where due to

109: a different symmetry of the wave function of the Cooper pairs one has

110: current phase relations that involve the higher harmonics \cite{lcti03}.

111: This different symmetry can also induce a negative coupling in (\ref{joseph})

112: leading to so-called $\pi$ junctions. Normal (0) and $\pi$ junctions can

113: be associated to form 0-$\pi$ junctions which have semifluxon solutions

114: that vary between 0 and $\pi$ and are attached to the discontinuity.

115: In the static regime the maximum DC current going through the system 

116: is strongly modulated for a short junction like for a SQUID while it

117: is constant for small field when the junction is longer \cite{l04}. The

118: IV characteristics give resonances that are similar to zero field steps

119: \cite{l04,gsgk04}. Note however the difference between the ballistic motion

120: of a kink in a (0) long Josephson junction and the hopping of semi-fluxons

121: in a (0-$\pi$) junction. As for 0 junctions, (0-$\pi$) junctions can

122: be organized as arrays \cite{r03,sg04} and these have interesting paramagnetic

123: behavior.

124: 

125: The model we introduce allows to describe accurately the coupling between 

126: small junctions in such an array. Both 0 and $\pi$ junctions can be considered

127: and disorder can be added. For simplicity we have considered the large

128: damping limit where it is possible to develop the analysis. We studied in

129: detail the cases of one and two junctions that remain fairly simple

130: and used our conclusions for the case of many junctions. For a single

131: junction, we found two limiting behaviors: the ohmic mode where the junction

132: acts as a pure resistor which stops waves and separates the cavity and 

133: the junction mode where the wave is homogeneous throughout the microstrip.

134: After introducing the model is section 2, we consider these limiting 

135: behaviors in sections 3,4 and 5.

136: These allow to understand in detail the features of the IV characteristics 

137: (section 6). The behavior of two junctions shown in section 7 

138: follows generally ohmic modes

139: and combined junction/ohmic modes. Using this, in section 8 we 

140: generalize the study to 

141: many junctions embedded in a microstrip and understand the influence of

142: the sign of the current density (0 or $\pi$ junctions) and external 

143: magnetic field. Conclusions are presented in section 9.

144: 

145: 

146: \section{The model}

147: 

148: The device we model shown in the top panels of Fig. \ref{f1} (top view 

149: on the left and section on the right) is a narrow microstrip in

150: which are embedded a number of small junctions in the window

151: design. We proceed from the description of the two superconductors 

152: as two inductance arrays connected by capacitive elements (resp. and

153: a resistor and nonlinear Josephson element) in the passive (resp. junction)

154: region. The equivalent circuit is shown in the bottom panel of Fig. 

155: \ref{f1}. In the following subsections we show that the continuum limit

156: yields an inhomogeneous 2D sine-Gordon equation which we reduce to 1D

157: when the microstrip is narrow. For small junctions we then justify our delta

158: function approach (\ref{sgd}).

159: 

160: 

161: \subsection{RSJ model for distributed Josephson junctions}

162: 

163: A simple mathematical model of the window Josephson junction is to 

164: describe

165: each superconductor by an array of inductances $L$. The coupling 

166: elements

167: between two adjacent nodes in each array are, a capacitor $C$, 

168: resistance $R$

169: and Josephson current $I_c$ \cite{Likharev}. The Kirchoff laws at 

170: each couple of nodes $(i_b,i_t)$ in the bottom and top superconducting 

171: layers can be 

172: combined to give the relation expressing the conservation of currents 

173: at node $i$ in the device

174: \begin{equation}

175: \label{discr}

176: C_i \ddot \Phi_i + \sum_j  \frac {\Phi_i - \Phi_j}{L} + 

177: I_i^c \sin \frac {\Phi_i}

178: {\Phi_0} +  \frac {\dot \Phi_i}{R_i} = 0 \; ,

179: \end{equation}

180: where $\Phi = \Phi_t - \Phi_b$ (resp. $\Psi = \Psi_t - \Psi_b$), 

181: is the phase difference between the two superconductors 

182: in the junction (resp. passive) part,  the summation $\sum_j$ is 

183: applied to 

184: the nearest neighbors, $I_i^c$ is the critical current which is non 

185: zero

186: only in the junctions and $R_i$ is the resistance associated to the

187: current of quasi-particles again non zero

188: only in the junctions. In agreement with experiments, we have 

189: assumed the same surface inductance for

190: the whole sample but have taken into account the variation of the 

191: capacity

192: $C$ and resistance $R$ from the junctions to the rest of the cavity.

193: Note that equation (\ref{discr}) is a discretisation of Maxwell's 

194: equations 

195: (wave equation part) and Josephson constitutive equations (sine term), 

196: assuming an electric field normal to the plates, a magnetic field in 

197: the 

198: junction plane and perfect symmetry

199: between the top and bottom superconducting layers. We can now obtain 

200: the model

201: in the continuum limit, more suitable for analysis.

202: 

203: \subsection{Continuum limit}

204: 

205: The continuum version of the system (\ref{discr})

206: can be derived by introducing the following quantities

207: per  unit area ($a^2$) of elementary cells 

208: of length $a$.

209: \begin{equation}

210: \label{normcjr}

211: \overline {C_{J}}=\frac {C_{J}} {a^2},~~~j_i^c= \frac {I_i^c} {a^2},~~~

212: \overline{r_i} = R_i a^2.~~~

213: \end{equation}

214: We normalize the phases by the flux quantum $\Phi_0$,

215: \begin{equation}

216: \label{normphi}

217: \varphi_i=\frac {\Phi_i}{\Phi_0} 

218: \end{equation}

219: and introduce the Josephson characteristic length 

220: \be \label{lambdaj}

221: \lambda_J^2=\frac {\Phi_0}{j_c L_J} .

222: \ee

223: Notice that in this 2D problem the inductance associated to each cell

224: is equal to the branch inductance $L$. 

225: To differentiate the junction regions from the passive regions we

226: introduce the indicator function $g_i$ such that $g_i=1$ in the

227: junctions and $0$ elsewhere. We can then write

228: \be\label{indic}

229: j_i^c =  g_i j^c~~,~~

230: {1\over \overline{r_i}}=g_i{1\over \overline{r}}

231: ~~{\rm and} ~~ 

232: {\bar C_i}= g_i{\bar C_J}  +(1-g_i){\bar C_I}

233: \ee

234: where $j^c$, $\overline{r}$, ${\bar C_J}$ and ${\bar C_I}$ 

235: are respectively the critical current densities, the quasi 

236: particle resistance in the junction, the capacitance per unit area in

237: the junction and the capacitance per unit area in

238: the passive region.

239: 

240: We  substitute relations (\ref{normcjr}) - (\ref{indic}) in 

241: (\ref{discr}) 

242: and obtain 

243: \begin{equation} \label{discre2}

244: L {\overline C_I} \ddot \varphi_i 

245: +  \sum_j \frac {\varphi_i - \varphi_j}{a^2}

246: + g_i \left [ L({\overline C_J}-{\overline C_I}) \ddot \varphi_i 

247: + {1 \over \lambda_J^2} \sin \varphi_i  

248: +\frac {L}{\overline{r}} \dot \varphi_i \right ] = 0 \; ,

249: \end{equation}

250: which in the continuum limit yields

251: \begin{equation} \label{cont1}

252: L{\overline C_I} \varphi_{tt}

253: -\Delta \varphi 

254: + {1 \over \lambda_J^2}g(x,y) \left [

255: L \lambda_J^2 ({\overline C_J}-{\overline C_I}) \varphi_{tt}

256: + \sin \varphi 

257: + \frac {L \lambda_J^2}{\overline{r}} \varphi_t \right ]=0 . 

258: \end{equation}

259: To obtain the final equation we rescale time by the inverse of the

260: plasma frequency $\omega_I^{-1}= \lambda_J \sqrt{ L {\overline C_I}} $

261: so that $\tilde t = \omega_I t$. As unit of space we use the

262: Josephson length, 

263: $ \tilde x = x/\lambda_J,~\tilde y =y/\lambda_J $. 

264: The final equation is then 

265: \be

266: \label{insg}

267: \varphi_{tt} -\Delta \varphi

268: + g(x, y)( c \varphi_{tt} + \sin \varphi + \alpha \varphi_{t})=0 ~,

269: \ee

270: where the

271: tildes have been omitted for simplicity

272: and where the coefficients $c$ and $\alpha$ are

273: \be\label{calpha} 

274: c = {{\overline C_J}\over{\overline C_I}} -1~~,~~ 

275: \alpha = {\lambda_J \over {\bar r}  }\sqrt{L \over {\overline C_I}}\ee 

276: 

277: The boundary conditions represent the input of an external

278: current $I$ or magnetic field $b$ on the device. They are

279: \be\label{overlap}

280: \varphi_x|_{x=0}=b -(1-\nu){I\over 2 w}~~,~~ 

281: \varphi_x|_{x=l}=b +(1-\nu){I\over 2 w}~~,~~

282: \varphi_y|_{y=0}= -\nu {I\over 2l}~~,~~

283: \varphi_y|_{y=w}= \nu {I\over 2l},\ee

284: where $0<\nu<1$. The case $\nu=1$ corresponds to a pure 

285: inline current feed and $\nu=0$ to a pure overlap current feed.

286: We will mostly consider the latter case in the discussion. 

287: 

288: 

289: \subsection{Reduction to a 1D problem}

290: 

291: 

292: When the microstrip is narrow and the junctions occupy the

293: whole width we can assume that $g$ is independent of $y$.

294: We can then write 

295: \be \varphi(x,y,t) = \nu I {y^2 \over 2s} 

296: + \sum^\infty_{n=0} \phi_n(x,t) \cos \left({n \pi y \over  w}\right), 

297: \ee

298: where the first term takes care of the current feed and the second

299: satisfies homogeneous Neumann boundary conditions for $y=0,w$.

300: 

301: 

302: The calculations detailed in \cite{bcf02} show that for

303: a narrow width and small current only the first mode $\phi_0$

304: needs to be taken into account. We then obtain the following

305: equation for $\phi_0$ where the 0 has been dropped for

306: simplicity

307: \be\label{final1d}

308: {\phi}_{tt} -{\phi}_{xx}

309: + g(x)( c {\phi}_{tt} + \sin {\phi} + \alpha {\phi}_{t})=\nu{I \over s} 

310: ~,

311: \ee

312: with the boundary conditions 

313: ${{\phi}_x}_{x=0}= -(1-\nu){I l \over 2 s}$  ~~

314: ${{\phi}_x}_{x=l}= (1-\nu){I l \over 2 s}$.  

315: In the following we will write

316: \be\label{curdens}

317: j = {I \over s},\ee

318: and assume an overlap current feed ($\nu=1$) except in section 7.

319: 

320: \subsection{The delta function model}

321: 

322: As the area of the junction is reduced the total super-current is

323: reduced and tends to zero. To describe situations where the area 

324: is small so that the variations of the phase can be neglected in the 

325: junction but the supercurrent is significant we introduce the

326: following delta function model.

327: Introduce the function $g(x) = d_a g_h(x)$\\

328: $g_h(x) = \left\{

329: \begin{array}{l r}

330: 1/(2h) & a-h<x<a+h \\

331: 0 & {\rm elsewhere} \\

332: \end{array} \right.$

333: 

334: \begin{equation} \label{sggh}

335: \phi_{tt}- \phi_{xx} + d_a g_h(x) \left(\sin(\phi)

336: + \alpha\phi_t + c \phi_{tt} \right) =j.

337: \end{equation}

338: In the following we assume $\phi$ to be $C^2$ in $x$ and $t$ and

339: $\phi_{ttx}$ to be continuous.

340: We now show that equation (\ref{sggh}) converges when

341: $h\rightarrow 0$ to

342: \begin{equation} \label{sg2}

343: \phi_{tt}- \phi_{xx} + d_a \delta(x-a)\left(\sin(\phi)

344: + \alpha\phi_t + c \phi_{tt} \right) = j.

345: \end{equation}

346: For that introduce the Taylor expansion of $\phi$ near $x=a$

347: \begin{equation}\label{taylor}

348: \phi(x,t)=\phi(a,t) + (x-a)\phi_x(a,t) + O((x-a)^2) .

349: \end{equation}

350: In the following for simplicity of writing we will omit the

351: $t$ and $x$ dependencies. We integrate (\ref{sggh}) from $a-h$ to $a+h$

352: (see details in appendix: Continuum limit). The

353: first term 

354: $$\int_{a-h}^{a+h} \phi_{tt} dx \rightarrow O ~~{\rm when} ~~h 

355: \rightarrow 0, $$

356: the second one

357: $$\int_{a-h}^{a+h} \phi_{xx} dx=\phi_x(a+h)-\phi_x(a),$$

358: and the third one

359: $$\lim_{h\rightarrow 0}{d_a \over 2 h}\int_{a-h}^{a+h}

360: (\sin(\phi)+ \alpha\phi_t + c \phi_{tt})dx = d_a

361: \left( \sin(\phi(a))+ \alpha\phi_t(a) + c \phi_{tt} (a) \right),$$

362: when $h \rightarrow 0$ which is consistent with the model

363: \begin{equation} \label{sgcd}

364: \phi_{tt}- \phi_{xx} + d_a \delta(x-a) \left( \sin(\phi) + \alpha\phi_t 

365: +

366: c \phi_{tt} \right)= j,

367: \end{equation}

368: together with the boundary conditions 

369: $\phi_x\left (x=0\right)=0~,~~~

370: \phi_x\left (x=l,t\right)=0.$

371: 

372: In the following we will consider the case $c=0$ for simplicity. We 

373: assume the

374: same capacity in the junction and in the passive region

375: so that the final equation is

376: \be\label{sgd}

377: \phi_{tt}- \phi_{xx} + d_a \delta(x-a) \left( \sin(\phi) + \alpha\phi_t 

378: \right)= j.

379: \ee

380: This equation can be considered as a 1D model of a point Josephson

381: junction embedded in a microstrip cavity.

382: 

383: \section{The single junction case}

384: 

385: The partial differential equation (\ref{sgd}) involves a distribution, 

386: however

387: its solution is differentiable except at the junction points.

388: To solve it numerically we integrate the equation

389: over reference intervals using the finite volume approach.

390: The resulting system of ordinary differential equations is then solved

391: using the DOPRI5 4th and 5th order Runge-Kutta variable step 

392: integrator developed by Hairer and Norsett \cite{hairer}.

393: Details of the numerical code are given in the second part of the appendix.

394: 

395: We now consider the two limiting cases of a static solution and

396: high voltage for which simple expressions of

397: the solution can be obtained. In particular these have been used to 

398: validate our numerical procedure.

399: 

400: \subsection {Static behavior}

401: 

402: We assume $\phi_t = \phi_{tt} \equiv 0$ in (\ref{sgd}) and get

403: \begin{eqnarray}

404: \label{statique}

405: -\phi_{xx} + \delta (x-a) d_a \sin (\phi) = j \\

406: \phi_x|_{x=0,l} =0

407: \end{eqnarray}

408: For $x\neq a$, $-\phi_{xx}=j $ so that for

409: $x<a$, $\phi= \phi_l \equiv -\frac{j}{2}x^2 + C_l x + D_l$ and for

410: $x>a$, $\phi= \phi_r \equiv -\frac{j}{2}x^2 + C_r x + D_r$.

411: Integrating (\ref{statique}) over the whole domain we get the

412: conservation of current

413: \be\label{conscur} d_a \sin(\phi(a))= j l,\ee

414: so that no solution exists for $|jl| > d_a$.

415: 

416: The solution is obtained as usual by assuming continuity of the 

417: phase and using the jump condition on the gradient obtained 

418: by integrating on a small domain including

419: the junction

420: \be\label{jump}[\phi_x]_{a^-}^{a^+}=d_a \sin(\phi(a)).\ee

421: Using these two constraints we obtain the static solution

422: \begin{equation}

423: \phi_s(x)=  \left\{\begin{array}{lr}

424: \frac{j}{2}(a - x)(a + x) + \arcsin \left ({j l \over d_a}\right) & 

425: x<a\\

426: \frac{j}{2}((a - x)(x+a+2l) + \arcsin\left({j l \over d_a} \right) & 

427: a<x

428: \end{array}\right .\end{equation}

429: 

430: \subsection{High voltage behavior}

431: 

432: Since the Josephson current is bounded, its influence in (\ref{sgd})

433: will decrease as the current $j $ is increased. In this limit

434: of high current the phase is growing very fast so that $\phi_t$

435: is close to the average voltage $V\equiv \left<\phi_t\right>_t$.

436: We can write approximately $\phi(x,t)= V t +\phi_v(x)$ so that

437: we can average (\ref{sgd}) assuming $\left<\phi_t(x,t)\right>_t = V$, 

438: $\left<\sin(\phi(a))\right>_t=0$ and $\left<\phi_{tt}\right>_t=0$, 

439: yielding

440: \be \label{hvolt}

441: -{\phi_v}{xx} + \delta\left(x-a\right)d_a \alpha V= j ,\ee

442: with the boundary conditions ${\phi_v}_x|_{x=0,l}=0$.

443: Integrating this equation over the whole domain 

444: we get $V= {j l \over d_a \alpha}$.

445: 

446: Using the same constraints as in the static case (continuity of the

447: phase and jump of its gradient at the junction) we get up to

448: a constant the high voltage solution $\phi_v$

449: \begin{equation}

450: \phi_v(x,t)= \left\{\begin{array}{lr}

451: {j\over 2 }(a-x)(a+x) + Vt + C & x<a \\

452: {j\over 2 }(a-x)(a+x-2l)+ Vt + C & a<x

453: \end{array}\right .\end{equation}

454: Notice that this and precedent solutions have the same spatial 

455: behavior;

456: $\phi_s-\phi_v = Vt + C$. We will not find this result with more 

457: junctions

458: (except in particulary cases).

459: 

460: \subsection{Fourier representation }

461: 

462: Because of the homogeneous Neumann boundary conditions one can write the

463: solution as a cosine Fourier series

464: \begin{equation}\label{four}

465: \phi(x,t) = \sum_{n=0}^{+\infty} A_n(t)\cos\left(\frac{n\pi x}{l} 

466: \right).

467: \end{equation}

468: The solution of (\ref{sgd}) has a discontinuous first spatial

469: derivative to that the amplitude of the mode $A_n$ decreases as $1/n^2$ 

470: \cite{carslaw}. This means that in general many modes are needed 

471: for the description of the solution. However since the system is

472: close to a linear one we will see that these modes provide insight into

473: the limiting behaviors.

474: 

475: 

476:  Plugging (\ref{four}) into (\ref{sgd}) and projecting we get the

477: evolution of the modes, 

478: \begin{equation}\label{a0}

479: l {A_0}_{tt} + d_a (\sin(\phi_a) + \alpha \phi_{at})

480: -j l =0 ,\end{equation}

481: \begin{equation}\label{an}

482: {A_n}_{tt} + \left({n \pi  \over l}\right)^2  A_n

483: + {2 d_a \over l } c_n ( \sin(\phi_a) + \alpha \phi_{at})

484: =0 ,  \end{equation}

485: where we have introduced the coupling coefficients 

486: associated to the junction at $x=a$,

487: $c_n=\cos\left(\frac{n\pi a}{l} \right)$.

488: Equations (\ref{a0}) and (\ref{an}) are coupled through the

489: term $\sin(\phi_a) + \alpha {\phi_a}_t$.  Due to this particular

490: feature, two limiting behaviors can be expected.

491: \begin{itemize}

492: \item The ohmic mode for which

493: \be\label{omode} 

494: d_a  \alpha \phi_a|_t - j l=0 ~.\ee

495: Then the equations for the $A_n$ have a right hand side so that

496: the cavity modes are driven by the junction.

497: 

498: \item The junction mode such that

499: \be \label{jmode} 

500: d_a ( \sin(\phi_a) + \alpha {\phi_a}_t)  -j l =0 ~.\ee

501: In this case the equations for the $A_n$ have a constant right hand 

502: side 

503: ($= j l$). We will show that in this case the cavity is synchronized 

504: with the junction.

505: \end{itemize}

506: 

507: The equations giving the evolution of $\phi_a$ in these two 

508: limiting behaviors can be solved yielding for the ohmic mode

509: \be\label{som}

510: \phi_a = { j l  \over d_a \alpha} t + C_1 ~, \ee

511: where $C_1$ is a constant and a voltage

512: \be\label{vohm}

513: V_o = { j l  \over d_a \alpha}.\ee

514: For the junction mode the

515: equation can be solved using separation of

516: variables and we get 

517: \be \label{sjm} \phi_a =2\arctan\left\{ 

518: {d_a \over j I } \left[ \alpha V_j    \tan

519: \left( {V_j \over 2 } \left(t+C_2\right)

520:  \right) +1 \right] \right\}~~,\ee

521: where the junction mode voltage $V_j$ is 

522:  \be\label{vjon}

523:  V_j = {1 \over \alpha}\sqrt{\left( {j l \over d_a}\right)^2 -1 }\ee

524: where $C_2$ is a constant.

525: 

526: Fig. \ref{f2} shows these two limiting behaviors for a junction

527: placed in $a=-1.57$ with $d_a=1$ $l=10$ and a current $j\approx 

528: 0.1374$.

529: The numerical solution is presented with the crosses while 

530: the analytical estimates (\ref{som}) and (\ref{sjm}) are shown 

531: in continuous line.

532: The left panel corresponds to the ohmic mode and one can

533: see the excellent agreement of the numerical solution with the

534: line given by (\ref{som}). The right panel presents

535: the junction mode and the extension of the estimate (see below), again 

536: the

537: agreement is very good. 

538: Notice that for the same time interval the variation of $\phi_t(a,t)$ 

539: is different for the two panels of Fig. \ref{f2} so that we will get 

540: a different average voltage for these two behaviors.

541: 

542: 

543: \section{The ohmic mode}

544: 

545: In this case $\phi_t(a,t)=V_o={j l \over \alpha d_a}$. To get

546: the field we need to solve on each subdomain $[0,a[$ and $ ]a,l]$

547: the following problems

548: $$\begin{array}{cc}

549: \left\{\begin{array}{l}

550: \phi_{tt}-\phi_{xx}= j  \\ 

551: \left .\phi_{x}\right|_{0}=0,~

552: \left .\phi_{t}\right|_{a}=\frac{j l }{\alpha d_a}

553: \end{array}\right.&

554: \left\{\begin{array}{l}

555: \label{onde1}

556: \phi_{tt}-\phi_{xx}= j \\ 

557: \left .\phi_{t}\right|_{a}=\frac{j l }{\alpha d_a},~

558: \left .\phi_{x}\right|_{l}=0 .

559: \end{array}\right.\\

560: \end{array}$$

561: Consider the left subdomain $[0,a[$, we introduce 

562: $\phi\left(x,t\right)=v\left(x,t\right)+\frac{ j l t}{\alpha d_a}$

563: so that if $\phi$ solves the left problem then $v$ solves

564: \begin{eqnarray}

565: \label{onde2}

566: v_{tt}-v_{xx}= j \\

567: \left. v_x \right |_{0}=0~\left. v_t\right|_{a}=0

568: \end{eqnarray}

569: A particular solution of (\ref{onde2}) is $v_l=

570: -\frac{j}{2}x^2 + D_l$ where

571: $D_l$ is an integration constant.

572: Then $w= v -v_l$ solves the homogeneous wave equation 

573: $$\partial_{tt} w-\partial_{xx} w=0 , $$ 

574: with $w_x|_0=0$ and $w_t|_a=0$.

575: We can assume without loss of generality that

576: $w\left(x,0\right)=0$.

577: The condition $w_x|_0 =0$ imposes a solution

578: of the type 

579: \begin{equation}

580: w\left(x,t\right)=a_l\sin\left(V_l t\right)

581: \cos\left(V_l x\right) ,

582: \end{equation}

583: where $V_l$ is a constant. Also, $w_t|_{a}=0$ imposes a

584: condition on $V_l$ because 

585: $$ w_t\left(x=a,t\right) =a_l \sin\left(V_l t\right) \cos(V_l a)=0$$ 

586: is possible for all $t$'s only if  $\cos\left(V_l a\right)=0$. So 

587: that we get $$V_l={ 2k_l+1 \over 2 }\frac{\pi}{a}.$$

588: The problem is identical for $x \in \left]a,l\right]$

589: and using similar arguments we 

590: get $v_r=-\frac{j}{2}\left(x-l\right)^2 + D_r$ and

591: $$V_r={2k_r+1 \over 2} \frac{\pi}{a-l} .$$

592: We obtain a solution for the equation (\ref{sgd}).

593: The phase on the whole domain is then

594: \begin{equation}

595: \label{totalcos}

596: \phi\left(x,t\right) =  \left\{ \begin{array}{l r}

597: \frac{j l t}{\alpha d_a} + a_l \sin\left(V_l t\right)\cos\left(V_l 

598: x\right) + \frac{j}{2} (a-x)(a+x)  & x<a \\

599: \frac{j l t}{\alpha d_a} + a_r \sin\left(V_r t\right)\cos\left(V_r

600: \left(x-l \right) \right) + \frac{j}{2} (a-x)(a+x-2l) & a<x

601: \end{array} \right. \end{equation}

602: From the problem (\ref{sgd}) integrating the equation on a small 

603: interval

604: centered on $x=a$ and using (\ref{omode}) we get the jump condition 

605: $$[\phi_x]_{a^-}^{a^+} -j l

606: = d_a\sin(\phi(a,t)) = d_a\sin (\frac{j l t}{\alpha d_a}).$$

607: On the other hand from (\ref{totalcos}) we get 

608: $$\begin{array}{ccl}

609: \left[\phi_x\right]_{a^-}^{a^+} -j l &=&  a_l 

610: \sin\left(V_l t\right)

611: \sin \left( V_l a\right)-a_r V_r \sin\left(V_r t\right)

612: \sin ( V_r\left(a-l\right)) \\

613:   & \equiv & C_1 \sin\left(V_l t\right)-C_2 \sin\left(V_r t\right)\\

614: \end{array}$$

615: Since $C_1$ and $C_2$ are independent of $t$, this implies that 

616: $\sin(V_l t)$, $\sin(V_r t)$ and  $\sin({j l t\over \alpha d_a})$ have 

617: the 

618: same period. We obtain, 

619: \be \label{VleftVright}

620: { 2k_l+1 \over 2 }\frac{\pi}{a}= {j l \over \alpha d_a}=

621: {2k_r+1 \over 2} \frac{\pi}{a-l}=V,\ee

622: if $a_l$ and $a_r$ are not equal to zero. Then one can write the

623: solution as

624: \be\label{leftaright}

625: \phi(x,t) =  \left\{ \begin{array}{l r}

626: \phi_v(x,t) + a_l \sin\left(V t\right)\cos\left(V x\right) & x<a \\

627: \phi_v(x,t) + a_r \sin\left(V t\right)\cos\left(V \left(x-l \right) 

628: \right) & a<x.

629: \end{array}\right .\ee

630: where $ \phi_v$ is the high voltage solution. Because of this the 

631: instantaneous

632: voltage $\phi_t$ has a discontinuous first derivative at $x=a$.

633: 

634: 

635: Assuming that $\alpha=1$ we see that the consistency equation 

636: (\ref{VleftVright}) 

637: implies 

638: $$(a-l) 2 k_l - 2a k_r = l.$$

639: There are very few solutions for $a$ rational et $l$ integer and for 

640: most $a$ there are none. Let us look at other possibilities. Notice 

641: first

642: that space homogeneous functions ($a_l=0$ or $a_r=0$) are solutions. 

643: Having said this

644: it is easy to see that one can obtain left only or right only 

645: oscillations. \\

646: left only: $\phi(x,t) =  \left\{ \begin{array}{l r}

647: \phi_v(x,t) + a_l \sin(V t)\cos(V x) & x<a \\

648: \phi_v(x,t) & a<x

649: \end{array}   \right .$\\

650: right only:

651: $\phi(x,t) =  \left\{ \begin{array}{l r}

652: \phi_v(x,t) & x<a \\

653: \phi_v(x,t) + a_r \sin(V t)\cos\left(V\left(x-l \right)\right) & a<x

654: \end{array}   \right .$\\

655: Again the instantaneous voltage at the junction has a discontinuous

656: first derivative.

657: 

658: Fig. \ref{f3} shows the instantaneous voltage as a function of $x$ for

659: different times for two such ohmic modes, the left only oscillation 

660: (left panel)

661: and right and left oscillations (right panel). The left panel shows

662: $\phi_t(x,t)$ for 60 values of time equidistributed between 0 and 10.

663: The parameters are the same as for Fig. \ref{f2}. In this case we have 

664: $k_l=1$

665: so that about 1.5 periods exist to the left of $x=a$. Notice how the 

666: amplitude

667: of $\phi_t$ is practically constant and equal to the average voltage

668: $V=j l =1.37$ on the right of $x=a$. The right panel corresponds to a 

669: right and left oscillation for $a=2$ and we show three consecutives 

670: times, 

671: $t=0, 0.125$ and $0.250$. Here ${ 2 k_l+1 \over 2 k_r +1} = {a \over 

672: l-a}=3/7$

673: so that we get about 1 period to the left of $x=a$ and two periods to

674: the right of $x=a$. We show for comparison

675: $V=\frac{\pi}{2}\left(=j l \right)$ the average voltage. Notice in both 

676: cases

677: the discontinuity of $\phi_t$ at $x=a$.

678: 

679: 

680: Let us show that this is consistent with

681: the whole equation (\ref{sgd}), for the intensity $I=V/ \alpha d_a$.

682: Consider the left only solution

683: \begin{equation}

684: \label{gauchecos}

685: \phi\left(x,t\right)=  \left \{ \begin{array}{l r}

686: Vt+ \frac{a_l}{V}\sin(V t)\cos(Vx)-\frac{j}{2}[x^2-a^2]& x < a \\

687: Vt-\frac{j}{2}\left[(x-l)^2+(a-l)^2\right]& x > a

688: \end{array}   \right .

689: \end{equation}

690: with, $V=\frac{2k_l+1}{2}\frac{\pi}{a}$.

691: On the interval $[0,a[$ (respectively $]a,l]$) $\phi$ satisfies

692: the linear wave equation 

693: $\phi_{tt}-\phi_{xx} = j$ with the boundary condition

694: $\phi_x|_{0}=0$ (respectively $\phi_x|_{l}=0$).

695: We now integrate (\ref{sgd}) on $[0,l]$ to obtain

696: \be

697: \int_{0}^a \phi_{tt}dx+\int_a^{l}\phi_{tt}dx +

698: \int_{0}^{l}\phi_{xx}dx+d_a\sin\left(Vt\right)+\alpha d_a V = jl, \ee

699: so that

700: $$-a_l \sin(V t)\left[\sin(Vx)\right]_0^a + 0 + 0 + 

701: d_a\sin\left(V t\right) = 0.$$

702: The end result is 

703: $$(-1)^{k_l+1} a_l \sin(V t)=d_a\sin(V t)$$

704: which implies $a_l=(-1)^{k_l+1} d_a$ and the equation is

705: balanced.

706: 

707: \section{The junction mode}

708: 

709: The expression (\ref{sjm}) a priori defined only for a given time 

710: interval

711: can be $C^{\infty}$ extended for all times using the continuation

712: $$f \left(t+k\frac{2\alpha \pi}{\sqrt{(j/d_a)^2-1}}  \right)= 

713: f \left(t\right)+k 2\pi ~,$$

714: where $k$ is an integer.

715: The pseudo-period $T$ of $\phi_a$ is $T=\frac{2\alpha 

716: \pi}{\sqrt{(j/d_a)^2-1}}$

717: so the voltage is $V=\frac{2\pi}{T}={1 \over 

718: \alpha}\sqrt{(j/d_a)^2-1}$. 

719: This can also be written

720: \be\label{ji}

721: j = {d_a \over l} \sqrt{\left(\alpha V\right)^2+1}\ee

722: So we deduce from $A^{''}_0(t)=0$ and the voltage expression 

723: (\ref{vjon}) 

724: that $A_0(t)=Vt+C$.

725: 

726: Let us now calculate the contribution of the

727: Fourier modes and 

728: give a solution to the whole equation (\ref{an}). From (\ref{an})

729: and the previous equality, we obtain: 

730: \be \label{a_pjunc}

731: A_p^{''}\left(t\right) +  \left(\frac{p \pi}{l}  \right)^2

732: A_p\left(t\right) = - 2 j c_p

733: \ee

734: The solution of this equation, $A_p$ is:

735: \be \label{solA_pjonc}

736: A_p \left(t\right)= \gamma_p \cos \left(\frac{p \pi}{l}t \right) 

737: + \beta_p\sin \left(\frac{p \pi}{l}t \right)

738: -\frac{2 j l^2}{(p \pi)^2} c_p

739: \ee

740: Now we have the expression of all the Fourier modes. 

741: We know that $\phi(a,t) = \sum_{n=0}^{+\infty}A_n(t) c_n^a$.

742: Substituting the Fourier modes (\ref{solA_pjonc}) into 

743: this expression we obtain:

744: \be \label{pjmi} 

745: \phi(a,t)=Vt+C+\sum_{p=1}^{+\infty}\left[\gamma_p \cos \left(\frac{p 

746: \pi}{l}t \right) 

747: + \beta_p \sin \left(\frac{p \pi}{l}t \right)-\frac{2 j l^2}

748: {(p \pi)^2} c_p\right]c_p ,

749: \ee 

750: The series in the last term is uniformly convergent so that the term 

751: can be

752: collected in the form of a constant yielding the final simpler 

753: expression

754: \begin{equation}\label{pjm} \phi(a,t)=Vt+C_2+\sum_{p=1}^{+\infty}

755: \left[\gamma_p c_p\cos \left(\frac{p \pi}{l}t \right) 

756: + \beta_p c_p \sin \left(\frac{p \pi}{l}t \right)\right]

757: \end{equation}

758: Introduce $\phi(a,t)-Vt+C_2 = u(t)$. We can choose $C_2$ as $\int_0^T 

759: u(t) dt =0$.

760: Notice that $u$ is a $C^{\infty}$ and $T$-periodic function. 

761: We can make Fourier projection on it. 

762: $$u(t) = \sum_{p=1}^{+\infty}\left[\gamma_p c_p\cos \left(\frac{p 

763: \pi}{l}t \right)

764: + \beta_p c_p \sin \left(\frac{p \pi}{l}t \right)\right] $$

765: Until $c_p \neq 0$, we have

766: \begin{eqnarray}

767: \gamma_p = \frac{2}{c_p T}\int_0^T u(t) \cos\left(\frac{p 

768: \pi}{l}t\right) dt \\

769: \beta_p = \frac{2}{c_p T}\int_0^T u(t) \sin\left(\frac{p 

770: \pi}{l}t\right) dt 

771: \end{eqnarray}

772: We can make the following remarks 

773: \begin{enumerate} 

774: \item $V=\frac{2\pi}{T}=\frac{n\pi}{l}$ so $\frac{1}{T}=\frac{n}{2l}$ 

775: and $u$ is $\frac{2l}{n}$ periodic.

776: \item If $V=\frac{n\pi}{l}$ the terms of the Fourier series that aren't

777: zero are the multiples of $n$.

778: \item If $c_n=0$ then the junction modes for $V=\frac{n\pi}{l}$ can't 

779: exist.

780: For example when the junction is in the middle of the circuit, then 

781: $c_n^a=0$

782: for $n=2k+1$ and junction modes exist only for $V=\frac{2k\pi}{l}$ 

783: where $k$ is 

784: an integer (see left panel of figure \ref{f5}).

785: \item When $|c_n|$ is very small, the circuit in the junction mode 

786: stores

787: a lot of energy but is unstable. We will see this in the next section.

788: \end{enumerate}

789: Now remember (\ref{four}), we can give the explicit solution of the 

790: phase 

791: in the $n$ th junction mode, up to a constant

792: \begin{equation} \label{modjonc}

793: \phi(x,t) = {n \pi \over l} t + \sum_{p=1}^{+\infty} 

794: \left [\gamma_p \cos \left(\frac{p n \pi}{l}t\right) + 

795: \beta_p\sin \left(\frac{p n \pi}{l}t \right) \right ]

796: \cos\left(\frac{p n\pi}{l} x \right) ~.

797: \end{equation}

798: Notice that the instantaneous voltage $\phi_t$ has a continuous 

799: derivative at the junction position $x=a$.

800: Fig.  \ref{f4} shows the instantaneous voltage $\phi_t(x,t)$ as a 

801: function of $x$ for 60 consecutive times for the same parameters as

802: in Fig. \ref{f2}. Notice how the solution oscillates over the whole

803: domain and that one cannot see the junction position. This type of

804: motion is similar to the transparency observed in the reflection

805: coefficient of a stripe \cite{lamb} when an integer number of 

806: half-periods

807: of the wave "fits" in the stripe.

808: 

809: \section{Single junction case: IV characteristics}

810: 

811: For a large enough time the system finds a stationary state where the

812: energy provided by the input current is balanced by dissipation. This 

813: state 

814: is described by the IV characteristic curve. This is measured

815: on real devices and it is therefore very important to understand its 

816: features. 

817: 

818: To compute it we fix the current and run the system for about 

819: $t_1=3000$ 

820: time units so that

821: all transients have died off. Then we compute the average voltage

822: \be\label{avolt}

823: \left<\phi_t(a,t)\right>_t\equiv 

824: {\phi(a,t_1+T)-\phi(a,t_1)\over T}\ee 

825: at the junction. The value $T$ is about 2000. The space average of the 

826: average voltage in time 

827: $\left<\frac{1}{l}\int_{0}^{l} \phi_t (x,t) dx\right>_t$

828: is compared to this quantity to ensure that the

829: system is completely thermalized. Only then is the voltage recorded.

830: 

831: The IV curves are computed for 99 steps of increasing current $j$ from 

832: 0 to 4

833: starting with an initial condition $\phi(x,0) \equiv 0$ and 

834: $\phi_t(x,0) \equiv 0$. We then decrease the current from 4 to 0.

835: All plots show these two curves for increasing and decreasing current.

836: 

837: \subsection{Bounds for IV curves}

838: 

839: We have isolated two characteristic behaviors of the circuit. These 

840: allow

841: us to bound the IV curves for a single junction in a

842: microstrip.  Fig. \ref{f5} shows two IV curves for 

843: a centered junction $a=5$ (left panel) and an off-centered

844: junction $a=9$ in a microstrip of total length $l=10$. 

845: For each case we plot the limiting behaviors given by the

846: ohmic mode and junction mode voltages (\ref{vohm}) and (\ref{vjon}).

847: $V, j$ are bounded by the relations 

848: $$ d_a \alpha V \leq jl \leq d_a \sqrt{(\alpha V)^2 + 1}~,~{\rm or}~~

849: {1 \over \alpha}\sqrt{\left({jl \over d_a}\right)^2-1}\leq V \leq 

850: {jl \over d_a \alpha}$$

851: The ohmic modes, (resp. the junction modes) only occur

852: for voltages such that $V=V_l=\frac{2 k_l +1}{2}\frac{\pi}{a}$ or

853: $V=V_r = \frac{2 k_r + 1}{2}\frac{\pi}{a-l}$ (resp.  $V={k\pi \over l}$).

854: The voltage (horizontal) axis is labeled with the indices corresponding

855: in each case to the junction modes. The ohmic modes are marked with

856: ticks but not labelled. 

857: We always go from an ohmic mode 

858: to a junction mode and vice versa. As the current is increased the 

859: junction mode ceases to exist and the

860: system then jumps to the closest ohmic mode. Notice 

861: the hysteresis  obtained between the curve for

862: increasing $j$ and the one for decreasing $j$.

863: Both the junction voltage and ohmic mode voltages are fixed by the length of 

864: the system. On the other hand the current for which one

865: obtains these voltages depends on the damping $\alpha$

866: and strength of the junction $d_a$. For fixed $d_a$, an decrease of

867: $\alpha$ will tilt the IV curves towards the horizontal voltage axis and 

868: this will separate the cavity modes and cause large voltage jumps from

869: a cavity mode to a far away ohmic mode and 

870: hysteresis. Conversely an increase of $\alpha$ will make the

871: IV curves more vertical so that the voltage jumps and hysteresis

872: will be reduced. 

873: Note also that as in the static case, a magnetic field has no

874: effect on the IV curves for a single junction.

875: 

876: When the junction is centered as in the left

877: panel of Fig. \ref{f5} the ohmic modes (resp. junction modes) 

878: correspond to the 

879: odd $V=(2 k + 1) \pi/l$ (resp. even $V=2 k \pi/l$) cavity modes so the 

880: voltage

881: interval between a junction mode and its corresponding ohmic mode

882: neighbors is constant. This does not happen when the junction is

883: off-centered as in the right panel of Fig. \ref{f5} so that one

884: can get very sharp resonances connecting an ohmic mode to its

885: junction mode neighbor like the ones for $n=2,3$ and 4. The resonance

886: for $n=3$ was obtained by decreasing current.

887: For larger values of $n$ the voltage separation becomes larger so that

888: the resonances are softer.

889: 

890: 

891: \subsection{Study of resonances}

892: 

893: We have been able to bound the IV curves for our system. Now let us

894: discuss the fine structure of the resonances, why some are sharp

895: (see resonances 3,4 and 6 in Fig. \ref{f5}) while others are not.

896: To explain this one should consider the energy of the system and

897: the Fourier modes. Multiplying (\ref{sgd}) by $\phi_t$ and integrating

898: from $0$ to $l$ we obtain the work equation

899: \be \label{work}

900: \partial_t \left[{1 \over 2}\int_{0}^{l} \phi_{t}^2  + \phi_{x}^2 dx 

901: + d_a \left(1-\cos\phi(a,t)\right) \right] =

902: j \int_{0}^{l} \phi_t dx -d_a \alpha\phi_t^2 (a,t).

903: \ee

904: We can write (\ref{work}) as

905: $\partial_t E_t\equiv \partial_t (E_p + E_j) = P_{dc} + P_q $ where

906: $E_p = {1 \over 2}\int_{0}^{l} ( \phi_{t}^2 + \phi_{x}^2) dx$ 

907: is the energy in the passive region, $E_j= d_a

908: \left(1-\cos\phi(a,t)\right) $ the Josephson energy,

909: $P_{dc} = j \int_{0}^{l} \phi_t dx$ is the power given by the direct 

910: current 

911: and $P_q= d_a \alpha\phi_t^2 (a,t)$ is power dissipated by the 

912: quasiparticles.

913: In figure \ref{f6}, we plot from top to bottom $E_t,E_p,P_q$ and $E_j$ 

914: for

915: a centered junction $a=5$ (bottom panels) and an off-centered junction

916: $a=9$ (top panels). The left panels show the junction mode $n=4$

917: and the right panels the corresponding ohmic mode for the same current.

918: In the junction mode the quasi-particle term $P_q$ oscillates strongly

919: and the Josephson energy $E_j$ is very anharmonic. In the ohmic mode

920: $P_q$ is almost constant and $E_j$ is sinusoidal. In all cases the 

921: energy in the cavity $E_p$ is much larger than the junction energy 

922: $E_j$.  

923: When the junction is off-centered (top left panel) the energy stored in

924: the cavity is much larger. This is due to the smallness of the coupling 

925: coefficient $c_n$ as we now show.

926: 

927: 

928: For that consider the phase $\phi(a,t)$ at the junction as given by 

929: (\ref{sjm}). It is independent of the junction position $a$. 

930: Since we are in the junction mode $V=n\pi/l$ we write from 

931: (\ref{modjonc}) 

932: $$\phi(a,t) = {n \pi \over l} t + \sum_{p=1}^{+\infty} 

933: \left [\gamma_p \cos \left(\frac{p n \pi}{l}t\right) + 

934: \beta_p\sin \left(\frac{p n \pi}{l}t \right) \right ]

935: c_{np}. $$

936: The first statement implies that the terms $\gamma_p c_{np}$ 

937: and $\beta_p c_{np}$ are

938: constants independent of the junction position $a$ so that 

939: $\gamma_p$ and $\beta_p$ are proportional to $1/c_{np}$.

940: As a consequence if $c_{np}$ is small the junction mode 

941: of index $n$ corresponds to a large amplitude in

942: the mode $A_{np}$. The system

943: accumulates this energy in the passive region and the resonance is very

944: sharp.

945: 

946: 

947: Fig. \ref{f6} shows a plot as a function of time, in each panel, 

948: from top to bottom of the total energy $E_t$, the energy in the

949: passive region $E_p$, the power dissipated by the quasi-particles $P_q$ 

950: and

951: the Josephson energy $E_j$ for

952: a centered junction $a=5$ (bottom panels) and an off-centered junction

953: $a=9$ (top panels). 

954: The left panels show the junction mode $n=4$

955: and the right panels the corresponding ohmic mode for the same current.

956: The coupling coefficient is small $c_4=0.3$ in the top panel

957: while it is maximum $c_4=1$ in the bottom panel where the junction

958: is centered. Therefore one expects a sharp resonance and a highly

959: excited passive cavity for $a=9$ as opposed to $a=5$.

960: This is indeed the case as one can see from the plot of $E_p$ in the

961: left bottom and left top panels. The Josephson energy $E_j$ and the

962: power dissipated by the junction are of the same order.

963: In the ohmic regime this accumulation of energy in the cavity is

964: absent. There is a small difference in $E_p$ between the two

965: configurations because they have different lengths of passive regions.

966: Note that the total energy is practically independent of time and so

967: is the dissipation. There is an exact balance between the Josephson

968: energy and the energy in the cavity $E_p$.

969: 

970: 

971: 

972: \section{The many junction case}

973: 

974: Assuming a general situation with magnetic field, a mixed current feed

975: and many junctions in the micro-strip we can generalize our model to

976: \be\label{manyj}

977: \phi_{tt}- \phi_{xx} + \sum_j d_j \delta(x-a_j) (\sin(\phi) +

978: \alpha\phi_t)=\nu j \ee

979: with the boundary conditions $\phi_x|_{x = 0}= b-{jl \over 2}(1-\nu)$

980: and $\phi_x|_{x = l}= b+{jl \over 2}(1-\nu)$.

981: If $\nu=1$, 

982: In the model (\ref{manyj}) $\nu=1$ (resp. $\nu=0$) corresponds to 

983: an overlap (resp. inline) current feed.

984: 

985: \subsection{Static case}

986: 

987: We have seen that for a single junction the maximum current $j l = d_a$

988: is reached for any value of the magnetic field. This is not true for

989: two junctions (a SQUID) where the maximum critical current if obtained 

990: for

991: $b=0 ~~{\rm mod.} {2 \pi n \over a_2-a_1}~~$ where $n$ is an integer

992: \cite{cg04}. Static solutions for many junctions in a 1D microstrip

993: will be discussed in a forthcoming paper \cite{many}.

994: 

995: \subsection{High voltage case}

996: 

997: As in the case of a single junction, when the stationary

998: state is reached we can assume $\phi_{t}=V$ and average out the

999: $\phi_{tt}$ and $\sin(\phi)$ terms yielding the following

1000: boundary value problem for the time independent part $\phi_v$ of $\phi$

1001: \begin{eqnarray} \label{hvolov}

1002: - {\phi_v}_{xx} + \sum_{k=1}^n d_k \delta (x-{a_k})\alpha V = \nu j ,\\

1003: {\phi_v}_{x} |_{0,l}=b \mp {jl \over 2}(1-\nu),

1004: \end{eqnarray}

1005: On each interval separating the junctions ${\phi_v}_{xx} = - \nu j$ so 

1006: that ${\phi_v}$ is a second degree polynomial 

1007: ${\phi_v}(x) =-(\nu j/2)x^2 + b x + d$. 

1008: 

1009: 

1010: Integrating (\ref{hvolov}) from $0$ to $l$ yields the voltage

1011: $$V = \frac{j l}{\alpha\sum_{k=1}^n d_k}.$$

1012: Integrating (\ref{hvolov}) over small domains containing the

1013: junctions we obtain as usual the jump conditions in the derivatives

1014: $$\left[ {\phi_v}_x \right]_{a_i^-}^{a_i^+} = 

1015: \frac{d_i j l}{\sum_{k=1}^n d_k}.$$

1016: 

1017: Using these remarks we can calculate $\phi_v$ for any number of 

1018: junctions

1019: (for simplicity one can assume $\phi_v(a_1)=0$). Let's introduce 

1020: $S(x)$, the

1021: spatial curvature of $\left<\phi\right>_t$. For two junctions we get

1022: \begin{equation}

1023: \phi_v(x) =

1024: \left\{\begin{array}{l r}

1025: \frac{-\nu j}{2}(x^2 -a_1^2) + \left(b-{jl \over 2}(1 - 

1026: \nu)\right)(x-a_1) 

1027: & x \leq a_1 \\

1028: \frac{-\nu j}{2}(x^2 -a_1^2) + \left({d_1 jl \over d_1+d_2} +

1029: b-{jl \over 2}(1 - \nu)\right)(x-a_1) & a_1<x<a_2 \\

1030: \frac{-\nu j}{2}(x^2 - a_2^2) + \left(b-{jl \over 2}(1+\nu)\right)

1031: (x-a_1) + S_2 & a_2 \leq x \\

1032: \end{array} \right .

1033: \end{equation}

1034: where $S_2 = \frac{-\nu j}{2}(a_2^2 -a_1^2) +

1035: \left({d_1 jl \over d_1+d_2} + b-{jl \over 2}(1 - \nu)\right)(a_2-a_1)$

1036: is the phase difference between the second and first junction 

1037: ($S_2 \equiv \phi(a_2,t) -\phi(a_1,t)$). The phase is then given by

1038: \begin{equation}\label{CurveHV}

1039: \phi(x,t) = \phi_v(x) + Vt + C,

1040: \end{equation}{manyj}

1041: where $C$ is an integration constant. Notice that it is possible to 

1042: find $\phi_v(x)$ for an arbitrary number of junctions and compute the

1043: phase difference $\phi_v(a_i)=S_i \equiv \phi(a_i,t)-\phi(a_{i-1},t)$.

1044: 

1045: In Fig. \ref{f7} we show $\phi(x,t)$ for 6 consecutive values of

1046: time separated by $\delta t=0.1$ for a system of two junctions

1047: located respectively at $a_1=2$ and $a_2=6.2$ in a domain of

1048: length $l=10$ with current $j=1$ fed in overlap geometry ($\nu=1$).

1049: One can see that the behavior

1050: is close to what is predicted by the above expression.

1051: 

1052: \subsection{Fourier representation} 

1053: 

1054: We can use the high voltage solution to simplify (\ref{manyj}).

1055: For that take $\psi = \phi + \phi_v$ so that (\ref{manyj}) becomes

1056: \begin{eqnarray}\label{simplj}

1057: \psi_{tt}-\psi_{xx}+\sum_{k=1}^n d_k \delta (x-{a_k})\left (

1058: \sin(\psi-S_k) + \alpha \psi_t-{jl \over \sum_{k=1}^n d_k} \right)=0,\\

1059: \left. \psi_x \right|_{x=0;l} = 0.

1060: \end{eqnarray}

1061: The transformation using the high voltage solution concentrates the

1062: current on the junctions. In this representation inline or overlap 

1063: configurations differ by the terms $S_k$.

1064: 

1065: 

1066: Let us go back with Fourier series. Because of the homogeneous Neumann 

1067: boundary conditions on (\ref{simplj}), we can decompose $\psi$ on 

1068: a cosine Fourier series

1069: $$\psi(x,t) = \sum_{n=0}^{+\infty} A_n(t) \cos\left({n\pi x \over 

1070: l}\right)$$ 

1071: and obtain the following equations for the modes $An$

1072: \begin{eqnarray}\label{manya0}

1073: l A_0^{''} + \sum_{j=1}^n d_j \left(\sin(\psi_j - S_j) + \alpha 

1074: \psi_{jt}

1075: -{jl \over \sum_{k=1}^n d_k}\right)=0, \\

1076: A_n^{''} + \left({n \pi \over l}\right)^2 A_n + {2 \over l}

1077: \sum_{j=1}^n d_j c^j_n \left(\sin(\psi_j - S_j) + \alpha \psi_{jt}

1078: -{jl \over \sum_{k=1}^n d_k}\right)=0,

1079: \end{eqnarray}

1080: where $c^j_n = \cos\left({n\pi a_j\over l}\right),~~\psi_j \equiv 

1081: \psi(a_j,t),~~

1082: \psi_{jt} \equiv \partial_t \psi(a_j,t)$.

1083: 

1084: \section{IV curves for two or more junctions}

1085: 

1086: \subsection{2 symmetrically placed junctions, limiting behaviors.}

1087: 

1088: In a symmetric circuit, we have $a_2=l-a_1$ and $d_1=d_2$. For 

1089: simplicity we assume no magnetic field so that

1090: $b=0$ too. With this symmetry, it is easy to show that 

1091: $c_{2k}^2=c_{2k}^1$

1092: and $c_{2k+1}^2=-c_{2k+1}^1$. 

1093: It is important to remark that when $b=0$, $S_2 = 0$. So there is no 

1094: phase shift between the two junctions. Let us now go back 

1095: to the Fourier modes, (\ref{manya0}) becomes:

1096: \begin{equation}\label{A_0_2jsym}

1097: d_1 \left( \sin(\psi_1) + \alpha \psi_{1t}+\sin(\psi_2)+ \alpha

1098: \psi_{2t}\right) -jl = - l A_0^{''}.

1099: \end{equation}

1100: For $p=2k$,

1101: \begin{eqnarray*}\label{A_2k_2j}

1102: A_{2k}^{''} + \left(\frac{2k \pi}{l}\right)^2A_{2k} &=&

1103: -\frac{2 c_{2k}^1}{l}[d_1 (\sin(\psi_1)+ \alpha \psi_{1t}+

1104: \sin(\psi_2)+\alpha \psi_{2t})-jl] \\

1105: &=& 2 c_{2k}^1 A_0^{''}

1106: \end{eqnarray*}

1107: We want to show that the following equation for $A_{2k+1}$ is decoupled 

1108: from the system of Fourier modes

1109: \begin{equation}\label{A_2k+1_2j}

1110: A_{2k+1}^{''}+\left(\frac{(2k+1) \pi}{l}\right)^2A_{2k+1}=

1111: \frac{-2 c_{2k+1}^1}{l}d_1 (\sin(\psi_1) - \sin(\psi_2)

1112: + \alpha (\psi_{1t} - \psi_{2t})).\end{equation}

1113: For that we group together the even terms and odd terms of the 

1114: Fourier series of $\psi_{1t}$ and $\psi_{2t}$ and define  

1115: $s_o(t)\equiv\sum^{+\infty}_{k=0}

1116: c_{2k}^1 A_{2k}(t)$ and 

1117: $s_e(t) \equiv \sum^{+\infty}_{k=0} c_{2k+1}^1 A_{2k+1}(t)$.

1118: Notice that:

1119: $$\psi(a_1,t)=s_o(t)+s_e(t)~~{\rm and }~~\psi(a_2,t)=s_o(t)-s_e(t)$$

1120: Replacing $\psi_1$ and $\psi_2$ in equation (\ref{A_2k+1_2j}), 

1121: we obtain: 

1122: \begin{equation}\label{decoupl_2j}

1123: A_{2k+1}^{''}+\left(\frac{(2k+1) \pi}{l}\right)^2A_{2k+1}=

1124: \frac{-4 c_{2k+1}^1}{l}d_1 \left(\cos(s_e)\sin(s_o) -

1125: \alpha s_e^{'}\right) 

1126: \end{equation}

1127: This equations show that the system performs like a single centered 

1128: junction with different coupling coefficients. So we can balance 

1129: the Fourier modes in the junction mode: with initial conditions 

1130: $\forall k$, $A_{2k+1}(0)=0$, $A^{'}_{2k+1}(0)=0$ we have 

1131: $S_2(0)=0$ and as a consequence $A_{2k+1}(t) \equiv 0$, 

1132: $\forall t \in \Re^+$. This shows that the equations for the 

1133: odd terms are decoupled from the system.

1134: 

1135: {\bf Junctions modes.}\\

1136: As we see, the system (\ref{manya0}) behaves like a single 

1137: junction: the odd terms are equal to zero and the even terms 

1138: give the junction modes 

1139: (see (\ref{sjm})):

1140: $$\phi_1=\phi_2=2\arctan\left\{{d_1 \over jl } \left [\tan

1141: \left({1 \over 2 \alpha} V_j \left(t+C_2\right)\right) V_j + 1\right] 

1142: \right\},$$

1143: with $V_j = \sqrt{\left({jl \over d_1}\right)^2-1} = \frac{n \pi}{l}$.

1144: 

1145: Notice that the two junctions are perfectly synchronized. In a 

1146: non-symmetric situation, junctions modes for both junctions

1147: are particular cases.

1148: 

1149: {\bf Ohmic modes.}\\

1150: Choosing $S_2 = 0$ allow us to show again analytical solutions. 

1151: Assuming $b=0$, there are three types of solution, given by different 

1152: oscillations of the phase:

1153: \begin{enumerate}

1154: \item The phase oscillates between the junctions so that $V=

1155: \frac{(2k+1)\pi}{a_2-a_1}$. Due to the symmetry of the problem, we 

1156: cannot find solutions with $V=\frac{2k\pi}{a_2-a_1}$.

1157: \item The phase oscillates outside the junctions: 

1158: $V=\frac{(2 p + 1)\pi}{2 a_1}$.

1159: \item The phase oscillates in all the passive parts so that

1160: $V=\frac{(2k+1)\pi}{a_2-a_1}=\frac{(2p+1)\pi}{2 a_1}$. This is

1161: only possible if $a_2-a_1$ is a fraction of $l-a_2$.

1162: \end{enumerate}

1163: To conclude in the symmetric case, we know limit IV curves with exact 

1164: solutions. See the left panel of Fig.\ref{f11}. In the general case, we cannot 

1165: find exact solutions, we have near ohmic mode solutions and peaks at the

1166: expected resonant voltages $V=\frac{n \pi}{l}$.

1167: 

1168: \subsection{The non-symmetric cases: IV curves }

1169: 

1170: The IV curve obtained for a circuit of two junctions placed in 

1171: a non symmetric fashion confirms the existence of ohmic modes 

1172: at the expected voltages. It is still possible to characterize the

1173: IV curves using ohmic modes and junction modes even if there are

1174: only approximate. Without symmetry and when $S_2=0~[\pi]$ we can obtain

1175: exactly any one of ohmic modes discussed in previous section. In 

1176: addition to the symmetric case there two new possibilities where $V$ can 

1177: be computed:

1178: \begin{enumerate}

1179: \item The phase oscillates between the junctions and to the

1180: right of $a_1$. See the left panel of Fig.\ref{f8} as an example.

1181: \item The phase oscillates to the left of $a_2$ and between the

1182: junctions.

1183: \end{enumerate} 

1184: In Fig.\ref{f8}, the middle panel shows the second point of the

1185: symmetric case and the right panel illustrates the case where

1186: the phase oscillates between the junctions for the non-symmetric case.

1187: 

1188: Peaks appear at the junction modes $V=\frac{k\pi}{l}$, but they are 

1189: in general smaller than in the symmetric cases. 

1190: Fig.\ref{f9} shows the IV curves for two non symmetrically placed 

1191: junctions and indeed one can see that the height of the resonances

1192: is lower than one for two decoupled junctions (the upper curve in

1193: the three panels). By varying the junction critical current 

1194: (the coefficients $d_1$ and $d_2$ ) we show that the resonances 

1195: are linked to one of the two junctions. To understand 

1196: this phenomenon, go back to the coupling coefficients $c_n^j$.

1197: For the second resonance in Fig. \ref{f9}:

1198: $|c^1_2|=1$, $|c^2_2|=0.73$ so that the junctions are strongly coupled to 

1199: the system. This is like in the symmetric case but it is an exception. The IV 

1200: curve shows that they are together in junction mode (or very close).

1201: Consider now the third resonance. The coupling coefficients are $c^1_3 

1202: = 0$ and $c^2_3 = 0.90$ so that the junction $a_1$ is decoupled from 

1203: the system, we can assume ohmic behavior for it and $a_2$ is in 

1204: junction mode. The fourth resonance corresponds to the reverse 

1205: situation, $c^1_4=1$, $c^2_4=0.06$ where $a_2$ is decoupled.

1206: 

1207: To confirm this, we plot in Fig. \ref{f10}, $\phi_1$ and $\phi_2$ 

1208: for the resonances $3$ and $4$ with $d_1=0.7$, $d_2=0.3$

1209: (left panel of Fig. \ref{f9}) and  $d_1=d_2=0.5$ (middle panel

1210: of Fig. \ref{f9}).

1211: Together with the numerical results we plot the junction mode solution

1212: (\ref{sjm}) in dashed line. For the top panel (third resonance) one

1213: can see that the junction $a_2$ follows closely this behavior as opposed 

1214: to $a_1$ which is in ohmic mode. The situation is reversed for the 

1215: fourth resonance as shown in the bottom panels. 

1216: For resonances 2, 3 and 4, the value of the critical current $d_i~,i=1,2$ does 

1217: not modify appreciably the behavior of the phase at the junctions 

1218: (remark that $S_2 \neq 0$ but stay small).

1219: 

1220: For a general resonance the junctions will work in different modes,

1221: one will be closer to a junction mode while the other will be closer to 

1222: an ohmic mode. We call this mode a resistor-junction mode.

1223: 

1224: {\bf Resistor-junction modes.}\\

1225: Resistor-junction modes exist if the circuit has one junction heavily 

1226: coupled while the other is decoupled. In this case, the system behaves almost 

1227: like a single junction circuit. 

1228: This is why these modes are associed to voltages 

1229: $V = \frac{k\pi}{l}$. 

1230: 

1231: If the junction $a_i$ is in ohmic mode, then $[\phi_x]_{a_i^-}^{a_i^+}=

1232: d_k \sin(Vt+C_1) + C_k$. So that we can only have only one harmonic 

1233: in the circuit. Whereas a junction 

1234: mode exists only with an infinity of harmonics. So the system reaches a 

1235: constructive equilibrium. 

1236: For example in all the panels of Fig. \ref{f10}, we plot the exact junction 

1237: mode to compare. The observation of the phases in the junction mode 

1238: indicates that it lingers more around $(2k+1)\pi/2$ than the 

1239: expression (\ref{sjm}). This gives for the IV curves 

1240: of Fig. \ref{f9} resonances that are slightly higher than expected.

1241: This summarizes the influence of one junction on the other.

1242: 

1243: \subsection{$\pi$-junctions and magnetic field effects}

1244: 

1245: We now consider the situation where the critical current $d_i$ for

1246: some $i$ can be negative. Then the ith junction is a so-called $\pi$ junction

1247: for which the Josephson relation is given by $\sin(\phi +\pi)$

1248: instead of $\sin(\phi)$. In this case we will show that the dynamics

1249: is very different than the one where all the $d_i$ are positive.

1250: Therefore combining $\pi$ and standard junctions provides a way 

1251: to change the resonant frequencies of the array.

1252: Another important parameter which we have not varied up to now is the

1253: magnetic field $b$. We will see that a change of $b$ changes the IV 

1254: characteristic in a way that is similar to changing the critical

1255: current density of some of the junctions in the array. This is

1256: another way to tune the array to resonate on specific

1257: frequencies. 

1258: 

1259: 

1260: We now illustrate these specific points on some examples.

1261: Consider an array of $N$ point junctions described by

1262: $$\phi_{tt}- \phi_{xx} + \sum_{j=1}^{N} d_j \delta(x-a_j) (\sin(\phi) +

1263: \alpha\phi_t + c \phi_{tt})=\nu j~,$$

1264: with the boundary conditions $\phi_x|_{x = 0}= b+{jl \over 2}(1-\nu)$

1265: and $\phi_x|_{x = l}= b-{jl \over 2}(1-\nu)$.

1266: First note that if $d_i<0$ for all i's then we can take 

1267: $\psi= \phi+\pi$ to get us back to the standard case $d_i>0$. We therefore

1268: obtain the same IV curve as when all the $d_i>0$. 

1269: 

1270: An interesting case corresponds to $N=2$. Let us consider the effect

1271: of an additional magnetic field $b'$ on the IV curve. For that

1272: introduce $\psi(x,t)=\phi(x,t)+f(x)$ where $f(x)= b'(x-a_1)$.

1273: The equation verified by $\psi$ is 

1274: $$\psi_{tt}- \psi_{xx} 

1275: + d_1 \delta(x-a_1) (\sin(\psi + f_1) + \alpha\psi_t )

1276: + d_2 \delta(x-a_2) (\sin(\psi + f_2) + \alpha\psi_t )

1277: = \nu j, $$

1278: with the boundary conditions 

1279: $\psi_x|_{x = 0}= b+ b'+{jl \over 2}(1-\nu)$

1280: and $\psi_x|_{x = l}= b+b'-{jl \over 2}(1-\nu)$.

1281: When $f_1=0$ and $f_2=2 \pi$ ie when $b'= 2 \pi / (a_2-a_1)$

1282: we obtain the same equation for 

1283: $\phi$ and $\psi$ apart from the boundary conditions.

1284: The IV curves

1285: for the two magnetic fields $b=0$ and $b= 2 n \pi / (a_2-a_1)$

1286: where $n$ is an integer are then identical. Now if one chooses

1287: a magnetic field $b= \pi / (a_2-a_1)$ then the phases are shifted

1288: by $\pi$ one from the other and this is like changing the sign of

1289: one of the $d_i$. 

1290: In practice

1291: these results only makes sense for magnetic fields for which the

1292: size of the junctions can be neglected. This is the limit of our

1293: approach.

1294: 

1295: 

1296: Fig. \ref{f11} shows two IV curves for two symmetrically placed junctions

1297: $a_1=2$ et $a_2=8$ such that

1298: $b=0$ $d_1=d_2=0.5$ corresponding to standard junctions

1299: (left panel) and $b=0$ $d_1=-d_2=0.5$ corresponding to a standard junction

1300: and a $\pi$ junction (right panel). 

1301: We observe in the left panel resonances (junction modes) every

1302: $\frac{2k\pi}{l}$ exactly as for a centered junction. In the right

1303: panel, resonances (junction modes) are found every 

1304: $\frac{(2k+1)\pi}{l}$. The $\pi$ phase shift between

1305: the two junctions implies that odd harmonics are highly coupled and 

1306: even harmonics are decoupled from the system. 

1307: We could have obtained the same

1308: change in the nature of the resonances by using the standard array (left panel)

1309: and changing the magnetic field to $b= \pi/(a_2-a_1)$.  As

1310: the field is changed one then goes continuously from an IV curve 

1311: with only even resonances to an IV curve with only odd resonances.

1312: 

1313: 

1314: Now let us consider how these results obtained for $N=2$ affect the

1315: behavior of an array of $N>3$ junctions. If there is a $d$ such that

1316: $a_{i+1}-a_i= n_i d$ where $n_i$ is an integer for all $i$ then

1317: the IV curves for $b=0$ will be the same as for $b= 2\pi/d$.

1318: If one chooses $b= \pi/d$ the junctions such that $f_i=\pi ~{\rm mod.}~ 2 \pi$

1319: will be reversed while the ones for which $f_i=2\pi ~{\rm mod.}~2 \pi$

1320: will not be affected. This is similar to a flute which is tuned

1321: by changing nodes. 

1322: 

1323: 

1324: \section{Conclusion}

1325: 

1326: We have introduced a simple and general model using a wave equation 

1327: with delta distributed sine nonlinearities to describe the

1328: dynamics of point Josephson junctions in a 1D microstrip. This

1329: can be extended to $\pi$ junctions for which the current density can

1330: be either positive or negative in the domain. We can also apply it

1331: to situations where the current density in each junction is different.

1332: 

1333: 

1334: For a single junction we have shown two limiting behaviors, the ohmic

1335: mode where the junction behaves as a resistor driven by the cavity, separating

1336: waves and the junction mode where the Josephson element is driving the cavity.

1337: These limiting behaviors have allowed us to bound the IV curves and 

1338: understand the observed resonances. When another junction is present 

1339: in the cavity, this simple classification is changed in 

1340: general and we observe ohmic modes and combined junction/ohmic modes. The 

1341: existence or not of the nth resonance is connected to the value of 

1342: the Fourier coefficient $c_n^a = \cos{n \pi a /l}$.

1343: 

1344: These results carry over to the case of an array with many junctions where

1345: it is possible to choose a voltage where one of the junctions is inactive and 

1346: another is in junction mode. We also use the analysis to understand the

1347: effect of an external magnetic field and the influence of having 0 or $\pi$

1348: junctions. We believe that such a device composed of normal and $\pi$

1349: junctions placed at specific locations in a microstrip 

1350: can be used to great advantage for specific applications like

1351: resonators and frequency mixers.

1352: 

1353: 

1354: \begin{thebibliography}{99}

1355: 

1356: \bibitem{josephson} B. D. Josephson, Phys. Lett. {\bf 1}, 251, (1962).

1357: 

1358: \bibitem{Likharev} K. Likharev, {\em Dynamics of Josephson junctions 

1359: and

1360: circuits}, Gordon and Breach, (1986).

1361: 

1362: \bibitem{Barone} A. Barone and G. Paterno, {\em Physics and 

1363: Applications

1364: of the Josephson effect}, J. Wiley, (1982).

1365: 

1366: \bibitem{Salez} M. H. Chung and M. Salez, in Proc. 4th European 

1367: Conference on

1368: applied superconductivity, EUCAS 99, 651, (1999).

1369: 

1370: \bibitem{bcf02} A. Benabdallah, J. G. Caputo and N. Flytzanis, Physica

1371: D 161, 79-101, (2002)

1372: 

1373: \bibitem{bc02} A. Benabdallah and J. G. Caputo, J. of Applied Physics,

1374: {\bf 92}, nb. 7, (2002).

1375: 

1376: \bibitem{lcti03} T. Lindstrom et al, Phys. Rev. Lett. {\bf 90}, 117021, (2003).

1377: 

1378: \bibitem{l04} N. Lazarides, Phys. Rev. B {\bf 69}, 212501, (2004).

1379: 

1380: \bibitem{gsgk04} E. Goldobin et al, Phys. Rev. Lett. {\bf 92}, 570051, (2004).

1381: 

1382: \bibitem{r03} G. Rotoli, Phys. Rev. B {\bf 68}, 052505, (2003).

1383: 

1384: \bibitem{sg04} H. Susanto and S. A. van Gils, Phys. Rev. B {\bf 69}, 92507, 

1385: (2004).

1386: 

1387: \bibitem{as01} E. Almaas and D. Stroud, Phys. Rev. B, {\bf 63}, 144522, 

1388: , (2001).

1389: 

1390: \bibitem{fp02} G. Filatrella and N. F. Pedersen, Physica C, 372-376, 

1391: 11, 

1392: (2002).

1393: 

1394: \bibitem{bb91} S. P. Benz and C. J. Burroughs, Appl. Phys. Lett. {\bf 

1395: 58},

1396: (1991).

1397: 

1398: \bibitem{bcsl99} P. Barbara, A. B. Cawthorne, S. V. Shitov and C. J. 

1399: Lobb,

1400: Phys. Rev. Lett. {\bf 82}, 1963, (1999).

1401: 

1402: \bibitem{vbsl02} S. V. Vasilic, P. Barbara, S. V. Shitov and C. J. 

1403: Lobb,

1404: Phys. Rev. B {\bf 65}, 180503, (2002).

1405: 

1406: \bibitem{hairer}  E. Hairer, S. P. Norsett and G. Wanner. S{\it olving

1407: ordinary differential equations I} (Springer-Verlag, 1987).

1408: 

1409: \bibitem{carslaw}  H. S. Carslaw, {\it An introduction to the theory of

1410: Fourier series and integrals }, 3rd revised edition, Dover (1950).

1411: 

1412: \bibitem{lamb} H. Lamb, {\it Elements of soliton theory}, Wiley, 

1413: (1983).

1414: 

1415: \bibitem{report} L. Maurice, {\it Statique de jonctions Josephson dans 

1416: un 

1417: microstrip}, Memoire de DEA, Insa de Rouen, (2003).

1418: 

1419: \bibitem{Boussaha} M.F. Boussaha, {\it D\'eveloppement d'un 

1420: d\'emonstrateur 

1421: de r\'ecepteur h\'et\'erodyne submillim\'etrique ultra-large bande \`a 

1422: base de jonctions Supraconducteur-Isolant-Supraconducteur (SIS), 

1423: refroidi \`a 4,2K.}, PhD thesis, Universit\'e de 

1424: Paris 6, (2003).

1425: 

1426: \bibitem{cg04} J. G. Caputo and Y. Gaididei, Physica C, {\it Two point 

1427: Josephson

1428: junctions in a superconducting stripline: static case.}, 

1429: Physica C {\bf 402}, 160-173, (2004).

1430: 

1431: \bibitem{many} J. G. Caputo and L. Loukitch, in preparation.

1432: 

1433: \end{thebibliography}

1434: 

1435: \section{Appendix}

1436: 

1437: \subsection{Continuum limit.}

1438: 

1439: We integrate (\ref{sggh}) from $a-h$ to $a+h$

1440: 

1441: The first term yields

1442: \begin{eqnarray*}

1443: \int_{a-h}^{a+h} \phi_{tt} dx & = & 

1444: \int_{a-h}^{a+h} \phi_{tt}(a)+(x-a)\phi_{ttx}(a)+O(|x-a|^2) dx\\

1445: & = & 2h\phi_{tt}(a) + 4h^2\phi_{ttx}(a) + O(h^3)\\

1446: \end{eqnarray*}

1447: The second one

1448: $$\int_{a-h}^{a+h} \phi_{xx} dx=\phi_x(a+h)-\phi_x(a-h)$$

1449: and the third one

1450: \begin{eqnarray*}

1451: \int_{a-h}^{a+h}\left [\sin(\phi)+ \alpha\phi_t + c \phi_{tt} \right]

1452: dx & = & \int_{a-h}^{a+h}\sin\left(\phi(a)+(x-a)\phi_x(a)+

1453: O_1(|x-a|^2)\right)\\

1454: & & + \alpha \phi_t(a)+ \alpha (x-a)\phi_{xt}(a) + O_2(|x-a|^2)\\

1455: & & + c\phi_{tt}(a) + c(x-a)\phi_{xtt}(a) + O_3(|x-a|^2)dx\\

1456: & =  & \int_{a-h}^{a+h}\sin\left(\phi(a)+(x-a)\phi_x(a) 

1457: + O_1(|x-a|^2)\right)dx\\

1458: & & +  \alpha \phi_t(a) + \alpha h \phi_{xt}(a)

1459:  + c\phi_{tt}(a)+ ch \phi_{xtt}(a) + O_4(h^2)\\

1460: \end{eqnarray*}

1461: We now expand the sine term

1462: \begin{eqnarray*}

1463: \sin\left(\phi(a)+(x-a)\phi_x(a)\right)&=&

1464: \sin(\phi(a))\cos((x-a)\phi_x(a))\\

1465: &&+\sin((x-a)\phi_x(a))\cos(\phi(a))

1466: \end{eqnarray*}

1467: so that

1468: \begin{eqnarray*}

1469: &&\int_{a-h}^{a+h}\sin\left(\phi(a)+(x-a)\phi_x(a)\right)\\

1470: &=&\int_{a-h}^{a+h}\sin(\phi(a))\cos((x-a)\phi_x(a))+

1471: \sin((x-a)\phi_x(a))\cos(\phi(a))dx\\

1472: &=&\sin(\phi(a)){2 \over \phi_x(a)} \sin(h\phi_x(a))-0

1473: \end{eqnarray*}

1474: Consider the limit of this term when $h\rightarrow 0$.

1475: We have:

1476: $$ d_a {1 \over 2 h}{2 \over \phi_x(a)}\sin(h\phi_x(a)) \rightarrow d_a,$$

1477: so that the equation for the point junction is

1478: $$ \phi_x(a+h)-\phi_x(a-h) + d_a(\sin(\phi(a))+ \alpha\phi_t(a) +

1479: c \phi_{tt} (a))=j,$$

1480: which is consistent with the model (\ref{sgcd}).

1481: $$\phi_{tt}- \phi_{xx} + d_a\delta(x-a) \left( \sin(\phi) + 

1482: \alpha\phi_t +

1483: c \phi_{tt} \right)=j $$

1484: 

1485: \subsection{Method of lines}

1486: 

1487: The basis of the method is to discretize the spatial part of the

1488: operator and keep the temporal part as such. We thereby transform

1489: the partial differential equation into a system of ordinary 

1490: differential equations. This method allows to increase the 

1491: precision of the approximation in time and space independently and

1492: easily. In our case the operator is a distribution so that the

1493: natural way to give it meaning is to integrate it over a volume.

1494: We therefore choose as space discretisation the finite volume

1495: approximation where the operator is integrated over reference volumes.

1496: The value of the function is assumed constant in each volume.

1497: 

1498: 

1499: As solver for the system of differential equations, we use

1500: the Runge-Kutta method of order 4-5 introduced by Dormand and Prince

1501: implemented as the Fortran code DOPRI5 by Hairer and Norsett 

1502: \cite{hairer}

1503: which enables to control the local error by varying the time-step.

1504: 

1505: 

1506: We first transform (\ref{sgd}) into a system of first order partial

1507: differential equations

1508: We write $\psi(x,t)=\phi_t(x,t)$.

1509: \begin{equation} \label{sysdiscret} 

1510: \left\{  \begin{array}{r c l}

1511: \psi(x,t) & = & \phi_t(x,t) \\

1512: \psi_t(x,t) & = & \phi_{xx}(x,t)-\delta(x-a) (d_a \sin(\phi(x,t)) + 

1513: \alpha \psi(x,t))+j 

1514: \end{array} \right. 

1515: \end{equation}

1516: with the boundary conditions : 

1517: $\phi_x|_{l \over 2}=\phi_x|_{-{l \over 2}}=0$. 

1518: 

1519: For simplicity we will describe the implementation of 

1520: the finite volume discretisation in the case of a single junction.

1521: We introduce reference

1522: volumes $V_k$ whose centers we call $x_k$, $1\leq k \leq nn$.

1523: The discretisation points are placed such that

1524: the point $x_{ng+1}$ is at the junction, ($x_{ng+1}=a$).

1525: We thus define $x_k$ and $V_k$ using the following identities

1526: $$V_k=\left ]x_k-{h_g \over 2},x_k+{h_g \over 2} 

1527: \right[~,~~~~0<k<ng+1$$

1528: with $(ng+1)h_g=a$

1529: $$V_k=\left ]x_k-{h_d \over 2},x_k+{h_d \over 2} 

1530: \right[~,~~~~ng+1<k<nn+1$$

1531: with $(nn-ng)h_d=l-a$. Finally at the junction, $k=ng+1$

1532: $$V_{k_{ng+1}}=\left ]x_{ng+1}-{h_g \over 2},

1533: x_{ng+1}+{h_d \over 2} \right[.$$

1534: $nn$, $ng$ and $nd$ are respectively the total number of discretisation

1535: points, the number of points to the left of the junction and the number 

1536: of points to the right.

1537: 

1538: For a fixed t, we assume $\phi(x,t)$ to be constant on each volume $V_k$, 

1539: so that

1540: $$\int_{x_k-{h \over 2}}^{x_k+{h \over 2}} \phi(x,t) dx = 

1541: h \phi(x_k,t)~{\rm,~~with}~h=hg~{\rm or}~h=hd$$

1542: Integrating over $V_k$ yields:

1543: \begin{enumerate}

1544: \item In the linear part of the PDE : $0<k<nn+1$ and $k \neq ng+1$: 

1545: \begin{equation} \label{spacediscr}

1546: \left \{  \begin{array}{r c l}

1547: \psi(x_k,t) & = & \phi_t(x_k,t) \\

1548: \psi_t(x_k,t) & = & 

1549: \frac{\phi(x_{k+1},t)-2\phi(x_k,t)+\phi(x_{k-1},t)}{h^2} + j

1550: \end{array} \right.

1551: \end{equation}

1552: with $h=hg$ for $0<k<ng+1$ or $h=hd$ for $k>ng+1$. We recognize the usual 

1553: discretisation of the second derivative. 

1554: 

1555: \item At the junction: $k=ng+1$, we obtain

1556: \begin{eqnarray*}\label{disconti}

1557: \int_{x_{ng+1}-{h_g \over 2}}^{x_{ng+1}+{h_d \over 2}}

1558: \delta(x-a)\left(d_a \sin(\phi(x,t)) + \alpha\phi_t(x,t)\right)=\\

1559: d_a \sin(\phi(x_{ng+1},t)) + \alpha\phi_t(x_{ng+1},t)

1560: \end{eqnarray*}

1561: So that the final system is:

1562: \begin{equation} \label{joncdiscr}

1563: \left \{  \begin{array}{r c l}

1564: \psi(x_{ng+1},t) & = & \phi_t(x_{ng+1},t) \nonumber\\

1565: \psi_t(x_{ng+1},t) & = & \frac{4}{hg+hd}\left (\frac{\phi(x_{ng+2},t)-

1566: \phi(x_{ng+1},t)}{hg/2} -\frac{\phi(x_{ng+1},t)-\phi(x_{ng},t)} 

1567: {hd/2} \right)\\

1568:  & & -\frac{2}{hg+hd}\left(d_a \sin(\phi(x_{ng+1},t))

1569:  +\alpha\phi_t(x_{ng+1},t)\right) + j

1570: \end{array} \right.

1571: \end{equation} 

1572: \end{enumerate}

1573: The ODE system (\ref{spacediscr}, \ref{joncdiscr}) is then integrated 

1574: numerically using the DOPRI5 integrator.

1575: 

1576: \begin{figure}

1577: %\centerline{ \epsfig{file=figs/microstrip.eps,height=7 cm,width=10 cm,angle=0}}

1578: \centerline{ \epsfig{file=f1.eps,height=7 cm,width=10 cm,angle=0}}

1579: \caption{Left top pannel, a schematic drawing of a narrow 

1580: 2D microstrip line containing

1581: two small Josephson junctions, right top panel section at one of the junction. 

1582: The bottom panel presents the equivalent 1D electric circuit.}

1583: \label{f1}

1584: \end{figure}

1585: 

1586: \begin{figure} 

1587: %\centerline{\epsfig{file=figs/phiajoncohm.ps,height=14 cm,width=5.5cm,angle=270}}

1588: \centerline{\epsfig{file=f2.ps,height=14 cm,width=5.5cm,angle=270}}

1589: \caption{plot of $\phi_a(t)$ for $j = 0.137414$ for a junction position

1590: $a=3.43$, $d_a=1$ and length $l=10$. The left panel shows 

1591: the ohmic mode regime and the

1592: right panel the junction mode regime. The numerical solution of

1593: (\ref{sgd}) is shown with the crosses while the analytical estimates

1594: (\ref{som}) and (\ref{sjm}) are in dashed line. }

1595: \label{f2}

1596: \end{figure}

1597: 

1598: \begin{figure} 

1599: %\centerline{\epsfig{file=figs/ohm1jonc.ps,height=15 cm,width=5cm,angle=270}}

1600: \centerline{\epsfig{file=f3.ps,height=15 cm,width=5cm,angle=270}}

1601: \caption{Instantaneous voltage $\phi_t\left(x,t\right)$

1602: as a function of $x$ for different times for the ohmic mode: left

1603: only oscillation for the junction shown in Fig. \ref{f2} placed at 

1604: $a=3.43$ (left panel) and left and right oscillation for a junction placed at 

1605: $a=2$ (right panel). The average voltage is indicated in both figures, $V=j l 

1606: =1.37$ in the left panel (same as in Fig. \ref{f2})

1607: and $V=\frac{\pi}{2}\left(=j l \right)$ on the right panel. 

1608: The other parameters are $\alpha=1$, and $d_a=1$.}

1609: \label{f3}

1610: \end{figure}

1611: 

1612: \begin{figure} 

1613: %\centerline{\epsfig{file=figs/phitjonc.ps,height=8 cm,width=5cm,angle=270}}

1614: \centerline{\epsfig{file=f4.ps,height=8 cm,width=5cm,angle=270}}

1615: \caption{Instantaneous voltage $\phi_t(x,t)$ in the

1616: junction mode as

1617: a function of $x$ for 60 values of time equidistributed between 0 and 

1618: 10.

1619: The parameters are the same as for Fig. 1. The average voltage $V$ is

1620: indicated on the vertical axis and the junction's position is indicated

1621: on horizontal axis.}

1622: \label{f4}

1623: \end{figure}

1624: 

1625: 

1626: \begin{figure}

1627: %\centerline{\epsfig{file=figs/a0_a4fig.ps,height=15. cm,width=6cm,angle=270}}

1628: \centerline{\epsfig{file=f5.ps,height=15. cm,width=6cm,angle=270}}

1629: \caption{Two IV characteristic for a single junction placed in the

1630: center of the microstrip at $a=5$

1631: (left panel) and near the right edge at $a=9$ (right panel).

1632: Only the positions of the junction modes are indicated in units of

1633: $\frac{\pi}{10}$. The other ticks correspond to the ohmic modes.

1634: For each panel two curves are presented, the top one is for increasing 

1635: current

1636: while the bottom one is for a decreasing current showing a clear 

1637: hysteresis.

1638: The minimum value of the current is 0, the maximum is 4 and

1639: the stepping is $\frac{4}{99}$. The other parameters are

1640: $d_a=\alpha=1$.

1641: }

1642: \label{f5}

1643: \end{figure}

1644: 

1645: \begin{figure}

1646: %\centerline{\epsfig{file=figs/nrja0_a4.ps,height=16 cm,width=9 cm,angle=270}}

1647: \centerline{\epsfig{file=f6.ps,height=16 cm,width=9 cm,angle=270}}

1648: \caption{Plot as a function of time, in each panel, from top to bottom 

1649: of the total energy $E_t$, the energy in the passive region $E_p$, the 

1650: power dissipated by the quasi-particles $P_q$ and the Josephson energy 

1651: $E_j$ for a centered junction $a=5$ (bottom panels) and an off-centered 

1652: junction $a=9$ (top panels). The left panels show the junction mode $n=4$

1653: and the right panels the corresponding ohmic mode for the same current.

1654: The averages of $E_t$ and $E_j$ are shown on the $y$ axis.}

1655: \label{f6}

1656: \end{figure}

1657: 

1658: 

1659: \begin{figure}

1660: %\centerline{\epsfig{file=figs/hautvolt2j.ps,height=6.5 cm,width=4.5 cm,angle=270}}

1661: \centerline{\epsfig{file=f7.ps,height=6.5 cm,width=4.5 cm,angle=270}}

1662: \caption{High voltage case: plot of $\phi(x,t)$ as a function of

1663: $x$ pour 6 consecutive times separated by $\Delta t=0.1$. The

1664: two junctions are located at $a_1=2$ and $a_2=6.2$. The other 

1665: parameters are $l=10, j=1,$ and $\alpha =0.5$.}

1666: \label{f7}

1667: \end{figure}

1668: 

1669: \begin{figure} 

1670: %\centerline{\epsfig{file=figs/modohmic.ps,height=16 cm,width=6. cm,angle=270}}

1671: \centerline{\epsfig{file=f8.ps,height=16 cm,width=6. cm,angle=270}}

1672: \caption{Three types of ohmic modes obtained for two junctions

1673: in a microstrip. The positions of the junctions are given

1674: in the pictures. We plot $\phi_t\left(x,t\right)$ $\forall x \in [0,l]$ 

1675: for six consecutive times. The parameters of equation 

1676: (\ref{sgd}) are $\lambda=0.5$ and $\alpha=1$. 

1677: All figures have been obtained with the same current $j l=2.6927937$.}

1678: \label{f8}

1679: \end{figure}

1680: 

1681: \begin{figure}

1682: %\centerline{\epsfig{file=figs/a0b1.23fig.ps, height=16 cm, width=6 cm,angle=270}}

1683: \centerline{\epsfig{file=f9.ps, height=16 cm, width=6 cm,angle=270}}

1684: \caption{IV characteristic curves for two junctions placed in $a_1=5$ 

1685: and $a_2=6.2$. The critical current densities are $d_1=0.7,~d_2=0.3$ for 

1686: the left panel, $0.5,~0.5$ for the middle panel and $0.3,~0.7$ for the

1687: right panel. We plot on all figures $j l=\sqrt{V^2+1}$, $j l =V$, 

1688: $j l =d_1\sqrt{V^2+1}+d_2 V$ and $j l =d_2\sqrt{V^2+1}+d_1V$.}

1689: \label{f9}

1690: \end{figure}

1691: 

1692: \begin{figure}

1693: %\centerline{\epsfig{file=figs/mix.ps,height=15 cm,width=10 cm,angle=270}}

1694: \centerline{\epsfig{file=f10.ps,height=15 cm,width=10 cm,angle=270}}

1695: \caption{Plots of the phases at the junctions $\phi(a_1,t)$ and

1696: $\phi(a_2,t)$ as a function of time for the IV curves in the

1697: two left panels of Fig. \ref{f9}. The left panels are

1698: for $d_1=0.7,d_2=0.3$ and the right panels for $d_1=d_2=0.5$.

1699: The top panels show the resonance of order 3 while the bottom

1700: panels show the resonance of order 4. Note how the junctions

1701: switch between ohmic mode and junction mode. The analytic

1702: expressions for the corresponding junction modes (\ref{sjm}) 

1703: are shown in dashed line.}

1704: \label{f10}

1705: \end{figure}

1706: 

1707: \begin{figure} 

1708: \centerline{

1709: %\epsfig{file=figs/a-3b3_0-0_0-pi.ps,height=16 cm,width= 6 cm,angle=270}}

1710: \epsfig{file=f11.ps,height=16 cm,width= 6 cm,angle=270}}

1711: \caption{IV curves for two symmetrically placed junctions

1712: $a_1=2$ et $a_2=8$.The scale on the $x$ axis is in units of 

1713: $\frac{\pi}{l}$. In the left panel the junctions are identical while

1714: in the right panel the second one is a $\pi$-junction.

1715: Note the resonance on the even harmonics $\frac{2k\pi}{l}$ in the left

1716: panel and on the odd harmonics $\frac{(2k+1)\pi}{l}$ in the right panel.

1717: We could have obtained the same change of IV curves for the

1718: same array of standard junctions by an appropriate choice of the

1719: magnetic field (see text).}

1720: \label{f11}

1721: \end{figure}

1722: 

1723: \end{document}

1724: 

1725: