European Space Agency

Producing the X-Ray Mirrors for ESA's XMM Spacecraft

D. de Chambure, R. Lainé, K. van Katwijk, J. van Casteren & P. Glaude

XMM Project Team, ESA Scientific Projects Department, ESTEC, Noordwijk, The Netherlands

The XMM satellite project has passed an important milestone with the Mirror Module's successful qualification review being held at ESTEC last October. This is therefore an opportune moment to present the developmental status of this challenging programme, this being the first time that such a large series of high-quality optics has had to be produced: more than 250 X-ray- quality mirrors with a total surface area of 200 m² and a surface micro-roughness of better than 0.5 nm.


The X-ray Multi-Mirror Mission (XMM) is the second 'Cornerstone' project of ESA's Horizon 2000 Long-Term Science Programme. This spaceborne observatory, due for launch in August 1999 by an Ariane-5 vehicle, has been designed as a high-throughput X-ray spectroscopy mission covering a broad energy range (0.1-10 keV). The payload will consist primarily of three telescopes, developed under direct ESA contract by a consortium of European firms (led by Media Lario, Bosisio Parini, Como, Italy), each with specific technical expertise in telescope manufacture. This approach of direct management by ESA was implemented in order to maintain the greatest possible flexibility in seeking solutions to the very considerable technical challenges inherent in the XMM mission.

The optics for each telescope (hereafter referred to as a Mirror Module, or MM) are made of 58 nested Wolter-1 grazing- incidence mirrors, chosen to maximise the effective collecting area within the allocated volume. This highly nested design calls for the manufacture of a large number of X-ray-quality mirror shells. In the early 1990s, therefore, a parallel development programme was conducted using two potentially interesting technologies: CFRP epoxy replication and nickel electro-forming. In 1993, based on the X-ray testing of several individual mirrors, nickel electro-forming technology was selected for the production of XMM's mirrors. Despite its lightness, the CFRP technology showed micro-roughness draw-backs, leading to unstable X-ray performance in vacuum.

Early in 1995, two MM demonstration models, each consisting of several X-ray-quality nickel mirror shells (ranging from the smallest to the largest) integrated onto a representative support structure, were built to validate the manufacturing and integration aspects of XMM mirror development. The test results obtained (a Half-Energy-Width resolution of better than 10 arcsec at 1.5 and 8 keV) confirmed that the nickel mirror technology was indeed able to fulfil the technical requirements of the XMM telescopes. At that point, however, the production yield ratio of high- quality mirrors was insufficient for manufacturing the 253 shells (total of 200 m² of optical surface) needed for the MM qualification- and flight-model production phases. A special programme was therefore initiated early in 1995 in order to deepen our understanding of the manufacturing processes and hence the yield ratio of the mirror-shell production process. By February 1996, this programme had led to the manufacture and on- time delivery of the MM qualification model and, after optical and environmental testing, the MM was pronounced qualified in October 1996.

The Mirror-Module Development Programme

The overall industrial organisation for the MM development programme is shown in Figure 1. The ESA XMM Project Team, located at ESTEC, is responsible for the management of the MM's development, design, manufacture and testing. It is advised in this task by the XMM Telescope Advisory Group, led by Dr. B. Aschenbach from MPE- Garching (D) who was a key member of the Rosat project.

XMM industrial organisation
Figure 1. The XMM industrial organisation

The project schedule for MM manufacture and testing (Fig. 2) shows their early delivery compared with the spacecraft itself, which is necessary for the timely testing and calibration of the scientific instruments, namely the European Photon Imaging Camera (EPIC) and the Reflection Grating Spectrometer (RGS).

Mirror schedule
Figure 2. The baseline development schedule for the Mirror Modules

The overall XMM development plan called for eight Mirror Module models:

The XMM Mirror-Module's design

The three Mirror Modules (Fig. 3), representing the heart of the XMM payload, are a major technological challenge, each consisting of 58 nested mirror shells bonded onto a spider (spoked-wheel) support structure. These grazing-incidence telescopes (Wolter-1 type) are designed to operate in the X-ray energy range 0.1-10 keV, with a focal length of 7.5 m and with resolutions (Half Energy Width) of 20 and 30 arcsec at 1.5 and 8 keV, respectively. The average grazing angle of the X-rays ranges from 20 arcmin for the smallest mirror to 40 arcmin for the largest.

Optical designof the XMM
Figure 3. Optical design of the XMM Mirror Module

Each mirror is a thin monolithic nickel shell which is shaped to a paraboloid surface in front and an hyperboloid surface at the rear. Incoming X-rays are reflected at grazing angles first from the paraboloid and then from the hyperboloid before converging to the focus.

The 58 mirror shells, with diameters of between 306 and 700 mm and a height of 600 mm, are mounted in a confocal and coaxial configuration. The thicknesses of the mirror shells range from 0.47 to 1.07 mm, in linear proportion to their diameters. The masses of the smallest and largest shells are 2.35 and 12.30 kg, respectively. The reflective surfaces of the shells are coated with a 250 nm layer of high-purity gold.

The 58 mirror shells are glued at their entrance plane to the 16 spokes of a spider (spoked wheel) which is made from Inconel (Fig. 4). This material was chosen because its thermal-expansion coefficient is close to that of the electrolytic nickel used for the mirrors themselves (confirmed by tests in ESTEC's Mechanical Systems Laboratory). The spider is connected to the XMM spacecraft platform via an aluminium Mirror Interface Structure (MIS, Fig. 4), which consists of a double conical surface reinforced by stiffeners, and an interface ring.

XMM Mirror module
Figure 4. Mechanical design of the XMM Mirror Module

To minimise mechanical deformation of the mirrors and therefore optical degradation during final assembly, the flatness of the mounting interface between the spider and the MIS (a surface with an inner diameter of 740 mm and an outer diameter of 770 mm) needs to be better than 5 microns. The MIS serves the following functions:

Heaters and thermistors, mounted on the spokes and the outer ring of the spider, provide the thermal control needed to maintain the Mirror Module at 20 ± 2°C with transverse and radial gradients not exceeding 2°C.

The production of the spiders, which are machined from solid blocks of Inconel by wire electro-erosion (Fig. 5), and the aluminium-alloy MIS is contracted to APCO (CH). Given the size of the MIS components and the high accuracy required, and due to the rivetted/ glued structure, heat treatment has to be introduced just before the final machining in order to eliminate residual stresses.

Figure 5. Electro-erosion of the spider

Table 1. Main characteristics of the XMM Mirror Module

Focal length                        7500 mm
Resolution  Half Energy Width       16 arcsec (0.1-10 keV)
            Full Width Half Max.    8 arcsec (0.1-10 keV)
Effective area                      1475 cm² at 1.5 keV
                                    580 cm² at 8 keV
Outermost mirror diameter           700 mm
Innermost mirror diameter           300 mm
Mirror length                       600 mm
Minimum packing distance            1 mm
Number of mirrors                   58
Mirror Module mass                  425 kg

Manufacture and integration of the mirrors

Mirror- shell manufacture is based on a replication process, which transfers a gold layer deposited on the highly-polished master mandrel to the electrolytic nickel shell, which is electro-formed on the gold layer. The process is fairly conventional but is complicated here by the tight tolerances required and the inherent flexibility of the elements. The production of the necessary 58 master mandrels has been contracted to Carl Zeiss (D). They are double-conical aluminium blocks coated with Kanigen nickel and then lapped to the exact shape needed, and finally super-polished in several cycles (Fig. 6) to a surface roughness of better than 4 angstrom.

Mandrel super-polishing
Figure 6. Mandrel super-polishing in progress

The lapping and polishing cycle typically takes about 10 to 12 weeks. The Kanigen nickel (8-10% phosphor content), deposited by electroless plating, has been selected for its hardness (necessary for good polishing), high adhesion to the aluminium, and low porosity.

The most critical steps in the manufacturing process remain the production of the mirror shells (Fig. 7) and their integration onto the spider, both of which are performed at Media Lario (I).

Mirror manufacturing
Figure 7. The mirror manufacturing and integration process

The mirror-shell production is divided into nine main steps:

(*) VITON® is a registered trademark of Dupont Dow Elastomers

The mirror-shell integration is performed in a Vertical Optical Bench (VOB), working from the smallest to the largest, with visible light (632 nm), in the following steps:

The total production time for one mirror, from the cleaning of the mandrel to its integration, including all verifications, is about 12 days. After the integration of all 58 mirrors (Fig. 8), and before the MM's shipment in an ultra-clean transport container to the testing facilities, integration is completed by the mounting of the MIS on the spider, the wiring of the electrical harness, and the adjustment of the alignment optics. To maintain the necessary high reflectivity, the particulate and molecular cleanlinesses of the mirrors have to be maintained (at 300 ppm and 2x10-7 g/cm², resp.) until the end of the mission lifetime.

miror module Mirrol module
Figure 8. Mirror Module integration in progress, with 30 (left) and all 58 (right) mirrors in place.

The mirror activities at Media Lario occupy an area of more than 3000 m² and employ about 45 persons. The metrology and integration activities are performed in temperature-controlled class-1000/100 clean rooms, which are about 10 m by 10 m and 6 m high. Six electro-forming baths (Fig. 9), two Vertical Optical Benches, two ultrasonic cleaning lines and two vacuum gold- deposition facilities have had to be built and operated simultaneously to maintain the required production rate.

The electro-forming
Figure 9. The electro-forming baths

Mirror technology development

Early in 1993, one of the first strategic decisions taken was to enforce a Total Quality Management system compliant with ISO 9000 standards in order to progress the nickel technology from an 'art' to an industrial production process, with the help of an external quality-control firm (Bureau Veritas). Each step in the manufacturing flow was therefore systematically and critically reviewed and every major parameter identified and monitored. Thereafter, all manufacturing procedures were validated and configured to ensure good repeatability, reliability and traceability of results. In parallel, the personnel were specially trained and safety precautions were introduced to further stabilise the processes.

To help achieve these goals, metrology equipment and software prediction tools were developed for measuring the geometry (roundness and axial profiles) and the surface quality (roughness) of the mirror shells and for characterising their X- ray optical performances. Also, several specialised items of equipment were developed to monitor the surface quality of the mandrels, including:

Experience showed that the geometry and surface quality of the master (i.e. the mandrel) can be transferred at least ten times to the electrolytic nickel shell without major degradation occurring. This means that theoretically mirror shells with an optical resolution of 5-6 arcsec (HEW) can be made, corresponding to the mandrel quality. In practice, however, this is thwarted by the relatively high adhesion coefficient between the gold layer of the mirror and the Kanigen nickel of the mandrel, and by the mechanical deformations of the electrolytic nickel layer of the shells. In other words, the mirror is a very flimsy shell and its overall shape is strongly affected by the way in which it is supported.

Most of the problems encountered during mirror development were associated with the gold plating, the nickel electro- forming, and the release, handling and integration of the mirror shells.

Gold plating
Some problems were encountered in getting the gold layer to separate from the Kanigen nickel of the mandrel. These were overcome by modifying the design of the gold-plating chamber and by improving the gold- plating procedure itself.

Nickel electro-forming
Some analyses and a dedicated metrology test campaign revealed that the mandrel was deformed at its edges during the mounting of the electro-forming tools (Fig. 10).

A gold-plated mandrel
Figure 10. A gold-plated mandrel ready for electro-forming

The observed 'trumpet' deformations could result in a loss in optical resolution (HEW) of about 10-20 arcsec. This required a redesign of the electro-forming bath and the associated tools, including deleting the conical mounting of the central shaft and replacing it by flat fixations at both ends of the mandrels, and changing the fixing-bolt position in order to reduce to a minimum the bending moment deforming the mandrel.

Cross-sectional cuts of several mirrors also showed that their thicknesses increased axially towards the edges by up to 20% of the nominal value. Similarly, the circumferential distributions at the edges showed deformations of up to 5%, despite the rotation of the mandrel in the nickel electro-forming bath. With the help of model-ling tools developed by ESTEC's Mathematics and Software Division, part of the bath geometry was modelled and the electrical field and current density were computed. As a result of the ensuing modifications to the design of the electro-forming tools, the excessive edge thickness was reduced to 5 10% and the variation in circumferential thickness was reduced to less than 0.5%.

Mirror-shell internal stress
Axial- profile measurements performed on the mirrors with the 3-D measuring machine indicated some very local deformations of 10-20 microns amplitude (propagating on 10% of the mirror surface) that were almost identical at both edges of the mirrors. After some investigations, these were traced to internal stresses in the mirror shells, leading to internal moments that show as radial displacements at the free edges.

Considerable effort was therefore devoted to monitoring and controlling these stresses using 'bent-strip' test samples and chemical analysis (liquid chromatography and atomic emission spectrometry) of the nickel electro-forming bath. These data, combined with axial profile measurements on the mirrors, allowed the internal stress to be satisfactorily controlled through a careful balancing of:

Mirror-shell release
This is definitely one of the most critical operations of the whole manufacturing process. The problems of distortions at the mirror edges due to excessive axial forces, and of adhesion between the mirror and the mandrel were encountered at the outset. This led to the complete redesign of the release stand, including the environmental control and the highly accurate guidance system to lift the mirror from the mandrel. The manufacturing procedure itself was also reworked with improved monitoring of several key parameters, and the final mirror-release step was also refined (Fig. 11). As a result of these efforts, final mirror quality was substantially improved.

Mirro-shell release
Figure 11. Mirror-shell during release

Mirror-shell integration

Testing on various mirror shells both at Media Lario and in ESTEC's Metrology Laboratory showed that the best way to obtain an optimum and repeatable mirror shape was to suspend a mirror shell at its exit plane (opposite to the integration side) using torsion-free wire. Given the design of the spider with 16 spokes, it was decided to suspend the mirror using 16 equally spaced hooks, placed in small holes drilled at the hyperboloid edge (exit plane) of the mirror and located in the shadow of the spokes of the spider to avoid optical-beam vignetting during subsequent mirror integration on the Vertical Optical Bench (VOB). The specially developed, actively con-trolled suspension system consists of 16 independent adjustable 'hanging devices' mounted symmetrically on a support ring.

As shown in Figure 13, this system allows the correction of the mirror shell's shape, and thus the optimisation of its optical performance, by changing the axial load distribution on each of the 16 'hanging devices'.

In most cases, equal load distribution to within an accuracy of 2% is sufficient for the production of an optimum mirror shell.

Mirro-shell shape
Figure 13. Mirror-shell shape correction through axial force distribution adjustment

In parallel, given the recurrent earlier problems (glue spills, poor wetting on spider and mirror surface, mirror displacement during injection), a series of tests were performed to investigate the injection technique, bonding strength and curing properties of the glue used in integrating the mirrors with the spider. These tests led to the following three-stage bonding procedure:

Mirro-shell in the process
Figure 12. Mirror shell in the process of final glueing

Mirror-shell production quality

The excellent results of this whole production-improvement programme are represented in the plots of mirror-shell optical performance versus development time presented in Figure 14. They show a production yield ratio of better than 80% after just a few months, which was the original goal. Also worthy of note is the fact that the performances of the large and small mirrors are almost identical despite the greater flexibility of the large ones. The second plot shows that the integration procedure (bonding) does not introduce any optical degradation.

Mirro-shell production Mirro-shell production
Figure 14. Mirror-shell production quality

Optical and environmental testing of the Mirror Module
After its delivery to ESA, the qualification model of the XMM Mirror Module, containing 21 X-ray-quality mirrors and 37 dummy mass-representative mirrors, was optically, mechanically and thermally tested between March and September last year at two test centres in Europe: the Max-Planck Institute's (MPE) X-ray facility (Panter) at Neuried in Germany, and the Centre Spatial de Liège (CSL) in Belgium (Fig. 15).

The qualification programme included:

Focal-X facility
Figure 15. Focal-X facility of Centre Spatial de Liège (CSL) with the qualification-model Mirror Module inside ready for testing

Their purpose was to demonstrate that the Mirror Module was meeting its performance requirements under simulated environmental conditions at least as severe as those to be expected during the service lifetime of the XMM spacecraft.

Test Facilities

The Max-Planck Institute (MPE) X-ray vacuum test facility (Panter) consists of a source chamber and an instrument (detector) test chamber connected by a 130 m-long tube. The instrument chamber (13 m long and 3.5 m in diameter) is equipped with an optical bench on which the XMM QM Mirror Module was mounted. The qualification Mirror Module (Fig.16) was tested in full illumination at X-ray energy levels of 1.5, 4.5 and 8 keV with two different detectors: the Position-Sensitive Proportional Counter (PSPC) developed by MPE for the Rosat satellite, and a Charge-Coupled Device (CCD) camera.

Figure 16. The qualification-model Mirror Module (above left: general view; below left: exit plane) ready for X-ray testing in MPE's Panter facility

The Centre Spatial de Liège (CSL) is an ESA-coordinated facility which provided the following test facilities for the XMM Mirror Module programme: the Extreme Ultraviolet Vertical (EUV) facility (Focal X), a thermal-vacuum test chamber (Focal 2), and a shaker for vibration testing. The Focal X optical facility provides a vertical collimated beam with a full aperture at 30 and 58 nm and several X-ray channels (pencil beam and collimated beam for reflectivity and scattering measurements). This large facility (4 m diameter and 10 m high vacuum chamber) was specially built for XMM under the leadership of the XMM Project Team and ESTEC's Testing Division.

Table 2. X-ray image quality of the QM Mirror Module at best focus

Test                          Before environmental test           After environmental test
Detector             Performance  1.5 keV  4.5 keV    8 keV      1.5 keV  4.5 keV    8 keV

CCD camera             FWHM         9.6"    not done   10.2"      10.5"     not done   10.9"
                       HEW         17.1"    not done   14.5"      17.8"     not done   15.3"

PSPC                   HEW*        16.8"    16.7"      <15"       17.3"    19.1"       14.7"
*PSPC intrinsic
resolution subtracted  W90         110"     216"       290"       120"     193"        236"
                 effective area    253cm²   179cm²     96cm²      249cm²   175cm²      93cm²

Table 3. Image quality of the QM Mirror Module at best focus from CCD measurement at CSL

   Pre-vibration              Post-vibration        Post thermal

  FWHM          HEW         FWHM        HEW         FWHM       HEW
  11.9"         20"         10.8"       19.6"       9.6"       20.6"

Optical results at MPE and CSL
The image- quality figures for the qualification-model Mirror Module are summarised in Tables 2 and 3.

All of the X-ray and UV measurements - resolution, focal length, and X-ray reflectivity - made at MPE and CSL before and after the environmental tests were identical to within the measurement accuracy of the facilities, indicating that there had been no optical degradation of the qualification-model MM due to that testing (Fig. 18).

X-ray focal image X-ray focal image
Figure 18. X-ray focal image of the QM Mirror Module (at 1.5 keV) before and after environmental testing (courtesy of MPE-Neuried)

Performance at 8 keV is better than at 1.5 keV because of the lower effect of the large mirror size at this energy level, and because of the better quality of the inner mirror shells.

The effective area is 15% lower than the theoretically achievable value, due not to the reflectivity of the mirrors (better than 95% of the theoretical value), but to the geometrical constraints of the Panter facility (finite source distance) vis-a-vis the geometry of the mirrors.

The UV results (HEW, FWHM) are slightly worse than for X-rays. The most probable cause is near-edge mirror deformation which cannot be detected in the Panter facility (finite source distance).

The tests were completed by optical tests performed with various thermal gradients applied to the Mirror Module. All of these tests confirmed the limited sensitivity of the Mirror Module to any radial or axial thermal gradients.

XMM spacecraft
Figure 17. Artist's impression of the XMM spacecraft, with its three X-ray telescopes (courtesy of Visulab)


The development work on XMM's X-ray mirrors is well on track to achieve the performance goals that have been laid down. The continuous efforts undertaken under direct ESA management to improve both the quality and consistency of the mirrors have clearly been successful:

Flight-model Mirror Module production has now been started as planned: two will be delivered early in 1997 and the remaining ones should be delivered by mid-1997 for acceptance and calibration testing. Current predictions are that the flight- model Mirror Modules will be of even higher quality than the qualification models.

Based on current knowledge, the metal electro-forming technology can certainly be further improved to deliver lightweight thin mirrors with still better resolutions than those required for XMM. Possibilities to be explored include the electro-forming of nickel alloys or other metals with better mechanical properties than pure nickel. The current state of the XMM X-ray mirror technology is already good enough to ensure the feasibility of missions like XIUS.


We would like to take this opportunity to congratulate the industrial team on the excellent results achieved so far, especially Dr. H. Rippel and G. Grisoni of Media Lario, D. Kampf of Kayser-Threde, W. Egle of Carl Zeiss, and A. Pugin and C. Jabaudon of APCO.

The members of the Telescope Advisory Group, Dr. B. Aschenbach and Dr. H. Bräuninger of MPE-Garching, Dr. P. de Korte of the Netherlands Space Research Organisation, Dr. R. Willingale of the University of Leicester, Dr. O. Citterio from Brera Astronomical Observatory and the many participating ESA personnel are also thanked for their continuous support.

The contributions by A. Hawkyard in the early phase of the mirror development effort are also gratefully acknowledged.

About | Search | Feedback

Right Left Up Home ESA Bulletin Nr. 89.
Published February 1997.
Developed by ESA-ESRIN ID/D.