University of Wisconsin-Madison Physical Sciences Lab

Learn More: X-ray Microbeam

Waisman Center Xray Microbeam

PSL first became involved in the project in the late 1970’s, during the proposal phase. At that time a precursor machine developed by JEOL in Tokyo was in operation. JEOL declined to develop a second machine, and PSL was awarded the development project (incorporating significant improvements over the JEOL design thanks to the advice of its users).

The instrument weighed about 20 tons, most of which resided on the second floor of the Waisman Center in Madison WI. Most of the weight was due to thick lead xray shielding and the support structures required. The shielding was up to 4.5″ (114mm) of lead plus 0.25″ (6.4mm) of steel cladding. The high voltage power supply was on the first floor, and stood 12 feet (3.7m) tall. At the other extreme of size was a gold/nickel coating about 0.8 millionths of an inch (200 Angstroms) thick, which is partially transparent yet conductive.

Challenging aspects of the microbeam’s development included ensuring subject safety, machine safety, and excellent long term performance. Specialized skills required included thin film deposition, UHV system design, precision mechanism design, very high voltage handling (up to 600,000 volts), radiation safety design and monitoring, electron optics in vacuum and simulation of scattering in solids, high speed analog and digital electronics design and fabrication, optical alignment, as well as many others.

Up to 3 kilowatts of electron beam power are concentrated into a 0.020″ (0.5mm) diameter area, striking a thin tungsten sheet. Electronics monitored the motion of the electron beam, since if it became stationary for more than a few tens of microseconds, it would melt through the tungsten sheet. The heating of the target was so rapid that the rate of thermal expansion is equivalent to running a tensile test to fracture 60 times per second. The power density in the beam was higher than that of a normal electron beam welder, and the maximum available power was comparable. On one side of the 0.020″ (0.5 mm) thick tungsten was 10-7 Torr vacuum; on the other was a thin layer of atmospheric pressure flowing cooling water. A closed loop cooling system with pressure regulation, ballast tank, and water purification was included to ensure that cooling continues for a short time after power failure, and to minimize the contamination of the vacuum system if a melt-through of the target does occur. Uninterruptible power was included at the control console to allow orderly shutdown.

When striking the target, the electrons produced xrays in all directions; a pinhole in the shielding allowed a thin pencil of xrays to pass through the experimental subject to a detector. By successively directing the electron beam to rectangular arrays of spots on the target, areas in the subject where pellets were predicted to have moved were illuminated by the xrays, somewhat like the raster scanning of a tv tube. By the combination of limited area scans near where the pellets were last seen, pattern recognition, and adaptive scheduling of the beam spot, xray dosage to the patient was minimized, allowing a wider variety of subjects and longer experiments than would otherwise be possible.

The instrument was intended to have a 10 year life. We were pleased that the reliability of this complex machine exceeded expectations, such that the design life was considerably exceeded. This instrument over time became a standard by which others were evaluated, thanks to its longevity, experimental database, resolution, tracking rate, and repeatability: Magnetometer and X-ray Microbeam Comparison

The xray generator’s electrical and mechanical components were designed and constructed in the early 1980’s. On April 13, 1984, it was assembled with a small radiation cooled target. By August 20, 1984, tests were under way with radiation cooled targets. In October 1985 the system was disassembled, shipped from PSL to the Waisman Center, and reassembled. After verification of safety and controls, and operation on static and moving mechanical phantoms, the first tests with live subjects were performed in early 1987.  After more than 20 years of operation, the system had outlived its need, and it was eventually decommissioned beginning in early 2009.

The project’s principal PSL electronics engineer, mechanical engineer, and physicist all remained at PSL, providing technical support to the project when required, without the much higher cost of maintaining such personnel on staff at Waisman Center.

[Xray Microbeam operator's console photo]
Above is the operator’s console for the Xray generator. Power supplies for the vacuum system’s pumps and gauges occupy a separate third rack placed next to the xray generator.

[Xray generator photo]

The xray generator was housed inside the blue painted steel clad lead shielding to provide the many millions of attenuation required to provide radiation safety. The shields’ steel cladding protected the soft lead and also partially shielded the beam from the effects of the earth’s magnetic field as well as an MRI machine located in the same building. The electron source is at left in the photo above. It contained a hairpin filament in a triode gun and electrostatic accelerator. At its outlet there were the first in a series of sets of static steering coils to compensate for the beam curvature caused by the earth’s magnetic field. Following the steering coils was a focus coil, ferrite magnetic field shields, and beam dump. In the taller central section there was an axially positionable 4-jaw independently adjustable collimator with a second set of steering coils. This was followed by a limiting aperture, second focus coil, and main deflection coils. The enlarged near section contained an aluminum vacuum tank with viewports and water cooled transmission xray target.

Finally the heavy metal pinhole collimated the xrays produced from the spot where the deflected electron beam struck the vacuum side of the tungsten target. The green chair was a dentist’s chair and was where the speech research subject sat, conveniently providing positioning of the subject’s vocal tract onto the machine’s optical axis. There were vacuum ion pumps near the source outlet, above the electron collimator, and on the target tank. Visible beneath the blue frame was a modified arc welder adapted to inexpensively power a Varian Ti-Ball sublimator pump. Many sections of the beamline provided current sensing and required water cooling. Visible in the lower left is a section of removed shielding. All shields weighed less than the room crane’s 1 ton limit. Each major shield was labeled with lift weight and instructions. Redundant microswitches sensed the absence or improper seating of any major shield and prevented electron beam power via interlock circuitry. Part of the machine maintenance protocol was to survey the machine exterior at low beam current for xray levels after any cycle of shield removal & reinstallation. The machine frame provided mounting for much of the shield tonnage even while removed from the beamline, since the load rating of the floor was low.

[Microbeam subject area photo]

After passing through the subject, with variable attenuation & scattering depending on the amount and atomic number of material encountered, the xrays reached the detector cabinet shown at left in the photo above, where a large area scintillator backed by an array of photomultiplier tubes detected them. The detector cabinet was movable on rails to vary the spacing from the subject while maintaining alignment to the source’s optical axis. Computers gathered the data and correlated it via timing with where the electron beam was steered and the pinhole camera properties on the xray side to map density, and via template recognition, estimate the new position of gold tracking pellets on the subject. Pellet velocities and accelerations were calculated to forecast future positions. The next set of raster positions for electron spot exposures were calculated from the pellet position forecasts and rescan schedules. Dwell times were selected to control subject dosage, provide adequate signal to noise ratio, and allow reasonable xray target life. Rescan schedules were set on a per-pellet basis; slowly moving locations such as skull and lower-jaw pellets were scanned less frequently than fast moving locations such as the tongue tip.

[Microbeam used target water side photo with corrosion evident]

Above is a photo of the wetted side of the water cooled tungsten target after it had been in use for some time & installed for years. It is about 305mm wide × 310mm high, and about 0.5mm thick. The electron-beam-addressable area is 256 × 256mm out of a slightly larger area exposed to vacuum. The sheet was clamped between two stainless steel flanges into a cylindrical shape. The wet side was the concave side of the bent target. Water flowed horizontally across this face of the entire beam-addressable area of the target sheet in a layer about 1.6mm (1/16″) thick. A thin layer of glass reinforced plastic provided the other side of the water flow channel, selected for adequate strength & stiffness & acceptable xray transparency. (Water & plastic must be minimized to control blurring of the very small xray source.) The colored deposits were believed to be oxides of tungsten, although some migration of copper from other parts of the cooling system may have occurred (despite the water purification system) and account for the greenish tinge. The horizontal transition line near center was the water level when not running. Note that the humid air environment in the top half produced a different oxide appearance. The irregular shaped area in the center was the area most used for speech tracking. Thin layers and frequent oxide shedding occur in areas that run hot during tracking. Boiling may sometimes be occurring at the more active areas. It has been cost effective to use bare tungsten targets, rather than plate them on the wet side to limit corrosion. Most platings would be less suitable for xray production on the vacuum side.

[Microbeam target water side photomicrograph]

In the photomicrograph above, scale is 1mm between large horizontal bars. Cracks & crevices in the (probably tungsten trioxide) bands can be seen at 1 to 2 mm separation, along with spalling of about 50 to 100 microns wide very thin gray oxide. In other areas these spalls are about 250 microns wide.

[Microbeam target vacuum side with numerous crack origins]

The thermal strain caused by electron beam heating of the vacuum side was considerable. Beam impact spots sustain damage over time as the same x,y addresses are digitally addressed repeatedly. Spots recrystallize, darken and develop central cracks which grow and cause delamination. The image above is about 4mm square, showing numerous cracks.

[Microbeam target peel photo]

With further use the surface layer would eventually peel. The peeling can be seen in xray production differences. Because the system performs background subtraction (a necessity to deal with the presence of teeth, bone and fillings in the tracking area), slow changes in the target that are essentially static during the course of a speech data run do not affect tracking as long as there are enough xrays passing through the subject to the detector. It was also practical to seat the subjects to use different areas of the target. A spent target may have several areas where peeling of about 1cm has occurred. In the example above, numerous arrays of spots can be seen many cm away from the 3 large peeled areas. Given time these too would crack and peel. The photo shows an area of about 11cm square, or 1/8 of the total area.

[Microbeam target separated wrinkled chip photo]
[Microbeam target separated chip closeup photo]
[Microbeam target separated chip curled up]

In an extreme case of peeling the peel would further distort and curl up because once delaminated its cooling was very poor. These could be seen by the local change in xray production. In effect the damaged target provided a radiograph of its own failed areas. This chip (shown in the 3 photos above) was observed as an area of reduced xray production, leading to scheduled target replacement. Note that it exhibits elongated spots, wrinkling, scale of over 1cm, and about 270 degrees curl. Previously buried layers will also peel. To prevent further weakening of the target and consequent introduction of water into the vacuum system, damaged targets were replaced.

Simulations indicated most of the electron beam’s power deposited in the target would be absorbed in the first 150 microns. About half the current (containing one third the power) eventually scattered back out the vacuum side of the target and strikes the aluminum vacuum chamber, creating a diffuse secondary source of xrays. Aluminum was chosen as an economical material to limit production of xrays at this secondary collision. Some of the back-scattered electrons are sufficiently energetic to create Cerenkov radiation visible through the target tank’s viewports. Several inches of high-lead-content shielding glass was used at the viewports to make viewing the vacuum side of the target safe.