astro-ph0603676/ms.tex
1: %%
2: % First draft W.Danchi - 03 November 2005
3: % Given to colleagues on  10 November 2005
4: % Close to final draft 22 November 2005, with  
5: % comments from Jay, Rich, and Jeremy.
6: % Additional comments from Jay, Jeremy, 23 Nov. 2005
7: % Fixed figures and improved discussion.
8: % Revised manuscript, 24 Feb. 2006
9: % 2nd revision, 17 Mar 2006
10: %%
11: %
12: \documentclass[12pt,preprint]{aastex}
13: %\documentclass[]{emulateapj}
14: %% manuscript produces a one-column, double-spaced document:
15: 
16: %\documentclass[manuscript]{aastex}
17: %\usepackage{amsmath}
18: %\usepackage{emulateapj5, apjfonts}
19: %\usepackage{epstopdf}
20: %% preprint2 produces a double-column, single-spaced document:
21: 
22: %% \documentclass[preprint2]{aastex}
23: 
24: %% Sometimes a paper's abstract is too long to fit on the
25: %% title page in preprint2 mode. When that is the case,
26: %% use the longabstract style option.
27: 
28: %% \documentclass[preprint2,longabstract]{aastex}
29: 
30: %% If you want to create your own macros, you can do so
31: %% using \newcommand. Your macros should appear before
32: %% the \begin{document} command.
33: %%
34: %% If you are submitting to a journal that translates manuscripts
35: %% into SGML, you need to follow certain guidelines when preparing
36: %% your macros. See the AASTeX v5.x Author Guide
37: %% for information.
38: 
39: \newcommand{\vdag}{(v)^\dagger}
40: \newcommand{\myemail}{William.C.Danchi@nasa.gov}
41: 
42: \slugcomment{Astrophysical Journal, in press, 10 July 2006, v645n 2 issue.}
43: 
44: \shorttitle{Phase in Nulling Interferometry}
45: \shortauthors{Danchi, Rajagopal, et al.}
46: 
47: \begin{document}
48: 
49: \title{The Importance of Phase in Nulling Interferometry  \
50: and a Three Telescope Closure-Phase Nulling Interferometer Concept}
51: %
52: \author{W. C. Danchi, J. Rajagopal\altaffilmark{1}, M. Kuchner, J. Richardson, D. Deming\altaffilmark{2}}
53: \affil{NASA Goddard Space Flight Center, Exoplanets and 
54: Stellar Astrophysics Laboratory, Code 667, Greenbelt, MD 20771}
55: \email{William.C.Danchi@nasa.gov}
56: %, jayadev@iri1.gsfc.nasa.gov, 
57: %mkuchner@milkyway.gsfc.nasa.gov, jeremy@iri1.gsfc.nasa.gov, 
58: %ddeming@pop600.gsfc.nasa.gov}
59: %
60: \altaffiltext{1}{University of Maryland, Astronomy Department, College Park,
61: MD 20742}
62: \altaffiltext{2}{Planetary Systems Laboratory, Code 693, Goddard
63: Space Flight Center, Greenbelt, MD 20771}
64: %
65: \begin{abstract}
66: We discuss the theory of the Bracewell nulling interferometer and 
67: explicitly demonstrate that
68: the phase of the ``white light" null fringe is the same as the 
69: phase of the bright output from an ordinary stellar interferometer.
70: As a consequence a ``closure phase" exists for a nulling interferometer
71: with three or more telescopes. 
72: We calculate the phase offset as a function of baseline length for an
73: Earth-like planet around the Sun at 10 pc, with a contrast ratio
74: of $10^{-6}$ at 10 $\mu$m.  The magnitude of the phase due to the
75: planet is $\sim 10^{-6}$ radians, assuming the star is at the phase
76: center of the array.  Although this is small, this phase may
77: be observable in a three-telescope nulling interferometer 
78: that measures the closure phase.
79: We propose a simple non-redundant
80: three-telescope nulling interferometer that can perform
81: this measurement.  This 
82: configuration is expected to have improved characteristics compared
83: to other nulling interferometer concepts,
84: such as  a relaxation of pathlength tolerances,
85: through the use of the ``ratio of wavelengths" technique, a 
86: closure phase, and better discrimination between exodiacal
87: dust and planets.
88: 
89: \end{abstract}
90: 
91: \keywords{telescopes --- techniques: interferometric ---
92: techniques: high angular resolution ---
93: stars: planetary systems --- stars: circumstellar material}
94: 
95: \section{Introduction}
96: %
97: Direct imaging
98: of Earth-like planets around nearby stars is extremely difficult
99: for two fundamental reasons. The first is the need for 
100: resolution well below one arcsec; the second is the extremely large
101: contrast ratio between the planet and star, which is  
102: $\sim 10^{-6} - 10^{-7}$ at mid-infrared wavelengths (e.g., 5-20 $\mu$m)
103: and  $\sim 10^{-9} - 10^{-10}$ at visible wavelengths (400-900 nm).  Generally, 
104: interferometric techniques are favored in the infrared, while coronagraphic techniques have received the most attention at visible wavelengths.
105: The extrasolar zodiacal dust around these stars provides an additional
106: complication, primarily 
107: because their intensity compared to the zodiacal dust in our
108: own Solar system, and also their spatial distribution, is largely unknown.
109: 
110: Interferometric methods for direct detection of extrasolar planets are
111: derived from the initial concept of Bracewell (1978) (see also 
112: Bracewell \& MacPhee 1979) of a rotating
113: two-telescope interferometer, in which a 180 degree phase shift was 
114: applied to the electric field from one of the two telescopes.  The 
115: net result of this phase shift is a response pattern with a minimum
116: or ``null" response on-axis (zero pathlength difference
117: between the two telescopes),
118: which suppresses the unwanted signal from
119: the starlight, and has a maximum response off-axis, at an angle 
120: proportional to the wavelength of light divided by the separation
121: between the two telescopes.  
122: 
123: More than a quarter century has passed since that initial paper, and
124: there has been a substantial body of work in which a variety of 
125: array configurations of interferometers has been proposed for planet detection, 
126: including the OASES (Angel \& Woolf 1997), 
127: Dual-Chopped Bracewell (Woolf et al. 1998), 
128: and Darwin (Leger et al. 1996, Mennesson \& Marriotti 1997)
129: configurations, respectively.  These interferometers were designed to 
130: improve the response of the array, principally to reduce the effect
131: of stellar leakage, such as in the OASES array, or to allow for subtraction
132: of a symmetrical distribution of extrasolar zodiacal dust, as in 
133: the Dual-Chopped Bracewell array.  Essentially all studies of the 
134: potential performance of these arrays have been based on calculations
135: of the response of the array in the 
136: far-field in which the interferometers are viewed as a phased array,
137: and only the intensity at the nulled 
138: output of the interferometer is considered (e.g., 
139: as described in Mennesson \& Marriotti 1997).
140: 
141: However, the literature on nulling interferometry
142: has not been explicitly connected to the large body of
143: work on conventional optical interferometers (in which the inputs from the 
144: individual elements are combined in phase at zero pathlength difference).
145: In particular, the observable quantity called the ``closure phase,''
146: was originally developed by Jennison (1958) for phase-unstable 
147: radio interferometers, and has been used successfully at visible 
148: and infrared wavelengths in the past few years.  For example,
149: images of complex sources, such as the surfaces of stars (Young et al. 2000),
150: and clumpy, dusty outflows around massive 
151: stars like the spiral-shaped outflow 
152: discovered around WR 104 (Tuthill, Monnier, \& Danchi 1999), have been synthesized at very high angular resolution (i.e., $\ll$ 0.1 arcsec),
153: at visible and infrared wavelengths, respectively, using this 
154: technique.  
155: 
156: In this paper we examine the fundamental connection 
157: between nulling interferometers and conventional stellar interferometers.
158: We begin by showing
159: that the phase of the ``white light" null fringe is the same as that 
160: of the ``white light" bright fringe 
161: from an ordinary stellar interferometer.
162: For two sources with very unequal intensities, such as a star and 
163: an earth-like planet, we show that the phase is
164: small but is observable. We demonstrate 
165: the existence of a ``closure phase" for nulling interferometers.
166: Finally, we propose a simplified non-redundant
167: three-telescope nulling interferometer for the detection of
168: Earth-like planets, which includes a ``null" closure phase.  This 
169: configuration is expected to have improved characteristics including 
170: essentially no variability noise, a relaxation of pathlength tolerances
171: through the use of the ``ratio of wavelengths" technique and a 
172: closure phase.
173: %
174: \section{The Phase of the Null Fringe}
175: Figure 1 displays a typical experimental situation for a simplified
176: two-telescope nulling interferometer.  Let $E_1$ be the electric field at the 
177: location $\mathbf{r}$, and $E_2$ be the complex electric field at the 
178: location $\mathbf{r + B}$, where $\mathbf{B}$ is the baseline vector 
179: separating the centers of the two telescopes (e.g., the central ray
180: in geometric optics), and $\mathbf{r} = (x,y,z)$ and
181: $\mathbf{B} = (B_x, B_y, B_z) $, are the cartesian coordinates of the 
182: position vector of the center
183: of the first telescope, and the baseline vector, respectively.  An achromatic
184: $\pi$ phase shift is applied to the electric
185: field $E_1$, and the two fields are combined on an ideal 50\% beamsplitter
186: labeled BS in the figure.  The intensities measured at the two 
187: output ports of the beamsplitter are labeled $I_1$ and $I_2$, respectively.
188: The time averaged correlation 
189: between the two electric fields is given by:
190: %
191: \begin{equation}
192: \tilde{\Gamma}_{12} = \langle E_1 (\mathbf{r},t) ~ {E_2}^* (\mathbf{r + B}, t) \rangle
193: \end{equation}
194:  %
195:  By the van Cittert-Zernike theorem of optics (see Thompson, Moran, 
196:  and Swenson 2001) the quantity $\tilde{\Gamma}_{12}$ is proportional
197:  to the complex visibility\footnote{ $\tilde{\Gamma}_{12}$ is the ``mutual 
198:  intensity function'' (Goodman 1985), which when normalized by the total 
199:  flux leads to the ``complex coherence factor.''  By the van Cittert-Zernike
200:  theorem (Thompson, Moran, \& Swenson 2001) this is the Fourier transform
201:  of the object intensity and is the classical visibility when $I_1 = I_2$.} 
202:  of
203:  the source intensity distribution in the far field of the array.  Thus,
204:  we can define $\tilde{\Gamma}_{12} \equiv | \Gamma _{12} ~ | e^{i \phi _{12}}$,
205:  where $ | \Gamma _{12} | $ is the visibility amplitude and $\phi _{12}$ 
206:  is the visibility phase.
207:  
208:  We assume the electric fields from 
209:  the two telescopes have additional
210:  phases $\phi _1$, and $\phi _2$ before they are combined at the beamsplitter,
211:  such as might be due to pathlength variations or mismatches
212:  between the two arms of the array.  By expanding Eqn. (1) the 
213:  resultant intensities are (in units where $c / 8 \pi = 1$):
214:  %
215:  \begin{eqnarray}
216:  I_1 = 1/2 ~ [ ~|E_1|^2 + |E_2|^2 - 2 |\Gamma _{12}| \cos ( \phi_{12} + \phi _1
217:  - \phi _2) ~] \\
218:   I_2 = 1/2 ~ [ ~|E_1|^2 + |E_2|^2 + 2 |\Gamma _{12}| \cos ( \phi_{12} + \phi _1
219:  - \phi _2) ~] 
220: \end{eqnarray}
221: %
222: where $I_1$ is the intensity at the null output port of the interferometer,
223: and $I_2$ is the intensity at the bright output.  If the output intensities
224: are normalized to the total power incident on the array, i.e., by dividing
225: $I_1	$ and $I_2$ by $I_T = |E_1|^2 + |E_2|^2 $ and defining the normalized
226: visibility, $|V_{12}| = 2 | \Gamma _{12} | / I_T$, then we can rewrite 
227: the above equations as:
228: %
229: \begin{eqnarray}
230: \tilde{I}_1 = 1/2 ~ [ 1 - |V_{12}| \cos ( \phi_{12} + \phi _1
231:  - \phi _2)~ ] \\ 
232: \tilde{I}_2 = 1/2 ~ [ 1 + |V_{12}| \cos ( \phi_{12} + \phi _1
233:  - \phi _2) ~]
234:  \end{eqnarray}
235: % 
236: where $\tilde{I}_1 = I_1 / I_T$ and $\tilde{I}_2 = I_2 / I_T$.
237: 
238: Thus we see that the phase terms for $\tilde{I}_1$ and $\tilde{I}_2$ are
239: identical and both contain the same visibility phase, $\phi_{12}$. 
240: This emphasizes the often overlooked fact that there is only one phase in 
241: any stellar interferometer.\footnote{Indeed,
242: the complementary output from the pupil plane combiner of all conventional 
243: interferometers is a null by the conservation of energy, and
244: this output is well established to have the same phase as the bright output.}
245: The fact that the null output is ``locked in'' at the position of the 
246: star has dominated thinking in terms of current designs  for 
247: the detection of earth-like planets around nearby stars, and as a 
248: result, phase measurements have not been considered (e.g., Angel \& Woolf 1997).
249: However, there is no fundamental obstacle to the measurement of the 
250: phase.  For example, a ditherless quadrature phase detection scheme 
251: (see Barry et al. 2005) at the bright output or a small dither at the 
252: null output should suffice.  In any case, some measurement of the null phase
253: (measured by delay offsets) is necessary 
254: in order to locate and track the minimum intensity at the null output
255: port relative to the delay set by the fringe tracking system, as in
256: the design of the Fourier-Kelvin Stellar Interferometer (FKSI) (Danchi et al. 2003, 2004; Hyde et al. 2004). Furthermore, as shown in the next section,
257: this phase is small but it can be measured and should not be 
258: ignored in the design of nulling interferometers for planet detection.
259:  
260: \section{Understanding the Null Phase}
261: Figure 2 displays a schematic diagram of the geometry of a binary
262: system, e.g., a star and planet if the ratios of the two intensities,
263: $I_A$, located at $(0, 0)$ and $I_B$, at $(x_0, y_0)$, 
264: are large.  The complex visibility is calculated by a simple Fourier transform of
265: \begin{equation}
266: I_T (x,y) = I_A \delta (x,y) + I_B \delta (x-x_0,y-y_0). 
267: \end{equation}
268: For the purpose of the present discussion we assume both sources are
269: point sources.  The visibility amplitude, $V(u,v)$, normalized to the total intensity, and 
270: phase, $\phi (u,v)$, are given by:
271: %
272: \begin{eqnarray}
273: V(u,v) & = & \biggl [ \frac{ {I_A}^2 + {I_B}^2 + 2 {I_A}{I_B} \cos ~ 2 \pi ( u x_0 + v y_0 )}{({I_A}+{I_B})^2} \biggr ] ^{1/2} \\
274: \phi (u,v) & = & - \arctan \biggl [ \frac{I_B \sin 2 \pi( u x_0 + v y_0 )}{I_A + I_B \cos 2 \pi ( u x_0 + v y_0)} \biggr ]
275: \end{eqnarray}
276: %
277: where $(u,v)$ denotes the usual coordinates in the Fourier plane.
278: 
279: For simplicity, let us take a one-dimensional example.  Let $y_0 = 0$, 
280: $u =  B_0 / \lambda$, and $I_A \gg I_B$.  Let $x_0$ be the angular
281: separation of the two point sources.  With these assumptions,
282: Eqn. (8) reduces to:
283: %
284: \begin{equation}
285: \phi \sim - \arctan ~[ ~ ( I_B / I_A ) \sin ( 2 \pi B_0 x_0 / \lambda ) ~]
286: \end{equation}
287: %
288: As a simple
289: numerical example, assume $x_0 =$ 0.1 arcsec and if $\lambda$ = 10 $\mu$m,
290: then the phase passes through zero
291: when $2 \pi B_0 x_0 / \lambda = \pi $, i.e., when $x_0 = \lambda / 2 B_0$.
292: This is exactly the same angular size given by the usual definition of
293: resolution in conventional stellar interferometry, which is that
294: two point sources are resolved if their separation $x_0 > \lambda / 2 B_0$,
295: or for the parameters in this example, $x_0 > $ 0.05 arcsec. 
296: The quantity $1 - V$ can be expanded, like the expression for the phase,
297: $\phi$, and is given by:
298: %
299: \begin{equation}
300: 1 - V \sim  ( I_B / I_A ) [1- \cos ( 2 \pi B_0 x_0 / \lambda )].
301: \end{equation}
302: %
303: From this result, we see that the output of the nulling interferometer
304: is essentially the intensity of the planet with the stellar flux removed.
305: We also observe that $1-V$ is maximized at  $x_0 = \lambda / 2 B_0$, which
306: is the same as the condition for the phase, $\phi$, to pass through zero.
307: 
308: Figure 3(a) (upper panel) 
309: displays the null phase, $\phi$, while Fig. 3(b) (lower panel)
310: displays the normalized intensity,$1 - V$,
311: of the null output port, i.e., from Eqn. (4),
312: respectively.   In this calculation we assume 
313: intensities, $I_B = 10^{-6}$, $I_A = 1 - I_B$,
314: and baselines ranging from 0 to 40 m at 10 $\mu$m, for an angular 
315: separation of 0.1 arcsec for the two sources (the Sun-Earth 
316: separation at 10 parsecs).  We clearly see that the object phase varies
317: and is of the order of $2 \times 10^{-6}$ radians, or about $10^{-4}$ degrees.
318: This is small but it can be measured, even with substantial
319: pathlength fluctuations within the array, by the use of the closure phase
320: concept, which we now discuss.
321: 
322: From Eqns. (2)-(5), we see that for a telescope pair, there is a phase
323: term, 
324: \begin{equation}
325: \tilde{\phi_{12}} = \phi_{12} + \phi_{1} - \phi_{2}
326: \end{equation}
327: which is the object phase, $\phi_{12}$, plus the telescope dependent phase errors,
328: $\phi_1$ and $\phi_2$, respectively.  For a three-telescope nulling
329: interferometer, shown schematically in Fig. 4 and described in the next
330: section, there are equivalent expressions to Eqn. (8) for each telescope,
331: \begin{eqnarray}
332: \tilde{\phi_{23}} & = & \phi_{23} + \phi_{2} - \phi_{3} \\
333: \tilde{\phi_{13}} & = & \phi_{13} + \phi_{1} - \phi_{3}
334: \end{eqnarray}
335: Adding Eqns. (11)-(13), it is easy to show there is a closure phase relation,
336: \begin{eqnarray}
337: \phi_C &  = & \tilde{\phi_{12}} + \tilde{\phi_{23}} - \tilde{\phi_{13}} \\
338:        &  = & \phi_{12} + \phi_{23} - \phi_{13}
339: \end{eqnarray}
340: which is exactly the same closure phase relation that is normally obtained in
341: ground-based stellar interferometry.
342: 
343: From ground-based interferometry we know that the closure phase can be measured
344: with good precision even in the presence of very large pathlength fluctuations
345: in the atmosphere. For example, at the IOTA interferometer at Mt. Hopkins,
346: closure phases are commonly measured to a precision of approximately 
347: 1 degree at 2 $\mu$m, for pathlength fluctuations on the order of 10 $\mu$m.
348: Thus there is an effective suppression of the fluctuations by about a 
349: factor of 1800 (Ragland et al. 2004).  
350: Similar results have been obtained with the 
351: Infrared Spatial Interferometer at Mt. Wilson (Hale, Weiner, \& Townes 2004).  
352: 
353: For a stellar interferometer operated in space, we assume that
354: the fringes are 
355: sensed and tracked at a wavelength of 2 $\mu$m.  If the 
356: science band of the instrument is at 10 $\mu$m, and if the fringes are 
357: tracked to an RMS precision of 1 degree at 2 $\mu$m (i.e., 0.2 degrees
358: at 10 $\mu$m), then in principle it should be
359: possible to obtain an improvement in precision by using the closure phase
360: at 10 $\mu$m to $ \sim 0.2/1800 \approx 10^{-4}$.  Thus, the closure 
361: phase technique can be used to obtain the precision required for 
362: the detection of Earth-like planets in the Solar neighborhood.
363: 
364: The fringe tracking precision quoted above is a very conservative one,
365: and is based on assuming a very small collecting area of about
366: 1 m$^2$, i.e., a two-telescope
367: interferometer like IOTA, consisting of 1/2 m in diameter telescopes.  In 
368: principle it is possible to track the fringes to a much higher precision,
369: as we show in the following calculation.  
370: 
371: The rms error in determining the fringe phase is given by 
372: \begin{equation}
373: {\sigma}_\phi = \sqrt { \frac {2}{N V^2} }, 
374: \end{equation}
375: where $N$ is the number of photons collected, and $V$ is the visibility
376: of the fringe amplitude (Goodman 1985).  Current designs for the Terrestrial
377: Planet Finder Interferometer (TPF-I) mission envision
378: much larger apertures, most likely in the 3-4 m diameter range
379: (Beichmann, Woolf, \& Lindensmith 1999).  For 
380: a pair of such telescopes, the collecting area is 14-25 m$^2$.  Assuming
381: an optical efficiency of 10\%, a bandwidth of 50\% at a center wavelength
382: of 2 $\mu$m, and an integration time of 0.01 sec, we can easily estimate
383: the attainable precision in tracking the fringe phase at 2 $\mu$m.  For 
384: a solar type star at 10 pc, it is easy to show that 
385: $ N \sim (0.7-1.3) \times 10^7$ photons for this short integration 
386: time.  Assuming $V \sim 1$, then 
387: the error in the phase measurement is $(1.6-2.0) \times 10^{-2}$ degrees.
388: Thus the phase error due to tracking the fringes at 10 $\mu$m is 
389: $(3-4) \times 10^{-3}$ degrees.  This means the suppression required 
390: from the closure phase technique is only about a factor of 10-20, much less
391: than the factor of 1800, which has been routinely achieved using current ground-based interferometers, like IOTA and ISI.  We conclude that for a large space
392: interferometer such as envisioned for TPF-I, only a modest suppression 
393: is needed from the closure phase technique.
394: 
395: In ground based radio interferometry, high dynamic range imaging 
396: is achieved through self-calibration and closure-phase calibration
397: as described by Perley (1999).   The 
398: technique of self-calibration, which is used to reduce antenna-based 
399: errors, employs a process of developing a model of a point source based
400: on the data itself.  The complex gains are computed from the point
401: source model and then used to correct the visibilities.  A new model is 
402: formed from the corrected visibilities and the process is repeated as 
403: necessary.  Following this, a closure correction can be applied by
404: additional measurements of strong point sources,
405: and this additional calibration gives dynamic ranges as high as $10^5$.
406: Part of the robustness
407: of the closure phase technique in radio astronomy is due to the fact
408: that there are a large number of antennas, e.g., 27 for the Very Large
409: Array, which allows for many closure phases to
410: be determined, as well as closure amplitudes (Pearson \& Readhead 1984).
411: This allows for recovery of most of the visibility phase and amplitude
412: information from the closure amplitude and phases.   
413: 
414: However, phase closure techniques are limited
415: in part by baseline-based errors, which are not calibrated out between 
416: source and calibrator measurements (see Perley 1999, and references therein).
417: Baseline dependent errors include correlator errors that cause non-factorable
418: gain errors (e.g., Cornwell \& Fomalont 1999), but these errors can be
419: made negligible using the best current correlator designs.  The accuracy and stability of the baseline measurements affects the 
420: dynamic range as well, as baseline variations that occur between source
421: and calibrator measurements will not be removed. 
422: 
423: For structurally connected
424: space-based  infrared interferometers
425: for planet detection will use passive cooling (e.g., using a sunshade) to reduce
426: thermal noise, and with careful design, thermal drifts which could
427: affect the baseline length due to motion
428: between source and calibrator can be substantially less than a degree.  
429: Composite structures also can have extremely small thermal expansion coefficients at the instrument temperature (expected to be around 35 K). 
430: 
431: Free-flying space based interferometers are more similar in nature 
432: to ground-based interferometers because of baseline drifts.  In space
433: these are likely due 
434: to imperfect formation control, however, metrology systems can be 
435: employed to retain knowledge of the baselines to a very high precision, 
436: i.e., much less than 1 $\mu$m, consequently this source of closure 
437: phase errors can be substantially reduced through calibration.
438: In ground-based interferometry, thermal drifts and mechanical imprecision,
439: for example, due to co-alignment errors between optical axes and mechanical
440: axes of rotation can contribute to baseline errors.  These errors 
441: can be significant at optical and infrared wavelengths but are much
442: less important at radio wavelengths.
443: 
444: Finally, a major difference between ground-based and space-based interferometers
445: is that there are likely to be far fewer telescopes employed in 
446: the latter as compared to the former.  Hence there will
447: be at most a few closure phases and a closure amplitude available,
448: e.g., for a 4 telescope configuration, there are six baselines (and
449: visibility amplitudes and phases), 4 closure phases (3 independent) and
450: one closure amplitude.  However, for planet detection, the
451: strong (essentially) point source from the star allows for multi-wavelength
452: techniques to be used as described above, such that the interferometer is 
453: stabilized at 2 $\mu$m and the planets are detected at 10 $\mu$m.  Thus
454: at any given time, there are closure phases at 2 $\mu$m available to 
455: calibrate the closure phases in the 10 $\mu$m science band, and this 
456: can be done essentially continuously, and both for source as well 
457: as calibrator.  This means that in space based planet detection, both
458: the intrinsic precision and calibration are expected to be much better than 
459: can be attained on the ground.  
460: 
461: In the next section we describe a possible implementation of a three
462: telescope closure phase nulling interferometer for planet detection. 
463: 
464: \section{A closure phase nulling interferometer concept}
465: 
466: Figure 4 displays a conceptual block diagram for a 
467: three telescope nulling interferometer
468: concept that is analogous to ground based interferometers that typically 
469: measure three visibilities and a closure phase.
470: In this concept, electric fields, $E_1$, $E_2$, and $E_3$ from the
471: three telescopes, with baselines $B_{12}$, $B_{23}$, and $B_{13}$ between
472: them, are first incident on ideal 50\% beamsplitters, labeled BS, in
473: the diagram.  After this, one of the two beams split from each telescope
474: is passed through an achromatic $\pi$ phase shifter, and is mixed with
475: a non-phase shifted counterpart from one of the other telescopes. 
476: This produces
477: three nulled and three bright outputs.
478: For example, 50\% of the light from telescope 1 (blue line) is mixed
479: with 50\% of the light from telescope 2 (dashed line), which had 
480: previously passed through a $\pi$ phase shifter, denoted by the 
481: purple rectangle in the drawing.  The resultant output beams from
482: this second beamsplitter are drawn with the blue dashed lines, and
483: have intensities $I_{A1}$ and $I_{A2}$ as labeled in the figure.
484: The other output beams are $I_{B1}$, $I_{B2}$, and $I_{C1}$, $I_{C2}$,
485: for baselines $B_{23}$ and $B_{13}$, respectively.  
486: 
487: Following the analysis of Section 1 (e.g. Eqn. (4)), we can write
488: the normalized intensities at the null output ports, as $\tilde{I}_{A1}$,
489: $\tilde{I}_{B1}$, $\tilde{I}_{C1}$, with corresponding bright output
490: intensities, $\tilde{I}_{A2}$, $\tilde{I}_{B2}$, $\tilde{I}_{C2}$.
491: As in Eqn. (4), the null intensities are given by:
492: %
493: \begin{eqnarray}
494: \tilde{I}_{A1} & = & 1/2 ~ [ 1 - |V_{12}| \cos ( \phi_{12} + \phi _1
495:  - \phi _2)~ ] \\ 
496:  \tilde{I}_{B1} & = & 1/2 ~ [ 1 - |V_{23}| \cos ( \phi_{23} + \phi _2
497:  - \phi _3)~ ] \\ 
498: \tilde{I}_{C1} & = & 1/2 ~ [ 1 - |V_{13}| \cos ( \phi_{13} + \phi _1
499:  - \phi _3)~ ] 
500: \end{eqnarray}
501: %
502: Consequently, the system described in Fig. 4 has three null outputs,
503: three bright outputs, and a closure phase.
504: 
505: Figure 5 displays a numerical example for a three telescope system 
506: with baselines $B_{12} = 5$m, $B_{23} = 25$m, and $B_{13} = 30$m.
507: We take the same star-planet parameters as the example in the previous
508: section, $I_{B} = 10^{-6}$, and $I_A = 1 - I_B$, and an angular
509: separation of 0.1 arcsec, the Earth-Sun separation at 10 pc.
510: We can see the substantial difference in null output intensities for
511: the three baselines, $B_{12}$ (red), $B_{23}$ (green), and $B_{13}$
512: (blue), and the behavior of the closure phase, which has substantial
513: variations with rotation angle.  The additional information from the 
514: closure phase added to the conventional intensity measurements from
515: the null and bright outputs will improve the detection performance
516: considerably.
517: 
518: \section{Discussion}
519: There are many advantages to this type of architecture, and we
520: summarize a few of them in this section.  One advantage is due to the
521: non-redundant nature of the baseline configuration.  If the telescopes
522: are movable, i.e., on free-flying satellites, the baseline spacing
523: can be varied according to the distance and expected characteristics
524: of the desired sources.  With one short and two longer baselines
525: it should be possible to discriminate with more certainty 
526: between contributions from the the star and planets compared with 
527: a non-uniform, asymmetric, or clumpy exozodiacal dust cloud.  This
528: is important because the planets are 
529: point sources, while the exozodiacal cloud itself and clumps 
530: are extended sources, and most
531: likely will be resolved on the longer baselines.  
532: 
533: The characteristic parameters of multiple
534: planets and planetary systems should be more easily deconvolved from
535: the data from this architecture, 
536: because the inner planets will be resolved on shorter baselines 
537: than the outer ones, so the nulled outputs will vary 
538: as a function of baseline rotation angle with much more unique
539: signatures than could be expected from redundant baseline
540: Dual-Chopped Bracewell type architectures.  As a result one would
541: expect to have fewer false positive and false negative detections
542: than in these other systems.  
543: 
544: The leakage of stellar light into the null outputs will also vary 
545: depending on baseline, which could be used 
546: to help reduce contamination of the 
547: planetary spectrum by the stellar spectrum, and hence extract cleaner
548: planetary spectra for all the planets in a planetary system.
549: 
550: This architecture will also allow for more strategies to deal effectively
551: with non-fundamental noise sources, particularly ones such as 
552: pathlength fluctuations,
553: gain variations, and baseline drifts.  For the pathlength fluctuations,
554: Danchi et al. (2003) developed a ``ratio of wavelengths" technique 
555: that allows substantial pathlength fluctuations to be cancelled out of the
556: data, and provides for a cleaner estimation of planet flux as a function
557: of wavelength.  Gain variations in 
558: the system, such as from mispointing will produce intensity fluctuations
559: at the bright outputs, which could be used in a feedback loop to 
560: reduce the variations at the nulled outputs, and thus reduce this type
561: of ``common mode" noise source.  
562: 
563: Given the reduction in unwanted noise sources
564: and the possible relaxation of dimensional tolerances and 
565: precision, it may be possible to reduce substantially the complexity of
566: TPF-I.  In particular, it may be possible to simplify or reduce
567: the need for complex metrology systems that may be necessary to 
568: ensure adequate baseline stability and knowledge.  This is a highly
569: desirable avenue of research as a reduction in complexity of any 
570: space system, increases overall system reliability, and reduces cost and risk
571: for the mission.
572: 
573: In summary, our analysis suggests the TPF-I missions could benefit
574: from the closure-phase nulling interferometer system suggested in
575: this paper.  Much more detailed analysis, beyond the scope of 
576: what is presented here, is necessary to understand the true benefits
577: associated with this new architecture for TPF and Darwin. 
578:  
579: \acknowledgments
580: 
581: This work was supported in part by a grant from NASA Goddard Space Flight Center Director's Discretionary Fund (DDF).
582: 
583: \begin{thebibliography}{}
584: 
585: \bibitem[Angel \& Woolf (1997)]{aw97} Angel, J.R.P., \& Woolf, N.J., 1997
586: \apj, 475, 373.
587: 
588: \bibitem[]{489} Barry, R.K., et al. 2005, Proc. SPIE, 5905, 311.
589: 
590: \bibitem[]{582} Beichman, C.A., Woolf, N., Lindensmith, C. 1999, 
591: [Pasadena: Caltech, Jet Propulsion Laboratory], JPL Publication 99-3.
592: 
593: \bibitem[Bracewell (1978)]{brace78} Bracewell, R.  1978, Nature,
594:     274, 780.
595: 
596: \bibitem[]{588}Bracewell, R., \& McPhee, 1979, Icarus, 38, 136.
597:     
598: \bibitem[]{590} Cornwell, T.J. \& Fomalont, E.B. 1999, in ASP Conf. Series 180,
599: 	187.
600:     
601: \bibitem[]{494} Danchi, W.C., Deming, D., Kuchner, M., \& Seager, S. 2003, \apjl, 597, L57.
602: 
603: \bibitem[]{496} Danchi, W.C., et al. 2004, Proc. SPIE, 5491, 236.
604: 
605: \bibitem[]{498}Hale, D.D., Weiner, J., \& Townes, C.H. 2004, Proc. SPIE,
606: 5491, 490.
607: 
608: \bibitem[]{501} Hyde, T.T., et al. 2004, Proc. SPIE, 5491, 553.
609: 
610: \bibitem[]{503} Goodman, J.W. 1985, in Statistical Optics, [New York: John Wiley \& Sons].
611: 
612: \bibitem[]{505}Jennison, R. 1958, \mnras, 118, 276.
613: 
614: \bibitem[]{507}Leger, A., Mariotti, J.M., Mennesson, B., Ollivier, M.,
615: Puget, J.L., Rouan, D., \& Schneider, J. 1996, Icarus, 12, 249.
616: 
617: \bibitem[]{510}Mennesson, B., \& Mariotti, J.M., 1997, Icarus, 128, 202.
618: 
619: \bibitem[]{611} Pearson, T.J., \& Readhead, A.C.S. 1984, ARAA, 22, 97.
620: 
621: \bibitem[]{613}Perley, R.A. 1999, ASP Conf. Series, 180, 275.
622: 
623: \bibitem[]{512} Ragland, S., Traub, W.A., Berger, J.-P., Millan-Gabet, R.,
624: Monnier, J.D., Pedretti, E., Schloerb, F.P., Carleton, N.P., 
625: Hagenauer, P., et al. 2004, Proc. SPIE, 5491, 1390.
626: 
627: \bibitem[]{516}Thompson, A.R., Moran, J.M., \& Swenson, G.W. 2001, 2nd edition, 
628: [New York: John Wiley \& Sons], p. 73.
629: 
630: \bibitem[]{519}Tuthill, P.G., Monnier, J.D., \& Danchi, W.C. 1999, \nat, 398, 487.
631: 
632: \bibitem[Woolf et al. (1998)]{woolf1998} Woolf, N.J., et al. 1998,
633:     Proc. SPIE, 3350, 683.
634: 
635: \bibitem[]{524} Young, J.S., Baldwin, J.E., Boysen, R.C., Haniff, C.A., Lawson,
636: P.R., Mackay, C.D., Pearson, D., Rogers, J., St.-Jacques, D., Warner, P.J.,
637: Wilson, D.M.A., Wilson, R.W. 2000, \mnras, 315, 635.
638: 
639: \end{thebibliography}
640: 
641: \clearpage
642: \begin{figure}
643: \epsscale{.40}
644: \plotone{f1.eps}
645: \caption{Conceptual block diagram of a simplified nulling interferometer.}
646: \end{figure}
647: 
648: %\clearpage
649: \begin{figure}
650: \epsscale{.40}
651: \plotone{f2.eps}
652: \caption{Binary geometry used in the discussion.}
653: \end{figure}
654: 
655: \clearpage
656: \begin{figure}
657: \epsscale{.50}
658: \plotone{f3a.eps}
659: \plotone{f3b.eps}
660: %\plotone{null-phase-3.eps}
661: \caption{(a) Phase in micro-radians for the
662: Earth-Sun system at 10 pc as a function of baseline length in meters. 
663: (b) Intensity of null output port as a function of baseline length.}
664: \end{figure}
665: 
666: %\begin{figure}
667: %\epsscale{.80}
668: %\plotone{Dual_chopped_bracewell.eps}
669: %\caption{Schematic diagram of a dual-chopped Bracewell interferometer
670: %currently in detailed studies for the TPF mission.}
671: %\end{figure}
672: 
673: \clearpage
674: \begin{figure}
675: \epsscale{.70}
676: \plotone{f4.eps}
677: \caption{Conceptual block diagram of a three-telescope closure-phase nulling interferometer.}
678: \end{figure}
679: 
680: \clearpage
681: \begin{figure}
682: \epsscale{.50}
683: \plotone{f5a.eps}
684: \plotone{f5b.eps}
685: %\plotone{null-phase-3.eps}
686: \caption{(a) Intensities of the three nulled outputs for the
687: Earth-Sun system at 10 pc for a 3 telescope closure phase nulling
688: interferometer with baselines of 5 m (red), 25 m (green), and 
689: 30 m (blue) as a function of baseline rotation angle, for an interferometer
690: whose baseline is rotated in a plane perpendicular to the line of
691: sight to the source.   
692: (b) Closure phase as a function of rotation angle for parameters in (a) 
693: above.}
694: \end{figure}
695: 
696: 
697: 
698: %\begin{figure}
699: %\includegraphics[angle=90,scale=.50]{f2.eps}
700: %\caption{Animation still frame taken from \citet{kim03}.
701: %This figure is also available as an mpeg
702: %animation in the electronic edition of the
703: %{\it Astrophysical Journal}.}
704: %\end{figure}
705: 
706: 
707: \end{document}
708: 
709: