Independent control of electrostatic chuck high-voltage power supplies in multi-station vacuum chambers and plasma cleaning
An electrostatic chuck supply serves four vacuum chambers in a coating line, with four independent high-voltage outputs. Each output delivers 5 kV at 10 mA, with a separate return line to the chamber ground. The chuck inside each chamber is a ceramic plate with embedded copper electrodes, and the wafer sits on the plate. The electrostatic force holds the wafer flat during plasma cleaning intervals, when the chamber pressure rises to 10 Pa and the wafer would otherwise deform. The supply outputs are normally on during the process, and ramp down to zero during the wafer exchange, when the robot arm enters the chamber.
The four outputs share a common input stage, with a single 540 V DC bus. Each output has a dedicated inverter, transformer, rectifier, and divider. The output voltage is independently adjustable from 0 to 5 kV, and the current limit is independently set from 1 to 10 mA. The four outputs are controlled by a single PLC, with a separate analog channel for each. The PLC reads the chamber state from the main line controller, and enables or disables each output based on the chamber sequence.
The chuck ceramic is alumina, 99.5 percent purity, 2 mm thick. The embedded electrodes are copper, 10 μm thick, screened onto the ceramic. The chuck surface is coated with a 50 μm layer of yttria, for plasma erosion resistance. The chuck resistance between the electrode and the wafer is measured after each wafer exchange, and the reading should be above 1 GΩ. A reading below 100 MΩ indicates a pinhole in the yttria coating, and the chuck is removed for refurbishment.
The plasma cleaning interval lasts 30 seconds, during which argon plasma at 100 W power cleans the wafer backside. The chuck voltage stays at 5 kV throughout the interval, and the wafer temperature rises from 20 to 40 degrees. The chuck has a helium backside cooling channel, with helium pressure at 5 mTorr. The helium flow is regulated by a mass flow controller, set at 10 sccm. The backside cooling keeps the wafer within the process temperature window, and the temperature is monitored by a thermocouple embedded in the chuck.
The output cable for each chamber is a 5 kV coaxial cable, 3 m long, with the outer jacket grounded at both ends. The cable connects to the chamber feedthrough, which is a ceramic-insulated vacuum feedthrough rated for 10 kV and 1 A. The feedthrough inner conductor is gold-plated, and the plating wears through after 500 chamber pump-down cycles. The plating once wore through, and the exposed copper oxidized, raising the contact resistance from 5 mΩ to 50 mΩ. The resistive drop at 10 mA was 0.45 V, small, but the local heating at 4.5 mW warmed the feedthrough by 3 degrees. The feedthrough was replaced at the next scheduled maintenance.
The supply cabinet is 800 mm wide, 1000 mm deep, 1800 mm tall, with IP54 protection. The cabinet contains the four output stages, the control board, and the input breaker. The cabinet cooling fans run on a thermal switch, starting at 35 degrees and stopping at 25 degrees. The fans once failed in summer, and the cabinet temperature reached 55 degrees. The electrolytic capacitors on the DC bus reached 65 degrees, above the 60 degree rating. The fan fault alarm now routes to the central monitoring room, and the fan is replaced within 10 minutes of alarm.
The control board uses a 16-bit ADC, sampling each output voltage at 1 kHz. The ADC reference is a buried-zener type, with a drift of 2 ppm per degree Celsius. The cabinet temperature is held at 25 plus or minus 2 degrees by the fan and heater, and the reference drift over a 24 hour operation stays below 10 ppm, equivalent to 0.05 V at 5 kV. The drift is within the 0.1 percent acceptance limit, and the reference is recalibrated annually against a 0.01 percent standard divider.
The safety interlock on each chamber door was tested during commissioning. A technician opened the door while the chuck was energized, and the output dropped to zero within 20 ms. The residual charge on the output capacitor bled through a 10 MΩ resistor within 100 ms. The interlock response time was logged at 18 ms, within the 25 ms design limit. The chamber door carries a high voltage warning label, and entry requires a sign-in log. The robot arm interlock also trips the output when the arm enters the chamber, and the arm position sensor is checked quarterly.
The maintenance log lists routine tasks. Each shift, the operator records the four output voltages and currents. Weekly, the technician inspects the cable jackets and the feedthrough contacts. Monthly, the insulation resistance is measured with a 1000 V megohmmeter, and the reading should exceed 1000 MΩ. Quarterly, the dividers are calibrated against a 0.1 percent standard. Annually, the ADC references are recalibrated. The log is signed at each step, and the record is filed in the plant archive.
The input side of the supply draws 400 V three phase at 10 A. A 30 kVA isolation transformer sits between the plant bus and the supply, to isolate the chamber grounds from the plant ground. The isolation transformer has a 2500 V insulation rating between primary and secondary. The input cable is copper, 6 mm², rated for 25 A, routed through a cable tray. The input breaker is a 32 A molded case type, with electronic trip set to 1.05 times rated current long delay. The input power factor is 0.96, measured with a power analyzer during commissioning.
The high-voltage transformer for each output is oil insulated, with a 1 L oil volume. The oil temperature reaches 40 degrees after 8 hours of continuous operation at 5 kV and 10 mA, against an ambient of 25 degrees. The silica gel breather on the transformer tank changes color after 12 months, and the gel is replaced before the color reaches half saturation. The oil breakdown voltage is tested annually, and the last measurement showed 40 kV per 2.5 mm gap, well above the 25 kV replacement threshold. The four transformers are mounted on a common rail inside the cabinet, with individual shields between the units.
The output divider for each channel is a series chain of 100 MΩ high-voltage glass glaze resistors, totaling 1 GΩ. The divider tap at 1:1000 feeds the control board ADC. The divider is calibrated quarterly against a 0.1 percent standard divider. The calibration drift over 12 months was 0.2 percent, equivalent to 10 V at 5 kV. The drift is within the 1 percent acceptance limit, but the dividers are scheduled for resistor replacement at the 24 month mark.
The wafer exchange sequence is controlled by the main line PLC. When a wafer is done processing, the PLC sends a signal to the chuck supply to ramp down the voltage to zero over 500 ms. Then the helium flow stops, and the robot arm enters the chamber. The arm places a new wafer on the chuck, and the PLC sends a signal to ramp up the voltage to 5 kV over 1 second. The helium flow restarts, and the process begins. The entire exchange takes 15 seconds, and the supply is offline for 2 seconds during the ramp down and up.
The chamber vacuum system uses a turbomolecular pump, rated at 60 L per second, backed by a rotary vane pump. The chamber base pressure is 5e-5 Pa, measured by a hot cathode ionization gauge. The gauge is calibrated annually, and the drift is checked quarterly. The gauge once drifted, reading higher than the actual pressure, and the chamber ran at 8e-3 Pa instead of the required 5e-3 Pa. The drift was caught during a quarterly check, and the gauge was recalibrated. The turbomolecular pump cooling water flow is monitored at 2 L per minute, and a low flow alarm routes to the central monitoring room.
The customer acceptance test was run over three days. Day one covered the chuck voltage stability, measuring ripple and drift at 5 kV. Day two covered the wafer exchange sequence, running 100 cycles and verifying the ramp times. Day three covered the plasma cleaning interval, running 50 intervals and checking the wafer temperature. The test report was signed by the customer representative, and the supply entered the 12 month warranty period.
The plant environment around the supply cabinet is held at 23 plus or minus 3 degrees Celsius, with 45 to 65 percent relative humidity. Winter heating raises the room temperature to 26 degrees, and the cabinet fan runs continuously. Summer cooling failure once pushed the room temperature to 32 degrees, and the cabinet internal temperature reached 50 degrees. The ADC reference drifted 0.3% during that week, equivalent to a 15 V error at 5 kV. The drift was corrected after the cooling unit was repaired, and the supply was recalibrated. The plant floor is epoxy coated, cleaned daily with a vacuum cleaner, and the illumination is 300 lux with LED fixtures at 24 V.
The safety interlock on the supply cabinet door was tested during commissioning. A technician opened the door while the supply was running, and all four outputs dropped to zero within 20 ms. The residual charge on the output capacitors bled through the 10 MΩ resistors within 100 ms. The interlock response time was logged at 18 ms, within the 25 ms design limit. The cabinet door carries a high voltage warning label, and entry requires a sign-in log. The non-authorized personnel are not permitted inside the power supply bay.
The coating process data is recorded on the line server, with the run number, the chuck voltage, and the wafer temperature. The data is stored for 5 years, and is available for customer audit. The quality department reviews the data weekly, and trends the chuck voltage stability. The data system generates a weekly report, showing the process yield and the defect rate. The report is distributed to the plant manager and the customer.
The coating bay safety audit is conducted annually, by an internal safety team. The audit covers the vacuum system, the high-voltage system, and the robot interlocks. The audit report lists the findings, and the corrective actions are scheduled within 30 days. The last audit found two minor findings: a worn light curtain, and a faded high voltage sign. Both findings were corrected within the scheduled time. The audit record is filed in the quality system.
The coating bay spare parts inventory includes: chuck feedthroughs, helium mass flow controllers, fuses, and thermocouples. The inventory is stocked at 20 percent of the annual consumption, and is reviewed monthly. The inventory is stored in a locked cabinet, and the key is held by the bay supervisor. The parts are logged out when used, and the log is reviewed by the quality department. The inventory was once depleted when a mass flow controller failed, and the stock was replenished within 3 days.
The coating bay operates three shifts per day, 7 days per week. The day shift runs the production runs, the evening shift runs the batch runs, and the night shift runs the maintenance and chamber cleaning. The shift handover includes a status report, listing the completed runs, the chuck voltage, and the chamber vacuum. The handover log is reviewed by the bay supervisor, who allocates resources for the next shift.
The vacuum coating process is qualified per SEMI Standard, with the chuck voltage stability and the wafer temperature tested. The test wafers are prepared alongside the production runs, and the wafers are tested after each batch. The chuck voltage stability is measured by a high-voltage divider, and the wafer temperature is measured by a thermocouple. The test results are recorded in the quality system, and the batch is released only when all tests pass. The process has maintained the qualification for 5 years, with no major findings in the last 2 audits.
The coating chamber robot system is maintained by the maintenance department, with weekly inspection and monthly preventive maintenance. The robot transfers wafers between the cassette and the chuck, and the robot alignment is checked quarterly. The robot system has a 99 percent uptime target, and the downtime is logged in the maintenance record. The system once had a robot alignment error, and the alignment was corrected within 2 hours, with no production loss. The robot teach pendant is backed up weekly, and the backup is stored offsite.
The supply has run for 14 months. The chuck feedthroughs have been replaced twice, after 600 and 1200 cycles. The output capacitors show a 3 percent capacitance loss, still within the 10 percent threshold. The unresolved issue is the fan reliability. The fans have failed three times in 14 months, and the mean time between failures is projected at 6 months. A more reliable fan model is under evaluation, and the replacement is scheduled for the next cabinet upgrade.
