State of the art in precision positioning

Laser interferometers represent the ultimate feedback device for high-precision semiconductor processing and inspection.

Laser interferometers represent the ultimate feedback device for high-precision semiconductor processing and inspection.

As feature sizes shrink, the accuracy of motion systems has had to increase.

As a result, linear interferometers are becoming the feedback device of choice in semiconductor inspection.

Now integral elements of wafer steppers, laser interferometers are showing up in an increasing number of wafer inspection applications.

A laser interferometer employs a highly stabilized light source and precision optics to accurately measure distances.

They are superior to glass encoders in having greater inherent accuracy and better resolution.

An additional advantage is that interferometers measure distances directly at the workpiece.

Mounting considerations often force linear encoders to be "'buried" inside the positioning stage" away from the workpiece, introducing an additional source of error.

In contrast, an interferometer can take measurements directly at wafer height, maximizing accuracy.

One interferometer widely applied in semiconductor processing comes from Aerotech Inc and is based on the Michelson Interferometer.

It includes a light source (a frequency stabilized He-Ne laser tube), a linear interferometer optic (composed of a polarizing beamsplitter and retroreflector), a moving linear retroreflector and detection electronics.

When the laser light reaches the interferometer optic, it divides into two distinct beams.

The first reflects back to the detectors and serves as a reference beam.

The second passes through the optic and reflects from a moving retroreflector to provide the measurement beam.

The motion of the moving retroreflector shifts the frequency of the second beam.

When the reference beam and measurement beam recombine, they create an interference pattern.

The interference fringe appears as a dark and bright pattern.

The intensity of this pattern is a sinusoid that can be treated like a standard A-quad-B encoder signal.

The Aerotech MXH-Series high-resolution multiplier can multiply such signals by factors of up to 1,024 to give an effective resolution as low as 0.3 nm, using a retroreflector-based system.

Two-dimensional systems, using plane mirror optics instead of retroreflectors, benefit from an optical doubling effect that improves the maximum resolution to 0.15 nm.

The simplest approach to detector electronics combines the detector and laser source in the same housing, to give a compact system which works best for single-axis applications.

For multi axis applications, a remote detector is preferred.

For example, the Aerotech LZR-Series remote-detection systems embed the detection photodiodes in the same housing as the interferometer optics for optimal beam stability.

When coupled with appropriate beam-splitting optics, this lets one laser head serve as the source for multiple axes.

Such a configuration is useful for XY systems, or even for XY systems with active yaw control.

Purchasing a single laser source cuts costs and saves valuable footprint space.

Two-dimensional implementations must ensure that there is a beam path at allocations throughout the stage travel.

This requires the use of plane-mirror optics.

This has the added benefit of optically doubling the laser signal, providing a fundamental resolution of a quarter wavelength.

A beam-splitting optic splits the single laser source to produce a signal for all axes of measurement.

These beams are steered to the interferometer optics and plane mirrors before they are measured at a remote detector.

For compactness, the detector electronics sit in the same housing as the interferometer optics.

Modern designs make the interferometer appear as a standard feedback device.

For example, output signals from the Aerotech LZR-Series laser interferometer are standard A-quad-B, electrically identical to the output of a traditional incremental encoder.

Today's motion controllers increasingly employ high-speed electronics that permit serial data rates as high as 32 MHz.

For a system with a resolution of 6 nm, the resulting speeds can be nearly 200 mm/sec.

All in all, interferometer signal-processing boards still have a high-speed niche, but the much simpler serial approach often proves to be optimal.

Several considerations enter into the design of a system that uses laser interferometer feedback.

Some of the most important issues include home marker implementation, handling loss of feedback, and reducing the effects of error sources.

Because the interferometer is strictly an incremental device, there is no way it alone can establish an accurate home reference.

The high precision involved in wafer measurements necessitates a highly accurate and repeatable home.

In many applications a registration mark is acquired directly from the wafer itself.

Once the motion controller acquires the mark, its counters reset to zero (software homed) and processing continues.

It is essential that the interferometer provides a "beam blocked" signal and that the motion controller has proper fault logic to process it.

Unlike a linear encoder, it is easy to block laser-feedback signals.

This condition requires the motion controller to immediately generate a fault condition and disable the axes, as loss of feedback in a servo system can lead to a runaway condition and potential damage.

Without proper regard for error sources, laser interferometers will perform no more effectively than a low-cost linear encoder.

Environmental conditions, mechanical design, and optical alignment all enter into the design.

In a vacuum, interferometers are accurate to +/-0.1 ppm.

However, most applications operate in atmospheric conditions and air's index of refraction can change the frequency of laser light to reduce accuracy which appears as a path length difference.

As the effects of temperature, pressure and humidity on the wavelength of light are well known, all high-accuracy interferometers incorporate a "weather station" that samples the environment to create a wavelength scale number that is used as a correction factor.

Environmentally corrected systems will have accuracies of ? ppm or better.

Final accuracy is largely a function of the stability of environmental conditions.

Although expensive, a wavelength tracker, or refractometer, is the most effective means of compensating for changes in the refractive index of air.

Because it compensates for relative changes only, initial environmental conditions must be known and computed to establish a baseline wavelength scale factor.

Mechanical vibration or air turbulence can cause perturbations in the positioning feedback system and limit its performance.

A well designed machine base and isolation system will limit vibration effects.

Air turbulence creates thermal gradients across the path, making the machine microenvironment critical to sub-nanometer performance.

A simple and effective means of minimizing such effects is either to "shield" the beam with a tube or to minimize airflow.

However, there is a trade-off because systems often need "downdraft" airflow to maintain wafer cleanliness.

For cutting-edge performance, an XY positioning system must use air bearings mounted to a granite base.

Air bearing stages have superior geometrical qualities, while granite provides an extremely flat reference surface as well as good thermal stability.

In the absence of outstanding linear stages, Abbe effects will drastically undermine the accuracy of the laser system.

Abbe errors are linear displacement errors caused by an angular deviation in the axis of motion.

A properly designed system will place the centre of the measurement mirror in the same plane and same axial orientation as the wafer under test.

The effect of any pitch/yaw deviations drops drastically by tracking the motion of the actual part under test, as opposed to the stage itself.

This practice combined with a linear stage sys-tem that is inherently geometrically accurate nearly eliminates Abbe errors.

A less obvious source of error arises from the environment and mechanical placement of the optics.

Known as dead-path error, it is caused by portions of the beam that are effectively uncompensated.

The moveable reflector translates throughout the measurement path, and environmental compensation electronics correct for the change in the index of refraction of air.

But the environmental compensation scheme only corrects for relative motion.

The distance over which the laser beam travels where it undergoes no relative motion remains uncorrected.

Absent any further correction, the dead-path error effectively moves the zero point (Xo) of the system as environmental conditions change.

The most straightforward means of addressing dead-path error is to compensate for it or eliminate it.

Software compensation for the dead-path error requires additional calculations that not only account for temperature, pressure, and humidity, but also for dead-path distances.

Mechanical compensation entails separating the interferometer's retroreflector from the beam-splitter by a distance equal to the dead-path error.

As a result, both the measurement beam and reference beam have equal dead-paths that cancel each other out.

This approach requires careful alignment of the optics and assumes that both dead-paths have identical environ-mental conditions.

Elimination of the dead-path requires placement of linear interferometer optics as close as possible to the zero point of the moveable reflector.

The final pieces of the puzzle consist of the optics themselves-and their alignment.

All optics have inherent inaccuracies in the form of optical non-linearity.

Users cannot control this, but high-quality optics can minimize the inaccuracies.

Cosine error is an optical misalignment error that the user can minimise through careful alignment of the optics to the stage, as it occurs when the laser beam path and the axis of stage motion are not completely parallel.

Back to top