Real-Time High-Voltage Correction for Proximity Effects in Electron Beam Lithography
Electron beam lithography remains the cornerstone of nanoscale patterning for advanced photomasks and for direct write applications in research and development. The ability to focus a beam of electrons to a spot just nanometers in diameter allows for the creation of features with extraordinary resolution. However, as feature sizes shrink below the wavelength of light and into the deep sub-100 nanometer regime, a fundamental physical limitation arises: the proximity effect. This phenomenon, caused by the scattering of electrons within the resist and the substrate, leads to unwanted exposure of areas adjacent to the intended pattern. After fifty years in high-voltage engineering, I have learned that while the proximity effect is often addressed through software correction of the dose, there is a powerful, complementary approach that involves real-time manipulation of the electron beam's high-voltage parameters.
The proximity effect has two main components. Forward scattering occurs as the primary electron beam passes through the resist, causing a slight broadening of the beam. Backscattering occurs when electrons penetrate deep into the substrate and then re-emerge, exposing the resist over a much wider area, sometimes microns away from the original point of impact. The result is that dense patterns receive extra dose from neighboring exposures, causing features to merge or become misshapen, while isolated features may be underexposed.
The standard correction method is dose modulation. A proximity effect correction algorithm calculates the pattern density and adjusts the exposure dose for each shape accordingly. A shape in a dense area receives a lower dose to compensate for the extra dose it will receive from its neighbors. This is typically implemented by varying the dwell time of the beam as it writes each feature. However, this method has limitations. It is a pre-compensation that assumes the pattern and the substrate are uniform. It cannot correct for dynamic effects or for variations in the resist or substrate.
A more advanced approach involves real-time correction through high-voltage manipulation. The key insight is that the range of the backscattered electrons is a strong function of the beam's accelerating voltage. A higher voltage beam penetrates deeper into the substrate, and the backscattered electrons emerge over a wider area. A lower voltage beam has a shorter range, and its backscatter is more confined. By dynamically varying the accelerating voltage during the write, we can, in principle, control the range of the proximity effect.
Consider a pattern with both dense and isolated features. While writing a dense area, we could use a lower accelerating voltage. The backscatter range is shorter, so the extra dose from neighbors is confined to a smaller area, which is precisely the area that is already being exposed. This reduces the unwanted background exposure. While writing an isolated feature, we could use a higher voltage to achieve a finer spot size and better resolution, accepting that the backscatter range is larger but its effect is minimal because there are no neighbors.
Implementing this requires an electron beam column with a high-voltage power supply capable of switching between different accelerating voltages on the fly, within microseconds, and with complete stability. The beam's focus, deflection, and astigmatism are all functions of the accelerating voltage. Changing the voltage would normally throw the entire column out of alignment. Therefore, the high-voltage supply must be tightly integrated with the lens and deflection supplies. A common approach is to use a dynamic correction system where the lens currents and deflection coil currents are adjusted in real-time, in lockstep with the accelerating voltage, to maintain a constant focus and position on the wafer.
This is a formidable challenge. The accelerating voltage supply must have a very fast settling time. When the pattern data indicates a change from a low-voltage to a high-voltage region, the supply must ramp to the new voltage without overshoot and settle to within parts-per-million of the final value in a few microseconds. Any overshoot would momentarily expose the resist at the wrong energy, blurring the pattern. Any ringing would cause focus and position errors. The supply must also be extremely low-noise, as any ripple on the voltage will be translated directly into beam energy spread and focus variation.
Furthermore, the beam blanker, which turns the beam on and off, must be synchronized with these voltage changes. The beam must be blanked during the voltage transition to prevent it from writing at an intermediate, uncontrolled energy. This requires a fast, high-voltage blanking supply that can also operate at the different accelerating potentials.
The control system for such a scheme is complex. The pattern data must be pre-processed to determine the optimal accelerating voltage for each shape. This is an additional computational load on the pattern generator. The system must then send commands to the high-voltage supply, the lens supplies, and the blanker, all with nanosecond-level synchronization. This is a real-time, multi-variable control problem of the highest order.
Another approach, less aggressive than full voltage switching, is to use a small, dynamic voltage modulation to fine-tune the focus or to correct for local charging effects. As the beam writes, it can implant charge into the resist and the substrate, creating local electric fields that deflect subsequent passes. By monitoring the beam position with a mark detection system, the control loop can apply a small correction voltage to a dedicated electrostatic deflector, or even a slight modulation of the main accelerating voltage, to steer the beam back onto its intended path. This is a form of real-time, closed-loop correction that can compensate for dynamic proximity effects caused by charging.
In conclusion, the real-time correction of proximity effects in electron beam lithography represents a new frontier in high-voltage engineering. Moving beyond static, pre-calculated dose modulation to dynamic, voltage-scaled exposure requires power supplies and control systems of unprecedented speed and precision. This approach promises to extend the resolution of electron beam lithography further into the nanometer regime, enabling the continued scaling of the devices that power our modern world. The high-voltage supply, once a silent, stable background component, becomes an active, agile participant in the writing process.
