1: \documentclass{rspublic}
2: \usepackage{amssymb,amsmath}
3:
4: \begin{document}
5:
6: \title[Integrability and Linearization]{On the complete integrability and
7: linearization of nonlinear
8: ordinary differential equations - Part II: Third order equations}
9:
10: \author[Chandrasekar, Senthilvelan and Lakshmanan]{V. K. Chandrasekar,
11: M. Senthilvelan and M. Lakshmanan}
12:
13: \affiliation{Centre for Nonlinear Dynamics, Department of Physics,
14: Bharathidasan Univeristy, Tiruchirapalli - 620 024, India}
15:
16: \label{firstpage}
17:
18: \maketitle
19:
20: \begin{abstract}{Integrability, Integrating factor, Linearization, Equivalence
21: problem}
22: We introduce a method for finding general solutions of third-order nonlinear
23: differential equations by extending the modified Prelle-Singer method. We
24: describe a procedure to deduce all the integrals of motion associated with the
25: given equation so that the general solution follows straightforwardly from these
26: integrals. The method is illustrated with several examples. Further, we propose
27: a powerful method of identifying linearizing transformations. The proposed
28: method not only unifies all the known linearizing transformations systematically
29: but also introduces a new and generalized linearizing transformation (GLT). In
30: addition to the above, we provide an algorithm to invert the nonlocal
31: linearizing transformation. Through this procedure the general solution for the
32: original nonlinear equation can be obtained from the solution of the linear
33: ordinary differential equation.
34: \end{abstract}
35:
36: \section{Introduction}
37: In a previous paper (Chandrasekar \textit{et al.} 2005) we have discussed the
38: complete integrability aspects of a class of
39: second-order nonlinear ordinary differential equations (ODEs) through a nontrivial
40: extension of the so called Prelle-Singer (PS) (Prelle \& Singer 1983;
41: Duarte \textit{et al.} 2001) procedure. We have illustrated the
42: procedure with several physically interesting nonlinear oscillator examples. We
43: have also developed a straightforward algorithmic way to transform the given
44: second order nonlinear ODE to a linear free particle equation, if it is
45: linearizable.
46:
47: One of the questions raised at the final stage of our earlier work
48: (Chandrasekar \textit{et al.} 2005) was that what are the
49: implications of the novel features which we introduced in the extended
50: Prelle-Singer procedure to obtain the second constant of motion (in the
51: case of second order ODEs) to third and higher order ODEs. To have
52: a closer look at the problem let us recall our earlier work briefly here. We
53: considered a second order ODE of the form
54: $\frac{d^2x}{dt^2}=\frac{P(t,x,\dot{x})}{Q(t,x,\dot{x})},
55: \;{ P,Q}\in \mathbb{C}{[t,x,\dot{x}]}$, and explored two
56: pairs of independent functions, say, $R_i$ and $S_i$, $i=1,2,$ associated with
57: the underlying
58: ODE. These functions are nothing but the integrating factors and null forms,
59: respectively. Once these
60: two pairs of functions are determined (by solving an overdetermined system
61: of first order PDEs) then each pair leads to an independent integral of motion,
62: which can then be used to find the general solution for the given equation.
63: Thus instead of integrating the first
64: integral and obtaining the general solution which is conventionally followed in the
65: literature we implemented some novel ideas in
66: the PS method such that one can construct the general solution for the given
67: equation in a self contained way and, in fact, our procedure works for a class
68: of problems.
69:
70: In the case of third order ODEs, one should have three independent integrals
71: of motion
72: in order to establish the complete integrability. To deduce these three integrals
73: one should have three pairs of independent functions ($R_i$, $U_i$ and $S_i$),
74: $i=1,2,3$.
75: When we extend the PS procedure to third order ODEs we find that the
76: determining equations
77: for the integrating factors and null forms provide either one or two integrals
78: of motion only straightforwardly. Again the hidden form of the functions
79: ($R_3,U_3,S_3$) should be
80: explored in order establish the complete integrability of the given equation
81: within the frame work of PS procedure. In this paper we describe a
82: procedure to capture the required set of functions. {\it With the completion of
83: this task we formulate a simple, straightforward and powerful method to
84: solve a wide class of third order ODEs of contemporary literature}.
85:
86: We stress at this point that the
87: application of PS procedure to third order ODEs is not a straightforward
88: extension of the second order case. In fact one has to overcome many faceted
89: problems. The first and foremost one is how to solve the determining
90: equations in such a way that one could obtain three sets of independent
91: functions, namely,
92: $(R_i,U_i,S_i),\;i=1,2,3,$ in a systematic way. In the present case we have six
93: equations for three unknown functions (in the case of second order equations we have three
94: equations for two unknowns). We overcome this problem by adopting suitable
95: methodologies,
96: the details which we present in \S3. Another obstacle one could face in higher
97: order ODEs, at least in some cases, is that one may be able to get only one
98: integral of
99: motion and in this situation how one would be able to generate the
100: remaining integrals of motion from the first integral is also tackled by us in
101: this paper.
102:
103: Our main goal, besides the above, is to bring out a novel and straightforward
104: way to construct linearizing transformations for third order ODEs. The latter
105: can be used to transform the given third order nonlinear ODEs to a linear
106: equation. We note that unlike the second order
107: equations, the third order ODEs can be linearized through different kinds of
108: transformations, namely, invertible point transformation
109: (Ibragimov \& Meleshko 2005), contact transformation
110: (Bocharov \textit{et al.} 1993; Ibragimov \& Meleshko 2005)
111: and generalised Sundman transformation
112: (Berkovich \& Orlova 2000; Euler \textit{et al.} 2003; Euler \& Euler 2004).
113:
114: In this paper we introduce a new kind of transformation which can be
115: effectively used to linearize a class of nonlinear third order ODEs. In fact,
116: one can linearize certain equations only through this
117: transformation alone and not by the known ones in the literature. We call this
118: transformation as generalized linearizing transformation (GLT). We note that
119: generalised Sundman transformation is a special case of this transformation.
120: In the generalised Sundman transformation the new independent variable is a
121: nonlocal one and so eventhough one is able to transform the given nonlinear
122: third order ODE to a linear one due to the nature of nonlocal independent variable
123: it is not easy to write down the general solution. In the case of generalized
124: linearizing transformation both the new dependent and independent
125: variables contain derivative terms also in addition to the
126: independent variable being nonlocal. Even for this general case, in this paper,
127: we succeed to present an efficient algorithm to deduce the general solution.
128:
129: Another fundamental problem regarding linearization is how to deduce the
130: linearizing transformations systematically. Generally, Lie symmetry analysis and
131: direct methods are often used to deduce the point and contact transformations
132: (Steeb 1993; Olver 1995; Bocharov \textit{et al.} 1993; Bluman \& Anco 2002;
133: Ibragimov \& Meleshko 2005). In this work we
134: propose a simple and straightforward method to
135: deduce linearizing transformations and we derive them from the first integral.
136: Our method of deducing linearizing transformations has several salient
137: features. Irrespective of the
138: form of the linearizing transformation (point/contact/generalised Sundman
139: transformation), it can be derived from the first
140: integral itself. We also note that one can also linearize a third order ODE to the
141: second order free particle equation
142: through our method. An added advantage of our method is that suppose a given
143: equation is linearizable through one or more kinds of transformations then our
144: procedure provides all these transformations in a straightforward way and as far as
145: our knowledge goes {\it no such single method has been formulated in the
146: literature}.
147:
148: The plan of the paper is follows. In the following section we extend the PS
149: procedure to third order ODEs and indicate new features in finding the three
150: independent integrals of motion. In \S3, we describe the methods of solving the
151: determining equations and how one can obtain compatible solutions from them. We
152: illustrate the procedure with several examples. In \S4, we propose a powerful
153: method of identifying linearizing transformations. This method not only brings
154: out all the known transformations systematically but also a new generalized
155: linearizing transformation (GLT) for the third order ODEs. We emphasize the
156: validity of the method
157: with several illustrative examples arising in different areas of mathematics
158: and physics. We present our conclusions in \S5.
159:
160: \section{Prelle-Singer method for third order ODEs}
161: \label{sec2}
162: \subsection{General Theory}
163: Let us consider a class of third order ODEs of the form
164: \begin{eqnarray}
165: \dddot{x}=\frac{P}{Q}, \quad
166: { P,Q}\in \mathbb{C}{[t,x,\dot{x},\ddot{x}]}, \;\;( ^{.}=\frac{d}{dt})
167: \label{met1}
168: \end{eqnarray}
169: where over dot denotes differentiation with respect to time and $P$ and $Q$
170: are polynomials in $t$, $x$, $\dot{x}$ and $\ddot{x}$ with coefficients in the
171: field of complex numbers, $\mathbb{C}$. Let us assume that the third order
172: ODE (\ref{met1})
173: admits a first integral $I(t,x,\dot{x},\ddot{x})=C,$ with $C$ being constant on the
174: solutions, so that the total differential of $I$ gives
175: \begin{eqnarray}
176: dI={I_t}{dt}+{I_{x}}{dx}+{I_{\dot{x}}{d\dot{x}}}+{I_{\ddot{x}}{d\ddot{x}}}=0,
177: \label{met3}
178: \end{eqnarray}
179: where each subscript denotes partial differentiation with respect
180: to that variable. Equation~(\ref{met1}) can be rewritten as
181: $\frac{P}{Q}dt-d\ddot{x}=0$. Now adding the null terms
182: $U(t,x,\dot{x},\ddot{x})\ddot{x}dt-U(t,x,\dot{x},\ddot{x})d\dot{x} $ and
183: $S(t,x,\dot{x},\ddot{x})\dot{x}dt-S(t,x,\dot{x},\ddot{x})dx $
184: to this we obtain that, on the solutions, the 1-form
185: \begin{eqnarray}
186: (\frac{P}{Q}+S\dot{x}+U\ddot{x})dt-Sdx-Ud\dot{x}-d\ddot{x} = 0.
187: \label{met6}
188: \end{eqnarray}
189:
190: Looking at equations~(\ref{met3}) and
191: (\ref{met6}) one can conclude that, on the solutions, these two forms are
192: proportional and the form of equation~(\ref{met6}) is equivalent to equation~(\ref{met3})
193: except for an overall multiplication factor. Thus multiplying equation~(\ref{met6}) by
194: the factor $ R(t,x,\dot{x},\ddot{x})$ which acts as the integrating factor
195: for (\ref{met6}), we have on the solutions that
196: \begin{eqnarray}
197: dI=R(\phi+S\dot{x}+U\ddot{x})dt-RSdx-RUd\dot{x}-Rd\ddot{x} =0,
198: \label{met7}
199: \end{eqnarray}
200: where $ \phi\equiv {P}/{Q}$. Comparing equations (\ref{met3})
201: with (\ref{met7}) we have the following relations, on the solutions,
202: \begin{align}
203: I_{t} = R(\phi+S\dot{x}+U\ddot{x}), \qquad
204: I_{x} = -RS, \qquad
205: I_{\dot{x}} = -RU,\qquad
206: I_{\ddot{x}} = -R.
207: \label{met8}
208: \end{align}
209: Now imposing the compatibility conditions,
210: $I_{tx}=I_{xt}$, $I_{t\dot{x}}=I_{{\dot{x}}t}$, $I_{t\ddot{x}}=I_{{\ddot{x}}t}$,
211: $I_{x{\dot{x}}}=I_{{\dot{x}}x}$, $I_{x{\ddot{x}}}=I_{{\ddot{x}}x}$,
212: $I_{\dot{x}{\ddot{x}}}=I_{{\ddot{x}}\dot{x}}$, which exist between the
213: equations~(\ref{met8}), we have the following equations which constitute three
214: determining equations ((\ref{met9})-(\ref{met11}) given below) for the
215: functions $S$, $U$ and $R$ along with three constraints
216: ((\ref{met12})-(\ref{met14}) given below) that they need to satisfy,
217: \begin{align}
218: D[S] & = -\phi_x+S\phi_{\ddot{x}}+US,\label{met9}\\
219: D[U] & = -\phi_{\dot{x}}+U\phi_{\ddot{x}}-S+U^2,\label{met10}\\
220: D[R] & = -R(U+\phi_{\ddot{x}}),\label{met11}\\
221: R_x & = R_{\ddot{x}}S+RS_{\ddot{x}},
222: \label{met12}\\
223: R_{\dot{x}}S & = -RS_{\dot{x}}+R_{x}U+RU_{x}, \label{met13}\\
224: R_{\dot{x}} & = R_{\ddot{x}}U+RU_{\ddot{x}}, \label{met14}
225: \end{align}
226: where
227: \begin{eqnarray}
228: D=\frac{\partial}{\partial{t}}+
229: \dot{x}\frac{\partial}{\partial{x}}+\ddot{x}\frac{\partial}{\partial{\dot{x}}}
230: +\phi\frac{\partial}{\partial{\ddot{x}}}.\label{opp}
231: \end{eqnarray}
232: The task of solving equations~(\ref{met9})-(\ref{met14}) can be
233: accomplished in the following way.
234: Substituting the given expression of $\phi$ into (\ref{met9})-(\ref{met10})
235: and solving
236: them one can obtain expressions for $S$ and $U$. With the known $U$,
237: equation~(\ref{met11}) becomes the determining equation for the function $R$.
238: Solving the latter one can get an explicit form for $R$.
239: Compatible solutions to equations~(\ref{met9})-(\ref{met11}) can also be obtained
240: in alternate ways, the details of which are given in \S3.
241:
242: Now the functions $R,U$ and $S$ have to satisfy an extra set of constraints,
243: that is, equations~(\ref{met12})-(\ref{met14}). Suppose a compatible solution
244: satisfying all the equations have
245: been found then the functions $R,\;U$ and $S$ fix the differential
246: invariant $I(t,x,\dot{x},\ddot{x})$ by the relation
247: \begin{align}
248: I(t,x,\dot{x},\ddot{x}) & = r_1 - r_2 - \int \left\{RU+\frac{d}{d\dot{x}} \left[r_1 -
249: r_2\right]\right\}d\dot{x}
250: \nonumber\\
251: &-\int\left\{R+\frac{d}{d\ddot{x}} \left[r_1 - r_2
252: -\int \left\{RU+\frac{d}{d\dot{x}} \left[r_1-
253: r_2\right]\right\}d\dot{x}\right]\right\}d\ddot{x},
254: \label{met15}
255: \end{align}
256: where
257: \begin{align}
258: r_1 = \int R(\phi+S\dot{x}+U\ddot{x})dt,\;\;\;
259: r_2 =\int \left( RS+\frac{d}{dx}\int r_1\right) dx. \nonumber
260: \end{align}
261: Equation~(\ref{met15}) can be derived straightforwardly by integrating the
262: equations~(\ref{met8}). Here it is to be noted that for every independent
263: set $(S,U,R)$, equation~(\ref{met15}) defines an integral.
264:
265: \subsection{Exploring the complete form of R: Theory}
266: From the above discussion, it is clear that equation~(\ref{met1}) may be
267: considered as completely integrable once we obtain
268: three independent sets of the solutions $(S_i,U_i,R_i),\;i=1,2,3,$
269: which provide three independent integrals of motion through the relation
270: (\ref{met15}). Here we note that since we are solving
271: equations~(\ref{met9})-(\ref{met11}) first and then check the compatibility of
272: this solution with equations~(\ref{met12})-(\ref{met14}), one often meets the
273: situation that all the solutions which satisfy
274: equations~(\ref{met9})-(\ref{met11}) need not satisfy the constraints
275: (\ref{met12})-(\ref{met14})
276: since equations~(\ref{met9})-(\ref{met14}) constitute an overdetermined system
277: for the unknowns $R$, $S$ and $U$.
278: In fact for a class of problems one often gets one or two sets of $S,U,R$
279: which satisfy all the equations~(\ref{met9})-(\ref{met14})
280: and another set(s) $(S,U,R)$ which satisfies only the first three equations
281: and not the other, namely, (\ref{met12})-(\ref{met14}). In this situation we
282: find an interesting fact that one can use the integral derived from the set(s)
283: which satisfies all the six equations (\ref{met9})-(\ref{met14})
284: and deduce the other compatible solution(s) $(S,U,\hat{R})$ (definition of
285: $\hat{R}$ follows). For example, let the set $(S_3,U_3,R_3)$ be a
286: solution of the determining equations~(\ref{met9})-(\ref{met11}) and not of the
287: constraints~(\ref{met12})-(\ref{met14}). After analysing several examples we
288: find that one can make
289: the set $(S_3,U_3,R_3)$ compatible by modifying the form of $R_3$ as
290: \begin{eqnarray}
291: \hat{R_3} & = F(t,x,\dot{x},\ddot{x})R_3,
292: \label{met101}
293: \end{eqnarray}
294: where $\hat{R_3}$ satisfies equation~(\ref{met11}), so that we have
295: \begin{eqnarray}
296: (F_t+\dot{x}F_x+\ddot{x}F_{\dot{x}}+\phi F_{\ddot{x}}) R_3+FD[R_3]
297: =-FR_3(U_3+\phi_{\dot{x}}).
298: \label{met102}
299: \end{eqnarray}
300: Further, if $F$ is a constant of motion (or a function of it), then the first
301: term on the left hand side
302: vanishes and one gets the same equation~(\ref{met11}) for $R_3$ provided $F$ is
303: non-zero. That is, whenever $F$ is a constant of motion or a function of it then
304: the solution to (\ref{met11}) may provide only a factor of the complete solution
305: $\hat{R_3}$ without the factor $F$ in equations~(\ref{met101}). This general form
306: of $\hat{R}_3$ along with $S_3$ and $U_3$ can now provide complete solution to
307: equations~(\ref{met9}) - (\ref{met14}) as discussed below.
308: \subsection{Exploring the complete form of R: Method}
309: Now if the sets $(S_i,U_i,R_i),\;i=1,2,$ are found to satisfy the equations~(\ref{met9}) -
310: (\ref{met14}) and the third set $(S_3,U_3,R_3)$ does not satisfy
311: equations (\ref{met12}) - (\ref{met14}) then $F$ may be
312: a function of the integrals $I_i,\;i=1,2,$ derived from the sets
313: $(S_i,U_i,R_i),\;i=1,2$.
314: We need to find the explicit form of $F(I_1,I_2)$ in order to obtain the
315: compatible solution $(S_3,U_3,\hat{R_3})$. To do so let us find the
316: derivatives of $\hat{R_3}$ with respect to $x$, $\dot{x}$ and $\ddot{x}$:
317: \begin{align}
318: &\hat{R_3}_x=(F_1'I_{1x}+F_2' I_{2x})R_3+FR_{3x},\;\;
319: \hat{R}_{3\dot{x}}=(F_1'I_{1\dot{x}}+F_2' I_{2\dot{x}})R_3
320: +FR_{3\dot{x}},\nonumber\\
321: &\hat{R}_{3\ddot{x}}=(F_1'I_{1\ddot{x}}+F_2' I_{2\ddot{x}})R_3
322: +FR_{3\ddot{x}},
323: \label{fint01}
324: \end{align}
325: where $F_1'=\frac{\partial F}{\partial I_1}$ and
326: $F_2'=\frac{\partial F}{\partial I_2}$. Substituting equation~(\ref{fint01})
327: into equations~(\ref{met12})-(\ref{met14}), we have
328: \begin{subequations}
329: \begin{eqnarray}
330: \frac{(f_1F_1'+f_2F_2')}{f_3}=\frac{F}{R_3},\;\;
331: \frac{(f_4F_1'+f_5F_2')}{f_6}=\frac{F}{R_3}, \;\;
332: \frac{ (f_7F_1'+f_8F_2')}{f_9}=\frac{F}{R_3},
333: \label{fint02}
334: %\end{align}
335: \end{eqnarray}
336: where
337: %\begin{align}
338: \begin{eqnarray}
339: &f_1=(I_{1x}-I_{1\ddot{x}}S_3),\quad
340: f_2=(I_{2x}-I_{2\ddot{x}}S_3),\quad
341: f_3=(S_3R_{3\ddot{x}}+R_3S_{3\ddot{x}}-R_{3x})\nonumber\\
342: &f_4=(I_{1\dot{x}}-I_{1\ddot{x}}U_3),\quad
343: f_5=(I_{2\dot{x}}-I_{2\ddot{x}}U_3),\quad
344: f_6=(U_3R_{3\ddot{x}}+R_3U_{3\ddot{x}}-R_{3\dot{x}})\nonumber\\
345: &f_7=(S_3I_{1\dot{x}}-I_{1x}U_3),\;
346: f_8=(S_3I_{2\dot{x}}-I_{2x}U_3),
347: \qquad\qquad\qquad\qquad\qquad\qquad\quad\nonumber\\
348: &f_9=(R_3U_{3x}+U_3R_{3x}-R_3S_{3\dot{x}}-S_3R_{3\dot{x}}).
349: \qquad\qquad\qquad\qquad\qquad\qquad\qquad\;\;
350: \label{fint05}
351: %\end{align}
352: \end{eqnarray}
353: \end{subequations}
354:
355: Equation~(\ref{fint02}) represents an overdetermined
356: system of equations for the unknown $F$. A simple way to solve this equation is to uncouple
357: it for $F_1'$ ($=\frac{\partial F}{\partial I_1}$) and
358: $F_2'$ ($=\frac{\partial F}{\partial I_2}$) and solve the resultant
359: equations. For example, eliminating $F_2'$ from equation~(\ref{fint02}) we
360: obtain equations for $F_1'$ in the form
361: \begin{align}
362: \frac{R_3F_1'}{F}=\frac{(f_3f_5-f_2f_6)}{(f_1f_5-f_2f_4)}
363: =\frac{(f_3f_8-f_2f_9)}{(f_1f_8-f_2f_7)}
364: =\frac{(f_6f_8-f_5f_9)}{(f_4f_8-f_5f_7)}.
365: \label{fint03}
366: \end{align}
367: On the other hand eliminating $F_1'$ from equation~(\ref{fint02}) we arrive
368: equations for $F_2'$ in the form
369: \begin{align}
370: \frac{R_3F_2'}{F}=\frac{(f_3f_4-f_1f_6)}{(f_2f_4-f_1f_5)}
371: =\frac{(f_3f_7-f_1f_9)}{(f_2f_7-f_1f_8)}
372: =\frac{(f_6f_7-f_4f_9)}{(f_5f_7-f_4f_8)}.
373: \label{fint04}
374: \end{align}
375:
376: It can be easily cheeked that the compatibility of the right three expressions
377: in equations~(\ref{fint03}) or (\ref{fint04}) give rise to relations which are
378: effectively nothing but the constraint equations~(\ref{met12})-(\ref{met14})
379: and so no new
380: constraint is added now. Consequently, equations (\ref{fint03}) and
381: (\ref{fint04}) can be written as
382: \begin{align}
383: \frac{\partial F}{\partial I_1}=g(I_1,I_2)F\quad \mbox{and}\quad
384: \frac{\partial F}{\partial I_2}=h(I_1,I_2)F,
385: \label{fint04a}
386: \end{align}
387: respectively, where $g(I_1,I_2)=\frac{1}{R_3}\frac{(f_3f_5-f_2f_6)}
388: {(f_1f_6-f_3f_4)}$
389: and $h(I_1,I_2)=\frac{1}{R_3}\frac{(f_3f_4-f_1f_6)}{(f_2f_4-f_1f_5)}$.
390: Now we can solve
391: equations (\ref{fint04a}) and obtain the form of $F(I_1,I_2)$. This is
392: demonstrated for several examples in the following sections explicitly.
393: Once $F$ is known we can obtain the complete solution
394: $\hat{R_3}$ from which, along with $S_3$ and $U_3$, the third
395: integral $I_3$ can be constructed. {\it Thus with the explicit forms of the
396: three integrals of motion, the complete integrability of equation~(\ref{met1})
397: is guaranteed}.
398:
399: Finally, if the set $(S_1,U_1,R_1)$ alone is found to satisfy the
400: equations~(\ref{met9}) - (\ref{met14}) and the second set $(S_2,U_2,R_2)$
401: also does not satisfy
402: equations (\ref{met12}) - (\ref{met14}), then $F$ may be
403: a function of the integral $I_1$ alone which was derived from the set
404: $(S_1,U_1,R_1)$.
405: We need to find the explicit form of $F(I_1)$ in order to obtain the
406: compatible solution $(S_2,U_2,\hat{R_2})$. To do so let us find the
407: derivatives of $\hat{R_2}$ with respect to $x$, $\dot{x}$ and $\ddot{x}$:
408: \begin{align}
409: \hat{R_2}_x=F_1'I_{1x}R_2+FR_{2x},\;\;
410: \hat{R}_{2\dot{x}}=F_1'I_{1\dot{x}}R_2
411: +FR_{2\dot{x}},\;\;
412: \hat{R}_{2\ddot{x}}=F_1'I_{1\ddot{x}}R_2+FR_{2\ddot{x}},
413: \label{fint01a}
414: \end{align}
415: where $F_1'=\frac{\partial F}{\partial I_1}$. Substituting
416: equation~(\ref{fint01a})
417: into equations~(\ref{met12})-(\ref{met14}), we have
418: \begin{subequations}
419: \begin{eqnarray}
420: \frac{R_2F_1'}{F}=\frac{s_2}{s_1}
421: =\frac{s_4}{s_3}
422: =\frac{s_6}{s_5},
423: \label{fint06}
424: \end{eqnarray}
425: where
426: \begin{eqnarray}
427: s_1&=(I_{1x}-I_{1\ddot{x}}S_2),\quad \quad
428: s_2=(S_2R_{2\ddot{x}}+R_2S_{2\ddot{x}}-R_{2x}),
429: \qquad \qquad \quad \nonumber\\
430: s_3&=(I_{1\dot{x}}-I_{1\ddot{x}}U_2),\quad\quad
431: s_4=(U_2R_{2\ddot{x}}+R_2U_{2\ddot{x}}-R_{2\dot{x}}),
432: \qquad \qquad\quad \nonumber\\
433: s_5&=(S_2I_{1\dot{x}}-I_{1x}U_2),\quad
434: s_6=(R_2U_{2x}+U_2R_{2x}-R_2S_{2\dot{x}}-S_2R_{2\dot{x}}).
435: \label{fint06a}
436: \end{eqnarray}
437: \end{subequations}
438: One can again check that the compatibility of the right three expressions in
439: equation (\ref{fint06}) leads to a condition, which is deducible from
440: (\ref{met12})-(\ref{met14}), and so no new condition is introduced in reality.
441: Then rewriting equation (\ref{fint06}) as
442: \begin{align}
443: \frac{\partial F}{\partial I_1}=\bigg(\frac{1}{R_2}\frac{s_2}{s_1}\bigg)F
444: =g(I_1)F,
445: \label{fint06b}
446: \end{align}
447: and solving it, one can obtain an explicit expression for $F$.
448: Once $F$ is known we can construct the complete form of $\hat{R_2}$ from
449: which, along with $S_2$ and $U_2$, the second integral of motion can be
450: obtained. Once $I_1$ and $I_2$ are known, we can proceed to find the third
451: compatible set $(S_3,U_3,\hat{R_3})$ as before and obtain
452: the third integral $I_3$
453: to establish complete integrability.
454:
455: \section{Methods of finding the explicit form of $R$}
456: In \S2 we outlined the method of solving the determining equations. However, in
457: practice, it is difficult to solve equations (\ref{met9})-(\ref{met11})
458: straightforwardly as they constitute a set of coupled first order nonlinear
459: partial differential equations (PDEs). So one has to look
460: for some intuitive ideas to solve these equations. We solve them and
461: obtain the forms $R,U$ and $S$ in the following way. For this purpose, we
462: observe the important fact that when
463: we rewrite the coupled equations (\ref{met9})-(\ref{met11}) into an equation
464: for a single variable, namely, $R$, the resultant equation turns out to be a
465: linear PDE. Then we solve this "$R$ equation" with a suitable ansatz
466: (say polynomial
467: or rational in $\ddot{x}$). Once '$R$' is found the remaining functions $U$ and
468: $S$ can be easily deduced.
469:
470: %\subsection{$R$ polynomial in $\ddot{x}$}
471:
472: As noted, rewriting equations~(\ref{met9})-(\ref{met11}) we arrive at a
473: third order linear partial differential equation for $R$ in the form
474: \begin{eqnarray}
475: D^3[R]-D^2[R\phi_{\ddot{x}}]+D[R\phi_{\dot{x}}]-\phi_x R=0,\quad
476: D=\frac{\partial}{\partial{t}}+\dot{x}\frac{\partial}{\partial{x}}
477: +\ddot{x}\frac{\partial}{\partial{\dot{x}}}
478: +\phi\frac{\partial}{\partial{\ddot{x}}}.\label{poly01}
479: \end{eqnarray}
480: Solving equation~(\ref{poly01})
481: with a suitable ansatz in $\ddot{x}$ is relatively easier in many cases than
482: solving equations~(\ref{met9})-(\ref{met11}). Once the explicit form of $R$ is
483: obtained, $U$ can be deduced from equation~(\ref{met11}) as
484: \begin{eqnarray}
485: U=-(\frac{D[R]}{R}+\phi_{\ddot{x}}),
486: \label{poly02}
487: \end{eqnarray}
488: from which $S$ can be fixed by using equation~(\ref{met10}). Now if this set
489: $(S,U,R)$ forms a compatible set for the remaining equations (\ref{met12})-
490: (\ref{met14}) then the corresponding integral $I$ can be found
491: using equation~(\ref{met15}). To illustrate this idea let us look into the
492: following examples.
493:
494: \subsubsection{Example:1}
495: Let us begin with a simple example, namely a linear third order ODE,
496: \begin{eqnarray}
497: \dddot{x}+\lambda x=0,
498: \label{cat101}
499: \end{eqnarray}
500: where $\lambda$ is an arbitrary parameter.
501: Substituting $\phi =-\lambda x$ into (\ref{poly01}), we get the following
502: equation for $R$
503: \begin{eqnarray}
504: D^3[R]+\lambda R=0,\;\;D=\frac{\partial}{\partial{t}}+
505: \dot{x}\frac{\partial}{\partial{x}}+\ddot{x}\frac{\partial}{\partial{\dot{x}}}
506: -\lambda x\frac{\partial}{\partial{\ddot{x}}}. \label{cat102}
507: \end{eqnarray}
508: We now assume an ansatz for $R$ in the form
509: \begin{eqnarray}
510: R = a(t,x,\dot{x})+b(t,x,\dot{x})\ddot{x},
511: \label{cat102a}
512: \end{eqnarray}
513: where $a$ and $b$ are arbitrary functions of $t,\;x$ and $\dot{x}$.
514: Substituting (\ref{cat102a}) into (\ref{cat102}) and equating the coefficients
515: of different powers of $\ddot{x}$ to zero we get a set of linear partial
516: differential
517: equations for the variables $a$ and $b$. Solving them one obtains three
518: nontrivial solutions,
519: \begin{eqnarray}
520: R_1=-e^{kt}, \quad R_2=-(2\ddot{x}+k\dot{x}-k^2x)e^{-kt},\quad
521: R_3=-\frac{\sqrt{3}k}{2}(\dot{x}+kx)e^{-kt},
522: \label{cat103}
523: \end{eqnarray}
524: where $k^3=\lambda$. Now substituting the form of $R_i$'s $,i=1,2,3,$
525: separately into the equation (\ref{poly02}), we get
526: \begin{align}
527: U_1=-k,\quad U_2=\frac{(2k^2\dot{x}+k\ddot{x}+k^3x)}{(2\ddot{x}+k\dot{x}-k^2x)},
528: \quad U_3=-\frac{(\ddot{x}-k^2x)}{(\dot{x}+kx)}.
529: \label{cat103a}
530: \end{align}
531: Now substituting the forms of $U_i$'s, $i=1,2,3,$ into (\ref{met10}) one can fix
532: the forms of $S_i$'s, $i=1,2,3,$ as
533: \begin{align}
534: S_1=k^2,\quad
535: S_2=\frac{2k^4x+k^3\dot{x}-k^2\ddot{x}}{2\ddot{x}+k\dot{x}-k^2x},\quad
536: S_3=-k\frac{(\ddot{x}+k\dot{x})}{(\dot{x}+kx)}. \label{cat103b}
537: \end{align}
538: As a consequence now we have three sets of independent solutions for the
539: equation
540: (\ref{met9})-(\ref{met11}). Now we check the compatibility of these solutions
541: with the remaining equations (\ref{met12})-(\ref{met14}).
542:
543: We find that the solutions $(S_1,U_1,R_1)$ and $(S_2,U_2,R_2)$
544: satisfy the equations~(\ref{met12}) - (\ref{met14}), and become
545: compatible solutions. Substituting the forms $(S_1,U_1,R_1)$ and $(S_2,U_2,R_2)$
546: separately into equation (\ref{met15}) and evaluating the integrals we get
547: \begin{align}
548: I_1& =\frac{1}{3k^2}\bigg(\ddot{x}-k\dot{x}+k^2x\bigg)e^{kt}, \label{cat111}\\
549: I_2&=\frac{2}{3k^2}\bigg(\ddot{x}^2+k^2\dot{x}^2+k^4x^2+k\dot{x}\ddot{x}
550: -k^2x\ddot{x}+k^3x\dot{x}\bigg)e^{-kt}.\label{cat112}
551: \end{align}
552: However, the set $(S_3,U_3,R_3)$ does not satisfy the extra
553: constraints (\ref{met12})-(\ref{met14}) which means that
554: the form of $R_3$ may not be the 'complete form'
555: but might be a factor of the complete form. As mentioned in \S2, in order
556: to recover the complete form $\hat{R_3}$ one may assume that
557: $\hat{R_3} = F(I_1,I_2)R_3$,
558: where $F(I_1,I_2)$ is a function of the integrals $I_1$ and $I_2$. Now we have
559: to determine the
560: form of $F(I_1,I_2)$ explicitly and for this purpose we proceed as follows.
561: Substituting
562: \begin{eqnarray}
563: \hat{R_3} = F(I_1,I_2)R_3 =
564: -\bigg[\frac{\sqrt{3}k}{2}(\dot{x}+kx)e^{-kt}\bigg]F(I_1,I_2),
565: \label{scat109a}
566: \end{eqnarray}
567: into equations~(\ref{fint03}) and (\ref{fint04}), along with (\ref{fint05}), we
568: obtain two equations for
569: $F$ as,
570: \begin{equation}
571: F_1'=0,\quad
572: I_2F_2'+F = 0,
573: \label{new1}
574: \end{equation}
575: where $F_1'$ and $F_2'$ denote partial derivatives of $F$ with respect to
576: $I_1$ and $I_2$, respectively. Upon integrating (\ref{new1}) we get,
577: $F = \frac {1}{I_2}$, (the integration constants are set to zero for simplicity)
578: which fixes the form of $\hat{R_3}$ as
579: \begin{eqnarray}
580: \hat{R_3} = \frac{R_3}{I_2}=-\frac{\sqrt{3}k}{2}\frac{(\dot{x}+kx)}
581: {(\ddot{x}^2+k^2\dot{x}^2+k^4x^2+k\dot{x}\ddot{x}-k^2x\ddot{x}+k^3x\dot{x})}.
582: \label{scat110}
583: \end{eqnarray}
584: Now one can easily check that this set $(S_3,U_3,\hat{R_3})$ is a compatible
585: solution for the equations~(\ref{met9})-(\ref{met14}) which in turn provides
586: $I_3$ through the relation (\ref{met15}) as
587: \begin{eqnarray}
588: I_3=-\frac{\sqrt{3}k}{2}t+tan^{-1}\bigg[\frac{\ddot{x}-k\dot{x}-2k^2x}
589: {\sqrt{3}(\ddot{x}+k\dot{x})}\bigg].
590: \label{scat111}
591: \end{eqnarray}
592: Using the explicit forms of the integrals $I_1,\; I_2$ and $I_3$, the
593: solution to equation~(\ref{cat101}) can be deduced directly as
594: \begin{eqnarray}
595: x(t)=I_1e^{-kt}+\sqrt{I_2}e^{\frac{k}{2}t}\cos\bigg(\frac{\sqrt{3}k}{2}t
596: +I_3\bigg).
597: \label{scat112}
598: \end{eqnarray}
599: The result exactly coincides with the solution presented in Polyanin
600: \& Zaitsev (2000).
601:
602: \subsubsection{Example:2}
603: The applicability of this method to nonlinear ODEs can be illustrated by
604: considering an equation of the form
605: \begin{eqnarray}
606: \dddot{x}=\frac{\ddot{x}^2}{\dot{x}}+\frac{\dot{x}\ddot{x}}{x}. \label {tex11}
607: \end{eqnarray}
608: Equation (\ref{tex11}) is a sub-case of the general form of a scalar third order
609: ODE which is invariant under the generators of time translation and rescaling
610: (Feix \textit{et al.} 1997; Polyanin \& Zaitsev 2000). A sub-case of equation
611: (\ref{tex11}), namely,
612: $\dddot{x}-c\frac{\ddot{x}^2}{\dot{x}}=0$, has been considered by both
613: Bocharov \textit{et al.} (1993) and Ibragimov \& Meleshko (2005) to show that
614: it can be linearized to a linear third order
615: ODE through a contact transformation. On the other hand, Euler \& Euler (2004)
616: have considered the equation $\dddot{x}-\frac{\dot{x}\ddot{x}}{x}=0$ and
617: showed that it can be linearized through Sundman transformation
618: (see \S5$\,c\,$ below).
619: Here we consider the combined form (\ref{tex11}) and derive integrating factors,
620: integrals of motion and the general solution of this equation. Further, we show
621: that the equation (\ref{tex11}) itself can be linearized by the generalized
622: linearizing transformation (GLT) (see \S5$\,d\,$ below).
623:
624: As before, substituting
625: $\phi =\frac{\ddot{x}^2}{\dot{x}}+\frac{\dot{x}\ddot{x}}{x}$ into
626: (\ref{poly01}), we get the following linear partial differential equation for
627: $R$ as
628: \begin{eqnarray}
629: D^3[R]-D^2[(\frac{2\ddot{x}}{\dot{x}}+\frac{\dot{x}}{x})R]
630: -D[(\frac{\ddot{x}^2}{\dot{x}^2}-\frac{\ddot{x}}{x})R]
631: +\frac{\dot{x}\ddot{x}}{x^2} R=0,\label{tex12}
632: \end{eqnarray}
633: where $D$ is defined by equation (\ref{opp}). Now substituting the ansatz
634: (\ref{cat102a}) into (\ref{tex12}) and proceeding as
635: before, we get
636: \begin{eqnarray}
637: R_1=-\frac{1}{\dot{x}x},\quad R_2=\frac{x}{\dot{x}},\quad
638: R_3=\frac{t\dot{x}^2-x(\dot{x}+t\ddot{x})}{2x\dot{x}^2}.
639: \label{tex13}
640: \end{eqnarray}
641: Following the ideas given in Example 1, one can deduce the corresponding forms
642: of $S_i$'s and $U_i$'s, $i=1,2,3,$ as
643: \begin{align}
644: S_1&=-\frac{\ddot{x}}{x},\quad
645: &U_1=-\frac{\ddot{x}}{\dot{x}},\qquad \qquad \qquad \qquad \\
646: S_2&=\frac{\ddot{x}}{x},\quad
647: &U_2=\frac{-2\dot{x}}{x}-\frac{\ddot{x}}{\dot{x}},\qquad \qquad \quad\;\;\\
648: S_3&=\frac{\dot{x}\ddot{x}(x+t\dot{x})}{x(-t\dot{x}^2
649: +x(\dot{x}+t\ddot{x}))}, \quad
650: &U_3=\frac{x\ddot{x}(2\dot{x}+t\ddot{x})}{\dot{x}(t\dot{x}^2
651: -x(\dot{x}+t\ddot{x}))}.\; \;\;\;\;
652: \label{tex14}
653: \end{align}
654: The solutions $(S_i,U_i,R_i),\;i=1,2,$ satisfy the constraints
655: (\ref{met12})-(\ref{met14}) so that they lead to first and second integrals
656: of the form
657: \begin{align}
658: I_1=\frac{\ddot{x}}{\dot{x}x}, \quad
659: I_2= \frac{2\dot{x}^2-x\ddot{x}}{\dot{x}}.\label{tex16}
660: \end{align}
661: In the present case also
662: the set $(S_3,U_3,R_3)$ does not satify the extra constraints and so one has
663: to explore the complete form of $\hat{R_3}$. To do so we proceed
664: as before and obtain the forms of $F$ and $\hat{R_3}$ as
665: $F = \frac {1}{\sqrt{I_1I_2}}$ and
666: \begin{eqnarray}
667: \hat{R_3} = \frac{t\dot{x}^2-x(\dot{x}+t\ddot{x})}
668: {2(\sqrt{I_1I_2})x\dot{x}^2},
669: \label{tex20}
670: \end{eqnarray}
671: where the explicit forms of $I_1$ and $I_2$ are given in equation (\ref{tex16}).
672: Now one can check that the set $(S_3,U_3,\hat{R_3})$ satisfies all the six
673: equations (\ref{met9})-(\ref{met14}) and furnishes the third integral $I_3$
674: through the relation (\ref{met15}) as
675: \begin{eqnarray}
676: I_3=-\frac{1}{2}(\sqrt{I_1I_2})t+tan^{-1}\sqrt{\frac{I_1}{I_2}}x.
677: \label{tex21}
678: \end{eqnarray}
679: Using the explicit forms of the integrals $I_1,\; I_2$ and $I_3$, the
680: solution to equation~(\ref{tex11}) can be deduced directly as
681: \begin{eqnarray}
682: x(t)=\sqrt{\frac{I_2}{I_1}}\tan \bigg[\frac{1}{2}(\sqrt{I_1I_2}t+2I_3)\bigg].
683: \label{tex22}
684: \end{eqnarray}
685: As can be seen from equation~(\ref{tex20}) the complete compatible solution
686: $\hat{R_3}$ has $\ddot{x}$ term which appear inside the square root sign. This
687: form of $\hat{R_3}$ can be explored only by making a suitable ansatz. Moreover
688: one may also face more difficulties in solving the determining equations~
689: (\ref{met9})-(\ref{met11}). In such complicated situations
690: the complete solution $\hat{R}$ can be obtained by using our procedure.
691:
692: \subsubsection{Example:3}
693: Let us consider a Chazy class of equation of the form
694: (Halburd 1999; Mugan \& Jrad 2002; Euler \& Euler 2004)
695: \begin{eqnarray}
696: \dddot{x}+4\alpha x\ddot{x}+3\alpha \dot{x}^2+6\alpha^2x^2\dot{x}
697: +\alpha^3x^4=0. \label {rex11}
698: \end{eqnarray}
699: Substituting $\phi =-(4kx\ddot{x}+3k\dot{x}^2+6k^2x^2\dot{x}+k^3x^4)$ into
700: (\ref{poly01}), we get
701: \begin{align}
702: D^3[R]+4\alpha D^2[xR]-6\alpha D[(\dot{x}+\alpha x^2)R]
703: +4\alpha (\ddot{x}+\alpha^2x^3+3\alpha x\dot{x})R=0.\label{rex12}
704: \end{align}
705: To solve the above equation we assume the same ansatz (\ref{cat102a}) for
706: the variable $R$. However, substituting this ansatz into equation~(\ref{rex12})
707: and solving the resultant equation we obtain only trivial solution. So we seek
708: a rational form of ansatz for $R$ in the form
709: \begin{align}
710: R=\frac{a(t,x,\dot{x})+b(t,x,\dot{x})\ddot{x}}
711: {(c(t,x,\dot{x})+d(t,x,\dot{x})\ddot{x})^r} \label{rex14b},
712: \end{align}
713: where $r$ is an arbitrary number. Substuting the equation~(\ref{rex14b}) into
714: equation~(\ref{rex12}) and solving the resultant partial differential equations
715: we get
716: \begin{align}
717: &R_1=\frac{\dot{x}+\alpha x^2}{(\alpha^2x^3+3\alpha x\dot{x}+\ddot{x})^2},\;\;
718: R_2=\frac{t(-2x+\alpha tx^2+t\dot{x})}
719: {(\alpha^2x^3+3\alpha x\dot{x}+\ddot{x})^2},\nonumber\\
720: &R_3=\frac{t(3-3\alpha tx+\alpha^2t^2x^2+\alpha t^2\dot{x})}{(\alpha^2x^3
721: +3\alpha x\dot{x}+\ddot{x})^2}.
722: \label{rex14e}
723: \end{align}
724: The corresponding forms of $S_i$'s and $U_i$'s $i=1,2,3,$ are
725: \begin{align}
726: &U_1=\frac{2\alpha^2x^3-\ddot{x}}{\alpha x^2+\dot{x}},
727: \qquad\;S_1=\frac{\alpha(\alpha^2x^4+3\dot{x}^2-2x\ddot{x})}{\alpha x^2+\dot{x}},
728: \nonumber\\
729: &U_2=\frac{2x-6\alpha tx^2+2\alpha^2t^2x^3-t^2\ddot{x}}
730: {t(-2x+\alpha tx^2+t\dot{x})},\label{rex13c}\\
731: &S_2=\frac{2\alpha x^2-4\alpha^2tx^3+\alpha^3t^2x^4-2\dot{x}
732: +3\alpha t^2\dot{x}^2+2t\ddot{x}(1-\alpha tx)}{t(-2x+\alpha tx^2+t\dot{x})},
733: \nonumber\\
734: &U_3=-\frac{3-12\alpha tx+9\alpha^2t^2x^2-2\alpha^3t^3x^3+\alpha t^3\ddot{x}}
735: { t(3-3\alpha tx+\alpha^2t^2x^2+\alpha t^2\dot{x})},\nonumber\\
736: &S_3=\frac{(\alpha(12\alpha tx^2-6\alpha^2t^2x^3+\alpha^3t^3x^4+3t(2\dot{x}
737: +\alpha t^2\dot{x}^2+t\ddot{x})-2x(3+\alpha t^3\ddot{x})))}{t(3-3\alpha tx
738: +\alpha^2t^2x^2+\alpha t^2\dot{x})}.\nonumber
739: \end{align}
740:
741: Now we proeced to find the integrals of motion. First we note that the functions
742: $(S_1,U_1,R_1)$ satisfy the
743: constraints~(\ref{met12})-(\ref{met14}) and hence they are compatible.
744: Thus substituting them into (\ref{met15}) the first integral $I_1$ is fixed easily
745: as
746: \begin{align}
747: I_1&=-t+\frac{\alpha x^2+\dot{x}}
748: {\alpha^2x^3+3\alpha x\dot{x}+\ddot{x}}.\label{rex15}
749: \end{align}
750: However, the set $(S_2,U_2,R_2)$ (and so also $(S_3,U_3,R_3)$) does not
751: satisfy the extra
752: constraints (\ref{met12})-(\ref{met14}) which means
753: the form of $R_2$ may not be the 'complete form'
754: but might be a factor of the complete form. As mentioned in \S2, in order
755: to recover the complete form $\hat{R_2}$ one may assume that
756: $\hat{R_2} = F(I_1)R_2$.
757: Here $F(I_1)$ is a function of the integral $I_1$. Now we have
758: to determine the
759: form of $F(I_1)$ explicitly and for this purpose we proceed as follows.
760: Substituting
761: \begin{eqnarray}
762: \hat{R_2} & = F(I_1)R_2 =
763: \frac{t(-2x+\alpha tx^2+t\dot{x})}{(\alpha^2x^3+3\alpha x\dot{x}+\ddot{x})^2}
764: F(I_1),
765: \label{rex15b}
766: \end{eqnarray}
767: into equation~(\ref{fint06}), we obtain $I_1F_1'+2F = 0$,
768: where $F_1'$ denotes differentiation with respect to
769: $I_1$. Upon integrating the latter we get $F = I_1^{-2}$,
770: (the integration constant is set to zero)
771: which fixes the form of $\hat{R_2}$ as
772: \begin{eqnarray}
773: \hat{R_2} = \frac{t(-2x+\alpha tx^2+t\dot{x})}
774: {(\alpha x^2-\alpha^2tx^3+\dot{x}-3\alpha tx\dot{x}-t\ddot{x})^2}.
775: \label{rex15e}
776: \end{eqnarray}
777: Now one can easily check that this set $(S_2,U_2,\hat{R_2})$ is a compatible
778: solution for the equations~(\ref{met9})-(\ref{met14}) which in turn provides
779: $I_2$ through the relation (\ref{met15}) as
780: \begin{eqnarray}
781: I_2=-\frac{-2\alpha tx^2+\alpha^2t^2x^3+x(2+3\alpha t^2\dot{x})+t(-2\dot{x}
782: +t\ddot{x})}{(\alpha x^2-\alpha^2tx^3+\dot{x}
783: -3\alpha tx\dot{x}-t\ddot{x})}.\label{rex16}
784: \end{eqnarray}
785:
786: As in the previous examples, the set $(S_3,U_3,R_3)$ does not satisfy the
787: constraints (\ref{met12})-(\ref{met14}) and hence one should seek a complete
788: form for $R_3$, which we denote as $\hat{R_3}$, in the form
789: \begin{eqnarray}
790: \hat{R_3} & = F(I_1,I_2)R_3 =
791: F(I_1,I_2)\frac{t(3-3\alpha tx+\alpha^2t^2x^2+\alpha t^2\dot{x})}
792: {(\alpha^2x^3+3\alpha x\dot{x}+\ddot{x})^2}.
793: \label{rex17}
794: \end{eqnarray}
795: Substituting (\ref{rex17}) into equations~(\ref{fint03}) and (\ref{fint04}),
796: we obtain the following equations for $F$, that is,
797: $I_1F_1'+2F=0,\;\;F_2'= 0$,
798: where again $F_1'=\frac{\partial F}{\partial I_1}$ and $F_2'=\frac{\partial
799: F}{\partial I_2}$.
800: Upon integrating the equations we get the explicit form of $F$ as
801: $F = \frac {1}{I_1^2}$,
802: which in turn fixes the form of $\hat{R_3}$ as
803: \begin{eqnarray}
804: \hat{R_3} = \frac{t(3-3\alpha tx+k^2t^2x^2+\alpha t^2\dot{x})}
805: {3\alpha(\alpha x^2-\alpha^2tx^3+\dot{x}-3\alpha tx\dot{x}-t\ddot{x})^2}.
806: \label{rex20}
807: \end{eqnarray}
808: Now the set $(S_3,U_3,\hat{R_3})$ satisfies all the six equations
809: (\ref{met9})-(\ref{met14}) and the relation (\ref{met15}) gives the form of
810: third integral $I_3$ as
811: \begin{eqnarray}
812: I_3=\frac{(6+3\alpha^2t^2x^2-\alpha^3t^3x^3+3\alpha t^2\dot{x}-3\alpha tx(2
813: +\alpha t^2\dot{x})-\alpha t^3\ddot{x})}{6\alpha(\alpha x^2-\alpha^2tx^3+\dot{x}
814: -3\alpha tx\dot{x}-t\ddot{x})}.
815: \label{rex21}
816: \end{eqnarray}
817: Thus we have obtained the explicit forms of the integrals $I_1,\; I_2$ and
818: $I_3$ and hence the
819: solution to equation~(\ref{cat101}) is obtained directly as
820: \begin{eqnarray}
821: x(t)=\frac{\frac{\alpha t^2}{2}+I_1t+I_1I_2}{\frac{\alpha^2t^3}{6}
822: +\alpha I_1\frac{t^2}{2}+\alpha I_1I_2t+I_1I_3}.
823: \label{rex22}
824: \end{eqnarray}
825:
826: Interestingly one can derive the solution (\ref{rex22}) through an alternate way
827: also. For example, instead of solving the '$R$ equation' with rational
828: ansatz,
829: one can look for equations in other variables, that is, either in $U$ or in $S$. For
830: example, from equation (\ref{met10}) we get
831: \begin{eqnarray}
832: S=-(D[U]+\phi_{\dot{x}}-U\phi_{\ddot{x}}-U^2).\label{rat01}
833: \end{eqnarray}
834: Substituting equation~(\ref{rat01}) into the equation~({\ref{met9}) we get a
835: nonlinear PDE for $U$:
836: \begin{eqnarray}
837: D^2[U]-3UD[U]-UD[\phi_{\ddot{x}}]+D[\phi_{\dot{x}}]
838: -\phi_{x}-\phi_{\dot{x}}\phi_{\ddot{x}}+\phi_{\ddot{x}}^2U \qquad \qquad
839: \nonumber\\ \qquad\qquad \qquad \qquad
840: +2\phi_{\ddot{x}}U^2-\phi_{\dot{x}}U+U^3=0.\label{rat02}
841: \end{eqnarray}
842: Now one can look for solutions of equation (\ref{rat02}) with polynomial in
843: $\ddot{x}$. Once $U$ is known one can make use of
844: equations~(\ref{rat01}) and (\ref{met11}) to get the forms of corresponding $S$
845: and $R$ respectively. It turns out that for some cases, like the present example
846: (\ref{rex11}) (see the actual forms of $U$ in equation (\ref{rex13c})), solving
847: equation~(\ref{rat02}) is
848: easier than solving equation~(\ref{poly01}) to solve. However, this can be
849: decided only by actual calculation.
850:
851: \section{Linearization}
852: In \S2 and \S3 we discussed the complete integrability of third order ODEs by
853: investigating sufficient number of integrals of motion. However, one can also
854: establish the complete integrability of the given nonlinear ODE by transforming
855: it into a linear free particle second order ODE or into a third order linear
856: ODE of the form $\frac{d^3w}{dz^3}=w'''=0$.
857: Unlike the second order ODEs, the third order nonlinear ODEs can be linearized
858: through a wide class of transformations, namely, invertible point
859: transformation (Steeb 1993; Ibragimov \& Meleshko 2005),
860: contact transformation
861: (Duarte \textit{et al.} 1994; Bocharov \textit{et al.} 1993;
862: Ibragimov \& Meleshko 2005),
863: generalised Sundman transformation
864: (Berkovich \& Orlova 2000; Euler \textit{et al.} 2003; Euler \& Euler 2004) and
865: their generalizations. In the
866: following, we describe a procedure to deduce these transformations
867: from the first integral itself and illustrate our ideas with relevant examples.
868: \subsection{Transformation from third order nonlinear ODEs to second order
869: free particle equation}
870: Let us suppose that the ODE (\ref{met1}) admits a first integral
871: \begin{eqnarray}
872: I_1 =F(t,x,\dot{x},\ddot{x}), \label{the01}
873: \end{eqnarray}
874: where $F$ is a function of $t,x,\dot{x}$ and $\ddot{x}$ only. Now extending our
875: earlier proposal for second order ODEs (Chandrasekar \textit{et al.} 2005) to
876: the third order equations (\ref{met1}), let us split the
877: function $F$ into a product of two functions such that one
878: involves a perfect differentiable function $\frac{d}{dt}G_1(t,x,\dot{x})$ and
879: another function $G_2(t,x,\dot{x},\ddot{x})$, that is,
880: \begin{eqnarray}
881: I_1=F\left(\frac{1}{G_2(t,x,\dot{x},\ddot{x})}\frac{d}{dt}G_1(t,x,\dot{x})
882: \right).\label{the02}
883: \end{eqnarray}
884: Suppose $G_2$ is a total time derivative of another function, say,
885: $z$, that is $dz/dt=G_2(t,x,\dot{x},\ddot{x})$, then (\ref{the02})
886: can be further rewritten in the form
887: \begin{eqnarray}
888: I_1=F\left(\frac{1}{\frac{dz}{dt}}\frac{dG_1}{dt}\right)
889: =F\bigg(\frac{dG_1}{dz}\bigg).
890: \label{the02b}
891: \end{eqnarray}
892: Now identifying the function $G_1(t,x,\dot{x})=w$ as the new dependent variable,
893: equation (\ref{the02b}) can be recast in the form
894: $I_1=F(\frac {dw}{dz})$.
895: In other words, we obtain
896: \begin{eqnarray}
897: \hat{I}_1=\frac {dw}{dz},
898: \label{the02d}
899: \end{eqnarray}
900: where $\hat{I}_1$ is a constant, from which one can get $\frac{d^2w}{dz^2}=0$.
901: Rewriting $w$ and $z$ in terms of old variables, namely,
902: \begin{eqnarray}
903: w = G_1(t,x,\dot{x}),\quad z = \int_o^t G_2(t',x,\dot{x},\ddot{x}) dt',
904: \label{the03}
905: \end{eqnarray}
906: we can get a linearizing transformation to transform the third order nonlinear ODE
907: into the second order free particle equation. Then to deduce the general
908: solution one has to carry out one more integration.
909: \subsection{Transformation from third order nonlinear equation (\ref{met1})
910: to third order linear ODE $w'''=0$}
911: Next, we aim to transform equation (\ref{met1}) into a third
912: order linear ODE and so we try to rewrite the first integral as a
913: perfect second order derivative.
914: We note that this can be done when one is able to evaluate the
915: following integral explicitly,
916: \begin{eqnarray}
917: \hat{w} = \int_o^t G_1(t',x,\dot{x})G_2(t',x,\dot{x},\ddot{x}) dt'
918: =G_3(t,x,\dot{x}),
919: \label{the03a}
920: \end{eqnarray}
921: where $G_1$ and $G_2$ are defined as in equation (\ref{the02}). Note that in the
922: function $G_3$, the $\ddot{x}$ dependence has been integrated out. The reason
923: for making such a specific decomposition is that in this case equation
924: (\ref{the02})
925: can be rewritten as a simple second order ODE for the variable $\hat{w}$ (see
926: equation (\ref{the003b}) below). We pursued a similar kind of approach in the
927: integrable
928: force-free Duffing van der Pol oscillator equation (Chandrasekar \textit{et al.}
929: 2004) which has now been generalized in the present case. Here one
930: can rewrite (\ref{the01}) as a perfect second order derivative
931: as follows.
932: Differentiating (\ref{the03a}) with respect to $t$, we obtain
933: $\frac {d\hat{w}}{dt}=G_1G_2$.
934: Rewriting the left hand side in the form
935: \begin{eqnarray}
936: \frac {dz}{dt}\frac {d\hat{w}}{dz}=G_1G_2,
937: \label{the003}
938: \end{eqnarray}
939: and using the identities already used in equation (\ref{the03}), namely
940: $\frac {dz}{dt}=G_2$ and $G_1=w$, in equation (\ref{the003}) the latter becomes
941: \begin{eqnarray}
942: \frac{d\hat{w}}{dz}=w.
943: \label{the003b}
944: \end{eqnarray}
945: Differentiating (\ref{the003b}) with repect to $z$ and using the
946: identity $\frac{dw}{dz}=\hat{I}_1$ (vide equation (\ref{the02d})), one gets
947: $\frac{d^2\hat{w}}{dz^2}=\hat{I}_1$.
948: In other words, we have
949: \begin{eqnarray}
950: \frac{d^3\hat{w}}{dz^3}=0, \label{the05b}
951: \end{eqnarray}
952: so that $\hat{w}$ and $z$ are the required transformation variables.
953:
954: \subsection{The nature of transformations}
955: In the previous two subsections, \S4$\,a\,$ and \S4$\,b\,$, we demonstrated
956: how to
957: construct linearizing transformations from the first integral and how they
958: effectively change the given third order nonlinear ODE to either second
959: or third order linear equation. Depending upon the explicit form of
960: the transformations we can call them as point, contact,
961: generalised Sundman or generalized linearizing transformations
962: (GLT). To
963: demonstrate how
964: different kinds of transformation arise let us consider the
965: transformation, $\hat{w} = G_3,\; z = \int_o^t G_2 dt'$ (vide
966: equations (\ref{the03a}) and (\ref{the03})), which takes the given nonlinear ODE
967: into a linear equation. Now, in the above transformation,
968: suppose $z$ is a perfect differentiable
969: function and $\hat{w}$ and $z$ do not contain the variable $\dot{x}$ and
970: $\ddot{x}$, then
971: we call the resultant transformation, namely, $\hat{w}=f_1(x,t)$ and $z=f_2(x,t)$,
972: as a point transformation. Further, if the transformation admits the variable
973: $\dot{x}$ also explicitly then it
974: becomes the contact transformation. In this case, we have $w=f_1(t,x,\dot{x})$
975: and $z=f_2(t,x,\dot{x})$. On the other hand, if
976: $\hat{w}=G_3(t,x)$ and $z = \int_o^t G_2(t',x) dt'$ then the
977: transformation is said to be a generalised Sundman transformation. Note that
978: in the latter case the new independent
979: variable is in an integral form. Besides the above, as pointed out earlier,
980: we find that
981: there exists another kind of transformation, namely, generalized linearizing
982: transformations (GLT), in which the new dependent and independent variables
983: take the form $\hat{w} = G_3(t,x,\dot{x})$ and $z = \int_o^t
984: G_2(t',x,\dot{x},\ddot{x}) dt'$, respectively.
985:
986: \subsection{Connection between the functions $G_1,\;G_2$ and $G_3$}
987: Finally we explore the connection between the functions $G_1,G_2$ and $G_3$.
988: As we have seen above, for the given equation to be linearizable it should be
989: transformable to
990: the form (\ref{the003b}). Rewriting the latter in terms of the variables, $G_1,G_2$
991: and $G_3$ we get,
992: \begin{eqnarray}
993: G_{3t}+\dot{x}G_{3x}+\ddot{x}G_{3\dot{x}}
994: =G_1(t,x,\dot{x})G_2(t,x,\dot{x},\ddot{x}).
995: \label{cont01}
996: \end{eqnarray}
997: Note that the left hand side of equation (\ref{cont01}) contains the variable
998: $\ddot{x}$ linearly. So the right hand side should
999: also be linear in $\ddot{x}$. Consequently, we can write
1000: \begin{eqnarray}
1001: G_2(t,x,\dot{x},\ddot{x})
1002: =G_{21}(t,x,\dot{x})\ddot{x}+G_{22}(t,x,\dot{x}).
1003: \label{cont04}
1004: \end{eqnarray}
1005: Using (\ref{cont04}) we can rewrite equation (\ref{cont01}) in the form
1006: \begin{eqnarray}
1007: \frac{\bigg(G_{3t}+\dot{x}G_{3x}\bigg)\bigg(1+\frac{\ddot{x}G_{3\dot{x}}}
1008: {G_{3t}+\dot{x}G_{3x}}\bigg)}{G_{22}\bigg(1
1009: +\frac{\ddot{x}G_{21}}{G_{22}}\bigg)} =G_1.
1010: \label{cont03}
1011: \end{eqnarray}
1012: Since the right hand side is independent of $\ddot{x}$, we have from
1013: (\ref{cont03}) that
1014: \begin{eqnarray}
1015: G_{21}(G_{3t}+\dot{x}G_{3x})=G_{22}G_{3\dot{x}}\quad
1016: \mbox{and} \quad G_1=\frac{G_{3\dot{x}}}{G_{21}},
1017: \label{cont02}
1018: \end{eqnarray}
1019: It may be noted that a similar condition has been derived by Bocharov
1020: \textit{et al.} (1993) and Ibragimov \& Meleshko (2005) for the case $G_2$ is a
1021: perfect differential function. In other words, our procedure indicates that more
1022: generalized transformations are possible in the case of third order ODEs. By
1023: imposing the condition (\ref{cont04}) it becomes clear that whatever be the type
1024: of linearizing transformation, the new independent variable should be at the
1025: maximum linear in $\ddot{x}$.
1026:
1027: \subsection{Transformation to fourth and higher order linear ODEs}
1028: In the above sub-sections \S4$\,a\,$, \S4$\,b\,$, \S4$\,c\,$ and \S4$\,d\,$ we
1029: can concentrated only on transforming a nonlinear third order ODE either to a
1030: second order or third order linear equations only. However, one could also
1031: linearize certain third order nonlinear ODEs to fourth order linear ODE. For
1032: example the equation (\ref{rex11}) is linearizable to a fourth order ODE of
1033: the form $\frac{d^4w}{dz^4}=0$, under the nonlocal transformation
1034: $x=\frac{\dot{w}}{\alpha w}$. This is not an isolated example and one can
1035: linearize
1036: a class equations through this procedure. Besides the above one can also consider
1037: linearizing transformations in which the new dependent variable, $\hat{w}$, is a
1038: non-local one. This choice leads us to classify another a large class of
1039: equations which we leave it for future work.
1040: \section{Application}
1041: In this section we consider specific examples to demonstrate the method given
1042: in \S4.
1043: \subsection{Example 1: Point transformation}
1044: Let us consider a nontrivial example which was discussed by
1045: Steeb (1993) in the context of invertible point transformations, namely,
1046: \begin{eqnarray}
1047: \dddot{x}+\frac{3\dot{x}\ddot{x}}{x}-3\ddot{x}-\frac{3\dot{x}^2}{x}+2\dot{x}=0.
1048: \label{fthe20}
1049: \end{eqnarray}
1050: The first integral, which can be obtained using the formulation in \S2, can be
1051: written as
1052: \begin{eqnarray}
1053: I_1 =\bigg(\dot{x}^2+x\ddot{x}-x\dot{x}\bigg)e^{-2t}. \label{the20}
1054: \end{eqnarray}
1055: Rewriting (\ref{the20}) in the form (\ref{the02}) we get
1056: $I_1 =e^{-t}\frac {d}{dt}(x\dot{x}e^{-t})$, so that
1057: \begin{eqnarray}
1058: w=x\dot{x}e^{-t},\quad z=e^{t}. \label{the22}
1059: \end{eqnarray}
1060: As we noted earlier, one could transform (\ref{fthe20}) to the second order free
1061: particle equation, $\frac{d^2w}{dz^2}=0$, by utilizing the transformation
1062: (\ref{the22}). Integrating the equation $\frac{d^2w}{dz^2}=0$ we get
1063: $w=I_1z+I_2$. Using (\ref{the22}) into this expression, the general solution
1064: of equation (\ref{fthe20}) can be obtained (after an integration) as
1065: \begin{eqnarray}
1066: x(t)=(I_1e^{2t}+I_2e^t+I_3)^{\frac{1}{2}},\label{the23c}
1067: \end{eqnarray}
1068: where $I_i,\;i=1,2,3,$ are the integration constants.
1069:
1070: Further, using the equation (\ref{the03a}) we get
1071: $\hat{w} = \int x\dot{x}dt = \frac{x^2}{2}$.
1072: Then we can directly check that the point transformation
1073: \begin{eqnarray}
1074: \hat{w}=\frac{x^2}{2}, \quad z=e^{t},\label{the23b}
1075: \end{eqnarray}
1076: transforms the equation (\ref{fthe20}) to the form $\frac{d^3\hat{w}}{dz^3}=0$.
1077: As we mentioned earlier, since $\hat{w}$ and $z$ involve only $x$ and $t$
1078: they are just point transformations. Integrating the linear equation
1079: $\frac{d^3\hat{w}}{dz^3}=0$ we get
1080: \begin{eqnarray}
1081: \hat{w}=\frac{I_1}{2}z^2+I_2z+I_3. \label{tmet01}
1082: \end{eqnarray}
1083: Rewriting $\hat{w}$ and $z$ in equation (\ref{tmet01}) in terms of the original
1084: variables
1085: $x$ and $t$ by using the transformation (\ref{the23b}) we get the same solution as
1086: equation (\ref{the23c}).
1087:
1088: \subsection{Example 2: Contact transformation}
1089: Let us consider an equation of the form
1090: \begin{eqnarray}
1091: \dddot{x}=\frac{x\ddot{x}^3}{\dot{x}^3}.
1092: \label{che20a}
1093: \end{eqnarray}
1094: Bocharov \textit{et al.} (1993) have shown that equation (\ref{che20a}) can be
1095: linearized through contact transformation. However, the explicit linearizing
1096: transformation is yet to be reported which is also a difficult problem. Here we
1097: derive the explicit form of the linearizing transformation through our
1098: procedure.
1099:
1100: The first integral can be easily deduced using the results of \S2 as
1101: \begin{eqnarray}
1102: I_1 =\frac{\dot{x}^2-x\ddot{x}}{\dot{x}\ddot{x}}. \label{che20}
1103: \end{eqnarray}
1104: Rewriting (\ref{che20}) in the form (\ref{the02}) we get
1105: $I_1 =\displaystyle{\frac{\dot{x}}{\ddot{x}}\frac {d}{dt}(\frac{x}{\dot{x}})}$
1106: so that we have $w=\frac{x}{\dot{x}}$ and $z=\log{\dot{x}}$. The latter transforms
1107: equation (\ref{che20a}) to the second order free particle equation
1108: $\frac{d^2w}{dz^2}=0$, so that $w=I_1z+I_2$,
1109: where $I_1$ and $I_2$ are integration constants. Rewriting $w$ and $z$ in terms
1110: of the old variables we get $x=(I_1\log(\dot{x})+I_2)\dot{x}$. Unlike the
1111: earlier example it is difficult to integrate this equation further and obtain
1112: the general solution. Therefore one can look for variables which
1113: transform the third order nonlinear ODE (\ref{che20a}) to a third order linear
1114: ODE so that the
1115: non-trivial
1116: integration can be avoided. Now using the equation (\ref{the03a}) we get
1117: $\hat{w} = \int \frac{x\ddot{x}}{\dot{x}^2}dt = t-\frac{x}{\dot{x}}$.
1118: The new variables
1119: \begin{eqnarray}
1120: \hat{w}=\frac{t\dot{x}-x}{\dot{x}}, \quad z=\log{\dot{x}},
1121: \end{eqnarray}
1122: transform the equation (\ref{che20a}) to the
1123: form (\ref{the05b}). Unlike the earlier example, $\hat{w}$ and $z$ admit
1124: the
1125: variable $\dot{x}$ explicitly and so they become contact transformation for the
1126: given equation.
1127:
1128: Integrating the linear third order equation (\ref{the05b}), we get
1129: $\hat{w}=\frac{I_1}{2}z^2+I_2z+I_3,$
1130: where $I_i,\;i=1,2,3,$ are integration constants. Now replacing $\hat{w}$ and $z$
1131: in terms of the old variables and using the previous result
1132: $x=(I_1\log(\dot{x})+I_2)\dot{x}$, one can obtain the general solution for the equation
1133: (\ref{che20a}) in the form
1134: \begin{eqnarray}
1135: x(t)=\bigg(-I_1\pm\sqrt{I_1^2+I_2^2-2I_1(I_3-t)}\bigg)
1136: e^{-\frac{I_1+I_2\mp\sqrt{I_1^2+I_2^2-2I_1(I_3-t)}}{I_1}}. \label{cmet02}
1137: \end{eqnarray}
1138:
1139: \subsection{Example 3: Generalised Sundman transformation}
1140: Next we consider the hydrodynamic type equation of the form
1141: (Berkovich \& Orlova 2000; Euler \& Euler 2004)
1142: \begin{eqnarray}
1143: \dddot{x}=\frac{\ddot{x}\dot{x}}{x}, \label {sex21}
1144: \end{eqnarray}
1145: which admits a first integral in the form $I_1=\frac{\ddot{x}}{x}$ and the
1146: latter can be rewritten as
1147: $I_1 =\frac{1}{x}\frac {d}{dt}\dot{x}=\frac{dw}{dz}$, from which we identify
1148: $w=\dot{x}$ and $dz=xdt$.
1149: By utilizing the new variables, one can transform (\ref{sex21}) to the second order free
1150: particle equation, $\frac{d^2w}{dz^2}=0$.
1151: However, from equation (\ref{the03a}) we get $\hat{w} = \int x\dot{x}dt = x^2$.
1152: Then the Sundman transformation
1153: \begin{eqnarray}
1154: \hat{w}=x^2, \quad dz=x dt,\label {sex22}
1155: \end{eqnarray}
1156: transforms the equation (\ref{sex21}) to the
1157: form (\ref{the05b}), namely $\frac{d^3\hat{w}}{dz^3}=0$.
1158:
1159: To derive the solution we proceed as follows. Rewriting the first integral $I_1$
1160: in the integral form we get, $I_1 =\frac{1}{x}\frac {d}{dt}\dot{x}\Rightarrow
1161: \dot{x}=I_1\int xdt$. Now using the identity (\ref{sex22}) in the latter expression
1162: we get $w=I_1z$.
1163: From equation (\ref{the03a}) (for the present case $G_1=w$ and
1164: $G_2=\frac{dz}{dt}$) we have
1165: \begin{eqnarray}
1166: \hat{w}=\int w dz=\int I_1zdz=\frac{I_1}{2}z^2+I_2, \label {sex25}
1167: \end{eqnarray}
1168: where $I_2$ is the integration constant. Using (\ref{sex22}) in (\ref{sex25})
1169: we obtain $x^2=\frac{I_1}{2}z^2+I_2$,
1170: which in turn
1171: leads to a differential equation which connects the variables $z$ and $t$ in
1172: the form (using the relation $dz=xdt$)
1173: \begin{eqnarray}
1174: dz=\sqrt{\frac{I_1}{2}z^2+I_2}dt. \label {sex27}
1175: \end{eqnarray}
1176: Integrating (\ref{sex27}), we obtain
1177: $z= \sqrt{\frac{I_2}{2}}(e^{\sqrt{I_1}(t+I_3)}-e^{-\sqrt{I_1}(t+I_3)})$, where
1178: $I_3$ is the integration constant.
1179: Substituting the latter in the relation
1180: $x^2=\frac{I_1}{2}z^2+I_2$, we arrive at the general solution for
1181: (\ref{sex21}) in the form
1182: \begin{eqnarray}
1183: x(t)=\sqrt{I_2} \cosh{\sqrt{I_1}(t+I_3)}.\label {sex28}
1184: \end{eqnarray}
1185: We note that the solution for the equation (\ref{sex21}) has been already
1186: derived in an alternate way from the Sundman transformation (\ref{sex22})
1187: by Euler \& Euler (2004). However {\it the
1188: procedure which we described in the above is new and can be used for more
1189: general linearizing transformations also}, as we see below.
1190:
1191: \subsection{Example 4: Generalized linearizing transformation (GLT)}
1192: As we noted earlier \textit{some nonlinear ODEs can be linearized only through
1193: more general
1194: nonlocal form of transformations}, which we designate here as generalized
1195: linearizing transformations (GLTs). We illustrate the GLT with the same example
1196: discussed as Example 2 in \S3, which admits a first integral of the form
1197: $I_1=\frac{\dot{x}x}{\ddot{x}}$ (vide equation (\ref{tex11})).
1198: Rewriting this first integral in the form (\ref{the02}) we get
1199: $I_1 =\frac{x}{\ddot{x}}\frac {d}{dt}(x),$
1200: so that
1201: \begin{eqnarray}
1202: w=x,\quad dz=\frac{\ddot{x}}{x}dt,\label {gex11c}
1203: \end{eqnarray}
1204: which can be effectively used to transform the nonlinear ODE (\ref{tex11}) to the
1205: equation $\frac{d^2w}{dz^2}=0$.
1206: Using the equation (\ref{the03a}) we get
1207: $\hat{w} = \int \ddot{x}dt = \dot{x}$.
1208: Then the GLT becomes
1209: \begin{eqnarray}
1210: \hat{w}=\dot{x}, \quad dz=\frac{\ddot{x}}{x}dt,\label {gex12}
1211: \end{eqnarray}
1212: which can be used to transform the equation (\ref{tex11}) to the form
1213: $\hat{w}'''=0$. Note that in the present case the new independent and
1214: dependent variables admit $\ddot{x}$ and $\dot{x}$ terms, respectively and that
1215: the transformation is nonlocal. {\it
1216: Indeed no
1217: such linearizing transformations have been reported in the literature atleast to
1218: our knowledge. We also now establish a method of finding the general solution for
1219: this case}.
1220:
1221: Integrating once the equation $\frac{d^2w}{dz^2}=0$ we get $w=I_1z$ from which
1222: we obtain
1223: \begin{eqnarray}
1224: x=I_1z.\label {gex12a}
1225: \end{eqnarray}
1226: On the other hand equation (\ref{the03a}) provides us a relation (after using
1227: (\ref{gex11c}) and (\ref{gex12}))
1228: \begin{eqnarray}
1229: \dot{x}=\frac{I_1}{2}z^2+I_2.\label {gex12b}
1230: \end{eqnarray}
1231: Now using (\ref{gex12a}) in (\ref{gex12b}), we obtain
1232: \begin{eqnarray}
1233: \bigg(\frac{2I_1}{I_1z^2+2I_2}\bigg)dz=dt. \label {gex17}
1234: \end{eqnarray}
1235: The variables are now separated out and one can integrate
1236: (\ref{gex17}) and obtain
1237: \begin{eqnarray}
1238: z=\sqrt{\frac{2I_2}{I_1}}\tan\sqrt{\frac{I_2}{I_1}}(t+I_3).\label {gex18a}
1239: \end{eqnarray}
1240: Now substutiting (\ref{gex18a}) into (\ref{gex12a}) we get
1241: \begin{eqnarray}
1242: x(t)=\sqrt{2I_1I_2}\tan\sqrt{\frac{I_2}{I_1}}(t+I_3).\label {gex18}
1243: \end{eqnarray}
1244: which is effectively the same as (\ref{tex22}).
1245:
1246: Finally, we note that the procedure given above can be profitably utilized for
1247: other examples also which are linearized by GLTs.
1248: \subsection{Example 5: An elementary nontrivial system of hydrodynamic type}
1249: Finally, to show the importance of the GLT and how this transformation gives
1250: additional information about the
1251: linearization of nonlinear third order ODEs we consider the following specific
1252: example which was discussed in the literature (Berkovich 1996;
1253: Berkovich \& Orlova 2000),
1254: \begin{eqnarray}
1255: \dddot{x}=\frac{\ddot{x}\dot{x}}{x}-4\alpha x^2\dot{x},\;\;\alpha :\;parameter.
1256: \label {gex21}
1257: \end{eqnarray}
1258: Equation (\ref{gex21}) is nothing but the dynamical equation of the
1259: Euler-Poinsot case of a rigid body written in terms of a single variable
1260: (Berkovich 1996; Berkovich \& Orlova 2000). As we have seen earlier this equation is
1261: linearizable in the case $\alpha=0$ through generalized Sundmann
1262: transformation. \textit{However, we wish to show here that the general equation
1263: (\ref{gex21}) itself is linearizable through the GLT}.
1264:
1265: From the first integral $I_1=\frac{\ddot{x}}{x}+2\alpha x^2$ associated with equation
1266: (\ref{gex21}),
1267: one can identify the GLT
1268: \begin{eqnarray}
1269: \hat{w}=x^2, \quad dz=\frac{2\dot{x}x}{\sqrt{\dot{x}^2+\alpha x^4}}dt,\label {gex22}
1270: \end{eqnarray}
1271: which transforms the equation (\ref{gex21}) to the form (\ref{the05b}).
1272: Note that for the choice $\alpha=0$ the independent variable becomes $dz=2xdt$ and so
1273: it becomes the generalized Sundman transformation, equation (\ref{sex22}),
1274: identified in the literature
1275: (Berkovich \& Orlova 2000; Euler \& Euler 2004). Now following the
1276: steps given in Example 4 one can deduce the general solution for the equation
1277: (\ref{gex21}) in terms of Jacobian elliptic function as
1278: \begin{eqnarray}
1279: x(t)=\bigg(I_1(c-(c-b)sn^2[\sqrt{\alpha
1280: I_1(c-a)}(t-t_0),m])+I_2\bigg)^{\frac{1}{2}},
1281: \label {gex28}
1282: \end{eqnarray}
1283: where $a+b+c=\frac{(2I_1-3\alpha I_2)}{4\alpha I_1},\;
1284: ab+ac+cb=\frac{(3\alpha I_2^2-2I_1I_2)}{4\alpha I_1^2},
1285: \;abc=-\frac{I_2^3}{4I_1^3}$, $m^2=\frac{b-c}{a-c}$ and $I_2$ and $t_0$ are
1286: integration constants.
1287:
1288: \section{Conclusion}
1289: In this paper we have discussed a method of finding the integrals of motion and
1290: general solution
1291: associated with third order nonlinear ODEs through the modified PS method
1292: by a nontrivial extension of our earlier work on second order ODEs
1293: (Chandrasekar \textit{et al.} 2005). We
1294: illustrated the validity of the method with suitable examples. Further, we introduced a
1295: technique which can be utilized to derive linearizing transformations from the
1296: first integral. Interestingly, we showed that different types of
1297: transformations, namely, point, contact, Sundman and generalized linearizing
1298: transformations can be derived in a unique way from the first integral.
1299: We also indicated a procedure to derive general solution for the third order
1300: ODEs when GLTs occur. We believe that the GLT introduced in this
1301: paper will be highly useful to tackle new systems such as equation
1302: (\ref{gex21}). Finally, the modified PS
1303: method can also be extended to higher order ODEs and coupled systems of
1304: ODEs. As far as the linearization of higher order ODEs are concerned
1305: it is still an open and challenging problem. As we pointed out in the
1306: introduction one can unearth a wide variety of linearizing transformations for
1307: the higher order ODEs besides formulating the necessary and sufficient condition
1308: for linearizing these equations in each form of transformation. We hope to
1309: address some of these aspects shortly.
1310:
1311: The work of VKC is supported by Council of Scientific and Industrial Research
1312: in the form of a Senior Research
1313: Fellowship. The work of MS and ML forms part of a Department of Science
1314: and Technology, Government of India, sponsored research project.
1315:
1316: \begin{thebibliography}
1317:
1318: \item
1319: Berkovich, L. M. 1996 The method of an exact linearization of
1320: $n$-order ordinary differential equations.
1321: \textit{J. Nonlinear Math. Phys.} \,\textbf{3}, 341-350.
1322:
1323: \item
1324: Berkovich, L. M. \& Orlova, I. S. 2000 The exact linearization of some classes of
1325: ordinary differential equations for order $n>2$. \textit{Proceedings of Institute
1326: of Mathematics of NAS of Ukraine}\, \textbf{30}, part 1, 90-98.
1327:
1328: \item
1329: Bluman, G. W. \& Anco, S. C. 2002 \textit{Symmetries and Integration Methods
1330: for Differential Equations}. New York: Springer.
1331:
1332: \item
1333: Bocharov, A. V., Sokolov, V. V. \& Svinolupov, S. I. 1993
1334: On some equivalence problems for differential equations. \textit{ESI Preprint}
1335: \,\textbf {54}, 1-12.
1336:
1337: \item
1338: Chandrasekar, V. K., Senthilvelan, M. \& Lakshmanan, M. 2004 New aspects of
1339: integrability of force-free Duffing-van der Pol oscillator and related
1340: nonlinear systems. \textit{J. Phys.} A \,\textbf {37}, 4527-4534.
1341:
1342: \item
1343: Chandrasekar, V. K., Senthilvelan, M. \& Lakshmanan, M. 2005 On the complete
1344: integrability and linearization of certain second order nonlinear ordinary
1345: differential equations. \textit{Proc. R. Soc. London A} \,\textbf{461},
1346: 2451-2476.
1347:
1348: \item
1349: Duarte, L. G. S., Moreira, I. C. \& Santos, F. C. 1994 Linearization under
1350: non-point transformations. \textit{J.Phys.} A \,\textbf{27}, L739-L743.
1351:
1352:
1353: \item
1354: Duarte, L. G. S., Duarte, S. E. S., da Mota, A. C. P. \& Skea, J. E. F.
1355: 2001 Solving the second-order ordinary differential equations by extending
1356: the Prelle-Singer method. \textit{J. Phys.} A \,\textbf {34}, 3015-3024.
1357:
1358: \item
1359: Euler, N., Wolf, T., Leach, P. G. L. \& Euler, M. 2003 Linearisable third-order
1360: ordinary differential equations and generalised Sundman transformations : The
1361: case $X'''=0$. \textit{Acta. Appl. Math.} \,\textbf{76}, 89-115.
1362:
1363: \item
1364: Euler, N. \& Euler, M. 2004 Sundman Symmetries of nonlinear second-order and
1365: third-order ordinary differential equations. \textit{J. Nonlinear Math. Phys.}
1366: \,\textbf{11}, 399-421.
1367:
1368: \item
1369: Feix, M. R., Geronimi, C., Cairo, L., Leach, P. G. L., Lemmer, R. L. \&
1370: Bouquet, S. 1997 On the singularity analysis of ordinary differential equations
1371: invariant under time translation and rescaling.
1372: \textit{J. Phys.} A \,\textbf{30}, 7437-7461.
1373:
1374: \item
1375: Halburd, R. 1999 Integrable relativistic models and the generalized Chazy
1376: equation. \textit{Nonlinearity} \,\textbf {12}, 931-938.
1377:
1378: \item
1379: Ibragimov, N. H. \& Meleshko, S. V. 2005 Linearization of third-order differential
1380: equation by point and contact transformations. \textit{J. Math. Anal. Appl}
1381: \,\textbf{308}, 266-289.
1382:
1383: \item
1384: Mugan, U. \& Jrad, F. 2002 Painlev\'e test and higher order differential
1385: equation. \textit{J. Nonlinear Math. Phys.} \,\textbf{9}, 282-310.
1386:
1387: \item
1388: Olver, P. J. 1995 \textit{Equivalence, invariants, and symmetry}. Cambridge:
1389: Cambridge University Press.
1390:
1391: \item
1392: Polyanin, A. D. \& Zaitsev, V. F. 1995 \textit{Handbook of exact solutions for
1393: ordinary differential equations}. London: CRC press, Inc.
1394:
1395: \item
1396: Prelle, M. \& Singer, M. 1983 Elementary first integrals of differential
1397: equations. \textit{Trans. Am. Math. Soc.}\,\textbf {279}, 215-229.
1398:
1399: \item
1400: Steeb, W. H. 1993 \textit{Invertible point transformations and nonlinear
1401: differential equations}. London: World Scientific.
1402:
1403: \end{thebibliography}
1404: \end{document}
1405:
1406:
1407: