1: %\documentclass[aps]{revtex4}
2: %\documentclass[article]
3: \documentclass[prd,twocolumn,floatfix,preprintnumbers,letterpaper]{revtex4}
4: %\documentclass[prd,nofootinbib,floatfix,twocolumn]{revtex4}
5:
6: \usepackage{graphicx}
7: \input{epsf}
8: \usepackage{epsf}
9: \usepackage{graphicx,epsfig}
10: \usepackage{bm}
11: \usepackage{latexsym,amssymb,amsmath,float}
12:
13: \newcommand {\ga} {\ {\raise-.5ex\hbox{$\buildrel>\over\sim$}}\ }
14: \newcommand {\la} {\ {\raise-.5ex\hbox{$\buildrel<\over\sim$}}\ }
15:
16: \begin{document}
17:
18: \title{Evolution of Oscillating Scalar Fields as Dark Energy}
19: \author{Sourish Dutta and Robert J. Scherrer}
20: \affiliation{Department of Physics and Astronomy, Vanderbilt University,
21: Nashville, TN ~~37235}
22:
23: \begin{abstract}
24: Oscillating scalar fields, with an oscillation frequency
25: much greater than the expansion rate, have been proposed as models for dark energy.
26: We examine these models, with particular emphasis on the evolution
27: of the ratio of the oscillation frequency to the expansion rate.
28: We show that this ratio always increases with time if the dark energy
29: density declines less rapidly than the background energy density. This
30: allows us to classify oscillating dark energy models in terms of
31: the epoch at which the oscillation frequency exceeds the expansion rate,
32: which is effectively the time at which rapid oscillations begin.
33: There are three basic types of behavior:
34: early oscillation models, in which oscillations begin during
35: the matter-dominated era, late oscillation models, in which oscillations
36: begin after scalar-field domination, and non-oscillating models.
37: We examine a representative set of models (those with power-law potentials)
38: and determine the parameter range giving acceptable agreement with
39: the supernova observations. We show that a subset of all three
40: classes of models can be consistent with the observational data.
41: \end{abstract}
42:
43: \maketitle
44:
45: \section{Introduction}
46:
47: It appears that
48: approximately 70\% of the energy density in the
49: universe is in the form of an exotic, negative-pressure component,
50: dubbed dark energy \cite{Knop,Riess1,Wood-Vasey,Davis}.
51: (See Ref. \cite{Copeland}
52: for a recent review).
53: The most likely possibility for this dark energy is a cosmological constant.
54: Another possibility,
55: dubbed quintessence, is a model in which the dark energy
56: arises from a scalar
57: field \cite{ratra,tw,caldwelletal,liddle,zlatev}.
58:
59: Although it is usually assumed that the evolution of the scalar
60: field is monotonic, a more exotic possibility is that the scalar
61: field oscillates. Such oscillating scalar fields were first
62: investigated by Turner \cite{turner}, and later reexamined in the context
63: of models for dark energy \cite{liddle,sahni,hsu,masso,gu}. (See
64: also Ref. \cite{KHS} for a phantom version of these models,
65: and Ref. \cite{DM} for applications to inflation).
66: Here we examine a key aspect of
67: such models: the interplay between the oscillation frequency
68: and the Hubble expansion rate. The discussions
69: in Refs. \cite{turner,sahni,hsu,masso,gu} assumed that the oscillation
70: frequency is large compared to the expansion rate.
71: However, if the ratio between these two
72: quantities crosses unity, it is possible to generate interesting
73: new behaviors for the dark energy. We explore this possibility here.
74: (Note that the models considered here, in which the scalar field oscillates,
75: are distinct from models in which the equation of state oscillates
76: \cite{dodosc,barosc1,barosc2,fengosc,linderosc}. In the
77: latter case, one is generally interested in oscillation frequencies
78: which are always
79: on the order of the Hubble parameter. In the models
80: considered here, oscillations with frequencies on the order of the Hubble
81: parameter are a transient, albeit important, phenomenon).
82:
83: In the next section, we review the evolution of oscillating
84: scalar fields in an expanding background, and we derive some
85: interesting general results on the evolution of the ratio
86: of scalar field oscillation frequency to the expansion rate.
87: In Sec. III, we compare numerical simulations of a representative
88: set of oscillating models (due to power-law potentials) with
89: the observations. Our conclusions are summarized in Sec. IV.
90: Our main new result is the existence of a class of oscillating
91: models in which the scalar field
92: is initially slowly rolling, but oscillates at late
93: times; we show that such models can be consistent with the
94: observations, but this result is very sensitive to the phase
95: of the scalar field oscillation at the present.
96:
97:
98: \section{Evolution of oscillating scalar fields}
99:
100: \subsection{Previous work}
101:
102: Consider a minimally-coupled
103: scalar field, $\phi$, in a potential $V(\phi)$.
104: The equation of motion for this field in an expanding background is
105: \begin{equation}
106: \label{motionq}
107: \ddot{\phi}+ 3H\dot{\phi} + \frac{dV}{d\phi} =0,
108: \end{equation}
109: where the Hubble parameter $H$ is given by
110: \begin{equation}
111: \label{H}
112: H = \left(\frac{\dot{a}}{a}\right) = \sqrt{\rho_T/3}.
113: \end{equation}
114: Here $a$ is the scale factor, $\rho_T$ is the total density, and we work in units
115: for which $8 \pi G = 1$.
116: The pressure and density of $\phi$
117: are given by
118: \begin{equation}
119: p_\phi = \frac{\dot \phi^2}{2} - V(\phi),
120: \end{equation}
121: and
122: \begin{equation}
123: \label{rhodense}
124: \rho_\phi = \frac{\dot \phi^2}{2} + V(\phi),
125: \end{equation}
126: respectively, and the equation of state parameter, $w_\phi$,
127: is defined to be
128: \begin{equation}
129: w_\phi = p_\phi/\rho_\phi.
130: \end{equation}
131: The density of the scalar field then evolves with the scale factor
132: as
133: \begin{equation}
134: \label{rhoevolve}
135: \rho_\phi \propto a^{-3(1+w_\phi)}.
136: \end{equation}
137:
138: Turner \cite{turner} was the first to consider the behavior of such
139: a field in the limit where it oscillates on a timescale much
140: shorter than the Hubble time. He examined power-law potentials
141: of the form
142: \begin{equation}
143: \label{power}
144: V(\phi) = k \phi^n,
145: \end{equation}
146: and he showed that the value
147: of $w_\phi$, averaged over the oscillation period, is
148: \begin{equation}
149: \label{wturner}
150: w_\phi = (n-2)/(n+2).
151: \end{equation}
152: Thus, the cases $n=2$ and $n=4$ behave like nonrelativistic matter
153: and radiation, respectively.
154:
155: Liddle and Scherrer \cite{liddle}
156: investigated oscillating scalar field models with power-law potentials in the context
157: of dark energy, noting an interesting set of behaviors:
158: for a universe dominated by a background fluid with equation of
159: state parameter $w_B$,
160: the oscillating solution is an attractor of the equations of motion
161: only for the case
162: \begin{equation}
163: \label{liddlelimit}
164: n < \frac{2(3+w_B)}{1-w_B},
165: \end{equation}
166: while potentials with $n > 2{(3+w_B)}/{(1-w_B)}$ produce a scalar
167: field trajectory that decreases monotonically to zero without
168: oscillating. Liddle and Scherrer provided no
169: physical explanation for this curious behavior; we give such an
170: explanation below.
171:
172: Sahni and Wang \cite{sahni} and Hsu \cite{hsu} examined the possibility that
173: an oscillating field could give rise to dark energy. From equation
174: (\ref{wturner}), it is clear that $w < 0$ requires $n< 2$, and
175: the production of a dark energy equation of state can only be
176: achieved if we take
177: $V = k |\phi|^n$, with $n \ll 1$. Although potentials
178: with very small $n$
179: might appear rather unnatural,
180: such a potential can arise from a Lagrangian with
181: a $\phi^2$ potential and
182: a non-standard kinetic term \cite{hsu}.
183:
184: Masso et al. \cite{masso}
185: performed a general investigation of oscillating scalar fields
186: in arbitrary potentials (as opposed to power laws). Using
187: the action-angle formalism, they derived
188: several very important results, which we will use here.
189: Consider a scalar field $\phi$ oscillating in an arbitrary
190: potential $V(\phi)$. The minimum and maximum values of $\phi$ are
191: $\phi_1$ and $\phi_2$ respectively. Then the energy density for this
192: field must be equal to $\rho_\phi = V(\phi_1) = V(\phi_2)$.
193: (Note that $\rho_\phi$ changes with time, but it
194: is assumed to evolve on a timescale longer than the inverse
195: oscillation frequency).
196: Masso et al. showed that all of the quantities of interest can
197: be expressed in terms of the action variable $J$, defined
198: by
199: \begin{equation}
200: \label{J}
201: J = 2 \int_{\phi_1}^{\phi_2}\sqrt{2(\rho_\phi-V(\phi))}.
202: \end{equation}
203: Then the oscillation frequency $\nu$ is
204: \begin{equation}
205: \label{nu1}
206: \nu = \frac{d\rho_\phi}{dJ},
207: \end{equation}
208: while the equation of state parameter is
209: \begin{eqnarray}
210: 1+w_\phi &=& \frac{J}{\rho_\phi}\frac{1}{dJ/d\rho_\phi} ,\\
211: &=& \frac{2}{\rho_\phi} \frac{\int_{\phi_1}^{\phi_2}[\rho_\phi-V(\phi)]^{1/2} d\phi}
212: {\int_{\phi_1}^{\phi_2}[\rho_\phi-V(\phi)]^{-1/2} d\phi}.
213: \end{eqnarray}
214: (Note that this last result was also derived by Turner \cite{turner}).
215: These results allow for the investigation of somewhat
216: more natural (non-power-law) potentials that
217: might serve as dark energy, and several of these are discussed
218: in Ref. \cite{masso}.
219:
220: Finally, Gu \cite{gu} revisited the power-law scenario, and calculated
221: the conditions necessary for the oscillation frequency to exceed
222: the expansion rate (our equation \ref{powernu} below). He used this
223: condition to place constraints on such power-law models.
224:
225: \subsection{Evolution of oscillation frequency}
226:
227: Implicit in all of these discussions is the assumption that
228: the oscillation frequency is much greater than the Hubble expansion
229: rate. We now consider this condition in more detail. The relevant
230: quantity is the ratio of oscillation frequency to Hubble parameter,
231: $\nu/H$. If $\nu/H \gg 1$, then the parameters of interest
232: can be averaged over the oscillation period, as in Refs.
233: \cite{liddle,turner,sahni,
234: hsu,masso,gu}. On the other hand, when $\nu/H \ll 1$, the field effectively
235: ceases to oscillate, and an alternative treatment is required.
236:
237: Consider the evolution of $\nu/H$. From equations (\ref{J})
238: and (\ref{nu1}) we can write
239: \begin{equation}
240: \nu = \left[\int_{\phi_1}^{\phi_2}\sqrt{\frac{2}{\rho_\phi - V(\phi)}}d\phi\right]^{-1}.
241: \end{equation}
242: This equation, along with the expression for $H$ given by
243: equation (\ref{H}), gives us
244: \begin{equation}
245: \label{nuH}
246: \nu/H = \left[\int_{\phi_1}^{\phi_2}\sqrt{\frac{2\rho_T}{3(\rho_\phi - V(\phi))}}d\phi\right]^{-1}.
247: \end{equation}
248: For the case of the power-law potential, $V(\phi) = k \phi^n$, the expression for
249: $\nu$ can be integrated exactly \cite{masso}, giving (see also \cite{gu})
250: \begin{equation}
251: \label{powernu}
252: \nu/H = \frac{\sqrt{3}n\Gamma(1/2 + 1/n)}{2 \sqrt{2\pi} \Gamma(1/n)}\sqrt{\frac{\rho_\phi}{\rho_T}}
253: \frac{1}{\phi_{max}},
254: \end{equation}
255: where $\phi_{max} = \phi_1 = \phi_2$, since the potential is symmetric.
256: For other symmetric potentials, we can use equation (\ref{nuH}) to derive an order
257: of magnitude estimate:
258: \begin{equation}
259: \label{nuapprox}
260: \nu/H \sim \sqrt{\frac{\rho_\phi}{\rho_T}} \frac{1}{\phi_{max}}.
261: \end{equation}
262: This result is clearly consistent with the exact result for power-law potentials
263: given by equation (\ref{powernu}).
264:
265: Consider first the case where the scalar field itself dominates
266: the expansion, so that $\rho_T \approx \rho_\phi$. In this
267: case, we have $\nu/H \sim 1/\phi_{max}$. Since $\rho_\phi$ always
268: decreases with time, $\phi_{max}$ is also a decreasing function
269: of time, so that $\nu/H$ always increases with time.
270:
271: The other possibility is that the universe is dominated by
272: a background fluid (e.g., matter or radiation) with density
273: $\rho_B$. In this case, we have $\nu/H \sim \sqrt{\rho_\phi/\rho_B}
274: (1/\phi_{max})$. We can again assume that $\phi_{max}$ decreases
275: with time. In order for $\phi$ to serve as a plausible dark
276: energy candidate, its density must decrease more slowly with
277: the expansion than the density of matter or radiation. Hence,
278: we can conclude that $\rho_\phi/\rho_B$ increases with
279: time, and once again we have that $\nu/H$ increases with time.
280:
281: Thus, for any model in which
282: $\phi$ serves as the dark energy today, $\nu/H$ is an increasing function of time,
283: regardless of whether the universe is background-dominated or
284: $\phi$-dominated.
285:
286: Before examining the implications of this result,
287: we reconsider the behavior discussed in Ref. \cite{liddle} and described
288: by equation (\ref{liddlelimit}). Equations (\ref{wturner}) and
289: (\ref{liddlelimit}), taken together, imply that scalar field oscillations
290: are an attractor in a universe dominated by a background fluid only for the case where
291: \begin{equation}
292: \label{wphiliddle}
293: w_\phi < 1/2 + w_B/2.
294: \end{equation}
295: Further, for a power-law potential given by equation (\ref{power}),
296: in the background-dominated case,
297: equation (\ref{nuapprox}) can be expressed as
298: \begin{equation}
299: \nu/H \sim \rho_B^{-1/2}\rho_\phi^{1/2 - 1/n}.
300: \end{equation}
301: Applying equation (\ref{wturner}), this becomes
302: \begin{equation}
303: \nu/H \sim \rho_B^{-1/2}\rho_\phi^{w_\phi/(1+w_\phi)},
304: \end{equation}
305: which can be rewritten in terms of the scale factor (using equation
306: (\ref{rhoevolve}) for $\rho_\phi$ and the corresponding equation
307: for $\rho_B$) as
308: \begin{equation}
309: \label{nua}
310: \nu/H \sim a^{3(1/2 + w_B/2 - w_\phi)}.
311: \end{equation}
312: A comparison of equations (\ref{nua}) and (\ref{wphiliddle})
313: provides a physical explanation for the bound noted by Liddle
314: and Scherrer \cite{liddle}. When this bound (equation \ref{liddlelimit})
315: is not satisfied, equation (\ref{nua}) indicates that
316: $\nu/H$ is a {\it decreasing} function of time. Thus,
317: at late times, the oscillating solution cannot be an attractor.
318: This is consistent with the behavior noted in Ref. \cite{liddle};
319: when equation (\ref{liddlelimit}) is not satisfied, the scalar
320: field decays asymptotically to the minimum of the potential without
321: oscillating. These models do not contradict our claim
322: that in any oscillating scalar field model for dark energy,
323: $\nu/H$ always increases with time. The reason is that models
324: that violate the bound in equation (\ref{liddlelimit}) have
325: an energy density that decays more rapidly than $\rho_B$; they
326: are not plausible models for dark energy.
327:
328: Now consider the possible oscillating scalar field models for dark energy.
329: The fact that $\nu/H$ increases with time simplifies the classification
330: of such models. If we assume that the universe
331: evolves from a matter-dominated state
332: to a scalar-field dominated state, there are basically three possibilities,
333: which depend on the redshift at which $\nu$ becomes greater than $H$:
334:
335: I. The oscillation frequency becomes greater than the expansion
336: rate during the matter-dominated era. Since $\nu/H$ increases
337: with time, we always have $\nu/H > 1$ during the quintessence-dominated era,
338: and we can
339: use time-averaged quantities to describe the equation of
340: state of the quintessence. We will call this the early-oscillation
341: case. This is basically the model described
342: in Refs. \cite{turner,sahni,hsu,masso,gu}, so these models have been exhaustively
343: explored.
344:
345: II. The oscillation frequency is less than $H$
346: at the present day. Again, since $\nu/H$ always increases with
347: time, this implies that $\nu$ is never larger than $H$, so these
348: models never reach the oscillatory stage. In the language
349: of Ref. \cite{CL}, these are thawing models in which
350: the field has not yet reached the minimum of the potential.
351: Such models are characterized by a value of $w_\phi$ very close
352: to $-1$ initially, which increases at late times. We will
353: call this the ``no oscillation" case. Again,
354: such models have been exhaustively explored (see, e.g., Ref.
355: \cite{SS} and references therein).
356:
357: III. The case intermediate between I and II occurs
358: when $\nu$ becomes greater than $H$ at late times, after
359: the quintessence density exceeds the matter density.
360: In this case, the scalar
361: field displays two distinct
362: behaviors during the quintessence-dominated era. At first, when
363: the oscillation frequency
364: is less than $H$, the scalar field rolls down the potential and $w_\phi$ increases.
365: Then,
366: when the oscillation frequency becomes greater than $H$, the
367: field oscillates rapidly, and $w_\phi$ takes its time-averaged value
368: as in Refs. \cite{turner,sahni,hsu,masso,gu}. This third possibility has
369: not been previously studied; we will refer to it as the ``late oscillation"
370: case.
371:
372:
373:
374:
375:
376:
377:
378: \section{Comparison with Observations}
379:
380: Although we believe that our classification scheme in the previous section
381: should describe most possible models for quintessence from oscillating
382: scalar fields, it is obviously impossible to compare all such models
383: with the observational data. We will therefore limit ourselves
384: to a representative set of models: the power-law oscillating
385: scalar field models discussed in Refs. \cite{sahni,hsu,masso,gu}. As we note
386: below, we believe that the general features of
387: the results we derive for these models
388: will be more generally applicable to other oscillating scalar field models.
389:
390: Following Ref. \cite{hsu}, we choose the potential to have
391: the form:
392: \begin{equation}
393: V\left( \phi \right)=k\vert\phi\vert^n,
394: \end{equation}
395: where $k$ and $n$ are constants. Then the evolution of the field
396: depends on three parameters: $k$, $n$, and the initial value
397: of $\phi$, which we denote $\phi_i$.
398: However, the parameter $k$ can be eliminated from
399: the equation of motion (equation \ref{motionq}) by a suitable rescaling
400: of $\phi$ and the time $t$. This rescaling means that we are no longer
401: using Planck units, but our units of time are irrelevant, since
402: we always define the ``present day" in terms of the value of $\Omega_\phi$,
403: rather then $t$.
404: (Another way to see this is
405: to note that all of our results depend only on the evolution of $w_\phi$
406: as a function of $\Omega_\phi$, and $w_\phi(\Omega_\phi)$
407: depends only on $V^\prime/V$, which
408: is independent of $k$ \cite{zlatev}). Thus, the evolution can be
409: described entirely in terms of $n$ and $\phi_i$.
410:
411:
412: For a given choice of values of the
413: two free parameters, $\phi_i$, and $n$,
414: we numerically evolve the system from a
415: starting point in the matter dominated
416: regime to the point when $\Omega_{m}=0.3$,
417: which we label as the present time $z=0$.
418:
419: The relation between the evolution of $\phi$ and the two input parameters
420: is illustrated in Fig. \ref{oscbehavior_plot}.
421: In our numerical simulations, we define oscillations to commence
422: when the scalar field crosses $\phi=0$. This corresponds to $w_\phi$
423: increasing from $-1$ to $+1$ and marks the point at which $dw_\phi/dz$
424: first changes sign. (Note that the oscillation frequency for $w_\phi$
425: is twice as large as the oscillation frequency for $\phi$).
426: For early oscillations, this
427: zero-crossing takes place during the matter-dominated era, while
428: for late oscillations it takes place during the quintessence-dominated era.
429: The no oscillation case corresponds to the parameter region in which
430: zero crossing never occurs up to the present.
431:
432: For a given $n$, the parameter $\phi_i$ determines when
433: the oscillations will begin. From equation (\ref{powernu}) (or
434: \ref{nuapprox}), it is clear that small $\phi_i$ corresponds
435: to a larger value of $\nu/H$ for a given ratio of $\rho_\phi$
436: to $\rho_T$. Thus, oscillations commence earlier if $\phi_i$
437: is small and later if $\phi_i$ is large. For fixed
438: $n$, sufficiently small
439: $\phi_i$ will produce the early oscillation case, while sufficiently
440: large $\phi_i$ gives no oscillations, with late oscillations arising
441: from intermediate values of $\phi_i$. This is apparent in Fig.
442: \ref{oscbehavior_plot}.
443: Fixing $n$, we see that increasing $\phi_i$ moves the evolution
444: from early oscillation, through late oscillation, and into no oscillation.
445: Since the late oscillation case has not been previously
446: explored, we present a graph, in Fig. \ref{hi_phi}, of
447: the evolution of $w(z)$ for some of these models. As claimed,
448: $w_\phi$ initially evolves slowly upward from $w_\phi=-1$, but then oscillates
449: rapidly at late times; this transition takes place after the quintessence
450: field becomes dominant.
451:
452:
453: Next, we compare the theoretical luminosity-vs-distance
454: results to the recent Type Ia Supernovae standard candle
455: data (ESSENCE+SNLS+HST from \cite{Davis}).
456: The theoretical distance modulus in our model is calculated in the standard
457: way:
458: \begin{equation}
459: \mu(z) = 5\log_{10}D_L(z) + 42.38 - 5 \log_{10} h,
460: \end{equation}
461: where $h$ is the Hubble parameter in units of 100 km sec$^{-1}$ Mpc$^{-1}$,
462: and $D_L(z)$ is the dimensionless luminosity distance, given by
463: \begin{equation}
464: D_L(z) = (1+z) \int_0^z \frac{H_0}{H(z^\prime)} dz^\prime,
465: \end{equation}
466: where $H(z^\prime)$, the Hubble parameter at redshift $z^\prime$,
467: also depends on $\Omega_{m0}$ and the parameters of our quintessence model.
468: We construct a $\chi^2$ likelihood plot
469: for $\phi_i$ and $n$ ranging from 0.01 to 0.40 (no further structure emerges from extending the parameter
470: space). We take $\Omega_{m0} = 0.3$ and marginalize over the nuisance
471: parameter $h$.
472: Fig. \ref{likelihood_plot} shows this
473: plot.
474: \begin{figure}
475: \epsfig{file=fig04.eps,height=55mm}
476: \caption
477: {\label{oscbehavior_plot}
478: For the potential $V(\phi) = k|\phi|^n$,
479: in the parameter space determined by $n$ and the initial magnitude of the scalar field,
480: $\phi_i$, this plot gives the
481: regions where the field $\phi$ oscillates prior to dominating the expansion (early
482: oscillations), after dominating the expansion (late oscillations), or does
483: not oscillate at all prior to the present (no oscillations).
484: }
485: \end{figure}
486:
487: \begin{figure}
488: \epsfig{file=fig06.eps,height=55mm}
489: \caption{\label{hi_phi}
490: Evolution of the quintessence equation of state parameter, $w_\phi$, as a
491: function of redshift $z$, in the late oscillation regime. Curves correspond
492: to the indicated values of the initial value of the field, $\phi_i$,
493: in the power-law potential $V = k|\phi|^n$ with $n=0.16$.
494: Diamonds mark the points at which the density
495: of dark energy equals the density of matter. The curve for $\phi_i = 0.15$
496: corresponds to the no oscillation regime and has been added for comparison.}
497: \end{figure}
498:
499:
500: \begin{figure}
501: \epsfig{file=fig02.eps,height=55mm}
502: \caption
503: {\label{likelihood_plot}
504: Likelihood plot for the parameters $\phi_i$ and $n$. The $95\%$
505: confidence region is green (dark shaded) and the $68\%$ confidence
506: region is green+yellow (dark shaded + light shaded). The diamonds
507: indicate
508: the models displayed in
509: Figs \ref{hi_phi}, \ref{mu1_plot}, \ref{mu2_plot}, and \ref{rhophi_plot}.}
510: \end{figure}
511:
512:
513:
514: \begin{figure}
515: \epsfig{file=fig04a.eps,height=55mm}
516: \caption
517: {\label{likelihood_phi}
518: The likelihood contours of Fig. \ref{likelihood_plot} superimposed
519: over the demarcation of the type of oscillatory behavior
520: from Fig. \ref{oscbehavior_plot}.
521: }
522: \end{figure}
523:
524: Some of the allowed regions can be understood in terms of our previous
525: discussion of the behavior of the field. In the limit of large $\phi_i$,
526: the field never oscillates, and it can be effectively treated as
527: a cosmological constant, which is known to be a good fit to the observational
528: data. For sufficiently small $\phi_i$, we are in the ``early oscillation" regime.
529: In this case, $w_\phi$ takes on a constant value determined by equation
530: (\ref{wturner}) and independent of $\phi_i$. It is clear from
531: equation (\ref{wturner}) that a value
532: of $w_\phi$ near $-1$ requires a value of $n$ close to zero.
533: Thus,
534: we see in Fig. \ref{likelihood_plot} that
535: for small $\phi_i$, the allowed region is confined to small values
536: of $n$ (which give $w_\phi$ near $-1$), and the likelihood contours
537: become vertical lines (indicating that the late-time value of $w_\phi$ is
538: independent of $\phi_i$ in this limit).
539:
540: To explore the allowed parameter space further, we
541: have superimposed the field behaviors outlined in Fig. \ref{oscbehavior_plot}
542: onto
543: the likelihood contours displayed in Fig. \ref{likelihood_plot}.
544: The results are displayed in Fig. \ref{likelihood_phi}. As noted earlier,
545: almost the entire ``no oscillation" parameter space is allowed, while
546: the ``early oscillation" parameter space is confined to the region
547: for which $w_\phi$ is close to $-1$. However, we also see that
548: there is an allowed region in the ``late oscillation" regime; this
549: is the subset of oscillating models that has not been previously explored.
550: In these models, the field is initially slowly rolling down the potential,
551: with increasing $w_\phi$; it then transitions into rapid oscillations
552: at late times. An interesting feature in this regime
553: is the non-smooth structure of the confidence contours. This is a real
554: effect.
555:
556: To illustrate the reason for the non-smooth
557: confidence contours, we have sampled
558: four representative models in this regime, denoted by the four diamonds
559: in Fig. \ref{likelihood_plot}, which correspond to $n=0.16$
560: and $\phi_i = $0.10, 0.12, 0.13, and 0.15. (These
561: are the four models previously displayed in Fig. \ref{hi_phi}).
562: This illustrates
563: the complex structure of the allowed region as $\phi_i = 0.10$ and 0.13
564: are ruled out at 95\% confidence, while $\phi_i = 0.12$ and 0.15 are not.
565: Next, we display the distance modulus for these four models, along
566: with the supernova data, in Fig. \ref{mu1_plot}. Although the individual
567: models cannot be distinguished here, it is clear that all of
568: these models lie near the edge of the allowed region because
569: they tend to produce a distance modulus that is lower than
570: the observations. In Fig. \ref{mu2_plot}, we show a blow-up
571: of Fig. \ref{mu1_plot}. The smallest values of the distance
572: modulus are produced for $\phi_i = 0.10$ and 0.13; these are the two
573: models that are ruled out at 95\% confidence. The important point
574: here is that the distance modulus is not a monotonic function
575: of $\phi_i$. The reason for this is illustrated in Fig. \ref{rhophi_plot},
576: where we give the redshift dependence of the
577: density of the scalar field (which determines
578: $H(z)$, and therefore, $\mu(z)$). Comparing
579: this figure with Fig. \ref{hi_phi}, we see that $H(z)$ depends
580: crucially on the oscillation phase at the present day. The two
581: models for which the field is at a maximum of the potential at present
582: (thus giving $w_\phi=-1$ at present)
583: are $\phi_i = 0.10$ and $\phi_i = 0.13$. These produce a larger
584: density at moderate redshift and therefore a
585: smaller integrated
586: value of $1/H(z)$. Thus, the complex behavior of our contours
587: in Figs. \ref{likelihood_plot} and \ref{likelihood_phi} is a
588: phase effect: the phase of $\phi$ at present is an oscillating
589: function of $\phi_i$, and this phase determines whether or not
590: a given model is excluded. For early oscillation models the phase
591: becomes irrelevant, since these models oscillate many times before
592: the present, and the equation of state parameter takes
593: on its period-averaged value.
594:
595: \begin{figure}
596: \epsfig{file=all_data.eps,height=55mm}
597: \caption
598: {\label{mu1_plot}
599: Distance modulus for the indicated values of $\phi_i$ along with the
600: supernova values used in our analysis, for $V(\phi) = k|\phi|^n$, with
601: $n = 0.16$.}
602: \end{figure}
603: \begin{figure}
604: \epsfig{file=fig07a.eps,height=55mm}
605: \caption
606: {\label{mu2_plot}
607: A blow-up of Fig. \ref{mu1_plot}, illustrating the difference
608: in behavior of
609: the four indicated models.}
610: \end{figure}
611: \begin{figure}
612: \epsfig{file=fig08.eps,height=55mm}
613: \caption
614: {\label{rhophi_plot}
615: Evolution of the scalar field density, normalized to unity
616: at the present, as a function
617: of redshift for the four models displayed
618: in Figs. \ref{hi_phi}, \ref{mu1_plot}, and \ref{mu2_plot}.}
619: \end{figure}
620:
621: \section{Conclusions}
622:
623: Our results indicate that oscillating quintessence models, even
624: for very simple potentials, can display more complex behavior than
625: has previously been considered. Previous discussions have centered
626: on the early oscillation case, in which the oscillation frequency
627: is always much larger than the expansion rate when the scalar
628: field dominates the expansion. Our results show
629: that the late oscillation case, in which the scalar field transitions
630: from a slowly rolling regime at early times into oscillations at late
631: times, can also be consistent with current observations, although
632: the behavior of these latter models depends crucially on the oscillation
633: phase of the scalar field at present. Johnson and Kamionkowksi
634: \cite{JK} have argued recently that rapidly-oscillating scalar fields
635: with a time-averaged value of $w_\phi$ less than $0$ are unstable
636: to the growth of inhomogeneities (see also the argument in
637: Ref. \cite{Kasuya}). In this case, only the
638: late oscillation models would remain as acceptable
639: oscillating scalar field models for dark energy.
640:
641:
642: How general are our results? The conclusion that $\nu/H$
643: increases with time depends only on the assumption that the scalar
644: field energy density decays more slowly than the background energy
645: density, so it should be
646: generally applicable to any plausible model for dark energy.
647: On the other hand,
648: our comparison of models with observations
649: was restricted to the case of power-law potentials. However, we would
650: expect qualitatively similar results for other bound potentials with
651: a single minimum at $V=0$. If the minimum of the potential were
652: at $V_0 > 0$, the analysis is almost identical, except that the
653: dark energy now has an additional component with constant density
654: $V_0$. Such a model would have a values of $w(z)$ closer to $-1$ than
655: the corresponding model with a minimum at $V=0$, and thus, would
656: tend to agree more closely with current observations. One type
657: of potential that could display completely different
658: behavior from those studied here is a $V(\phi)$ with multiple minima,
659: such as the Mexican hat potential discussed in Ref. \cite{masso}.
660: In this case, the field evolution can be considerably more complex,
661: as the field can initially oscillate over the entire range of
662: the potential, but will eventually become trapped in a local
663: minimum at late times. In this case the simple analysis presented
664: here will not apply.
665:
666:
667: \acknowledgments
668:
669:
670: R.J.S. was supported in part by the Department of Energy (DE-FG05-85ER40226).
671:
672: \begin{thebibliography}{99}
673:
674: \bibitem{Knop}
675: R.A. Knop, et al., Ap.J. {\bf 598}, 102 (2003).
676:
677: \bibitem{Riess1}
678: A.G. Riess, et al., Ap.J. {\bf 607}, 665 (2004).
679:
680: \bibitem{Wood-Vasey}
681: W.M. Wood-Vasey, et al., \apj {\bf 666}, 694 (2007).
682:
683: \bibitem{Davis}
684: T.M. Davis, et al., \apj {\bf 666}, 716 (2007).
685:
686: \bibitem{Copeland}
687: E.J. Copeland, M. Sami, and S. Tsujikawa, Int. J. Mod. Phys. D
688: {\bf 15}, 1753 (2006).
689:
690: \bibitem{ratra}
691: B. Ratra and P.J.E. Peebles,
692: \prd {\bf 37}, 3406 (1988).
693:
694: \bibitem{tw}
695: M.S. Turner and M. White,
696: \prd {\bf 56}, 4439 (1997).
697:
698: \bibitem{caldwelletal}
699: R.R. Caldwell, R. Dave, and P.J. Steinhardt,
700: \prl {\bf 80}, 1582 (1998).
701:
702: \bibitem{liddle}
703: A.R. Liddle and R.J. Scherrer,
704: \prd {\bf 59}, 023509 (1998).
705:
706: \bibitem{zlatev}
707: P.J. Steinhardt, L. Wang, and I. Zlatev,
708: \prd {\bf 59}, 123504 (1999).
709:
710: \bibitem{turner}
711: M.S. Turner, \prd {\bf 28}, 1243 (1983).
712:
713: \bibitem{sahni}
714: V. Sahni and L. Wang, \prd {\bf 62}, 103517 (2000).
715:
716: \bibitem{hsu}
717: S.D.H. Hsu, Phys. Lett. B {\bf 567}, 9 (2003).
718:
719: \bibitem{masso}
720: E. Masso, F. Rota, and G. Zsembinszki,
721: \prd {\bf 72}, 084007 (2005).
722:
723: \bibitem{gu}
724: J.-A. Gu, arXiv:0711.3606.
725:
726: \bibitem{KHS}
727: A. Kurek, O. Hrycyna, and M. Szydlowski, Phys. Lett. B {\bf 659},
728: 14 (2008).
729:
730: \bibitem{DM}
731: T. Damour and V.F. Mukhanov, \prl {\bf 80}, 3440 (1998).
732:
733: \bibitem{dodosc}
734: S. Dodelson, M. Kaplinghat, E. Stewart, \prl {\bf 85},
735: 5276 (2000).
736:
737: \bibitem{barosc1}
738: G. Barenboim, O. Mena, and C. Quigg, \prd
739: {\bf 71}, 063533 (2005).
740:
741: \bibitem{barosc2}
742: G. Barenboim and J. Lykken, Phys. Lett. B {\bf 633},
743: 453 (2006).
744:
745: \bibitem{fengosc}
746: B. Feng, M. Li, Y.-S. Piao, and X. Zhang,
747: Phys. Lett. B {\bf 634}, 101 (2006).
748:
749: \bibitem{linderosc}
750: E.V. Linder, Astropart. Phys. {\bf 25}, 167 (2006).
751:
752: \bibitem{CL}
753: R.R. Caldwell and E.V. Linder, \prl {\bf 95}, 141301 (2005).
754:
755: \bibitem{SS}
756: R.J. Scherrer and A.A. Sen, \prd {\bf 77}, 083515 (2008).
757:
758: \bibitem{JK}
759: M.C. Johnson and M. Kamionkowski, arXiv:0805.1748.
760:
761: \bibitem{Kasuya}
762: S. Kasuya, Phys. Lett. B {\bf 515}, 121 (2001).
763:
764: \end{thebibliography}{}
765:
766:
767:
768: \end{document}
769: