What works in the microworld very often doesn’t work in nanopositioning applications such as nano-imprinting, scanning microscopy, microlithography and automated alignment. End users will often need to rethink positioning strategies, including the way they define basic device parameters such as resolution. When the term “nanopositioning” was coined, some companies advertised open-loop, stepper-driven leadscrew devices as nanopositioners. Although such simple drives have since been replaced by more sophisticated systems, how some companies define accuracy and resolution still can be elusive. However, regardless of a device’s stated resolution, repeatable nanometer-range motion is impossible as long as there is friction, which all sliding or rolling bearings produce. True nanopositioning devices provide frictionless motion, virtually instantaneous response, high linearity and stiffness, and trajectory control on top of subnanometer resolution. They do so with help from flexure design, multiaxis low-inertia parallel kinematics, and active trajectory and high-bandwidth control. Eliminating friction The first nanopositioning design rule is to eliminate friction. This rules out all devices with ball, roller or sliding bearings, leaving air bearings and flexures. Air bearings are suitable for long travel ranges but can be bulky, feature high inertia and are costly to operate. They also do not work in a vacuum, as is increasingly required. While flexures work only over short travel ranges, this is hardly a disadvantage in nanopositioning. These frictionless, stictionless, hingelike devices rely on elastic deformation (flexing) of a solid material to permit motion (Figure 1). Available in multiaxis designs if needed, they provide trajectory control with excellent straightness and flatness. The devices also exhibit no wear, involve no operating costs, and are very stiff and maintenance-free. Figure 1. Anti-arcuate-motion flexure designs provide guiding precision in the low-nanometer range. Drive friction also is unacceptable. Lead- or ball screws — even friction-based, ultrasonic, linear piezo motor drives — cannot surpass submicron precision. Electromagnetic linear motors, voice-coil drives and solid-state piezo actuators are the most common frictionless drives. The first two options are fine for larger distances but have the disadvantages of magnetic fields (not tolerable in applications such as electron-beam lithography), heat generation and moderate stiffness and acceleration, resulting in low bandwidth. Although piezoelectric drives are limited to small distances, they are extremely stiff, and they can achieve accelerations up to 10,000 g, a requisite for millisecond or submillisecond step-and-settle and high scanning rates. Such devices neither produce magnetic fields nor are they influenced by them. A recent breakthrough in production technology has eliminated the need for polymer insulation, thereby increasing drive lifetime even under extreme conditions (Figure 2). Figure 2. Ceramic-insulated piezoelectric actuators have extended lifetimes, even under extreme conditions, and exhibit no outgassing in vacuum applications. Measuring motion End users also need to examine other features of nanopositioning systems. For example, indirect motion metrology devices, such as motor-mounted rotary encoders and actuator or flexure-mounted piezoresistive strain sensors, are cheap but do not qualify for state-of-the-art nanopositioning. High-performance nanopositioning systems employ noncontact direct metrology, placed to measure motion where it matters most to the application. Examples include capacitive sensors, laser interferometers and noncontact optical incremental encoders. Incremental encoders are excellent for long-distance measurements. Most are based on a grating pitch of 20, 10 or 2 μm. To get from there to the published 10- or 5-nm resolution usually requires interpolation. Although many encoders are very linear at multiples of the pitch, linearity at the nanometer scale can be as poor as 20 percent. In addition, if they are not mounted coaxially with the drive, any tilt in the guiding system caused by motion reversal will further increase error. Often overlooked are the small forces induced by the moving cable of an encoder read head, which can cause friction and hysteresis on the order of several tens of nanometers. For processes requiring repeatable nanometer-scale step widths, there are better solutions. Laser interferometers are the accepted standard in position measurement. However, the output of a heterodyne interferometer is not perfectly linear. This nonlinearity, caused mainly by polarization ellipticity or nonorthogonality of laser beams, is also influenced by imperfections in optics. The best commercial interferometers thus offer linearity of 2 to 5 nm — not good enough in some high-end nanopositioning applications. Users also must have profound knowledge of interferometry and special equipment to maximize performance of an interferometer, as either a feedback or calibration device. The highest performance is delivered by absolute-measuring, two-plate capacitive sensors. Working best over small ranges, these devices provide a perfect match for flexure-guided piezoelectric drives. Capacitive sensors are compact, vacuum-compatible and insensitive to electromagnetic interference. Good designs also provide extremely high linearity with resolution of 0.1 nm and below. The absolute measuring principle eliminates the need for a homing procedure, and there is no bandwidth-limiting interpolator or counter circuit prone to lose motion in high-speed applications or when there is ringing at the end of a fast step. Exploring motion In industrial production and testing processes, throughput and time matter more than they ever have before. Head/media test applications, for example, require movement in the form of subnanometer steps to reach a new position and to hold it to nanometer tolerances in a matter of milliseconds or less. It does not matter how rapidly the positioning stage can stop, but how fast the load reaches a stable position — something often overlooked. Piezoelectric drives can respond to input in less than 0.1 ms, often more than the payload or the supporting structures are designed for. The ultrafast step time of the nanopositioning stage, however, can excite vibrations in its load or in neighboring components. One way to prevent structural resonances involves real-time feedforward technology called Input? Shaping, which was commercialized by Convolve Inc. of New York. Now an integrated option for Polytec PI digital piezo nanopositioning controllers, it eliminates motion-driven excitation of resonances throughout the system, including all fixturing and ancillary components (Figure 3). Input Shaping also requires no feedback because it works with a priori knowledge of multiple resonances throughout the system (see “Eliminating Vibration in the Nano-World,” July 2002, p. 60). Figure 3. Piezo devices are capable of millisecond-scale step-and-settle. At left, a Polytec laser vibrometer visualizes external resonances. At right, an Input Shaping control process eliminates motion-driven ringing of components outside the servo-loop. Settling after rise time completes by t ~ Fres—1. Ultimately, it is resolution, linearity and accuracy that qualify the static performance of a motion system. However, in dynamic applications such as scanning or tracking, static specifications are meaningless. One way to measure dynamic behavior is bandwidth, which specifies the amplitude response of a system in the frequency domain. The problem is that static accuracy and bandwidth together still do not indicate a system’s dynamic accuracy. To qualify a system in such an application, end users must record and evaluate target data and actual position data for a given waveform, with the difference indicating following or tracking error. In conventional piezoelectric nanopositioning systems with proportional, integral, derivative servo controls, tracking error can reach double-digit percentages even at scanning rates below 10 Hz. It also increases with frequency. For these reasons, tracking error is a key parameter in dynamic nanopositioning applications. Recent advances in digital controller design have led to sophisticated adaptive digital linearization methods that reduce dynamic errors in repetitive waveforms from the micron realm to indiscernible levels, even with high-frequency dynamic actuation under load (Figure 4). Figure 4. The top image shows the response to a triangular scan signal with a conventional PID controller and piezoelectric nanopositioning system. Blue = target position; red = actual position; green = tracking error (10x for better visibility). In the bottom image, the same system with adaptive digital linearization has tracking error (100x for better visibility) reduced by several orders of magnitude. Serial or parallel kinematics? In applications such as scanning microscopy, small areas must be scanned in two dimensions, with a third axis controlled by an external input; e.g., force in atomic force microscopes or current in atomic tunneling microscopes. Subnanometer line spacing and scanning rates of hundreds of hertz are desirable, and these are feasible only with parallel-kinematics, multiaxis closed-loop piezo-driven flexure stages. Rather than stacking single-axis subassemblies, parallel-kinematics stages are monolithic, with actuators operating in parallel on a central moving platform (Figure 5). This not only significantly reduces inertia, but also yields identical resonant frequencies and dynamic behavior in both the X and Y directions. Alternative, stacked assemblies always result in different X vs. Y behavior (though published specifications sometimes fail to reflect this). Figure 5. A monolithic, parallel-kinematics nanopositioning stage with piezoelectric drives and flexures uses capacitive position sensors to directly measure the central moving platform, compensating for the slightest off-axis motion in real time (left). Stacked serial-kinematics two-axis nanopositioning stages (right) cannot correct off-axis errors. Consistent X vs. Y dynamic behavior is desirable for accurate and responsive scanning and tracking. The use of capacitive sensors to measure the monolithic moving platform means that orthogonal axes automatically compensate for each other’s runout and crosstalk (active trajectory control or multiaxis direct metrology), whereas with serial kinematics, runout errors of the individual axes accumulate. For example, tilt errors of only ±10 μrad — caused by the bottom platform of a hypothetical 4-in. multiaxis stack of stages — would cause a 2-μm linear error at the top platform. Other shortcomings of serial kinematics include high inertia, a high center of gravity, and up to five moving cables that can cause friction and hysteresis. State-of-the-art nanometer scanning systems based on parallel kinematics control all six degrees of freedom, automatically compensating for unwanted out-of-plane motion and rotational errors. Best specs or best performance? The above discussion illustrates the complexity often involved in quantifying performance of a nanopositioning system. To find the highest-performing device for an application (not the one with the best specifications on paper), the user should engage in a dialogue with the manufacturer and ask the questions relevant to his or her application. Answers sounding too good to be true usually are just that. It always helps to find out how long a manufacturer has been involved in nanopositioning, what quality control system is in place, how specifications have been measured and what equipment was used. In the aftermath of the telecom crash, analysts and investors are looking for promising markets, and nanotechnology seems to be one of them. This is why we will see new companies trying to make a fortune in this field. Start-ups that claim to have revolutionary nanopositioning solutions may lure millions of dollars in investor funding. Let’s not forget that, in telecom, more than 99 percent of the revolutionary concepts and ideas soon proved worthless. The real challenge lies not in the concept, but in production, yield and consistent quality, where delivered unit after delivered unit performs as well as the gently assembled prototype, fine-tuned by the chief engineer. Because nanopositioning is not as simple as one, two, three, only companies that have experienced, well-equipped design and production teams as well as proven quality control systems will be able to satisfy the ever-growing demands of the market. Their product specifications may not always seem revolutionary, but they will hold up in the application environment. Suggested sources 1. Scott Jordan (2000). Repealing Moore’s law; sub-0.25μm linewidths drive metrology, trajectory-control advancements for positioning subsystems. Semiconductor FABTECH, 12th Edition. 2. R. Gloess (1998). New methods of signal preshaping strongly increase bandwidth of closed-loop PZT actuators. ACTUATOR International Conference on New Actuators. 3. Ping Ge and Musa Jouaneh (1996). Tracking control of a piezoceramic actuator. IEEE Transactions on Control Systems Technology, 4, 3. Meet the author Stefan Vorndran is director of corporate product marketing communications/nanopositioning technologies for Polytec PI Inc. in Auburn, Mass. He holds an MS in electrical engineering from FH Dieburg in Germany.