Professional Documents
Culture Documents
Mirror Analysis - 1
Mirror Analysis - 1
Design Development of a
20-Metre Segmented Mirror with
Lightweighted Silicon
Carbide Segments
w Revision 0 w
By
AMEC Dynamic Structures Ltd.
Tables
Table 1: Properties of SiC................................................................................................................... 6
Table 2: Calculation using Couder Theory........................................................................................... 10
Table 3: Properties of SiC used in FE Analysis................................................................................... 11
Table 4: Results Summary ............................................................................................................... 23
Table 5: Actuator Performance......................................................................................................... 25
Michael Gedig
AMEC Dynamic Structures, Ltd.
Port Coquitlam, B.C. Canada
Introduction
Astronomers around the world are now considering next generation large optical-infrared telescopes in the
15 to 30-Meter range that will follow the current generation of 8 to 10-Meter telescopes. Mirror design and
fabrication is one important issue in scaling the dimensions of existing telescopes. Monolithic primary
mirrors made of a single piece of glass have many advantages. The surfaces of such mirrors can be made
smooth and continuous and can be polished using a procedure which takes advantage of the axisymmetric
properties of the mirror. Scaling the size of a monolithic primary mirror becomes extremely costly, and at
some point, virtually impossible. An important consideration in the scaling of telescope dimensions is that
the optical goals are set by the wavelength of light, therefore larger mirrors must have similar surface
accuracy to smaller ones. A consequence of this required surface accuracy is that problems of mirror
support become increasingly difficult [4].
The concept of constructing a large mirror from segments rather than a single monolithic blank has been
used for many years in the fabrication of large radio and millimeter wave telescopes. There have been
considerably fewer segmented optical telescopes, primarily because of the more stringent tolerances
required in the optical regime. The 10 metre telescopes of the Keck Observatory are an example of large,
segmented mirror optical telescopes. Each 10 metre telescope has a primary mirror with 36 hexagonal
segments. For a segmented 20-Metre primary mirror, approximately 350 segments of diameter 1 metre
would be required. This number of segments represents a large increase in complexity considering existing
segmented-mirror telescopes. By sheer virtue of the multiplicity involved, a 20-Metre segmented telescope is
expected to cost much more than existing telescopes.
It is recognized that a significant portion of the cost of the Keck telescopes lay in the mirror support system.
Each Keck segment is supported by a complex whiffle tree arrangement fabricated to precise tolerances. A
simpler mirror support arrangement will lead to significant potential cost savings in a next generation
telescope. This report summarizes research by AMEC Dynamic Structures into different mirror segment
support configurations and mirror materials.
It is well known that the deformation patterns of mirrors subject to gravity loads can be improved by the
process of lightweighting, where material is removed from parts of the mirror. This process has been used for
Zerodur mirrors, where material is removed by machining. Due to the hardness of Zerodur, the machining
process is relatively expensive. Using lightweighted Zerodur for a 20-Metre class primary would lead to large
mirror segment costs. Silicon carbide (SiC), is a potential mirror segment material that offers some promise
in the fabrication of lightweight mirror segments for large segmented telescopes.
SiC is significantly stiffer than mirror materials such as Zerodur. The high stiffness of SiC may allow for a
simpler passive support system than would be needed for more classical materials as Zerodur. SiC also has
favourable thermal properties which should allow for better thermal control of the primary mirror while using a
simpler active temperature control system.
This report summarizes research and development work that will lead to design options for a 20-Metre
segmented mirror using lightweighted hexagonal SiC segments. The work covers the following areas of
design:
• a passive support system
PROJECT XLT LARGE TELESCOPE
TITLE LIGHTWEIGHT SILICON CARBIDE SEGMENT REV. 0
DOC. NAME XLT-MIRSIC DATE/TIME 30-Oct-02 9:28 AM
FILE NAME xltmirsic0.doc CHECKED SHEET 5
• an active support system (6-degrees of freedom)
• mirror thermal behaviour
• mirror cell design
A primary goal of this work is to develop preliminary SiC mirror segment designs compatible with a simple
three-point support system. In addition, the work addresses the relationship between mirror segment size
and mirror support requirements. Several different mirror configurations are investigated and compared, and
preliminary designs are proposed for both a primary mirror cell configuration, as well as for the active and
passive supporting elements that connect the mirror segments to the cell.
1. Silicon Carbide
Silicon Carbide has several advantages as a primary mirror substrate material. SiC is isotropic,
homogeneous, and tests indicate that it is dimensionally stable when subjected to repeated temperature
variations. SiC has a relatively high strength to weight ratio compared to other viable mirror materials (see
Table 1). A major advantage of SiC is that it can be formed into complex shapes, allowing for optimal
material usage. SiC is also useful as a mirror substrate because some forms of SiC can be polished to a
high quality mirror finish using diamond grit.
There are several different manufacturing processes for producing SiC. The processes include CVD
(Chemical Vapour Deposition), Reaction Bonded SiC, Sintered SiC, Hot Pressed SiC, Foamed SiC, and
C/SiC.
Figure 1 shows a 630mm diameter parabolic lightweight mirror for a non-astronomical application, the ATLID
atmospheric lidar. This mirror is coated with a 150 micron SiC CVD layer, and polished to a surface
roughness of less than 20nm peak to valley. The mass of the finished mirror is approximately 6kg.
Astrium SAS, Toulouse France, has been contracted by the European Space Agency (ESA) to
manufacture a 3.5 metre SiC primary mirror for the Herschel Space Telescope, to be launched in 2007 [1].
As part of the development of the Herschel telescope, a 1.35 metre prototype mirror was fabricated using
the Sintered SiC process. In this process, Silicon Carbide powder is subjected to a very high pressure and
temperature. To make the 1.35 metre prototype, nine petals were made separately and joined using a
brazing process to form a monolithic mirror (Figure 2). The mirror was then machined, ground and polished,
then coated with a thin layer of metal to create the reflective surface. The 3.5 metre mirror is expected to
weigh about 240kg, and will have a thickness of 200mm. Similar to the prototype, the 3.5 metre mirror will
consist of 12 petals that have been joined by brazing. The petals each measure roughly 900 by 1600mm.
Three different sizes of composite mirrors were made: 100mm (4”), 250mm (10”), and 450mm (18”).
According to Ultramet,
“Based on current results further scale up to 1-m (39.37-in) diameter is expected to be reasonably
straightforward in commercial application. The process allows extremely stiff, lightweight, and
directly contoured optical structures to be fabricated quickly and efficiently from abundant,
inexpensive, and non-strategic raw materials.”
SUPPORT a
RING
b
The Couder Law equations were programmed using a spreadsheet (see Table 2). The variation of deflection
w with support radius b was calculated for a number of different discrete radii b. It was found that deflection
was minimized with b = 0.678a. Using material properties ρ = 3140kg/m3, E = 420GPa and ν = 0.25, a
cylindrical SiC blank of radius a = 500mm is found to have minimum deflection with b = 0.578a = 339mm.
For this mirror and support arrangement, it can be shown that a mirror thickness t = 78mm is required to
achieve peak-to-valley deflections of 20nm (see Table 2). A mirror with the dimensions given in Table 2
weighs about 192kg.
DESCRIPTION
-Determine the deflections of a circular mirror supported on a ring
-Material assumed to be Silicon Carbide
PROJECT XLT LARGE TELESCOPE
TITLE LIGHTWEIGHT SILICON CARBIDE SEGMENT REV. 0
DOC. NAME XLT-MIRSIC DATE/TIME 30-Oct-02 9:28 AM
FILE NAME xltmirsic0.doc CHECKED SHEET 9
-reference Reflecting Telescope Optics II, Wilson, pp.242-247
INPUT
MATERIAL
-density rho = 3140 [kg/m^3]
-Poisson's ratio nu = 0.25 []
-Young's modulus E = 4.20E+11 [Pa]
GEOMETRY
-mirror radius a = 0.500 [m]
-support ring radius b = 0.339 [m]
-mirror thickness t = 0.078 [m]
CONSTANTS
-acceleration due to gravity g = 9.81 [m/s^2]
EVALUATION POINT
-radius to evaluation point r = 0.300 [m]
CALCULATIONS
-dimensionless zonal radius rze = r/a = 0.600 []
-dimensionless annular rae = b/a = 0.678 []
radius
-Couder parameters V = rze^4-(6+2*nu)/(1+nu)*rze^2+(5+nu)/(1+nu) = 2.45760 []
Uo = 2*(1-rze^2)+(1-nu)/(1+nu)*(1-rze^2)*(1- = 0.65005 []
rae^2)+2*(rze^2+rae^2)*ln(rze)
Ui = 2*(1-rae^2)+(1-nu)/(1+nu)*(1-rze^2)*(1- = 0.65104 []
rae^2)+2*(rze^2+rae^2)*ln(rae)
-deflection at r w = -3*g*rho/(4*E)*(1-nu^2)*(V/4- = 1.94E-08 [m]
if(r<=b,Ui,Uo))*a^4/t^2
wm = w*1000000 = 0.02 [µm]
wn = wm*1000 = 19 [nm]
-Couder parameters (r=0) Vo = (5+nu)/(1+nu) = 4.20000 []
-deflection at centre (r=0) Uio = 2*(1-rae^2)+(1-nu)/(1+nu)*(1- = 1.04755 []
rae^2)+2*rae^2*ln(rae)
-deflection at centre (r=0) wo = -3*g*rho/(4*E)*(1-nu^2)*(Vo/4-Uio)*a^4/t^2 = -1.30E-09 [m]
wmo = wo*1000000 = 0.00 [µm]
wno = wmo*1000 = -1 [nm]
-Couder parameters (r=b) Vb = rae^4-(6+2*nu)/(1+nu)*rae^2+(5+nu)/(1+nu) = 2.02095 []
-deflection at support (r=b) Uib = 2*(1-rae^2)+(1-nu)/(1+nu)*(1-rae^2)*(1- = 0.54125 []
rae^2)+2*(rae^2+rae^2)*ln(rae)
-deflection at support (r=b) wb = -3*g*rho/(4*E)*(1-nu^2)*(Vb/4-Uib)*a^4/t^2 = 1.91E-08 [m]
wmb = wb*1000000 = 0.02 [µm]
wnb = wmb*1000 = 19 [nm]
-Couder parameters (r=a) Va = 1-(6+2*nu)/(1+nu)+(5+nu)/(1+nu) = 0.00000 []
-deflection at edge (r=a) wa = -3*g*rho/(4*E)*(1-nu^2)*(Va/4)*a^4/t^2 = 0.00E+00 [m]
wma = wa*1000000 = 0.00 [µm]
wna = wma*1000 = 0 [nm]
-deflection range wr = max(wno,wnb,wna)-min(wno,wnb,wna) = 20 [nm]
Geometrically, the face of the mirror is assumed to be planar. The face of the actual mirror segments is
expected to form part of the surface of a hyperboloid. This assumption is warranted because axial
deformations of the mirror are assumed to be the most critical deflection component.
Three dimensional elastic shell elements with both bending and membrane capabilities are used to model
the mirror segment. The element has either three or four nodes with six degrees of freedom at each node.
The element is classified as a thin shell element, which means that the shell thickness is assumed to be
zero in constructing the model geometry. Shell thickness is only considered insofar as it affects shell mass
and stiffness. Shear deflection is not included in the particular element formulation used for the analysis
described in this report.
The material properties assumed in the finite element analysis are detailed in Table 3. Further assumptions
regarding the finite element model boundary conditions are given in the sections describing each particular
model.
Design optimization was performed on the finite element models described in this report. The finite element
program ANSYS V5.7 was used for the analysis. This program has built in first-order optimization
procedures. The general approach was to define the mirror model in terms of parameters. Typical
parameters are mirror size, support location and mirror thickness. The optimization procedure performs
repeated finite element analysis runs, automatically modifying the model parameters based on the results of
each run until some predefined criterion is satisfied. The approach for mirror analysis generally was to
attempt to minimize the weight of the mirror while satisfying some goal for the amount of mirror deflection
permitted.
SUPPORT Ø1000
RING
rs
2.3.1 Description
In order to verify the operation of the finite element solution procedure, the circular plate mirror used in the
Couder Law calculations was modeled (Figure 5).
There are many different segmentation patterns that may be considered for a large mirror. In the past,
hexagonal arrays and annular rings of wedges have been the most common. Figure 6 shows a proposed
hexagonal array of 1-metre hexagonal segments for a 20-metre mirror.
11547
Ø20000
19998
A simple hexagonal mirror segment of uniform thickness supported at 3 discrete points is shown in Figure
7. This model was created in order to set an upper bound for the weight of any lightweighted hexagonal
mirror segment of a given size. The boundary conditions at the support points have the effect of supporting
the mirror in the axial direction only. The centre point of the mirror was further constrained against radial
translation and rotation about the mirror axis in order to prevent rigid body translation.
2.3.2 Results
The Couder model of Figure 5 was subjected to a gravity load in the axial direction. The resulting peak-to-
valley (p-v) deflection was 19nm (Figure 8), found at an optimal support ring radius of 339mm (b = 0.678a).
The peak-to-valley deflection was used to evaluate the model during each optimization iteration.
2.4.1 Description
The lightweighted hexagonal mirror with radial and straight circumferential ribs shown in Figure 10 was
modeled using finite elements. The geometric parameters controlling the model were the support radius
along with the overall segment depth at the centre (1 parameter), edges (3 parameters), support points (1
parameter) and at intermediate points (2 parameters). In addition, the face sheet thickness was specified as
a separate parameter. The thickness of each rib was specified as a dependent function of the overall mirror
depth at the midpoint of the rib. The rib thickness was set to 1/10th of the overall mirror depth based on
several trials using different rib aspect ratios.
The root mean square (RMS) deflection was used to evaluate the model during each optimization iteration.
The procedure used to calculate RMS deflection used a command in ANSYS which allows an arbitrary path
to be specified on a model. In this case, the face of the mirror was selected, and a series of rectangular
paths was superimposed on the face based on a predefined grid spacing. At mirror edges the rectangular
paths were adjusted as required to follow the edge. As a result, the mirror face was fully covered with a
pattern of touching tiled areas. The finite element program calculates the average deflection along the path,
which is used in the RMS calculation along with a weighting factor based on the area enclosed by the path.
The boundary conditions applied at the support points have the effect of supporting the mirror in the axial
direction only. The centre point of the mirror was further constrained against radial translation and rotation
about the mirror axis in order to prevent rigid body translation.
2.5.1 Description
The lightweighted mirror with hexagonal cells shown in Figure 12 was modeled using finite elements. The
geometric parameters controlling the model were the support radius along with the overall segment depth at
the centre (1 parameter), edges (3 parameters), support points (1 parameter) and at intermediate points (2
parameters). In addition, the face sheet thickness and rib aspect ratio (depth : width) were specified as
separate parameters. Optimization was conducted using RMS deflection as an evaluation function.
The boundary conditions applied at the support points have the effect of supporting the mirror in the axial
direction only. The centre point of the mirror was further constrained against radial translation and rotation
about the mirror axis in order to prevent rigid body translation. This model was subjected only to gravity load
in the axial direction.
2.5.2 Results
Figure 12 shows the shape resulting from optimization of the SIC-2 parametric model subject to gravity load
in the axial direction. The mirror segment shown in Figure 12 weighs 118kg, resulting in an areal density of
182kg/m2 (118/0.65). The RMS deflection was found to be 11nm, and the peak-to-valley deflection 68nm
(Figure 13). Rib aspect ratio was optimized during the analysis process, which led to a ribs with a depth to
thickness ratio of about 8.9.
2.6.1 Description
The lightweighted mirror with diagonal ribs intersecting support points (Figure 14) was modeled using finite
elements. The geometric parameters controlling the model were the support radius along with the overall
segment depth at the edges (2 parameters) and support points (1 parameter). In addition, the face sheet
thickness and rib aspect ratio were specified as separate parameters. Optimization was conducted using
RMS deflection as an evaluation function.
2.6.2 Results
Figure 14 shows the shape resulting from optimization of the SIC-3 parametric model subject to gravity load
in the axial direction. The RMS deflection was found to be 14nm, and the peak-to-valley deflection 78nm
(Figure 15). The mirror segment shown in Figure 14 weighs 68kg, resulting in an areal density of 105kg/m2
(68kg/0.65m2). In comparison, the Keck mirror segments (900mm edge length) have areal density 190kg/m2
(400kg/2.1m2). The weight of SIC-3 is similar to the proposed 1-metre mirror segments for the GSMT 30-
metre telescope [10], which have weigh 74kg and have areal density 114kg/m2.
The SIC-3 model was analyzed for varying degrees of tilt, where the mirror axial ranged from 0 to 90 degrees
above horizon. Several different sets of boundary conditions were tried during the tilted mirror analysis. The
behaviour of the mirror was found to be very sensitive to the types of boundary conditions applied at the
support points.
An alternative set of boundary conditions used with the SIC-3 model restrained only three points on the
mirror; lateral forces were taken out through the 3 axial support points. This set of boundary conditions
resulted in relatively large deflections in the tilted mirror analysis. A common feature of the lightweighting
patterns studied here is that the overall mirror segment depth is large compared to customary glass mirrors.
In addition, the lightweight mirrors have open backs which cause the centre of gravity to be located close to
the face of the mirror. As a result of the eccentricity between the gravity vector applied at the centre of
gravity and the reaction vectors at the support points, considerable deflections were generated.
2.7.1 Description
The impact of thermal gradients on the geometry of the individual mirror segments was studied using finite
element analysis. The 1-metre triangular rib segment (SIC-3) was used in the analysis. A linear temperature
gradient from the front face to the back surface of the mirror was imposed on the model. As shown in Figure
17, the mirror segment temperature was assumed to be zero degrees Celsius at the back (t back= 0.0) face
and 0.1 degrees at the front face (t front= 0.1).
2.7.2 Results
Figure 18 shows the shape resulting from optimization of the SIC-3 parametric model subject to a 0.1ºC
temperature gradient load. The RMS deflection was found to be 31nm, and the peak-to-valley deflection
121nm. This result shows that the mirror segment is extremely sensitive to temperature gradients. A 0.1ºC
gradient causes axial deformations in the same order of magnitude as gravity. Further thermal studies are
required to understand how well the temperature of the mirror segments can be controlled. Silicon Carbide
has relatively high thermal conductivity compared to other mirror materials, and the lightweighting process
creates a large surface to volume ratio for the mirror segment. These two factors help to decrease the
magnitude of potential thermal gradients.
A 2-metre version of the triangular rib model SIC-3 was created in order to study the relationship between
mirror size, mirror weight, and deflections.
2.8.1 Description
The lightweighted mirror with diagonal ribs intersecting support points (Figure 19) was modeled using finite
elements. The geometric parameters controlling the model were the support radius along with the overall
segment depth at the edges (2 parameters) and support points (1 parameter). In addition, the face sheet
thickness and rib aspect ratio were specified as separate parameters. Optimization was conducted using
RMS deflection as an evaluation function.
2.8.2 Results
Figure 20 shows the shape resulting from optimization of the SIC-4 parametric model subject to gravity load
The studies presented here suggest that smaller mirrors in the range of 1-metre size are preferable to larger
mirrors because of lower weight and better thermal performance. Of course, the costs of fabrication,
handling and shipping are also important factors in the choice of mirror segment size, and must be
considered in any decision.
Three active supports are provided, one at each axial support point. These three supports allow for active
control of mirror segment tip, tilt, and piston degrees of freedom. A small actuator may be required to
actively control rotation about the mirror axis. This system does not allow for in-plane active control of the
mirror segments. The mirror segments would be expected to be accurately located on the mirror cell using a
rigid, adjustable base frame structure. The mirror cell is envisioned to be a relatively rigid steel monocoque
structure with predictable thermal behaviour. It is believed that active control of the tip, tilt and piston
degrees of freedom should be able to compensate for in-plane positional variations of the segments, but
further optical and structural studies are required before this can be proven.
Coupling between the active and passive support elements and the mirror segment is of critical importance
in the design of optical telescopes. The efforts described in this report have been concentrated on the design
of a SiC mirror segment; the detail design of the support elements and their connection to the mirror has not
been studied in detail. Some innovative work in mirror support details has been done by Jerry Nelson and
the people responsible for the design of the existing Keck telescopes. Recent efforts by Nelson in the
design of a mirror support system for the 30-metre California Extremely Large Telescope (CELT) [10].
Although the concepts proposed for the CELT mirror segment supports use more conventional whiffle tree
supports, several design details are compatible with the 3-point supported SiC segments proposed in this
report. For example, piano wire flexures are suggested for axial supports, resulting in large axial stiffness
with small lateral stiffness. In addition, a central support is suggested for the lateral support system.
centre spindle
base frame actuator mirror segment
mirror cell
Description Performance
Axial Stiffness 12N/µm or higher
Lateral Stiffness 1.2N/µm or higher
Total Stroke Range 10mm
Position Resolution 5nm or better
Response Time 0.2 sec at a 50 nm step increment or faster
Total Mass 5 kg or less
The weight of the mirror cell is strongly influenced by the weight of the mirror and mirror support systems.
One goal of the research described here is to compare the weight of a SiC mirror system with a conventional
Zerodur mirror system. The Keck mirror segments have an edge length of 900mm, resulting in a segment
area of 2.1m2. The Keck mirrors weigh about 400kg; the associated whiffle trees and sensors weigh an
additional 50kg. The areal density of the Keck mirror system is therefore 450kg/2.1m2 = 214kg/m2. In
comparison the 1-metre SiC segment SIC-3 has area 0.650m2 and weighs 68kg, leading to an areal density
of 105kg/m2. The 2-metre SiC segment SIC-4 is considerably heavier, weighing 400kg/m2. The 1-metre
segments proposed for CELT weigh 74kg, and the axial support system weighs an additional 9kg (for
comparison purposes, the weight of the lateral support system is excluded), giving a total of 83kg for an
area of 0.650m2. This works out to an areal density of 128kg/m2. These rough comparisons indicate that
there is no significant weight saving in using SiC mirrors as opposed to conventional Zerodur.
5. Conclusions
Silicon Carbide has several advantages as a primary mirror substrate material. Silicon Carbide is isotropic,
homogeneous, dimensionally stable, has a high strength to weight ratio and good thermal properties
compared to other mirror materials. Silicon Carbide can be formed into complex shapes, allowing for optimal
material usage. Of further benefit is that Silicon Carbide can be polished to a high quality mirror finish.
Existing Silicon Carbide fabrication processes are capable of producing lightweight mirrors as large as 3m in
diameter. Silicon Carbide is now not widely used as a mirror material for ground-based large optical
telescopes. The optical figuring of steep aspheric Silicon Carbide mirrors is a non-trivial problem that needs
to be investigated further.
The work described here indicates that Silicon Carbide mirror segments of reasonable weight and rigidity
can be constructed so as to eliminate the need for conventional whiffle tree support mechanisms. Hexagonal
mirror segments of 1m and 2m size were studied; this work indicates that mirror segments closer to 1m in
size represent more practical solutions than larger segments. The larger mirror is has roughly four times the
areal density of the smaller mirror and is much more sensitive to thermal gradients in the segment.
The effect of thermal gradients on the behaviour of the proposed Silicon Carbide mirror segment designs was
studied. The results of these studies indicate that thermal control will be an important design consideration
for such segments. A temperature gradient of 0.1ºC along the mirror segment axis was found to cause
mirror deformations comparable to those cause by gravity.
The weight of a Silicon Carbide primary mirror system was estimated and compared to existing, more
conventional Zerodur mirror systems. At this point there, Silicon Carbide segments offer no clear weight
saving compared to more conventional designs.
Several areas of further study have been made apparent by the work described here. Some of these areas
are listed here:
• Mirror segments are non-axisymmetric optical components. The deviations between the required surface
and a spherical surface may be considerable, especially for small -fnumbers. The fabrication and
polishing of SiC mirrors to the required surface profile should be investigated.
• Further work is required to look at the mechanical design of mirror support components, especially at
the support to SiC mirror interface. Performance requirements should be established for the mirror
actuators.