1: \documentclass[12pt,preprint]{aastex}
2:
3: \shorttitle{Are High-Redshift Quasars Blurry?}
4: \shortauthors{Steinbring}
5:
6: \input epsf
7: \def\plotone#1{\centering \leavevmode
8: \epsfxsize=1.0\columnwidth \epsfbox{#1}}
9: \def\plottwo#1#2{\centering \leavevmode
10: \epsfxsize=0.5\columnwidth \epsfbox{#1}
11: \epsfxsize=0.5\columnwidth \epsfbox{#2}}
12: \def\plotonehalf#1{\centering \leavevmode
13: \epsfxsize=0.5\columnwidth \epsfbox{#1}}
14: \def\plotonetwothirds#1{\centering \leavevmode
15: \epsfxsize=0.666667\columnwidth \epsfbox{#1}}
16:
17:
18: %\received{2006 February 1}
19: \begin{document}
20:
21: \title{Are High-Redshift Quasars Blurry?}
22:
23: \author{Eric Steinbring\altaffilmark{1}}
24:
25: \altaffiltext{1}{Herzberg Institute of Astrophysics, National Research Council Canada, Victoria, BC V9E 2E7, Canada}
26:
27: \begin{abstract}
28: It has been suggested that the fuzzy nature of spacetime at the Planck scale may
29: cause lightwaves to lose phase coherence, and if severe enough
30: this could blur images of distant point-like sources sufficiently that they do
31: not form an Airy pattern at the focal plane of a telescope.
32: Blurring this dramatic has already been observationally ruled out by images
33: from {\it Hubble Space Telescope} ({\it HST}), but I show that the underlying phenomenon
34: could still be stronger than previously considered. It is harder to detect, which may
35: explain why it has gone unseen. A systematic search is made in archival {\it HST} images
36: of among the highest known redshift quasars. Planck-scale induced blurring may be
37: evident, but this could be confused with partially resolved sources.
38: \end{abstract}
39:
40: \keywords{time --- gravitation --- quasars:general}
41:
42: \section{Introduction}\label{introduction}
43:
44: Characterizing the microscopic properties of spacetime using images of distant sources was
45: first proposed in a Letter by \cite{Lieu2003}.
46: This compelling possibility is a consequence of a phenomenon
47: likely to be found at the Planck scale, where
48: lengths shrink to $l_{\rm P} \sim 10^{-35}$ m and time intervals diminish to $t_{\rm P} \sim 10^{-44}$ s.
49: Following \cite{Ng2003}, here distance measurements $l$
50: are uncertain by $\delta l \gtrsim l_{\rm P}(l/l_{\rm P})^{1 - \alpha}$ (with similar uncertainties for time), where
51: the parameter $\alpha$ specifies different quantum gravity models.
52: A natural choice for $\alpha$ is 1, but ${1\over2}$ is also possible \citep{Amelino-Camelina2000},
53: and ${2\over3}$ is preferred \citep{Ng2002}, because it is consistent with the holographic
54: principle of \citet{tHooft1993} and established black-hole theory \citep{Beckenstein1973, Hawking1975}.
55: An electromagnetic wave travelling a distance $L$ from source to observer would
56: be continually subjected to these random spacetime fluctuations, which following \cite{Ng2003}
57: leads to a cumulative statistical
58: phase dispersion of
59: $$\Delta \phi_0 = 2 \pi a_0 {l_{\rm P}^{\alpha}\over{\lambda}} L^{1-\alpha}, \eqno(1)$$
60: where $\lambda$ is the observed wavelength, and $a_0$ is close to unity.
61: This presents a means of
62: detection in an image of a point source.
63: \cite{Ragazzoni2003} point out that $\Delta \phi_0$ can
64: be interpreted as an apparent angular broadening of the source, and over a sufficient distance might
65: grow enough to obliterate the diffraction pattern at the focal plane
66: of a telescope.
67:
68: Because this has not been seen in {\it Hubble Space Telescope} ({\it HST}) data, Lieu et al. and other
69: authors \citep{Ragazzoni2003, Ng2003} have
70: concluded that the effect, if present, is too weak to be observed.
71: In Section~\ref{limits} I expand on previous work to account for the shorter
72: wavelength of emitted light, which predicts stronger blurring. But observing this is still
73: problematic, because it must be disentangled from a partially resolved
74: source. A test is outlined in Section~\ref{selection}
75: which makes use of the best available archival data, a {\it HST} snapshot survey of $z>4$ quasars.
76: These data are discussed in Section~\ref{reductions}, after which a
77: new analysis in the context of blurring by Planck-scale effects is presented in Section~\ref{results}.
78:
79: \section{Limits on Blurring Induced by the Planck Scale}\label{limits}
80:
81: Equation 1 assumes that the emission and detection wavelengths are the same, and so
82: represents a minimal estimate for blurring.
83: For cosmological distances, photons left the source at the shorter wavelength $\lambda\over{1 + z}$, for which the
84: blurring action of the Planck scale should be stronger.
85: Consider the following:
86: $$\Delta \phi_{\rm max} = 2 \pi a_0 (1 + z) {l_{\rm P}^{\alpha}\over{\lambda}} L^{1 - \alpha} = (1 + z) \Delta \phi_0. \eqno(2)$$
87: One might call this a maximal estimate, because it assumes the bluest
88: possible photons propagate a distance $L$ and are then detected at wavelength $\lambda$.
89: It is instructive to write equation 2 in a different form. By taking its differential with
90: respect to $z$, it is possible to rewrite it as
91: $$\Delta \phi_{\rm max} = 2 \pi a_0 {l_{\rm P}^{\alpha}\over{\lambda}} \Big{\{}\int_0^z L^{1 - \alpha} dz + {{(1-\alpha)c}\over{H_0 q_0}} \int_0^z (1+z) L^{-\alpha} \Big{[}1 - {{1 - q_0}\over{\sqrt{1 + 2 q_0 z}}}\Big{]} dz \Big{\}}$$
92: $$ = \Delta \phi_{\rm l.o.s.} + \Delta \phi_z. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\eqno(3)$$
93: where the luminosity distance $L$ is given by
94: $$L = {c\over{H_0 q_0^2}}\Big{[}q_0 z - (1 - q_0)(\sqrt{1 + 2 q_0 z} - 1)\Big{]} \eqno(4)$$
95: and $q_0={{\Omega_0}\over{2}} - {{\Lambda c^2}\over{3 H_0^2}}$ is the deceleration parameter.
96: By comparison with equation 1, the first integral can be recognized as the observed phase dispersion of
97: photons of wavelength $\lambda$ arriving from all redshifts up to the source.
98: In practice these could conceivably come from intervening sources along the line of sight, hence the label $\Delta \phi_{\rm l.o.s.}$.
99: The second integral must then be the remaining phase dispersion $\Delta \phi_z$ associated exclusively with
100: photons redshifted to the observer's wavelength. The latter scenario seems the more plausible.
101: This is plotted in Figure~\ref{figure_multiwavelength} for a $z=4$ point source as a function
102: of observed wavelength using the standard LCDM cosmology ($\Omega_\Lambda=0.7$, $\Omega_{\rm M}=0.3$, and $H_0=70~{\rm km}~{\rm s}^{-1}~{\rm Mpc}^{-1}$)
103: assumed throughout this paper. Three other possibilities are also plotted: minimal blurring $\Delta \phi_0$, the combination
104: of minimal blurring of the source plus that of intermediate sources along its line of sight $\Delta \phi_0 + \Delta \phi_{\rm l.o.s.}$, and maximal blurring $\Delta \phi_{\rm max}$, for a total of four.
105: Following \cite{Ng2003} I use $\alpha={2\over3}$ and
106: $a_0=1$. Reducing (increasing) $a_0$ shifts these lines in parrallel
107: towards lower left (upper right) in this plot, as does increasing (decreasing) $\alpha$.
108: The most dramatic changes come with variation in $\alpha$, by virtue of it appearing as an exponent in equations 1 and 3.
109:
110: Can blurring be constrained by current observations?
111: Figure~\ref{figure_multiwavelength} illustrates at what wavelengths an answer may be most convincing.
112: The highest spatial resolutions of some current telescopes
113: are indicated.
114: Chandra routinely observes sources to a resolution limit better than that imposed
115: by either equations 1 or 3 for $\alpha={2\over3}$. But this conflict is easily avoided, by either
116: suitably increasing $\alpha$ (0.7 would be sufficient) or reducing $a_0$ - or both.
117: This flexibility is allowed because current telescopes operating shortward of the
118: optical (including Chandra) do not form diffraction-limited images, and so it may be more difficult to
119: distinguish blurring induced by the
120: Planck scale from simply an unresolved source such as a galaxy. The dashed lines indicate
121: the angular size $\Delta \theta$ of
122: extended $z=4$ sources for a range of
123: physical diameters.
124: Thus, it may be best to look for blurring in the optical.
125: {\it HST} probes a regime predicted to be blurry, which is indicated
126: by the shaded regions. It can also resolve sources confirmed to be physically small at
127: other wavelengths. For example, if $\alpha={2\over3}$ and subject to the maximum blurring limit, no
128: $z=4$ source should appear sharper
129: than about 0\farcs1 across at 800 nm, broad enough to affect the {\it HST} diffraction pattern.
130: An object with $z=4$ would need to be 500 pc across to mimic this level of broadening.
131: This is smaller than a typical galaxy, but not quasars. Even if it were a factor of about 2.5 less,
132: {\it HST} could still marginally discriminate between Planck-scale-induced blurring and the effects of a
133: 200 pc source. But a $\sim 25$-m space telescope would be needed to rule out the minimum set by equation 1, by
134: resolving objects as small as 50 pc.
135:
136: \section{Observational Plan and Sample Selection}\label{selection}
137:
138: \cite{Ragazzoni2003} have already confirmed that the most severe blurring discussed in Section~\ref{limits} is
139: not seen in {\it HST} images of a $z=5.4$ Hubble Deep Field galaxy. It is worth considering how weak
140: blurring could be and still be ruled out for {\it HST}.
141: Ragazonni et al. point out that small perturbations of the phase should only cause a drop in image Strehl ratio $S$.
142: This is the ratio of the image peak to that of the diffraction spike of an unabberated telescope.
143: In this case, and where $\Delta \phi$ is comparable to the telescope optical aberrations, the Marechal
144: approximation to the Strehl ratio applies, and
145: $$S\approx\exp{\Big{[}-\Big{(}\Delta \phi {D\over{\lambda}}\Big{)}^2\Big{]}}. \eqno(5)$$
146: Figure~\ref{figure_models} is a plot of point-source Strehl ratio as a function
147: of redshift for an unaberrated telescope with $D=2.4$ m observing
148: at $\lambda=800$ nm. The shaded regions have the same meaning as in Figure~\ref{figure_multiwavelength}, again
149: plotted for $\alpha={2\over3}$ and $a_0=1$.
150: If the object is not a true point source but is actually extended, equations 1 and 3 do not strictly apply.
151: In this case the Strehl ratio of the image would also drop due to the partial
152: resolution of the source. For an object of angular size $\Delta \theta$ close to the
153: diffraction limit of the telescope one would expect a Strehl ratio of
154: $$S\approx \exp{\Big{[}-\Big{(}\Delta \theta {D \over{\lambda}}\Big{)}^2\Big{]}} \eqno(6)$$
155: in the absense of Planck-scale effects. The dashed lines in Figure~\ref{figure_models} represent equation 6 for
156: various source diameters.
157:
158: Judging from Figure~\ref{figure_models}, to test if high-redshift sources are indeed blurred, one could look at a
159: large
160: homogeneous sample of compact high-redshift objects imaged with {\it HST}.
161: In principle, the test is simple: Is there
162: a downward trend in Strehl ratio with increasing source redshift?
163: In practice, two factors must be well controlled. First, the sample must be of
164: uniformly small sources of very high redshift.
165: Figure~\ref{figure_models} indicates that for any Planck-scale effect to
166: be detected with {\it HST} the sample must extend to redshifts well beyond $z=4$.
167: Second, the point-spread function (PSF) must be
168: well characterized because the Planck-scale-blurring signal is
169: relative to the expectation for diffraction-limited imaging.
170:
171: As it turns out, there is already at least one archival
172: {\it HST} dataset that meets these strict critera: the Advanced Camera for Surveys (ACS) High-Resolution Channel
173: (HRC) snapshot survey of SDSS quasars.
174: This was a program of imaging to detect lensed companions among a sample of $4<z<6.3$ quasars.
175: None were found despite the excellent spatial resolution afforded by ACS HRC.
176: Because depth was not as important as spatial resolution, these images are relatively shallow, which
177: should minimize confusion by any extended host galaxy. Also, \cite{Kaspi2005} suggest that none of the
178: SDSS quasars should have a broad-line region larger than a few parsecs, which would make equation 6 negligibly
179: less than unity for {\it HST}.
180: Thus, this sample should satisfy the first condition. That they also satisfy the second is
181: a property of both {\it HST} PSF stability and the Nyquist-sampled ACS HRC pixel scale. Neither the
182: Space Telescope Imaging Spectrograph (STIS) nor the Wide Field and Planetary Camera 2 (WFPC2) can
183: fulfill this second constraint.
184:
185: \section{Data and Reductions}\label{reductions}
186:
187: The archival {\it HST} ACS snapshot survey
188: of Sloan Digital Sky Survey (SDSS) sources (Proposal: 9472, PI: Strauss) includes all
189: SDSS quasars with $z>4$; some of which have $z>6$. They were obtained in snapshot mode
190: through either the F775W filter (central wavelength 761 nm, 95 sources, $3.9<z<5.4$) or
191: F850LP (869 nm, 4 sources, $5.8<z<6.3$). All were imaged with the HRC, which has a
192: pixel scale of 0\farcs0246.
193:
194: These data have already been discussed in detail in \cite{Richards2004, Richards2006}, and
195: those reductions are closely followed here.
196: Pipeline-processed images were downloaded from the {\it HST} archive. This provides
197: standard corrections for overscan, bias, and dark-current, flatfield division, bad-pixel masking,
198: cosmic-ray removal, and photometric calibration.
199: The position of the quasar was determined to the nearest pixel using the IRAF\footnote[1]{IRAF is
200: distributed
201: by the National Optical Astronomy Observatory, which is operated by the Association of Universities
202: for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation.}
203: program IMEXAM. Next, the Tiny Tim (v6.3) program \citep{Krist1995} was used to
204: generate an appropriate PSF for each. Due to the strong
205: positional dependence of the HRC field distortion this produces a more accurate
206: PSF than can be determined by images of stars. Color dependence
207: is also properly accounted for, by inputing
208: a spectral energy distribution (SED) and convolving this with the {\it HST} filter curve.
209: Synthetic PSFs were generated for each quasar using a redshifted template SED based on
210: a composite SDSS QSO spectrum \citep{VandenBerk2001}. By shifting in the Fourier domain,
211: the sub-pixel position and peak of the PSF were allowed to float
212: relative to the image, and minimized based on the sum of the square of residuals. The result is
213: an unique PSF for each quasar, correct for its detector position, filter bandpass, and SED.
214:
215: All of the images were then co-aligned to the center of the nearest pixel.
216: Figure~\ref{figure_images} shows the result for SDSS J0836+0054 ($z=5.82$), which will serve as an example.
217: Each box is 3\arcsec $\times$ 3\arcsec~with the same orientation as in \cite{Richards2004} (their Figure 2).
218: All images have the same grey-scale stretch.
219: The non-circular first Airy ring is clearly evident, and further rings would be visible in
220: a harder stretch. The diffraction pattern of a circularly symmetric $2.4$-m diameter pupil
221: with a $0.7$-m central obstruction is also shown. Note that this excludes any aberrations
222: associated with the ACS camera, which explains its circular symmetry. The same SED as the quasar was used.
223: It is relative to this ideal diffraction pattern that - by definition - the Strehl ratio of the quasar and PSF is
224: to be determined.
225: Next to this is the PSF residual (the model residual will be discussed later, in Section~\ref{results}). This residual
226: is robust against variation in the template SED, both in relative
227: line strengths and choice of continuum power law.
228: It is also, as can be expected, comparable to that
229: obtained by Richards et al. and reveals no obvious host galaxy or lensed component. But underlying effects
230: due to microlensing cannot be ruled out, although it is not clear what effect they may have.
231: Results are similar for the other sources.
232:
233: Strehl ratio was then measured, which is not reported by
234: \cite{Richards2004, Richards2006}.
235: Each image and PSF was normalized to a flux of unity based on synthetic aperture photometry, and the height of its
236: peak determined
237: relative to the ideal diffraction pattern. The accuracy of the final
238: measurement is to within a few percent, which is the level of photometric uncertainty.
239:
240: \section{Analysis and Results}\label{results}
241:
242: Figure~\ref{figure_strehl} is a plot of Strehl ratio; crosses indicate the PSF.
243: The unfortunate division between the F775W and F850LP data at $z=5.5$ is evident.
244: Even if one ignores this, a clear downward trend with increasing source redshift can be seen, despite some scatter in the data.
245: Overplotted is a linear least-squares fit to these, which has slope of $-0.05\pm0.01 z$
246: (1-$\sigma$ errors).
247: That some of the lowest redshift quasars have Strehl ratios higher than
248: the PSF indicates that the limits of the telescope have been reached here. The
249: differences of a few percent are consistent with the uncertainties in the measurements.
250: Thus, the decrement below the PSF Strehl ratio for the $z>5$ redshift sources is probably real.
251: The situation for the $z>5.5$ sources is more secure. Limits predicted by equations 1 and 3 assuming $\alpha={2\over3}$ and $a_0=1$
252: are overplotted.
253: The break is due to the change in filter effective wavelength, from 761 nm to 869 nm.
254: For this choice of $a_0$ and $\alpha$, both $\Delta \phi_{\rm max}$ and $\Delta \phi_0 + \Delta \phi_{\rm l.o.s.}$ are clearly ruled out.
255: But blurring associated with just redshifted sources is not. It is striking how closely the maximum Strehl ratio
256: follows this curve.
257: And of those quasars for which the PSF should be sharp enough to allow it, none are found in the region that
258: it borders.
259: If this is correct, it places a tight constraint on $\alpha$.
260: Even if $a_0=10$, $\alpha$ need only grow
261: by 2\% (to 0.68) and still be in agreement with the data.
262: Although also possible, this result is not easily explained by
263: the size of the sources alone. It would seem that the intrisic sizes of these quasars (either their narrow-line or broad-line regions - or both) would need to occupy a very narrow range for this to happen, between about 200 pc and 300 pc.
264:
265: In a further simple test of the internal consistency of the $z>5.5$ results, a simple ``blurred" quasar
266: model was generated. Guided by the good fit of $\Delta \phi_z$ in Figure~\ref{figure_strehl}, the existing PSF was
267: cropped slightly below
268: its peak, at a cutoff given by ${\rm cutoff}={\rm peak}\times\exp{\Big{[}-\Big{(}\Delta \phi_z {D\over{\lambda}}\Big{)}^2\Big{]}}$. A
269: value $\lambda = 869$ nm was used, for $D=2.4$ m.
270: This naive model is indeed a better fit to the observed Strehl
271: ratio, indicated by the open triangles in Figure~\ref{figure_strehl}. The effects on the residual are benign, which
272: can be seen in Figure~\ref{figure_images}. This is also evident in slices along the x and y axes, which are plotted
273: in Figure~\ref{figure_profiles}. This confirms that the blurred quasar model is a better fit than the original PSF,
274: accomplished without adversely affecting the residual. This is reassuring, as it demonstrates that these results are not
275: in conflict with those of \cite{Richards2004}.
276:
277: In summary,
278: I have searched for blurring induced by the effects of the Planck scale in {\it HST} images of
279: high-$z$ quasars. Although blurring may be seen, if real, it is just at the observable threshold.
280: The test here is more sensitive than previous ones, which only looked for compact sources lacking diffraction rings.
281: The shorter wavelength of emitted light has also been accounted for, which if correct,
282: restricts $\alpha$ to be 0.68 even if $a_0$ is as big as 10.
283:
284: The implications of blurring are significant, so it is worth looking for.
285: A true detection could point to a successful quantum gravity theory.
286: But detection is elusive because
287: the signal is confused with that of a partially resolved source.
288: It is hoped that this work will encourage further tests for the
289: effect with {\it HST}. A first step would be to
290: re-observe the four $z>5.5$ quasars in this sample with the HRC and the F775W filter.
291: Observations of higher-$z$ quasars (as they become known) could ultimately confirm or refute
292: the current result.
293:
294: \begin{thebibliography}{}
295: \bibitem[Amelino-Camelina(2000)]{Amelino-Camelina2000} Amelino-Camelina, G. 2000 in Lect. Notes
296: Phys. 541, Towards Quantum Gravity, ed. J. Kowalski-Glikman (Berlin: Springer), 1
297: \bibitem[Beckenstein(1973)]{Beckenstein1973} Beckenstein, J.D. 1973, Phys. Rev. D, 7, 2333
298: \bibitem[Becker et al.(2001)]{Becker2001} Becker, R.H. et al. 2001, \aj, 122, 2850
299: \bibitem[Kaspi et al.(2005)]{Kaspi2005} Kaspi, S., Maoz, D., Netzer, H., Peterson, B.M., Vetergaard, M., \& Jannuzi, B.T. 2005, \apj, 629, 61
300: \bibitem[Hawking(1975)]{Hawking1975} Hawking, S. 1975, Commun. Math Phys., 43, 199
301: \bibitem[Krist(1995)]{Krist1995} Krist, J. in ASP Conf. Ser. 77, Astronomical Data Analysis Software and
302: Systems IV, ed. R. A. Shaw, H. E. Payne, \& J. J. E. Hayes (San Francisco: ASP), 349
303: \bibitem[Lieu \& Hillman(2003)]{Lieu2003} Lieu, R. \& Hillman, L.W. 2003, \apjl, 585, L77
304: \bibitem[Ng(2002)]{Ng2002} Ng, Y.-J., 2002, Int. J. Mod. Phys., D11, 1585
305: \bibitem[Ng, Christiansen, \& van Dam(2003)]{Ng2003} Ng, Y.J., Christiansen, W.A., \& van Dam, H. 2003, \apjl, 591, L87
306: \bibitem[Ragazzoni, Turatto, \& Gaessler(2003)]{Ragazzoni2003} Ragazzoni, R., Turatto, M., \& Gaessler, W. 2003, \apjl, 587, L1
307: \bibitem[Richards et al.(2004)]{Richards2004} Richards, G.T., Strauss, M.A., Pindor, B., Haiman, Z., Fan, X., Eisenstein, D., Schneider, D.P.,
308: Bahcall, N.A., Brinkman, J. \& Brunner, R. 2004, \aj, 127, 1305
309: \bibitem[Richards et al.(2006)]{Richards2006} Richards, G.T., Haiman, Z., Pindor, B., Strauss, M.A., Fan, X., Eisenstein, D., Schneider, D.P.,
310: Bahcall, N.A., Brinkman, J. \& Fukugita, M. 2006, \aj, 131, 49
311: \bibitem[t'Hooft(1993)]{tHooft1993} t'Hooft, G. 1993, CQGra, 10, 8, 1653
312: \bibitem[Vanden Berk et al.(2001)]{VandenBerk2001} Vanden Berk, D. E., et al. 2001, \aj, 122, 549
313: \end{thebibliography}
314:
315: \begin{figure}
316: \plotonetwothirds{f1.eps}
317: \caption{Accumulated phase dispersions in arcseconds as a function
318: of wavelength for a point source with $z=4$, according to equations 1 and 3 with
319: $\alpha={2\over3}$ and $a_0=1$. By comparison, the resolution limts of some
320: telescopes are indicated: Chandra X-ray Observatory (Chandra), Roentgen Satellite (ROSAT),
321: Far Ultraviolet Spectroscopic Explorer (FUSE), adaptive optics on an 8-m-class telescope (AO), Smithsonian Submillimeter Array (SMA),
322: Very Large Array (VLA), and Very Long Baseline Interferometry (VLBI). The sizes of extended objects of
323: various physical diameters are shown
324: as dashed lines. Shaded regions bordered by equations 1 and 3 (and two intermediate subregions formed by either excluding or combining their terms) are potentially precluded from observation.}
325: \label{figure_multiwavelength}
326: \end{figure}
327:
328: \clearpage
329:
330: \begin{figure}
331: \plotonetwothirds{f2.eps}
332: \caption{Precluded Strehl ratio for blurred point sources (shaded regions) and
333: predicted Strehl ratio for partially-resolved extended sources (dashed lines) as a
334: function of redshift for {\it HST} observing at 800 nm. Shaded regions have the same
335: meanings as in Figure~\ref{figure_multiwavelength}.}
336: \label{figure_models}
337: \end{figure}
338:
339: \clearpage
340:
341: \begin{figure}
342: \plotonetwothirds{f3.eps}
343: \caption{Images of the quasar, PSF, ideal diffraction pattern, and residuals for one of the QSOs in the sample.
344: Results are similar for the others.}
345: \label{figure_images}
346: \end{figure}
347:
348: \clearpage
349:
350: \begin{figure}
351: \plotonetwothirds{f4.eps}
352: \caption{A plot of Strehl ratio (open squares) for the sample. The crosses are the PSF for each
353: quasar, giving an optically corrected Strehl ratio via the Tiny Tim software.
354: There is a noticable trend towards lower observed Strehl ratio with increasing redshift. A linear
355: least-squares fit to all data is overplotted. No correction for the change in filter at $z=5.5$ has been
356: made. Models of Planck-scale induced blurring are shown; the shaded regions have the same meanings as
357: in Figures~\ref{figure_multiwavelength} and ~\ref{figure_models}.
358: The break is due to the shift in filter central wavelength. Open triangles indicate the ``blurred" quasar model
359: of Section~\ref{results}. The region bordered by $\Delta \phi_z$ is avoided, and so cannot
360: be ruled out by the observations.}
361: \label{figure_strehl}
362: \end{figure}
363:
364: \clearpage
365:
366: \begin{figure}
367: \plottwo{f5a.eps}{f5b.eps}
368: \caption{Slices along the x and y axes of the same quasar shown in Figure~\ref{figure_images} (dotted line). Overplotted are
369: the PSF residual (thin solid line) and model residual (thick solid line). All are normalized
370: to the peak intensity in the quasar image. The scale has been restricted to 0.6 and below to better display the
371: residuals. The model, although
372: still not perfect, is a better fit to the quasar.}
373: \label{figure_profiles}
374: \end{figure}
375:
376: \end{document}
377: