math0509261/cttt.tex
1: \documentstyle[11pt]{article}
2: 
3: \raggedbottom
4: \setlength{\textwidth}{6.0in}
5: \setlength{\textheight}{8.5in}
6: \setlength{\oddsidemargin}{0.2in}
7: \setlength{\topmargin}{-0.5in}
8: \title{A CIRCUIT-THEORETIC ANOMALY RESOLVED BY NONSTANDARD ANALYSIS}
9: \author{A. H. Zemanian}
10: \date{}
11: 
12: \begin{document}
13: \newcommand{\R} {I \kern -4.5pt R}
14: \newcommand{\N} {I \kern -4.5pt N}
15: \maketitle
16: \baselineskip21pt
17: 
18: {\ Abstract --- An anomaly in circuit theory is 
19: the disappearance of some of the stored energy when two capacitors, 
20: one charged and the other uncharged, are connected together through 
21: resistanceless wires.  Nonstandard analysis shows that, when the wires are 
22: taken to have infinitesimally small but nonzero 
23: resistance, the energy dissipated 
24: in the wires equals that substantial amount of energy that 
25: had disappeared, and that all but an infinitesimal amount of this
26: dissipation occurs during an infinitesimal 
27: initial time period.  This provides still another 
28: but quite simple model of what is in fact a multifaceted physical 
29: phenomenon. It also exemplifies the efficacy of at least 
30: one application of nonstandard analysis to circuit theory.\\
31: 
32: Key Words: Capacitive circuits, capacitive energy anomaly, 
33: nonstandard circuit analysis
34: } 
35: 
36: \section{Where Did the Energy Go?}
37: 
38: A much-discussed
39: anomaly concerning the flow of energy in purely capacitive networks
40: with switches and resistanceless wires is illustrated perhaps in
41: its simplest form by the circuit of Figure 1.  Here, two 
42: capacitors of equal capacitance $c>0$ are initially disconnected 
43: from each other during time $t<0$ because of an open switch.
44: Let the charge on the left-hand capacitor be of constant value
45: $q_{0}=q_{1}(t)$ and the charge on the 
46: right-hand capacitor be $0=q_{2}(t)$ for $t<0$.
47: Thus, during that time, $v_{0}=q_{0}/c$ is the voltage on the left-hand
48: capacitor, and the total stored energy in the circuit is $q_{0}v_{0}/2$. 
49: With the switch being thrown closed at $t=0$, the charge on each capacitor 
50: becomes $q_{0}/2$, and the total stored energy becomes
51: $q_{0}v_{0}/8+q_{0}v_{0}/8=q_{0}v_{0}/4$. Energy in the amount 
52: of $q_{0}v_{0}/4$ has disappeared.  Where did it go?
53: 
54: \section{No Problem When the Circuit Has Resistance}
55: 
56: Assume that the wires have some resistance $r>0$, as shown 
57: in lumped fashion in Figure 2.  Then, the disappearing 
58: energy is accounted for
59: by dissipation in $r$.  Indeed, for $t>0$, the current 
60: $i_{r}(t)$ in the circuit is 
61: \begin{equation}
62: i_{r}(t)\;=\;\frac{q_{0}}{rc}\,e^{-2t/rc},  \label{2.1}
63: \end{equation}
64: and the corresponding power $p_{r}(t)\;=\;i(t)^{2}r$ is
65: \begin{equation}
66: p_{r}(t)\;=\;\frac{q_{0}v_{0}}{rc}\,e^{-4t/rc}.  \label{2.2}
67: \end{equation}
68: Hence, the total energy $E_{r}(0,\infty)$ dissipated in $r$ is
69: \begin{equation}
70: E_{r}(0,\infty)\;=\;\int_{0}^{\infty}\,p_{r}(t)\,dt\;=\;
71: \frac{q_{0}v_{0}}{4}  \label{2.4}
72: \end{equation}
73: This is exactly the difference between the initial capacitively
74: stored energy $q_{0}v_{0}/2$ and the final capacitively stored energy 
75: $q_{0}v_{0}/4$ occurring in the limit as $t\rightarrow\infty$.
76: 
77: \section{Standard Distribution Theory Does Not Help}
78: 
79: Perhaps we can account for the discrepancy noted in  Section 1
80: by using standard distribution theory.  After all, as
81: $r\rightarrow 0$, $i_{r}(t)$ approaches $\delta(t)q_{0}/2$, 
82: a delta function of size $q_{0}/2$.
83: Indeed, for each $r>0$, $i_{r}(t)=0$ for $t<0$ and $\int_{0}^{\infty}
84: i_{r}(t)\,dt\,=\, q_{0}/2$, whereas $\int_{t>T} i_{r}(t)\,dt\rightarrow 0$
85: for every $T>0$ as $r\rightarrow 0$.  So, why not use this delta function 
86: in the first place, assume $r$ is not zero (perhaps it is 
87: a nonzero infinitesimal), 
88: and calculate the energy dissipation in $r$ as follows:
89: \[ \int_{-\infty}^{\infty}(i_{r}(t))^{2}\,r\,dt\;=\;\frac{q_{0}^{2}}{4}r
90: \int_{-\infty}^{\infty} (\delta(t))^{2}\,dt\;=\;? \]
91: Unfortunately $\delta^{2}$ is not well-defined as a distribution.  
92: Singular distributions cannot be multiplied together,
93: according to standard distribution theory.\footnote{However, nonstandard
94: generalized functions may circumvent this trouble in multiplication;
95: in this regard, see the Final Remark in Section 7.}
96: It appears that the
97: difficulty of Section 1 persists if we try this approach.
98: 
99: \section{More Complicated Models Can Account for the Vanishing Energy}
100: 
101: One can set up in the laboratory the circuit of Figure 1 using
102: real components and can note that the discrepancy between the 
103: initial and final stored energies truly occurs.  One must conclude
104: that the ideal circuit of Figure 1 with perfect elements is 
105: too simplified to account for the real phenomenon.
106: The vanishing stored energy might be accounted for in several ways.
107: There is indeed some resistance in any wire, which will dissipate energy.
108: The dielectric within the capacitors has nonlinear resistance 
109: as well, producing more dissipation.  The current produces a magnetic field
110: and thereby inductance, which will affect the transient state and
111: thereby the resistive power dissipation over time.
112: There may be arcing at the switch as it is being closed, 
113: producing thereby more heat.  Moreover, the varying electric fields in
114: the capacitors produce magnetic fields and thereby radiation.
115: A search on the internet\footnote{For instance, 
116: search ``capacitance energy loss'' in google or yahoo.} 
117: yields a number of references discussing all this, some examples of which are
118: \cite{j-s}, \cite{a-w}, \cite{fo}, \cite{pa}.  
119: Some of that discussion occurs as internet ``chat.''
120: 
121: So, a variety of different models are suggested by this 
122: multifaceted physical phenomenon.  Which model is 
123: preferred depends upon a compromise between 
124: accuracy and simplicity. 
125: The objective of this brief note is to suggest one more, but quite simple,
126: model.  It uses nonstandard analysis to extend the $rc$ circuit of 
127: Figure 2 to the case where $r$ is a positive infinitesimal, a quantity that 
128: is less than any real positive number but is not negative, and thus is 
129: effectively zero from the perspective of standard analysis.
130: Another purpose of the note is to demonstrate the efficacy 
131: of nonstandard analysis in circuit theory and perhaps 
132: encourage its use in other engineering analyses.
133: 
134: \section{How Nonstandard Analysis is Used in This Case}
135: 
136: Up to now, we have tacitly restricted all our variables and 
137: parameters to real numbers. We will continue to take the capacitance $c$
138: and the initial charge $q_{0}$ and voltage $v_{0}=q_{0}/c$ as
139: fixed real positive numbers.  But, we will allow all other quantities to
140: be nonstandard, namely hyperreal numbers---with real numbers being a 
141: special case of hyperreal numbers.\footnote{As is conventional, we will simply 
142: write ``hyperreal'' for ``hyperreal number'' and ``real'' 
143: for ``real number.''}  The hyperreals comprise the enlargement 
144: $^{*}\!\R$ of the real line $\R$.  $^{*}\!\R_{+}$ and  $\R_{+}$
145: denote the positive parts of $^{*}\!\R$ and $\R$.  Each real
146: $a\in\R$ is contained in a set of hyperreals that are infinitesimally 
147: close to $a$.  That set is called the halo---or synonymously 
148: the monad---for $a$.  Moreover, for any two reals $a$ and $b$ where $a< b$,
149: the halos for $a$ and $b$ do not overlap.  Furthermore, the 
150: hyperreal line $^{*}\!\R$ extends into unlimited (synonymously, 
151: infinitely large) hyperreals.  
152: 
153: Here is how these ideas can be used to resolve the anomaly.
154: Instead of setting $r$ exactly equal to 0, we can let $r$ be a 
155: positive infinitesimal in Figure 2.  It can then be 
156: shown that, at each real positive time $t$, 
157: the current $i_{r}(t)$ and power $p_{r}(t)$
158: dissipated in $r$ are infinitesimals.  On the other hand, 
159: it can be shown that, at all sufficiently small positive 
160: infinitesimal time $t$, the current $i_{r}(t)$ and 
161: power $p_{r}(t)$ are positive unlimited hyperreals.
162: It can also be seen that the total energy $E_{r}(0,\infty)$
163: dissipated in the positive infinitesimal $r$ during the real time 
164: interval $0\leq t <\infty$ is equal to the real value 
165: $q_{0}v_{0}/4$, and this is so no matter 
166: how small this infinitesmal $r$ is chosen.  
167: No longer does that dissipated energy 
168: disappear---as it did under standard analysis with $r=0$.
169: Moreover, it can be seen that the energy $E_{r}(\tau,\infty)\;=\;
170: \int_{\tau}^{\infty}\, p_{r}(t)\,dt$ dissipated in the 
171: infinitesimal $r$ during the real time interval $\tau\leq t <\infty$, where 
172: $\tau$ is any real positive time, is also infinitesimal.
173: Consequently, we can assert that the energy dissipated
174: in $r$ during the positive part of the time halo around $t=0$
175: is infinitesimally close to $q_{0}v_{0}/4$.  Done.
176: 
177: \section{The Details}
178: 
179: To explicate all of this, we invoke a few results from nonstandard analysis.  
180: There have been many expositions of that theory during the 45 years 
181: since its inception \cite{ro}.  The book \cite{go} lists 41 such
182: sources appearing before 1998, and \cite{go} is itself a 
183: well-written introduction to the subject.  A concise 
184: listing and explanation of the ideas used herein
185: can be found in \cite[Appendix A]{ze}.  We shall now present
186: derivations of the thoughts of the preceding section and will refer 
187: to \cite[Appendix A]{ze} for certain definitions and results of 
188: nonstandard analysis.
189: 
190: As above, $\R$ and $^{*}\!\R$ denote the real and hyperreal lines
191: respectively \cite[Appendix A.5]{ze}, and $\R_{+}$ and $^{*}\!\R_{+}$
192: denote their positive parts.  Also, $\N=\{0,1,2,\ldots\}$ 
193: denotes the set of natural numbers.
194: $\langle a_{n}\rangle$ will denote a sequence of real numbers 
195: $a_{0},a_{1},\ldots,a_{n},\ldots$ indexed by the natural numbers.
196: One way (but not the only way) of introducing the hyperreals is 
197: to define them as equivalence classes of sequences
198: of real numbers.  To specify that equivalence relation,
199: we choose a nonprincipal ultrafilter $\cal F$.  This is a set of subsets of 
200: $\N$ satisfying certain conditions \cite[Appendix A.4]{ze}.
201: Two sequences $\langle a_{n}\rangle$ and $\langle b_{n}\rangle$
202: are taken to be equivalent if $\{n\in\N: a_{n}=b_{n}\}\in{\cal F}$.
203: A hyperreal will be denoted by $[a_{n}]$, where the $a_{n}$ 
204: are the elements of one (i.e., any one) of 
205: the sequences in the equivalence class.  Thus, if $\langle a_{n}\rangle$
206: and $\langle b_{n}\rangle$ are equivalent sequences, then 
207: $[a_{n}]$ and $[b_{n}]$ denote the same hyperreal (i.e., $[a_{n}]=[b_{n}]$).
208: Each hyperreal is either positive (i.e., $\{n\!: a_{n}>0\}\in {\cal F}$),
209: or negative (i.e., $\{n\!: a_{n}<0\}\in {\cal F}$),
210: or 0 (i.e., $\{n\!: a_{n}=0\}\in {\cal F}$).  Only one of these 
211: conditions will hold \cite[Appendix A.6]{ze}.
212: 
213: An infinitesimal $[a_{n}]$ is a special kind of hyperreal defined
214: as follows.  If for every $x\in\R_{+}$ we have 
215: $\{n\!: |a_{n}|< x\}\in{\cal F}$, then $[a_{n}]$ is an infinitesimal.
216: Similarly, if for every $x\in\R_{+}$ we have 
217: $\{n\!: |a_{n}|> x\}\in{\cal F}$, then $[a_{n}]$ is called 
218: an unlimited hyperreal (synonymously, 
219: an infinitely large hyperreal).  If $[a_{n}]$ is neither infinitesimal nor 
220: unlimited, it is called appreciable.  The product of an unlimited 
221: hyperreal and an appreciable hyperreal is an unlimited hyperreal.
222: We emphasize that these definitions do not depend upon the 
223: choice of the representative sequence $\langle a_{n}\rangle$
224: for $[a_{n}]$;  this is consequence of the properties of the 
225: nonprincipal ultrafilter $\cal F$.
226: 
227: Now, let $r=[r_{n}]$ be a positive infinitesimal.  Consider Figure 2
228: with $r$ replaced by $r_{n}$.  Then, for 
229: any natural number $n$ for which $r_{n}>0$ and for any 
230: $t\in\R_{+}$, the current $i_{r_{n}}(t)$ is given by (\ref{2.1}) and 
231: the power dissipated in $r_{n}$ is given by (\ref{2.2}), but with $r$ replaced 
232: by $r_{n}$ of course.
233: 
234: We shall now show that the hyperreal power $p_{r}(t)=[p_{r_{n}}(t)]$
235: is infinitesimal at each $t\in\R_{+}$.  Remember that $r_{n}$ and 
236: $p_{r_{n}}(t)$ are both real positive numbers.  Given any $t\in\R_{+}$ and 
237: given any $\epsilon\in\R_{+}$, 
238: there exists a $\rho\in\R_{+}$ such that $r_{n}<\rho$ implies 
239: that $p_{r_{n}}(t)<\epsilon$.  Thus, 
240: \begin{equation}
241: \{n\!: p_{r_{n}}(t)<\epsilon\}\;\supseteq\;\{n\!: r_{n}<\rho\}  \label{5.1}
242: \end{equation}
243: But, $\{n\!: r_{n}<\rho\}\in{\cal F}$ because $[r_{n}]$ is
244: infinitesimal.  By the properties of an ultrafilter (in this 
245: case, any filter) \cite[Appendix A.4]{ze}, it follows that 
246: $\{n\!:p_{r_{n}}(t)<\epsilon\}\in{\cal F}$, and this 
247: is so for every $\epsilon\in\R_{+}$. 
248: So truly, $[p_{r_{n}}(t)]$
249: is infinitesimal.
250: 
251: In a similar way, it can be shown that the hyperreal
252: current $i_{r}(t)=[i_{r_{n}}(t)]$ is infinitesimal for each $t\in\R_{+}$.
253: 
254: Furthermore, $p_{r}(t)=[p_{r_{n}}(t)]$ can be shown to be positive
255: unlimited for every sufficiently small infinitesimal time
256: $t=[t_{n}]$ by examining sets in $\cal F$ as above.  But,
257: let us now use a somewhat more concise argument.  For $r=[r_{n}]$
258: being a positive infinitesimal, $[q_{0}v_{0}/r_{n}c]$
259: is a positive unlimited hyperreal.  On the other hand,
260: if the infinitesimal time $t=[t_{n}]\in\,^{*}\!\R_{+}$ is chosen 
261: so small that $t<r$, then $[e^{-4t_{n}/r_{n}c}]$
262: is appreciable, being larger than $e^{-4/c}\in\R_{+}$ and
263: less than 1.  (To do this, just choose the $t_{n}$ such that
264: $\{n\!: t_{n}/r_{n}<1\}\in{\cal F}$.  Now, $p_{r}(t)=[p_{r_{n}}(t)]$
265: is the product of the unlimited $[q_{0}v_{0}/r_{n}c]$ and the appreciable
266: $[e^{-4t_{n}/r_{n}c}]$
267: and therefore must be positive unlimited.
268: 
269: Again, in the same way we can show that the hyperreal current 
270: $i_{r}(t)=[i_{r_{n}}(t)]$ is also positive unlimited for all 
271: sufficiently small infinitesimal $t=[t_{n}]$.  
272: 
273: Consider now the energy $E_{r_{n}}(\tau,\infty)$ dissipated 
274: in $r_{n}\in\R_{+}$ from real time $t=\tau\in\R_{+}$ to $t=\infty$.
275: \[ E_{r_{n}}(\tau,\infty) \;=\;\int_{\tau}^{\infty} p_{r_{n}}(t)\,dt
276: \;=\;\frac{q_{0}v_{0}}{4}\,e^{-\tau/r_{n}c} \]
277: Again an argument similar to that given for 
278: $p_{r}(t)=[p_{r_{n}}(t)]$ (see the argument regarding (\ref{5.1}))
279: shows that the hyperreal energy $E_{r}(\tau,\infty)\,=\,
280: [E_{r_{n}}(\tau,\infty)]$  dissipated in the infinitesimal
281: resistor $r=[r_{n}]$ from $\tau$ to $\infty$ is also infinitesimal, whatever 
282: be the choice of $\tau\in\R_{+}$.  More particularly,
283: for any hyperreal $t=[t_{n}]>\tau\in\R_{+}$ (possibly, $t$ is positive 
284: unlimited), the hyperreal energy
285: \[ E_{r}(\tau,t)\;=\;\left[\int_{\tau}^{t} p_{r_{n}}(x)\,dx\right] \]
286: remains infinitesimal and less than $E_{r}(\tau,\infty)$
287: no matter how large the hyperreal time 
288: $t=[t_{n}]$ is chosen.
289: 
290: Since all these results hold for every $\tau\in\R_{+}$, we can 
291: interpret them as follows:  {\em The total hyperreal energy 
292: dissipated in the infinitesimal resistance
293: $r=[r_{n}]$ during the real time
294: interval from 0 to $\tau$ is infinitesimally close to (but less than) 
295: the real value $q_{0}v_{0}/4$, and this is so no matter how small we
296: choose $\tau\in\R_{+}$.  On the other hand, that infinitesimal difference 
297: in the energy is dissipated in $r=[r_{n}]$ during all time 
298: larger than $\tau\in\R_{+}$, again no matter small $\tau$ is.}
299: Thus, we see that by using an infinitesimal resistor $r$, 
300: we have a way of accounting for all 
301: of the initial energy.
302: 
303: \section{Conclusions}
304: 
305: This solves---by means of nonstandard analysis---the ``mystery
306: of the vanishing energy'' in the following way:
307: From the perspective of standard mathematics, the infinitesimal 
308: resistance $r$ is equivalent to zero resistance because it is less than any 
309: real positive resistance but not negative.  
310: Nonetheless, all but an infinitesimal part of the missing 
311: appreciable energy $q_{0}v_{0}/4$ is found as dissipation in the 
312: infinitesimal $r$ due to the infinitely large, hyperreal, power dissipation 
313: $p_{r}(t)$ occurring during some initial positive part of the time
314: halo at $t=0$.  The remaining infinitesimal part of that missing energy 
315: occurs as infinitesimal power dissipation occurring during all 
316: subsequent hyperreal time.
317: 
318: {\em A Final Remark}:  The ``solution'' being proposed here may not
319: be the only nonstandard way of accounting for the anomaly.  
320: Effectively, our solution
321: it is saying that, since there is no anomaly when the resistor 
322: $r$ is positive, there should be no anomaly when $r$ is infinitesimally small
323: but positive
324: (a reflection of the transfer principle of nonstandard analysis
325: \cite[Appendix A.22]{ze}). 
326: But, one cannot jump to the limit as $r\rightarrow 0$ because 
327: the anomaly then jumps into view.  
328: Perhaps there are other nonstandard solutions available.
329: For instance, perhaps the newer theories concerning 
330: the multiplication of distributions coupled with nonstandard analysis 
331: and that justify the square $\delta^{2}$ of the delta function $\delta$
332: multiplied by an infinitesimal $r$
333: may be used to account for the vanishing energy 
334: at the instant $t=0$.  See, for example, \cite[Section 23]{ob}.
335: 
336: 
337: \begin{thebibliography}{99}
338: 
339: \bibitem{j-s}  S.M. Al-Jaber and S.K. Salih, Energy consideration in the 
340: two-capacitor problem, Eur. J. Phys., vol. 21, pp, 341-345, 2000.
341: \bibitem{a-w} M. Andelman and W.L. Warren, Energy recovery
342: schemes for the flow through capacitor technology 
343: (preprint---available via email: dragonfly@flowtc.com).
344: \bibitem{fo} J. Foutz, Energy loss in charging a capacitor, SMPS Technology,
345: Alta Loma, CA, March 2004
346: \bibitem{go} R. Goldblatt, {\em Lectures on the Hyperreals}, 
347: New York, Springer, 1998.
348: \bibitem{ob} M. Oberguggenberger, {\em Multiplication of Distributions
349: and Applications to Partial Differential Equations},  
350: Harlow, England, Longman Scientific 
351: \& Technical, 1992.
352: \bibitem{pa} C. Parton, Conservation and capacitance, 
353: {\em Phy. Educ.}, vol. 23, p. 138, 1990.
354: \bibitem{ro} A. Robinson, {\em Non-standard Analysis}, Amsterdam,
355: North-Holland Publishing Co., 1970.
356: \bibitem{ze} A.H. Zemanian, {\em Graphs and Networks: Transfinite and 
357: Nonstandard}, Boston, Birkhauser, 2004.
358: 
359: \end{thebibliography}
360: 
361: \end{document}
362: