1: \documentclass[12pt,preprint]{aastex}
2:
3: \newcommand{\mycitet}[1]{[\citet{#1}]}
4: \newcommand{\myeq}[1]{Eq. \ref{#1}}
5: \newcommand{\myfig}[1]{Fig. \ref{#1}}
6:
7: %\usepackage{subfigure}
8:
9: \begin{document}
10:
11: \title{Optimal Occulter Design for Finding Extrasolar Planets}
12:
13: \author{Robert J. Vanderbei}
14: \affil{Operations Research and Financial Engineering, Princeton
15: University}
16: \email{rvdb@princeton.edu}
17: \author{Eric Cady}
18: \affil{Mechanical and Aerospace Engineering, Princeton University}
19: \email{ecady@princeton.edu}
20: \and
21: \author{N. Jeremy Kasdin}
22: \affil{Mechanical and Aerospace Engineering, Princeton University}
23: \email{jkasdin@princeton.edu}
24:
25: \begin{abstract}
26: One proposed method for finding terrestrial planets
27: around nearby stars is to use two spacecraft---a telescope and a specially
28: shaped occulter that is specifically designed to prevent all but a
29: tiny fraction of the starlight from diffracting into the telescope.
30: As the cost and observing cadence for such a mission will be driven largely by
31: the separation between the two spacecraft, it is critically important to
32: design an occulter that can meet the observing goals while flying as close to
33: the telescope as possible. In this paper, we explore this tradeoff between
34: separation and occulter diameter. More specifically,
35: we present a method for designing the shape
36: of the outer edge of an occulter that is as small as possible and gives
37: a shadow that is deep enough and large enough for a $4$m telescope
38: to survey the habitable zones of many stars for Earth-like planets.
39: In particular, we show that in order for a $4$m telescope to detect in
40: broadband visible light a planet $0.06$ arcseconds from a star
41: shining $10^{10}$ times brighter
42: than the planet requires a specially-shaped occulter $50$m
43: in diameter positioned about $72,000$ km in front of the telescope.
44: \end{abstract}
45:
46: \keywords{Extrasolar planets, occulter, apodization, petal-shaped mask,
47: Babinet's principle}
48:
49: \section{Introduction} \label{sec:intro}
50:
51: Since the first extrasolar planet was discovered by
52: \cite{May95}, various techniques have been used to
53: infer the presence of large planets, but none have the capability to
54: image Earth-like planets directly. Finding terrestrial planets is
55: difficult because the difference in brightness between the star and
56: the planet is so large; in the visible spectrum, the intensity
57: difference, or contrast, is approximately $10^{10}$ (see \cite{Des02}).
58: Just as importantly, the maximal angular separation between planet and star
59: is on the order of $0.1$ arcseconds for a star 10 parsecs from
60: Earth. The problem becomes one of reducing the intensity of
61: starlight at the planet's location by a factor of $10^{10}$. Several
62: methods (see \cite{Kas03,Kuc02,Van06,Guy03})
63: have been proposed to do this within the telescope
64: by adjusting the point spread function so that there is very little
65: starlight at the location of the planet in the image plane of the
66: telescope. While these techniques have demonstrated the potential
67: to provide the necessary contrast (\cite{Sid06}), they have the
68: intrinsic difficulty that they require an adaptive-optics system
69: within the telescope to correct aberrations in the wavefront (induced largely
70: by imperfect optics), which
71: tend to spill unwanted light into the search area.
72:
73: One solution to this problem is to remove the starlight before it
74: reaches the telescope by using a second spacecraft, an
75: \emph{occulter}, positioned between the telescope and the target
76: star. Such a concept was first proposed by
77: \citet{Spi62}; since then, a number of proposals
78: (see \cite{Cop00,Sch03,Cas06}) have appeared
79: that use occulters to look for planets, both of Jupiter and
80: Earth size.
81:
82: Simple ray optics would suggest that a circular disk occulter of
83: diameter $D$ would be adequate to block all of the starlight from entering a
84: telescope of aperture $D$.
85: %Ray optics predicts that the physical shape of the occulter blocks
86: %the light from the star while allowing the light of a nearby planet
87: %to pass by the edge.
88: %Such a telescope would then be able to see
89: %everything in the system at once, up to the edge of the disk.
90: %Everything behind the disk, including the entire star, would be
91: %stopped by the opaque occulter.
92: Unfortunately, this analysis
93: neglects diffraction, which is a significant factor in propagations
94: involving narrow angles. It was known as early as 1818 that diffraction
95: around objects could produce light in areas that geometric optics
96: would predict to be dark; the most notable example of this is Poisson's
97: spot, which earned Fresnel a prize from the French Academy of
98: Sciences (see \cite{Goo96}). A proper design of an occulter-based
99: mission thus requires careful consideration of diffraction
100: effects.
101:
102: \citet{Spi62} noted that it was sufficient to change the
103: transmission function with radius in order to suppress this central
104: spot. Subsequent papers have suggested some specific mechanisms for
105: accomplishing this. One concept, called the {\em Big Occulting Steerable
106: Satellite (BOSS)} (\cite{Cop00}),
107: is based on a transmissive apodization defined by polynomials.
108: %
109: A more recent entry into this field is the {\em New Worlds Observer} proposed
110: by Webster Cash and funded by NIAC. Originally conceived as a pinhole camera
111: in space (\cite{2004SPIE.5487.1634S}), it was eventually reincarnated as a
112: space-based occulter (\cite{Sim04,Cas06}).
113: There is some hope that an occulter mission, if technically feasible, could
114: overcome the significant challenges that more traditional coronagraphic
115: approaches to planet finding must confront.
116: %The proposed New Worlds Observer mission has generated much excitement
117: %recently in the planet-finding community.
118: The purpose of this paper is to
119: explore the trade-off between inner-working-angle
120: and telescope-occulter separation. We show that an occulter capable of
121: detecting in broad-band visible light
122: an Earth-like planet at 60 mas separation from its parent star will
123: need to be 50m in diameter (tip-to-tip) and fly 72,000 km in front of the
124: telescope.
125: %
126: %Subsequently, \citet{2004SPIE.5487.1634S} proposed building
127: %a large space-based pinhole camera, called the {\em New Worlds Observer},
128: %using a shaped pupil similar to those proposed in \citet{KVSL02,VSK02,VSK03}.
129: %\citet{Sim04}, along with the first and third authors of this paper,
130: %demonstrated that such a pinhole camera would require an
131: %enormous screen in order to mitigate diffraction effects around the screen's
132: %outer edge and that it would be operationally impossible to search for targets
133: %through the pinhole. The suggestion was made to use a shaped
134: %occulter in place of a shaped pupil. \citet{Cas06} adopted this view and
135: %the New Worlds Observer was reincarnated as an occulter based on
136: %hypergaussian functions. While the hypergaussian allows a simple analysis of
137: %the on-axis performance, it does not provide a wide enough shadow and the
138: %contrast degrades rapidly with wavelength.
139: %In this paper, we use
140: %optimization methods to design the best possible occulters according to given
141: %objectives and design constraints.
142:
143: \section{Babinet's principle} \label{sec:prin}
144:
145: An occulter is complementary to a pinhole
146: camera; instead of allowing light only through a small hole, an
147: occulter allows all light except for the light blocked by the occulter which
148: now replaces the small hole. This
149: complementarity allows us to calculate the downstream electric field produced
150: by an occulter using Babinet's principle; that is, the sum of the light
151: passing around the occulter and the light passing through an
152: occulter-shaped hole is a free-space plane wave. The electric field
153: past the occulter is thus given by
154: \begin{equation}\label{oc1}
155: E_{o} = E_{u} -E_{h}
156: \end{equation}
157: where $E_o$ is the field produced by an occulter,
158: $E_u$ is the electric field of an unobstructed plane wave, and
159: $E_h$ is the field produced by a complimentary pinhole.
160: %At the
161: %occulter, $E_{u}(r, \theta) = E_0$, and at the telescope,
162: %$E_{u}(\rho, \phi) = E_0e^{\frac{2 \pi i z}{\lambda}}$, the plane
163: %wave solution of the Helmholtz equation.
164: From the Helmholtz equation it follows that a plane wave having
165: complex amplitude $E_0$ at the plane of the occulter would, if unimpeded by an
166: occulter, be given by
167: $E(\rho,\phi) = E_0 e^{2 \pi i z / \lambda}$ at the telescope's pupil plane,
168: which is located a distance $z$ behind the occulter. Here, and
169: throughout the paper, we use polar coordinates $(\rho, \phi)$ to represent the
170: pupil plane of the telescope. We assume that $\rho = 0$ corresponds to the
171: center of the pupil.
172: %Thus, finding the
173: %propagated field due to the occulter transforms to the easier
174: %problem of finding the field past an aperture.
175:
176: Before we investigate simple shaped occulters, it is instructive to consider a
177: more general setting in which an occulter (or a hole) need not be purely opaque
178: or transparent. Instead, we introduce the possibility for partial
179: attenuation. To this end, we introduce a function $A(r,\theta)$ to denote the
180: attenuation profile for the occulter (we use $r$ and $\theta$ to denote polar
181: coordinates in the plane of the occulter). If $A(r,\theta)$ takes the value $1$
182: somewhere, then no light gets through at that point. On the other hand, if
183: it takes the value zero, then all light gets through. All values in between
184: are allowed.
185: Under circular symmetry, the attenuation profile $A(r,\theta)$ does not depend
186: on $\theta$ and so we can write
187: $A(r)$ for the attenuation profile.
188: Of course, when thinking about transmission through a ``tinted'' hole, the
189: function $A(r)$ represents the level of transmission rather than attenuation.
190: That is, $A(r)$ describes an apodization of the hole.
191: With these assumptions, $E_{h}$ at the occulter plane can be written as
192: \begin{equation} \label{oc13}
193: E_{h}(r) = E_{u}A(r) = E_0A(r) .
194: \end{equation}
195: %
196: Assuming that the function $A(r)$ is zero for $r$ larger than some threshold
197: $R$, the Fresnel integral for the propagation of the field from the hole
198: a distance $z$ can then be written in polar coordinates
199: (\cite{Goo96}) as
200: \begin{equation}\label{oc15}
201: E_{h}(\rho)
202: =
203: E_0
204: \frac{2 \pi }{i \lambda z}
205: e^{\frac{2 \pi i z}{\lambda}}
206: e^{\frac{\pi i \rho^2}{\lambda z}}
207: \int^R_0
208: J_0\left(\frac{2 \pi r \rho}{\lambda z}\right)A(r)
209: e^{\frac{\pi i}{\lambda z}r^2}
210: r dr
211: \end{equation}
212: and the field due to an occulter can be expressed as
213: \begin{equation}\label{oc16}
214: E_{o}(\rho)
215: =
216: E_0e^{\frac{2 \pi i z}{\lambda}}
217: \left(
218: 1
219: -
220: \frac{2 \pi e^{\frac{\pi i \rho^2}{\lambda z}}}{i \lambda z}
221: \int^R_0
222: J_0\left(\frac{2 \pi r \rho}{\lambda z}\right)A(r)
223: e^{\frac{\pi i}{\lambda z}r^2}
224: r dr
225: \right) .
226: \end{equation}
227:
228: \section{Optimal attenuation functions} \label{sec:optim}
229:
230: We find $A(r)$ by minimizing the ``extent'' of $A(r)$ subject to the
231: constraint that the intensity, which is the square of the magnitude of the
232: electric field, in a specified dark region is no more
233: than $10^{-10}$. To be precise, we
234: %
235: %For example, we could choose as our cost function the throughput of
236: %the apodization due to an on-axis plane-wave source:
237: %\begin{eqnarray}
238: % \mathcal{T} &=& \int \int |E_h|^2 dx dy \label{op1} \\
239: % &=& \int^R_0 \int^{2 \pi}_0 |E_h(r)|^2 r dr d\theta \label{op2} \\
240: % &=& |E_0|^2\int^R_0 A(r)^2 2 \pi r dr \label{op3}
241: %\end{eqnarray}
242: %\* NOTE \* First paragraph is mine, second is Jeremy's. Input
243: %requested.
244: %
245: %Maximizing this subject to contrast constraints produces an
246: %apodization with the highest stellar throughput; Babinet's principle
247: %then implies that the light due to the star at the telescope plane
248: %will be minimized.
249: %
250: %Maximizing this would produce an apodization with minimum taper and
251: %thus, when converted to a shaped occulter, petals with minimum
252: %extent (desirable from a manufacturing viewpoint).
253: %
254: %\* end NOTE \*
255: %
256: %The complete optimization problem is thus written:
257: \begin{eqnarray}
258: \mbox{minimize } & \int^R_0 A(r) r dr & \nonumber \\
259: \mbox{subject to }
260: & |E_o(\rho)|^2 \leq 10^{-10}|E_0|^2,
261: & \quad 0 \leq \rho \leq \rho_{\mathrm{max}} \nonumber \\
262: & 0 \leq A(r) \leq 1, & \quad 0 \leq r \leq R.
263: \label{op4}
264: \end{eqnarray}
265: This is an infinite-dimensional, quadratic programming problem,
266: which would produce a shadow from $0$ to $\rho_{\mathrm{max}}$ with
267: $10^{10}$ contrast at the telescope's pupil plane. Unfortunately, it is
268: computationally intractable. To make it solvable, we introduce
269: certain simplifications to reduce it to a finite-dimensional, linear
270: programming problem.
271:
272: First, we rewrite the constraint on $E_o(\rho)$ as:
273: \begin{equation} \label{op5}
274: |E_o(\rho)| \leq 10^{-5}|E_0|
275: \end{equation}
276: Since $E_o(\rho)$ is complex, we can constrain
277: the magnitude of the real and imaginary parts of $E_o(\rho)$ to be
278: less than or equal to $10^{-5} |E_0|/\sqrt{2}$ to get a more
279: conservative, but linear, constraint on $E_o(\rho)$. Finally, we
280: discretize $[0, R]$ and $[0, \rho_{\mathrm{max}}]$ to get a
281: finite-dimensional program.
282:
283: %If the apodization is binary, then $A(r)^2 = A(r)$, and \myeq{op3}
284: %becomes:
285: %\begin{equation} \label{op6}
286: % \mathcal{T} = |E_0|^2\int^R_0 A(r) 2 \pi r dr
287: %\end{equation}
288: %For a binary mask, the integral is the open area of the mask, and so
289: %directly proportional to the throughput. It can serve as an
290: %optimization cost function for a continuous $A(r)$, as well, and is
291: %known as a pseudoarea. \cite{Van03} If we use the pseudoarea as
292: %our cost function in place of the throughput, and approximate the
293: %integrals as Riemann sums, the cost function becomes linear and
294: %finite-dimensional. For the Riemann sum, we can consider dividing
295: %the integral into $N$ subintervals and evaluating the function at
296: %the midpoint, $m$, of each subinterval.
297: %
298: %Let:
299: %\begin{eqnarray}
300: % E'_o &=& E_o e^{-\frac{2 \pi i z}{\lambda}} \label{op7} \\
301: % E'_o(\rho) &=& E_0\left(1-\frac{2 \pi e^{\frac{\pi i \rho^2}{\lambda z}}}{i \lambda
302: % z}\int^R_0 J_0\left(\frac{2 \pi r \rho}{\lambda z}\right)A(r)e^{\frac{\pi i}{\lambda z}r^2} r
303: % dr\right) \label{op8} \\
304: % |E'_o|^2 &=& |E_o|^2 \label{op9}
305: %\end{eqnarray}
306: %Since \myeq{op9} holds, we can use $E'_o(\rho)$ in place of
307: %$E_o(\rho)$ in \myeq{op4}, to simplify the real and imaginary parts
308: %of the constraints. Again, we replace the integrals over $r$ with
309: %the appropriate Riemann sums.
310: %\begin{eqnarray}
311: % \mathfrak{R}(E'_o(\rho)) &=& 1-\frac{2 \pi R}{N \lambda
312: % z}\sum^N_{j=1} J_0\left(\frac{2 \pi m_j \rho}{\lambda z}\right)
313: % A_j\sin{\left(\frac{\pi i}{\lambda z}(m_j^2+\rho^2)\right)} m_j
314: % \label{op10} \\
315: % \mathfrak{I}(E'_o(\rho)) &=& \frac{2 \pi R}{N \lambda
316: % z}\sum^N_{j=1} J_0\left(\frac{2 \pi m_j \rho}{\lambda z}\right)
317: % A_j\cos{\left(\frac{\pi i}{\lambda z}(m_j^2+\rho^2)\right)} m_j
318: % \label{op11}
319: %\end{eqnarray}
320: %
321: %We choose a set $S$ of discrete $\rho$-values in the plane of
322: %the telescope at which we wish the constraints to be met. The
323: %problem in \myeq{op4} is then rewritten as a finite-dimensional,
324: %linear programming problem:
325: %\begin{eqnarray}
326: % \mbox{Maximize } \mathcal{T} = \frac{2 \pi R |E_0|^2}{N} \sum^{N}_{j=1} A_j m_j
327: % \nonumber \\
328: % \mbox{subject to } -\frac{10^{-5}}{\sqrt(2)}|E_0| \leq \mathfrak{R}(E'_o(\rho_i)) \leq \frac{10^{-5}}{\sqrt(2)}|E_0|
329: % \nonumber \\
330: % -\frac{10^{-5}}{\sqrt(2)}|E_0| \leq \mathfrak{I}(E'_o(\rho_i)) \leq \frac{10^{-5}}{\sqrt(2)}|E_0|
331: % \nonumber \\
332: % \rho \in S; \forall \quad j = 1..N, \quad 0 \leq A_j \leq 1
333: % \label{op12}
334: %\end{eqnarray}
335: %and the $A_j$ can be interpolated to produce a smooth A(r).
336: %This linear program can solved quickly using a variety of commercially
337: % available optimization tools; we use LOQO \cite{Van94}.
338:
339: As formulated, this optimization model produces the desired shadow only
340: at a single selected wavelength. For such an optimization model,
341: the ``optimal'' function $A(r)$ turns out to take on only two values: zero and
342: one.
343: In other words, the solution is a {\em concentric ring mask} (see
344: \cite{VSK02}). Such a solution
345: achieves the desired contrast at the specified wavelength, but its
346: performance degrades quickly as one moves either to longer or
347: shorter wavelengths. To find a design that works over a broad band of
348: wavelengths, we make a few simple
349: changes to our optimization model. Obviously, the first change is to
350: stipulate that the function $A(r)$ provide a
351: dark shadow at multiple wavelengths. Hence, the contrast
352: constraints are repeated for a discrete set of wavelengths that span the set
353: of wavelengths for which a shadow is
354: desired. Of course then one needs to worry about the gaps between the chosen
355: discrete set of wavelengths. One possibility is simply to require the shadow
356: to be darker than necessary at the chosen wavelengths with the thought that
357: the performance can't degrade with arbitrary abruptness as one moves to
358: intermediate wavelengths. But, a better solution is to impose smoothness
359: constraints on the function $A(r)$. If this function is smooth, then one
360: expects the shadow to remain deep longer as one moves away from the specified
361: wavelengths. A simple way to impose smoothness is to place a bound on
362: the magnitude of the second derivative
363: \[
364: -\sigma \le A''(r) \le \sigma , \qquad 0 \le r \le R .
365: \]
366: Such constraints help, but
367: it turns out that the best thing to do is to let $\sigma$ be an optimization
368: variable and minimize this bound on the smoothness:
369: \begin{eqnarray}
370: \mbox{minimize } & \sigma & \nonumber \\
371: \mbox{subject to }
372: % & -10^{-5}/\sqrt{2} \le \Re(E_o(\rho)/E_0) \le 10^{-5}/\sqrt{2},
373: % & \quad 0 \leq \rho \leq \rho_{\mathrm{max}} \nonumber \\
374: % & -10^{-5}/\sqrt{2} \le \Im(E_o(\rho)/E_0) \le 10^{-5}/\sqrt{2},
375: % & \quad 0 \leq \rho \leq \rho_{\mathrm{max}} \nonumber \\
376: & -1 \le \Re(E_o(\rho)/(10^{-5}E_0/\sqrt{2})) \le 1,
377: & \quad 0 \leq \rho \leq \rho_{\mathrm{max}} \nonumber \\
378: & -1 \le \Im(E_o(\rho)/(10^{-5}E_0/\sqrt{2})) \le 1,
379: & \quad 0 \leq \rho \leq \rho_{\mathrm{max}} \nonumber \\
380: & -\sigma \le A''(r) \le \sigma , & \quad 0 \leq r \leq R \nonumber \\
381: & 0 \leq A(r) \leq 1, & \quad 0 \leq r \leq R
382: \label{op10}
383: \end{eqnarray}
384: (the original objective of minimizing the integral of $A(r)$ actually has
385: little effect on the problem as long as $R$ is small enough).
386: %
387: Of course, once we introduce a shadow constraint for each
388: of several wavelengths, we
389: have the freedom to let the depth and width of the shadow be wavelength
390: dependent.
391:
392: Practical considerations also provide further constraints. For
393: a realistic binary occulter, the innermost section should be opaque
394: out to some radius $a$ to accommodate the spacecraft. This is
395: expressed as:
396: \begin{equation} \label{op13}
397: A(r) = 1, \quad 0 \leq r \leq a
398: \end{equation}
399: We might also wish to impose the constraint that
400: $A'(r) \leq 0$ as this will ensure that the petal-mask to be described next
401: will have petals that get monotonically narrower as one moves out to the tip.
402: Such petal shapes are probably easier to manufacture.
403: %\leq R_{max}$, with $R_{max}$ being a
404: %value limiting the curvature. These can be converted into discrete
405: %constraints:
406: %\begin{eqnarray}
407: % A'(r) &\approx& (A_{j+1} - A_j)\frac{R}{N} \label{op14} \\
408: % A''(r) &\approx& (A_{j+1} -2A_j + A_{j-1})\frac{R^2}{N^2}
409: % \label{op15}
410: %\end{eqnarray}
411:
412: \section{Adding petals} \label{sec:petal}
413:
414: Unfortunately, it is not currently possible to build an apodized occulter
415: to the required precision. So, instead, we replace the apodized occulter
416: with a binary occulter of a particular shape.
417: For instance, inspired by \citet{Van03}, \citet{Cas06} suggested using
418: an occulter made up of a
419: set of N identical evenly spaced ``petals'' as shown in Figure \ref{fig:1}.
420: These petals are
421: wedges of the circle whose width varies with radius such that the
422: fractional angular extent of the occulter at a given radius
423: is the attenuation profile $A(r)$.
424: %We convert to a
425: %petal occulter via the mapping $A(r) = \frac{N}{2 \pi}\alpha(r)$,
426: %where $\alpha(r)$ is the width of a petal as a function of radius,
427: %using Eq. \ref{pet9} to determine the needed number of petals $N$.
428: Except for Babinet's principle,
429: this petal-shaped occulter is identical to the starshaped pupil masks
430: described in \citet{Van03}. The resulting propagated field for
431: such an occulter is thus found via the same procedure using the
432: Jacobi-Anger expansion. The result is
433: \begin{eqnarray}
434: E_{o\mathrm{,petal}}(\rho, \phi) &=& E_{o\mathrm{,apod}}(\rho) \nonumber \\
435: &&- E_0 e^{\frac{2 \pi i z}{\lambda}}\sum^{\infty}_{j = 1}\frac{2 \pi (-1)^j}{i \lambda z}
436: \left(\int^R_0 e^{\frac{\pi i}{\lambda
437: z}(r^2+\rho^2)} J_{j N}\left(\frac{2 \pi r \rho}{\lambda
438: z}\right) \frac{\sin{(j \pi A(r))}}{j \pi} r dr\right) \nonumber \\
439: &&\qquad \qquad
440: \times \left(2\cos{(j N (\phi-\pi/2))}\right)\label{pet9}
441: \end{eqnarray}
442: where $E_{o\mathrm{,apod}}(\rho,z)$ is the field from the smooth
443: apodization and N is the number of petals (assumed even). For large $N$,
444: all of the $J_{jN}$ ($j > 0$) become small exponentially fast
445: near the center of the telescope and so the field approaches
446: that of the smooth apodization as $N$ increases.
447:
448: \section{Results} \label{sec:results}
449:
450: One consideration that must be taken into account when designing
451: optimized occulters is angular size of the shade. As mentioned in
452: Sec. \ref{sec:intro}, the maximum angular separation between
453: Earth-like planets and their stars is 0.1 arcsecond for a star 10
454: parsecs distant. The angular size of the shade is $R/z$. For
455: example, for a $25$ m radius shade, the shade must be at least
456: $51600$ km distant. If we want to see planets at smaller angular separations,
457: i.e., further from Earth, the shade must be shrunk or the distance
458: increased. We present a series of shades optimized with different
459: sizes and at different distances.
460:
461: Radially-symmetric apodizations were created to provide $10^{10}$
462: contrast out to a given radius for four separate occulter profiles:
463:
464: $\cdot$ 18m occulter, 18000 km distance, 3m shadow radius
465:
466: $\cdot$ 20m occulter, 36000 km distance, 3m shadow radius
467:
468: $\cdot$ 25m occulter, 72000 km distance, 2.5m shadow radius
469:
470: $\cdot$ 30m occulter, 100000 km distance, 2.5m shadow radius
471:
472: These occulters were designed to provide the specified contrast over
473: a band from $400$nm to $1100$nm with contrast constraints specified in $100$nm
474: increments across this band. Radial profiles of the shadow
475: at the telescope are shown in \myfig{fig:2}. The profiles are shown for three
476: wavelengths: $400$nm, $750$nm, and $1100$nm. Note that these wavelengths
477: correspond to the shortest and longest wavelengths at which high contrast was
478: dictated as well as an intermediate wavelength which happens to fall midway
479: between the two nearest wavelengths at which high contrast was constrained
480: ($700$nm and $800$nm).
481:
482: Once a profile is created by optimization, we use \myeq{pet9} to
483: calculate the effect of converting a smooth apodization to petals;
484: this petalization tends to reduce the width of the shadow at certain
485: angles. In a forthcoming paper, we will present a method of
486: optimizing the petal shape directly, to prevent this degradation. Each
487: of the four occulters was converted to a binary occulter with $16$
488: petals; the performance of these occulters at $400$nm, $700$nm, and $1100$nm
489: is shown in \myfig{fig:3}.
490:
491: Finally, some may suggest that it is overly conservative to insist on
492: $10^{10}$ contrast at the telescope's pupil plane since additional contrast is
493: generated by the telescope itself as it forms an image. The residual
494: starlight, being roughly flat across the telescope's pupil, forms something
495: similar to an Airy pattern in the image plane. The planet will be slightly
496: off-axis and therefore offset slightly from the on-axis Airy pattern. Since
497: the first diffraction ring in an Airy pattern is almost two orders of
498: magnitude suppressed relative to its main lobe, one can expect some benefit.
499: To test this, we modified our optimization code to
500: minimize an upper bound on the intensity of the light over a $6$m diameter
501: shadow. We ran tests assuming various separations $z$. The tip radius $R$
502: was fixed so that the a planet appearing at the tip is $0.060$ arcseconds
503: off-axis (i.e., we set $R/z$ radians equal to $60$ milliarcseconds and solved
504: for $R$). The smallest value of $z$ that provides a sufficiently dark hole
505: for the planet to be detectable in the image plane turns out to be $66000$km.
506: For this case, the shadow at the telescope's pupil is slightly brighter than
507: $10^{-8}$ times the unattenuated brightness. In the image plane, a planet
508: at $60$ milliarcseconds has about the same brightness as the residual
509: starlight falling in the same location in the image (a $Q=1$ detection in TPF
510: parlance). Figure \ref{fig:4} shows image plane images for the $66000$km
511: design described here. Also shown in the figure for comparison is the image
512: plane image for the $72000$km design described earlier.
513:
514: \section{Final Remarks} \label{sec:rem}
515:
516: Whenever one uses optimization for engineering design, an important question
517: to address is this: how sensitive is the optimal design
518: to small deviations from the given design scenario?
519: We have already discussed some of our efforts to ensure that our design is
520: robust. Namely, we have discussed the issue of specifying shadow depth at
521: several wavelengths spread across the desired waveband and we have
522: discussed using smoothness of $A(r)$ as a surrogate for solution robustness.
523: Furthermore, we have shown plots that verify the shadow depth at two
524: contrast-specified wavelengths ($400$nm and $1100$nm) as well as at a
525: wavelength at which contrast was not specifically constrained but instead is
526: midway between two such wavelengths. In all three of these cases the depth of
527: the shadow proves to be more than adequate.
528:
529: There are further robustness issues that need to be investigated.
530: For example, how deep will the dark shadow be if the occulter-telescope
531: separation deviates from the design value by a few percent? Also, to what
532: precision do the petals need to be manufactured and then deployed? Finally,
533: how much can the occulter's orientation be tilted relative to the
534: occulter-telescope axis?
535: Regarding the second question,
536: preliminary analyses in which we randomly perturbed $A(r)$ by one part in
537: $100,000$ and recomputed the shadow profiles showed that perturbations at this
538: level do not degrade the depth or size of the shadow. On the other hand,
539: perturbations at the level of one part in $10,000$ do start to affect
540: performance.
541: Anyway, these are just very preliminary results.
542: All of the above questions are important and will be
543: addressed in detail in a forthcoming paper.
544:
545: In this paper we have used optimization techniques to investigate the
546: trade-off between inner working angle and size/distance of the occulter.
547: For terrestrial planet finding, it seems that the inner working angle should
548: be no more than 60 mas. The number of Earth-like planets one can hope
549: to find drops quickly as one moves to larger separations. We have shown that,
550: for an inner working angle 60 mas, the occulter needs to be about 50 m
551: tip-to-tip and it must be positioned about 72,000 km in front of the
552: telescope. Future studies should be directed at determining
553: whether such a size and
554: distance combination can be achieved within a reasonable mass and fuel budget.
555:
556: %One thing that should not be separated from occulter edge design is
557: %systems engineering. Occulters close to the telescope take less
558: %fuel to hold in place, but must be smaller than occulters further
559: %out, and in the past these smaller shades have often proved
560: %difficult to optimize without losing science capability in the
561: %process. Shades that are too large would simply blot out a close-in
562: %planet. Shades further away from the telescope can be larger, but
563: %the orbits are more costly to maintain, and limits on fuel can
564: %constrain the number of stars that can be visited. Moreover, larger
565: %occulters may prove more vulnerable to deformation and damage.
566: %
567: %Decreasing the number of petals decreases the quantity of the
568: %equipment required to unfold them, reducing both mass and risk for
569: %the occulter. It also tends to reduce the total perimeter of the
570: %occulter, bringing down the amount of light scattered off the edges.
571: % However, petalizing causes the edges of the shadow to encroach
572: % inward, and the edges move in further as petal number shrinks.
573: % This in turn requires tighter tolerances on attitude and position
574: %for both the shade and telescope, increasing the requirements on the
575: %control systems. For this reason, it is useful to be able to
576: %develop feasible designs at a variety of shade sizes and distances,
577: %in order for the engineering tradeoffs to be accommodated.
578:
579: \acknowledgments{The authors would like to thank R. Lyon for a
580: number of fruitful discussions. We acknowledge support from the
581: Goddard Space Flight Center and Sigma Space Corporation, contract
582: \#NNG06EE69C. R. Vanderbei acknowledges support from the ONR
583: (N00014-05-1-0206).}
584:
585: %\bibliography{refs} %>>>> bibliography data
586: %\bibliographystyle{plainnat}
587:
588: \begin{thebibliography}{16}
589: \providecommand{\natexlab}[1]{#1}
590: \providecommand{\url}[1]{\texttt{#1}}
591: \expandafter\ifx\csname urlstyle\endcsname\relax
592: \providecommand{\doi}[1]{doi: #1}\else
593: \providecommand{\doi}{doi: \begingroup \urlstyle{rm}\Url}\fi
594:
595: \bibitem[Cash(2006)]{Cas06}
596: W.~Cash. 2006,
597: \newblock Detection of earth-like planets around nearby stars using a
598: petal-shaped occulter.
599: \newblock \emph{Nature}, 442, 51
600:
601: \bibitem[Copi and Starkman(2000)]{Cop00}
602: C.J. Copi and G.D. Starkman. 2000,
603: \newblock The \emph{Big Occulting Steerable Satellite} [{B}{O}{S}{S}].
604: \newblock \emph{ApJ}, 532, 581
605:
606: \bibitem[Goodman(1996)]{Goo96}
607: J.W. Goodman. 1996,
608: \newblock \emph{Introduction to Fourier Optics}.
609: \newblock McGraw-Hill
610:
611: \bibitem[Guyon(2003)]{Guy03}
612: O.~Guyon. 2003,
613: \newblock Phase-induced amplitude apodization of telescope pupils for
614: extrasolar terrestrial planet imaging.
615: \newblock \emph{A \& A}, 404, 379
616:
617: \bibitem[Kasdin et~al.(2003)Kasdin, Vanderbei, Spergel, and Littman]{Kas03}
618: N.J. Kasdin, R.J. Vanderbei, D.N. Spergel, and M.G. Littman. 2003,
619: \newblock Extrasolar planet finding via optimal apodized pupil and shaped pupil
620: coronagraphs.
621: \newblock \emph{ApJ}, 582, 1147
622:
623: \bibitem[Kuchner and Traub(2002)]{Kuc02}
624: M.J. Kuchner and W.A. Traub. 2002,
625: \newblock A coronagraph with a band-limited mask for finding terrestrial
626: planets.
627: \newblock \emph{ApJ}, 570, 900
628:
629: \bibitem[Marais et~al.(2002)Marais, Harwit, Jucks, Kasting, Lin, Lunine,
630: Schneider, Seager, Traub, and Woolf]{Des02}
631: D.J.~Des Marais, M.O. Harwit, K.W. Jucks, J.F. Kasting, D.N. Lin, J.I. Lunine,
632: J.~Schneider, S.~Seager, W.A. Traub, and N.J. Woolf. 2002,
633: \newblock Remote sensing of planetary properties and biosignatures on
634: extrasolar terrestrial planets.
635: \newblock \emph{Astrobiology}, 2\penalty0 (2), 153
636:
637: \bibitem[Mayor and Queloz(1995)]{May95}
638: M.~Mayor and D.~Queloz. 1995,
639: \newblock A jupiter-mass companion to a solar-type star.
640: \newblock \emph{Nature}, 378, 355
641:
642: \bibitem[Schultz et~al.(2003)Schultz, Jordan, Kochte, Fraquelli, Bruhweiler,
643: Hollis, Carpenter, Lyon, DiSanti, Miskey, Leitner, Burns, Starin, Rodrigue,
644: Fadali, Skelton, Hart, Hamilton, and Cheng]{Sch03}
645: A.B. Schultz, I.J.E. Jordan, M.~Kochte, D.~Fraquelli, F.~Bruhweiler, J.M.
646: Hollis, K.G. Carpenter, R.G. Lyon, M.~DiSanti, C.~Miskey, J.~Leitner, R.D.
647: Burns, S.R. Starin, M.~Rodrigue, M.S. Fadali, D.~Skelton, H.M. Hart,
648: F.~Hamilton, and K.-P. Cheng. 2003,
649: \newblock {U}{M}{B}{R}{A}{S}: A matched occulter and telescope for imaging
650: extrasolar planets.
651: \newblock In \emph{Proceedings of SPIE--High-Contrast Imaging for Exo-Planet
652: Detection}, volume 4860
653:
654: \bibitem[Sidick et~al.(2006)Sidick, Shi, Basinger, Moody, Lowman, Kuhnert, and
655: Trauger]{Sid06}
656: E.~Sidick, F.~Shi, S.~Basinger, D.~Moody, A.E. Lowman, A.C. Kuhnert, and J.T.
657: Trauger. 2006,
658: \newblock Performance of {T}{P}{F}'s high-contrast imaging testbed: modeling
659: and simulations.
660: \newblock In \emph{Proceedings of SPIE--Space Telescopes and Instrumentation I:
661: Optical, Infrared, and Millimeter}, volume 6265
662:
663: \bibitem[{Simmons} et~al.(2004){Simmons}, {Cash}, {Seager}, {Wilkinson},
664: {Kasdin}, {Vanderbei}, {Chow}, {Gralla}, and
665: {Kleingeld}]{2004SPIE.5487.1634S}
666: W.~L. {Simmons}, W.~C. {Cash}, S.~{Seager}, E.~{Wilkinson}, N.~J. {Kasdin},
667: R.~J. {Vanderbei}, N.~{Chow}, E.~{Gralla}, and J.~{Kleingeld}. 2004,
668: \newblock {The New Worlds Observer: a mission for high-resolution spectroscopy
669: of extra-solar terrestrial planets}.
670: \newblock In \emph{Microwave and Terahertz Photonics. Edited by Stohr, Andreas;
671: Jager, Dieter; Iezekiel, Stavros. Proceedings of the SPIE, Volume 5487, pp.
672: 1634-1645 (2004).}, pages 1634
673: \newblock \doi{10.1117/12.552069}.
674:
675: \bibitem[Simmons(2005)]{Sim04}
676: W.L. Simmons. 2005,
677: \newblock A pinspeck camera for exo-planet spectroscopy.
678: \newblock Technical report, M.S. Thesis, Department of Mechanical and Aerospace
679: Engineering, Princeton University
680:
681: \bibitem[Spitzer(1962)]{Spi62}
682: L.~Spitzer. 1962,
683: \newblock The beginnings and future of space astronomy.
684: \newblock \emph{American Scientist}, 50, 473
685:
686: \bibitem[Vanderbei(2006)]{Van06}
687: R.J. Vanderbei. 2006,
688: \newblock Diffraction analysis of two-dimensional pupil mapping for high
689: contrast imaging.
690: \newblock \emph{ApJ}, 636, 528
691:
692: \bibitem[Vanderbei et~al.(2003{\natexlab{a}})Vanderbei, Spergel, and
693: Kasdin]{Van03}
694: R.J. Vanderbei, D.~Spergel, and N.J. Kasdin. 2003a,
695: \newblock Circularly symmetric apodization via star-shaped masks.
696: \newblock \emph{ApJ}, 599, 686
697:
698: \bibitem[Vanderbei et~al.(2003{\natexlab{b}})Vanderbei, Spergel, and
699: Kasdin]{VSK02}
700: R.J. Vanderbei, D.N. Spergel, and N.J. Kasdin. 2003b,
701: \newblock Spiderweb masks for high contrast imaging.
702: \newblock \emph{ApJ}, 590, 593
703:
704: \end{thebibliography}
705:
706: \clearpage
707:
708: \begin{figure}
709: \begin{center}
710: %\includegraphics[width=3in]{C:/usr/rvdb/public_html/JAVA/tpf/tpf-o/petals/16/AA.eps}
711: \includegraphics[width=3in]{f1.pdf}
712: \end{center}
713: \caption{An optimally-shaped sixteen-petal occulter.} \label{fig:1}
714: \end{figure}
715:
716: \clearpage
717:
718: \begin{figure}
719: \begin{center}
720: %\mbox{\subfigure{\includegraphics[width=3in]{f2a.pdf}} \quad
721: % \subfigure{\includegraphics[width=3in]{f2b.pdf}}}
722: %\mbox{\subfigure{\includegraphics[width=3in]{f2c.pdf}} \quad
723: % \subfigure{\includegraphics[width=3in]{f2d.pdf}}}
724: \includegraphics[width=3in]{f2a.pdf}
725: \includegraphics[width=3in]{f2b.pdf}
726: \includegraphics[width=3in]{f2c.pdf}
727: \includegraphics[width=3in]{f2d.pdf}
728: \end{center}
729: \caption{These plots show the radial profile of
730: the shadow at the telescope, in different wavelengths.
731: {\em Top left.} This plot is for an $18$m occulter at
732: $18000$km. {\em Top right.} This plot is for a $20$m
733: occulter at $36000$km.
734: {\em Bottom left.} This plot is for an $25$m occulter at
735: $72000$km. {\em Bottom right.} This plot is for a $30$m
736: occulter at $100000$km.} \label{fig:2}
737: \end{figure}
738:
739: \clearpage
740:
741: \begin{figure}
742: \begin{center}
743: %\mbox{\subfigure{\includegraphics[width=1.6in]{f3a.pdf}}
744: % \subfigure{\includegraphics[width=1.6in]{f3b.pdf}}
745: % \subfigure{\includegraphics[width=1.6in]{f3c.pdf}}
746: % \subfigure{\includegraphics[width=1.6in]{f3d.pdf}}}
747: %\mbox{\subfigure{\includegraphics[width=1.6in]{f3e.pdf}}
748: % \subfigure{\includegraphics[width=1.6in]{f3f.pdf}}
749: % \subfigure{\includegraphics[width=1.6in]{f3g.pdf}}
750: % \subfigure{\includegraphics[width=1.6in]{f3h.pdf}}}
751: \includegraphics[width=1.6in]{f3a.pdf} \hspace*{-0.2in}
752: \includegraphics[width=1.6in]{f3b.pdf} \hspace*{-0.2in}
753: \includegraphics[width=1.6in]{f3c.pdf} \hspace*{-0.2in}
754: \includegraphics[width=1.6in]{f3d.pdf} \\
755: \includegraphics[width=1.6in]{f3e.pdf} \hspace*{-0.2in}
756: \includegraphics[width=1.6in]{f3f.pdf} \hspace*{-0.2in}
757: \includegraphics[width=1.6in]{f3g.pdf} \hspace*{-0.2in}
758: \includegraphics[width=1.6in]{f3h.pdf}
759: \end{center}
760: \caption{
761: The shadow cast at the telescope pupil plane for four different occulter
762: distances, which are from left to right
763: $18,000$km, $36,000$km, $72,000$km, and $100,000$km.
764: The top row shows linear stretch plots whereas the bottom row shows
765: logarthmic stretches with $10^{-10}$ set to black.
766: These are RGB images composited using $\lambda =
767: 1.0\mu$m for the red channel, $\lambda = 0.7\mu$m for the green
768: channel, and $0.4\mu$m for the blue channel.
769: } \label{fig:3}
770: \end{figure}
771:
772: %\begin{figure}
773: %\begin{center}
774: %\mbox{\subfigure{\includegraphics[width=3in]{pet18at11.pdf}} \quad
775: %\subfigure{\includegraphics[width=3in]{pet36at11.pdf}}}
776: %\mbox{\subfigure{\includegraphics[width=3in]{pet72at11.pdf}} \quad
777: %\subfigure{\includegraphics[width=3in]{pet100at11.pdf}}}
778: %\end{center}
779: % \caption{These plots are the results of petalizing each of the
780: % apodizations used to create \myfig{fig:1} with 16 petals.
781: % Each displays the response at $1100$nm.
782: % {\em Top left.} This plot is for an $18$m occulter at
783: % $18000$km. {\em Top right.} This plot is for a $20$m
784: % occulter at $36000$km.
785: % {\em Bottom left.} This plot is for an $25$m occulter at
786: % $72000$km. {\em Bottom right.} This plot is for a $30$m
787: % occulter at $100000$km.} \label{fig:3}
788: %\end{figure}
789:
790: \clearpage
791:
792: \begin{figure}
793: \begin{center}
794: %\mbox{\subfigure{\includegraphics[width=3in]{f4a.pdf}}\quad
795: % \subfigure{\includegraphics[width=3in]{f4b.pdf}}}
796: \includegraphics[width=3in]{f4a.pdf}
797: \includegraphics[width=3in]{f4b.pdf}
798: \end{center}
799: \caption{These plots are simulated (noiseless) images at
800: the telescope's image plane.
801: The RGB images were computed using $\lambda =
802: 1.0\mu$m for the red channel, $\lambda = 0.7\mu$m for the green
803: channel, and $0.4\mu$m for the blue channel.
804: In both images, the off-axis planet is positioned at the tip of the the
805: occulter.
806: {\em Left.} This image is for an $R=22$m occulter at $66000$km.
807: The planet shown here is at $60$ milliarcseconds.
808: {\em Right.} This plot is for an $R=25$m occulter at $72000$km.
809: The planet shown here is at $72$ milliarcseconds.
810: } \label{fig:4}
811: \end{figure}
812:
813: \end{document}
814: