An online calibration method and system for excimer laser pulse energy detectors
By incorporating a beam-splitting sampling unit and a reference detector into the excimer laser, online calibration curve updates were achieved, resolving the calibration curve drift problem, improving the accuracy and stability of energy measurements, and reducing downtime.
Patent Information
- Application Number
- CN202611494415.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-23
- Publication Date
- 2026-11-17
AI Technical Summary
In the existing online energy measurement chain of excimer lasers, the calibration curve drift cannot be detected between two shutdown calibrations, resulting in a decrease in the accuracy of the energy closed loop and affecting production stability.
The optical path is divided into two branches using a beam splitting sampling unit. One branch is connected to the main detector for pulse-by-pulse energy measurement, and the other branch is connected to a reference detector with a low response speed but stable performance. Bias correction and gain correction are performed during the burst signal interval or within the calibration window to update the calibration curve. The drift of the main detector is calibrated by utilizing the stability of the reference detector.
It enables explicit output of the energy closed-loop reference during online calibration, reducing downtime, improving the accuracy and stability of energy measurement, avoiding energy deviation caused by drift, and not affecting normal production.
Smart Images

Figure CN123111488A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of laser technology, specifically relating to an online calibration method and system for an excimer laser pulse energy detector. Background Technology
[0002] Deep ultraviolet excimer lasers (ArF 193 nm, KrF 248 nm, etc.) are the mainstream light source for integrated circuit lithography. The stability of their single-pulse energy directly determines the accuracy of exposure dosage and the uniformity of critical dimensions. Therefore, an online energy measurement branch must be set up inside the laser: a small proportion of the beam is taken from the output optical path by a beam splitter and sent to an energy detector. The detector output is converted into pulse energy by a calibration curve, which serves as the feedback for the closed-loop adjustment of the discharge voltage. The accuracy of this calibration curve is the metrological benchmark for the entire energy closed loop. Once the benchmark drifts, the more accurate the closed loop, the more stably the actual output energy deviates from the target value.
[0003] In actual production, the aforementioned calibration curve is usually calibrated once by an external standard energy meter before the machine leaves the factory, and then recalibrated by connecting an external standard during planned downtime or periodic maintenance. This approach has three shortcomings: First, recalibration requires interrupting light output and connecting an external standard, occupying the entire downtime and directly reducing machine output; second, the timing of calibration is determined by the maintenance schedule, while detector degradation is driven by the cumulative dose of deep ultraviolet photons received, and the two are not synchronized, resulting in sparse calibration during heavy-load operation and dense calibration during light-load operation; third, the drift in detector responsivity between the two calibrations cannot be detected and can only be extrapolated after degradation occurs in the critical dimensions of the exposure process, by which time batch losses have already occurred.
[0004] As a document reflecting the aforementioned background technology, Chinese invention patent publication CN108352674B discloses an excimer laser device. In this device, a beam splitter and an optical sensor are used to form an energy monitor in the output optical path of the optical resonator to measure the pulse energy of the pulsed laser. The control unit detects energy reduction based on the measurement results, adjusts the partial pressure of the halogen gas in the cavity, and outputs a cavity replacement request when further improvement is impossible. Chinese invention patent application CN113783097A discloses a method for controlling the energy of a dual-cavity laser. In burst mode, the reading of an online laser energy measurement module is used as feedback, and the discharge voltage is adjusted by proportional-dose calculation. In the above documents, the online energy measurement process is assumed to be accurate, and the drift and online correction of its own responsivity are not addressed. Summary of the Invention
[0005] The present invention aims to solve the problem that the drift of the calibration curve itself in the online energy measurement link of existing excimer lasers cannot be detected between two shutdown calibrations.
[0006] A first aspect of this invention provides an online calibration method for an excimer laser pulse energy detector. The excimer laser's output optical path includes a beam-sampling unit, a main detector, and a reference detector. The main detector is located on a first sampling branch branched by the beam-sampling unit. The peak voltage output of the main detector is converted into single-pulse energy via a calibration curve. The reference detector has a lower response speed than the main detector and is located on a second sampling branch branched by the beam-sampling unit, which includes a shutter. The shutter is opened during the interval of the burst signal or within a preset calibration window, and the dark level of the main detector is measured in a zero light input state and used as the bias correction amount. The excimer laser is instructed to emit a preset number of calibration pulses at a constant discharge voltage, and the peak voltage of each pulse of the main detector and the cumulative energy reading of the reference detector are collected simultaneously. The peak voltage of each pulse is subtracted from the bias correction amount and then converted using the calibration curve. The energy of each pulse obtained from the conversion is summed to obtain the total energy of the main detector. The gain correction coefficient is determined by the ratio of the splitting ratio of the first sampling branch to the second sampling branch, which is then reduced to the aperture of the first sampling branch, to the total energy of the main detector based on the cumulative energy reading of the reference detector. The calibration curve is updated with the gain correction coefficient and the offset correction amount. The shutter is then closed to end the current calibration. The nonlinear shape of the calibration curve remains unchanged.
[0007] In some embodiments, the main detector includes a photodiode and an amplification and integration circuit, the reference detector is a pyroelectric energy detector, and the responsivity stability of the reference detector is higher than that of the main detector; the shutter is closed outside the calibration period and blocks the second sampling branch.
[0008] In some embodiments, measuring the dark level of the main detector includes: In the zero-light-input state, the output of the main detector is continuously sampled a preset number of times and the arithmetic average is taken. The baseline reading of the reference detector is acquired simultaneously under the zero light input state, and the baseline reading is subtracted from the cumulative energy reading of the reference detector.
[0009] In some embodiments, the updated calibration curve and the gain correction coefficient are respectively: In the formula, The pulse energy is calculated based on the updated calibration curve. This is the gain correction coefficient. This is the bias correction amount. The peak voltage output by the main detector. The calibration curve represents the nonlinear mapping before the update. For the first The peak voltage of the calibration pulse. The preset number of calibration pulses. The difference between the cumulative energy reading of the reference detector and the baseline reading is calculated by subtracting the baseline reading from the cumulative energy reading of the reference detector. It is the ratio of the splitting ratio of the first sampling branch to that of the second sampling branch.
[0010] In some embodiments, before opening the shutter during the interval of the burst signal or within a preset calibration window, the method further includes: The current calibration is initiated when the cumulative number of pulses emitted by the excimer laser since the last calibration reaches the preset calibration interval pulse number and the calibration window appears. The calibration window is the excimer laser's wafer change interval, batch change interval, or standby period. When the number of pulses that can be accommodated in the interval of a single burst signal is less than the preset number, the calibration is divided into multiple segments. Each segment emits a portion of the calibration pulses in the interval between adjacent burst signals, and the peak voltage of each pulse and the cumulative energy reading of the reference detector are collected respectively. The total energy of the main detector in each segment is summed, and the cumulative energy reading of the reference detector in each segment is summed and then compared.
[0011] In some embodiments, updating the calibration curve includes: The product of the previously calibrated gain correction factor and the total energy of the main detector is used as the predicted value. The cumulative energy reading of the reference detector is reduced to the first sampling branch aperture and used as the measured value. The ratio of the absolute value of the difference between the predicted value and the measured value to the measured value is used as the comparison residual. When the comparison residual is not greater than the preset residual upper limit, the update is performed using the gain correction coefficient and the bias correction amount determined in this instance. When the comparison residual is greater than the upper limit of the residual, the update is performed using the gain correction coefficient and the bias correction amount of the previous calibration to keep the calibration curve unchanged and output an alarm.
[0012] In some embodiments, after determining the gain correction coefficient, the method further includes: The relative change of the gain correction coefficient relative to the previous calibration is normalized to the cumulative number of pulses emitted by the excimer laser between the current calibration and the previous calibration to obtain the responsivity drift per unit pulse: In the formula, The responsivity drift rate, This is the gain correction coefficient. This is the gain correction factor from the last calibration. The cumulative number of pulses; When the responsivity drift rate is greater than a preset first threshold and this condition is met for a preset number of consecutive times, the window of the main detector is determined to be contaminated, a cleaning request or replacement request for the main detector is output, and the light output of the excimer laser is kept uninterrupted.
[0013] In some embodiments, it also includes: When the deviation of the gain correction coefficient from the initial gain correction coefficient is greater than a preset second threshold, the main detector is determined to be faulty, a replacement request for the main detector is output, and a degraded operation flag is set, wherein the initial gain correction coefficient is the gain correction coefficient corresponding to the calibration curve at the factory calibration time. When the degraded operation flag is set, the shutter is kept open, and the reference for adjusting the discharge voltage of the excimer laser using the single pulse energy as feedback is switched from the main detector to the reference detector. The pulse-by-pulse energy closed loop is then switched to a dose sliding window closed loop using the cumulative energy of a preset number of pulses as feedback.
[0014] A second aspect of the present invention provides an online calibration system for an excimer laser pulse energy detector, comprising a beam-splitting sampling unit disposed on the output optical path of the excimer laser, and further comprising: The main detector is located on the first sampling branch of the beam-splitting sampling unit, and its peak voltage output is converted into single-pulse energy by the calibration curve. The reference detector, located on the second sampling branch of the beam-splitting sampling unit, has a lower response speed than the main detector and outputs the cumulative energy reading of a batch of calibration pulses. The shutter is located on the second sampling branch, and it is opened during the calibration period and closed and blocks the second sampling branch outside the calibration period. The synchronous acquisition unit is used to acquire the peak voltage of each pulse of the main detector and the cumulative energy reading of the reference detector under the same trigger time base; The controller is used to open the shutter during the interval of the burst signal or within a preset calibration window, measure the dark level of the main detector in a zero-light input state and use it as a bias correction, control the excimer laser to emit a preset number of calibration pulses with a constant discharge voltage, subtract the bias correction from the peak voltage of each pulse, convert and sum them through the calibration curve to obtain the total energy of the main detector, determine the gain correction coefficient by the ratio of the splitting ratio of the first sampling branch to the second sampling branch, and then calculate the ratio of the first sampling branch aperture to the total energy of the main detector based on the cumulative energy reading of the reference detector, and update the calibration curve with the gain correction coefficient and the bias correction while maintaining the nonlinear shape of the calibration curve unchanged, and then close the shutter.
[0015] In some embodiments, the controller includes a calibration control module, a calibration parameter calculation module, a health assessment module, and an energy closed-loop module; The calibration control module is used to determine the calibration window and arrange the opening and closing timing of the shutter and the dark level acquisition timing. The calibration parameter calculation module is used to perform summation and comparison operations and solve for the gain correction coefficient and the bias correction amount; The health assessment module is used to record the gain correction coefficient, the bias correction amount, the relative change of the gain correction coefficient relative to the previous calibrated gain correction coefficient, and the responsivity drift rate obtained by normalizing the number of pulses emitted by the excimer laser between the current calibration and the previous calibration. The energy closed-loop module is used to calculate the peak voltage output of the main detector according to the updated calibration curve, and to adjust the discharge voltage of the excimer laser according to the calculated single-pulse energy.
[0016] The beneficial effects of this invention include: (1) The closed-loop energy reference is changed from an unobservable quantity to a quantity that is explicitly output in each calibration. Existing energy closed loops use the converted energy value as the true value, and no quantity inside the loop can expose the drift of the reference itself; the present invention checks the accuracy with a highly stable reference detector inside the machine every set cumulative pulse count, so the closed-loop accuracy no longer deteriorates unidirectionally with the number of running pulses.
[0017] (2) The downtime required for calibration is reduced to zero. The calibration action is compressed into the interval of the burst signal or the existing idle window, and a constant discharge voltage and a fixed number of pulses are used without changing the timing of the normal light output task; in continuous production conditions with no long idle period, the calibration can also be divided into multiple segments and completed in adjacent intervals.
[0018] (3) The summation eliminates the requirement for the single-pulse resolution capability of the reference detector, allowing pyroelectric devices with low response speed but long-term stable response to enter the role of in-machine online reference. The advantages of the two types of devices are utilized respectively, while their disadvantages are mitigated by the configuration.
[0019] (4) The additive error and multiplicative error are respectively carried by the bias correction amount and the gain correction coefficient. The two types of degradation mechanisms are separated mathematically, and the calibrated undetermined degrees of freedom are reduced to one, avoiding ill-conditioned solutions due to the narrow sample energy range when the two parameters are combined. Only two low-dimensional parameters are updated online while retaining the shape of the original nonlinear mapping. The unsampled energy range will not deteriorate due to online updates.
[0020] (5) Degradation diagnosis becomes a zero-cost byproduct of calibration. All inputs to health assessment come from intermediate quantities already generated during calibration, without adding any measurement hardware or measurement actions; normalizing the cumulative pulse count makes the drift rate an intrinsic degradation rate of the device that is independent of production scheduling, and the same threshold can be applied to machines with different loads.
[0021] (6) The laser can continue to emit light even after the main detector is judged to have failed. The reference detector serves as the online metrological benchmark during normal operation, the source of health observations during the evaluation period, and a redundant closed-loop benchmark after failure, without the need to add any additional devices for redundancy. Attached Figure Description
[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0023] Figure 1 This is a flowchart illustrating an online calibration method for an excimer laser pulse energy detector, as shown in an embodiment of the present invention. Figure 2 This is a structural block diagram of an online calibration system for an excimer laser pulse energy detector, as shown in an embodiment of the present invention. Figure 3 This is a flowchart illustrating the measurement of the dark level of the main detector, as shown in an embodiment of the present invention. Figure 4 This is a flowchart illustrating the calibration startup and segmented execution in an embodiment of the present invention; Figure 5 This is a flowchart illustrating the determination of the validity of calibration results in an embodiment of the present invention. Detailed Implementation
[0024] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0025] The following will clearly and completely describe the concept, specific structure, and technical effects of the present invention in conjunction with embodiments and accompanying drawings, so as to fully understand the purpose, features, and effects of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the scope of protection of the present invention. Furthermore, all connections / linkages involved in the patent do not simply refer to direct contact between components, but rather to the ability to form a better connection structure by adding or reducing connecting accessories according to specific implementation conditions. The various technical features in the present invention can be combined interactively without contradicting each other.
[0026] like Figure 1 As shown in the figure, the online calibration method for an excimer laser pulse energy detector provided by the embodiments of the present invention is applied to, for example, Figure 2 The online calibration system for an excimer laser pulse energy detector shown includes a beam-splitting sampling unit, a main detector, and a reference detector on the output optical path of the excimer laser. The main detector is located on the first sampling branch branched by the beam-splitting sampling unit. The peak voltage output of the main detector is converted into single-pulse energy using a calibration curve. The reference detector has a lower response speed than the main detector and is located on the second sampling branch branched by the beam-splitting sampling unit, which is equipped with a shutter. The method includes: S1. Open the shutter during the interval of the burst signal or within the preset calibration window, and measure the dark level of the main detector in the state of zero light input as the bias correction amount.
[0027] A burst signal is a working instruction from an excimer laser to continuously output a set of pulses at a fixed repetition frequency, pause for a period of time, and then output the next set of pulses. The time between two adjacent sets of pulses without discharge is called the interval, which typically lasts on the order of tens of milliseconds. The calibration window is an idle period in the production cycle that is significantly longer than the interval. Shutter 5 is a light-blocking mechanism located on the second sampling branch where reference detector 6 is located. Normally, it is closed and blocks this branch. The purpose of opening shutter 5 is to connect reference detector 6 to the same beam source as the main detector 4, so that both can observe the same batch of pulses within the same time period. The zero-light-input state refers to the state where the discharge cavity and resonant cavity 1 are not discharging and there is no laser injection into the photosensitive surface of the detector in the optical path. At this time, the output of the main detector 4 is not zero. Its residual value comes from the dark current of the photodiode, the offset voltage of the amplification and integration circuit, and temperature drift. This residual value is the dark level. The purpose of using the dark level as the bias correction is that the peak voltage of the main detector 4 is always superimposed with this additive component that is independent of the incident light. If it is not subtracted, the calculated single pulse energy will be too high overall, and this deviation will account for a larger proportion in the low energy range.
[0028] S2. The excimer laser emits a preset number of calibration pulses at a constant discharge voltage, and the peak voltage of each pulse of the main detector and the cumulative energy reading of the reference detector are collected simultaneously.
[0029] A constant discharge voltage means that the charging voltage applied by the high-voltage power supply 9 to the discharge cavity and resonant cavity 1 remains constant throughout the entire calibration pulse sequence, and is no longer subject to pulse-by-pulse adjustment of the energy closed loop. Maintaining a constant voltage is necessary to limit the source of energy fluctuations to the laser's own inter-pulse statistical fluctuations, so that the difference between the two readings can be uniquely attributed to changes in the responsivity of the main detector 4 rather than changes in the emission conditions. The preset number of calibration pulses refers to the total number of pulses emitted in one calibration cycle. Its value must ensure that the total energy of this batch of pulses falls within the linear range of the reference detector 6, and that the batch of pulses can be accommodated within the selected interval or calibration window in time. Synchronous acquisition means that the peak voltage sequence of the main detector 4 and the cumulative energy reading of the reference detector 6 are driven by the same trigger time base, and the start and end points of the acquisition strictly correspond to the same batch of pulses. The peak voltage is the highest point voltage of the response waveform of a single pulse to the amplification and integration circuit, while the cumulative energy reading is the output corresponding to the total energy absorbed by the reference detector 6 over a period of time.
[0030] S3. After deducting the bias correction amount from the peak voltage of each pulse, convert it using the calibration curve, and sum the converted pulse energies to obtain the total energy of the main detector.
[0031] Subtracting the bias correction means subtracting the bias correction determined by the dark level from the peak voltage of each pulse acquired by the synchronous acquisition unit 7, thus shifting the zero point of the voltage domain back to the true zero point before inputting it into the calibration curve for conversion. The order of subtraction followed by conversion cannot be reversed because the bias component acts in the voltage domain; conversion before subtraction would amplify the additive error into an error related to the operating point through nonlinear mapping, which cannot be compensated for by a constant. Calibration curve conversion means substituting the bias-subtracted voltage value into the nonlinear mapping obtained from factory calibration to obtain the energy value corresponding to the pulse. The shape of this mapping remains unchanged in this calibration. Summing the converted pulse energies is a crucial step for successful comparison, aiming to actively reduce the time resolution of the main detector 4 to the same dimension as the reference detector 6. The two readings cannot be directly divided; after summation, they simultaneously become two independent measurements of the total energy of the same batch of pulses, thus achieving comparability. The summed value represents the total energy of the main detector.
[0032] S4. Based on the ratio of the splitting ratio of the first sampling branch to the second sampling branch, calculated by subtracting the first sampling branch aperture from the cumulative energy reading of the reference detector, determine the gain correction coefficient by the ratio of the splitting ratio of the first sampling branch to the total energy of the main detector. Update the calibration curve with the gain correction coefficient and the offset correction amount, close the shutter, and end the calibration. The nonlinear shape of the calibration curve remains unchanged.
[0033] The gain correction factor is the ratio of the cumulative energy reading of the reference detector 6 to the total energy of the main detector. Physically, it represents the reciprocal of the current actual responsivity of the main detector 4 relative to its factory-calibrated responsivity. When the incident window of the main detector 4 is contaminated with deposits or suffers irradiation damage, the peak voltage generated by a pulse of the same energy decreases, resulting in a smaller total energy calculated from the original curve. Meanwhile, the reference detector 6, protected by shutter 5, retains its readings reflecting the true energy. Therefore, the ratio between the two increases monotonically and quantitatively characterizes the degree of degradation of the main detector 4. Updating the calibration curve with both means superimposing a voltage domain shift and an energy domain scaling on top of the original nonlinear mapping. Maintaining the nonlinear shape of the calibration curve means that the curvature and inflection point positions of the curve determined by factory calibration are completely preserved. The online process only adjusts two low-dimensional parameters without refitting the entire curve. This is necessary because the available online sampling window can only provide small samples within a narrow energy range; retraining the entire nonlinear curve with such samples would render the parameters unidentifiable. Closing the shutter 5 is the final step in this calibration, returning the reference detector 6 to its shaded state, ensuring that its cumulative radiation dose remains far lower than that of the main detector 4 throughout the entire lifespan of the device.
[0034] Specifically, in the energy measurement and closed-loop control subsystem of the deep ultraviolet excimer laser used for photolithography, a complete online calibration process is as follows: The deep ultraviolet pulses output from the discharge cavity and resonant cavity 1 are sampled twice by the beam-splitting sampling unit 2. The first sampling branch is normally sent to the main detector 4, and the second sampling branch is sent to the reference detector 6 via the shutter 5. The sampled main beam is then directed to the exposure device through the output port 3. The controller 8 monitors the cumulative number of pulses emitted by the laser since the last calibration. When this number reaches the set calibration interval pulse number and the current period is a wafer change idle period, a calibration request is set. The controller 8 first sends an opening command to the shutter 5. The shutter 5 takes about 30 milliseconds to go from fully closed to fully open. Then, it waits for about 20 milliseconds to allow the optomechanical vibration to decay and the pyroelectric baseline to stabilize. During this period, the high-voltage power supply 9 does not trigger discharge. The synchronous acquisition unit 7 continuously acquires 256 samples from the main detector 4 and takes the arithmetic average to obtain a bias correction of about 3.5 millivolts. At the same time, the baseline reading of the reference detector 6 is read. Subsequently, controller 8 sends a constant voltage calibration trigger to high-voltage power supply 9, continuously emitting 500 pulses at a repetition frequency of 500 Hz, lasting approximately 1 second, based on the current operating point voltage. Simultaneously, the synchronous acquisition unit 7 records 500 peak voltages under the same trigger time base and reads the cumulative energy reading of reference detector 6 within this second. Controller 8 subtracts the bias from each of the 500 peak voltages, inputs the factory nonlinear curve for conversion, and sums the results to obtain the total energy of the main detector. Simultaneously, it subtracts the baseline from the cumulative reading of reference detector 6 and calculates the aperture of the branch where main detector 4 is located according to the splitting ratio. Dividing the two yields the gain correction coefficient for this operation, which, along with the bias correction, is written into the calibration curve, maintaining the nonlinear shape unchanged. Finally, controller 8 closes shutter 5, and the energy closed loop immediately resumes normal light output adjustment according to the updated calibration curve. The entire process falls within the film changing idle period, ensuring uninterrupted normal light output.
[0035] The determination of the calibration window can be achieved by the controller 8 timing the burst signal and determining that the window is established when it detects that the current state is in the non-discharge segment between two adjacent pulse groups. Alternatively, it can be achieved by receiving the chip replacement, batch replacement, or standby status word from the host computer and using that status word as the criterion for window establishment. The acquisition of the dark level under zero-light input state can be achieved by continuously sampling the output of the main detector 4 and taking the arithmetic mean to suppress random noise by the square root of the sampling count. Alternatively, it can be achieved by taking the median or a moving average of the sampling sequence to suppress occasional spike interference.
[0036] The above steps transform the closed-loop energy reference from an unobservable quantity into an explicitly output quantity for each calibration. Existing energy closed-loop systems use the converted energy value as the true value; no quantity within the loop can expose the reference's drift. Even if the responsivity decreases, the closed loop can still stably lock the measured energy at the set value, while the actual output energy has systematically deviated. This scheme checks the accuracy with the highly stable internal reference detector 6 every set cumulative pulse count, preventing the closed-loop accuracy from deteriorating unidirectionally with the number of operating pulses. The calibration process is compressed into the burst interval or an existing idle window, using a constant discharge voltage and fixed pulse count, without altering the timing of normal output tasks, thus reducing the downtime required for traditional stop-type calibration to zero. The summation eliminates the requirement for the single-pulse resolution of the reference detector 6, allowing pyroelectric devices with low response speed but long-term stable responsivity to function as an internal online reference.
[0037] In some embodiments, the main detector 4 includes a photodiode and an amplification and integration circuit, and the reference detector is a pyroelectric energy detector. The reference detector has a higher responsivity stability than the main detector. The shutter is closed outside the calibration period and blocks the second sampling branch.
[0038] A photodiode is a semiconductor device that converts incident deep ultraviolet photons into photogenerated charge carriers. Its photosensitive surface receives light from the first sampling branch, converting the luminous flux of a single pulse into a photocurrent pulse of extremely short duration under reverse-biased operation. The function of the amplification and integration circuit is to integrate this photocurrent pulse over time and amplify it into a voltage waveform that is easy to sample. The purpose of integration is to convert the charge, which is proportional to the pulse energy, into a voltage quantity, and the purpose of amplification is to raise this voltage to the effective quantization range of the analog-to-digital converter. After the two are connected in series, a one-to-one correspondence is formed between the peak voltage of the output waveform and the energy of the pulse. The main detector 4 thus has the ability to resolve pulses one by one and can act as a pulse-by-pulse feedback source for energy closed loop. The reason for choosing this configuration is that the duration of deep ultraviolet pulses is on the order of nanoseconds, and only photoelectric devices with sufficiently fast response speeds, combined with an integration circuit, can completely collect the energy of a single pulse.
[0039] Reference detector 6 is a pyroelectric energy detector. Its working principle involves the sensitive element absorbing incident light, causing its temperature to rise. The spontaneous polarization intensity of the pyroelectric material changes with temperature, inducing a charge on the electrodes proportional to the absorbed light energy. This conversion chain does not involve interband transitions or surface state processes in the semiconductor, thus making it insensitive to photochemical aging caused by deep ultraviolet irradiation. Its responsivity is mainly determined by the absorptivity of the absorption layer, the pyroelectric coefficient of the material, and its heat capacity; these quantities change very little over the entire lifespan of the device. Reference detector 6 has higher responsivity stability than the main detector 4, meaning that under the same cumulative irradiation and operating time conditions, its relative responsivity change is significantly smaller than that of the main detector 4, thus allowing it to be used as an internal metrological standard. The trade-off is that its thermal response time is much longer than the pulse interval, providing only a cumulative energy reading over a period of time and not the ability to distinguish individual pulses.
[0040] Outside of the calibration period, shutter 5 is closed, blocking the second sampling branch where reference detector 6 is located. In the closed state, its light-blocking element completely obstructs the light path, so the cumulative radiation dose to reference detector 6 only accumulates during the brief opening periods of each calibration, several orders of magnitude lower than that of the main detector 4 under normal light. This arrangement serves two purposes. First, it maintains the responsivity stability of the reference device itself. If reference detector 6 is under normal light, it will also experience deep ultraviolet cumulative radiation and drift. Comparing two drifting devices will not yield any absolute information. Second, it maintains the proportional constant in the comparison link. After beam-splitting sampling unit 2, only a small number of optical elements unique to the reference branch are located after shutter 5. These elements are not irradiated outside the calibration period, and their transmittance changes negligibly over the entire lifespan of the device. Therefore, the ratio of the splitting ratios of the two sampling branches can be considered constant over the long term.
[0041] Specifically, in a deep ultraviolet excimer laser used for photolithography, the coordination process of the above-mentioned devices is as follows: During normal light output, shutter 5 is closed, reference detector 6 is in a light-shielded sleep state, and main detector 4 continuously receives sampled light through the first sampling branch. Its photodiode converts each pulse into a photocurrent, and the amplification and integration circuit integrates and amplifies it into a peak voltage waveform. The energy closed-loop module 804 converts it into single-pulse energy according to the calibration curve and adjusts the discharge voltage of the next pulse accordingly. When the accumulated pulse count reaches the set value and an idle window appears, shutter 5 opens, and reference detector 6 is connected to the second sampling branch, simultaneously observing the same batch of constant voltage calibration pulses with main detector 4. Main detector 4 provides 500 peak voltages, and reference detector 6 provides one energy accumulation reading. The two observe the same object but have different time resolutions, and after summing, they become comparable quantities. After calibration, shutter 5 closes again, reference detector 6 returns to the sleep state, and main detector 4 continues to receive normal light and continue to accumulate degradation.
[0042] The reference detector 6 can be a pyroelectric energy detector, which has a relatively fast thermal response and can read the incremental cumulative reading segment by segment when the calibration is split into multiple executions. Alternatively, a thermopile energy detector can be used, which also relies on the temperature rise after absorbing light energy and is not sensitive to deep ultraviolet photochemical aging, but its thermal response time is usually on the order of seconds, making it suitable for equipment that completes calibration in one go with a long idle window. The acceptable level of the reference detector 6's responsivity stability can be determined by the fact that its relative change in responsivity over the entire lifespan of the equipment is less than the upper limit of the allowable relative change in responsivity between two calibrations, so that the drift of the reference detector 6 itself is not included in the degradation of the main detector 4. Among them, shutter 5 can be a normally closed electromagnetic shutter that is kept closed by a reset spring when power is off. It automatically returns to the blocking state when the controller 8 loses power or malfunctions, so that the reference detector 6 will not be accidentally illuminated due to control failure. Alternatively, a rotating light shield driven by a stepper motor can be used. Its opening and closing position is determined by the number of motor steps and can be confirmed by the position sensor to be fully open. This eliminates the possibility of low light flux in the reference branch caused by shutter 5 not being fully open before issuing the calibration pulse.
[0043] The aforementioned device configuration allows the strengths of both types of devices to be utilized while mitigating their weaknesses through configuration. The main detector 4 trades stability for speed, undertaking real-time pulse-by-pulse feedback; its degradation is accepted as normal and compensated for by calibration. The reference detector 6 trades speed for stability, undertaking long-term metrological traceability; its speed deficiency is mitigated by summation comparison. This division of labor enables the system to possess both pulse-by-pulse energy resolution and a long-term, drift-free metrological reference—two requirements that cannot be simultaneously met by a single device. The introduction of shutter 5 widens the radiation dose of reference detector 6 by several orders of magnitude from that of main detector 4, thus giving the comparison results absolute significance. Furthermore, the same shielding ensures that the transmittance of the unique optical elements in the reference branch remains constant.
[0044] In some embodiments, such as Figure 3 As shown, measuring the dark level of the main detector in step S1 includes: S101. In the zero-light-input state, continuously sample the output of the main detector a preset number of times and take the arithmetic average.
[0045] The output of the main detector 4 is sampled continuously a preset number of times under zero-light input conditions, and the arithmetic mean is taken. The purpose is to separate the deterministic additive component of the dark level from the random noise superimposed on it. The instantaneous output of the main detector 4 under no-light input conditions consists of two parts: a deterministic component contributed by dark current and amplification / integration circuit misalignment, and a random component with a mean of zero contributed by thermal and electronic noise. After taking the arithmetic mean of the continuous sampling sequence, the random component decays by the square root of the number of samplings. The larger the number of samplings, the closer the mean is to the deterministic component. Therefore, the number of samplings is selected to suppress the residual noise below the allowable uncertainty of the bias correction. The bias correction is expressed as: In the formula, This is the bias correction amount. The number of samples under zero light input condition. For the first The output voltage of the main detector 4 is obtained from the second sampling. The usable range of the number of samplings is 64 to 1024, with the lower limit determined by the noise suppression requirements and the upper limit limited by the duration that can be occupied in the zero-light input state.
[0046] S102. Simultaneously acquire the baseline reading of the reference detector under zero light input state, and subtract the baseline reading from the cumulative energy reading of the reference detector.
[0047] The baseline reading of reference detector 6 is acquired simultaneously under zero light input conditions, and this baseline reading is subtracted from its accumulated energy reading. The output of pyroelectric devices is also non-zero under no light input conditions. This non-zero output originates from the temperature rise caused by the slow change in ambient temperature on the sensitive element, the offset voltage of the preamplifier, and the leakage current of the charge amplification link. These quantities are collectively referred to as the zero drift of reference detector 6. Simultaneous acquisition means that the baseline reading and the subsequent accumulated energy reading are taken from the same thermal state after the opening of shutter 5. The ambient temperature of both is consistent with the operating point of the amplification link, therefore the zero drift contained within them is the same quantity, which can be accurately eliminated after subtraction. If the baseline is acquired at a different time, the temperature drift between the two acquisitions will remain in the difference and directly mix into the gain correction coefficient. The baseline reading and the dark level are acquired in parallel within the same zero-light input time period, thus not occupying additional calibration window time.
[0048] Specifically, during calibration within a single wafer change idle period, controller 8 enters a zero-light input state after shutter 5 is fully open and waits approximately 20 milliseconds. High-voltage power supply 9 does not trigger discharge during this phase, and no laser is injected into the optical path. Synchronous acquisition unit 7 continuously samples the output of main detector 4 256 times at a fixed sampling rate. The arithmetic mean of the sampled sequence is the bias correction for this operation, which is approximately 3.5 millivolts, falling within a reasonable range of ±20 millivolts. Simultaneously, synchronous acquisition unit 7 reads the output of reference detector 6 and records it as the baseline reading, reflecting the zero-drift level of the pyroelectric sensitive element at this moment. After the zero-light input phase ends, controller 8 issues a constant-voltage calibration trigger. The accumulated energy reading after the calibration pulse is completed thus contains both the actual pulse energy contribution and the zero-drift contribution. Controller 8 subtracts the previously recorded baseline reading from this accumulated energy reading, obtaining a difference contributed by only 500 pulses of energy. This difference is then reduced according to the splitting ratio and divided by the total energy of the main detector to obtain the gain correction coefficient for this operation.
[0049] The sampling count under zero-light input can be calculated by rounding down the ratio of the standard deviation of the random noise output by the main detector 4 to the allowable uncertainty of the bias correction, and limiting it to an acceptable range of 64 to 1024. This directly corresponds to the accuracy index of the bias correction. Alternatively, during the overall system debugging period, the sampling count can be increased sequentially, and the dispersion of the measured mean can be measured. The sampling count corresponding to the point where the dispersion no longer decreases significantly can be taken into account. This method can account for the influence of low-frequency components in the actual noise. The baseline reading of the reference detector 6 can be read synchronously with the dark level sampling of the main detector 4, sharing the same zero-light input duration, which simplifies the timing arrangement. Alternatively, it can be read at the end of the zero-light input period, with the reading time close to the constant voltage calibration trigger. The baseline reading is then closest to the thermal state of the subsequent energy accumulation reading.
[0050] Since the bias correction is measured separately under zero-light input conditions, it is completely decoupled from the solution of the gain correction coefficient. The two undetermined parameters no longer need to be solved simultaneously from the same set of data. The calibrated undetermined degrees of freedom are reduced to one, and the gain correction coefficient can be uniquely determined by a single summation and comparison. This avoids ill-conditioned solutions that occur when solving two parameters simultaneously due to the narrow sample energy range. Continuous sampling and averaging reduce random noise by the square root of the number of samplings, preventing the uncertainty of the bias correction from being amplified to the energy domain through nonlinear mapping, thus ensuring the conversion accuracy at low-energy operating points. Simultaneous acquisition and subtraction of the baseline readings of the reference detector 6 removes the zero drift of the pyroelectric device from the comparison link, ensuring that both quantities involved in the comparison reflect only the true energy of the same batch of pulses. The comparison results are no longer affected by gradual changes in ambient temperature.
[0051] In some embodiments, the updated calibration curve and gain correction coefficient are as follows: In the formula, The pulse energy is calculated based on the updated calibration curve. This is the gain correction factor. This is the bias correction amount. The peak voltage output by the main detector. This represents the nonlinear mapping that the calibration curve represents before the update. For the first The peak voltage of the calibration pulse. To determine the preset number of calibration pulses, The difference between the cumulative energy reading of the reference detector and the baseline reading is used. It is the ratio of the splitting ratio of the first sampling branch to that of the second sampling branch.
[0052] The above two equations mean that the updated calibration curve retains the nonlinear mapping unchanged from the previous version, only adding a low-dimensional parameter to both the input and output sides of the mapping. The bias correction is placed on the input side, corresponding to the additive degradation caused by the superposition of dark current and amplification / integration circuit misalignment on the peak voltage, which is compensated for by translation in the voltage domain. The gain correction coefficient is placed on the output side, corresponding to the multiplicative degradation caused by incident window contamination and irradiation damage causing a proportional decrease in response, which is compensated for by overall scaling in the energy domain. The two parameters are located on opposite sides of the mapping. The bias correction only needs to be determined by the output under zero-light input conditions and does not involve the gain correction coefficient. The two can be mathematically separated. The denominator of the second equation is a known quantity after the bias correction is determined first. The numerator of the second equation is the energy calculated by subtracting the baseline reading from the reference detector 6 and reducing it according to the ratio of the splitting ratio. The denominator is the sum of the energies obtained by subtracting the bias correction from each calibration pulse in the same batch and converting them according to the nonlinear mapping before the update. The order of subtraction and conversion is the same as in the first equation. Therefore, the conversion method used to obtain the gain correction coefficient is consistent with the method used in subsequent routine conversions. Substituting the gain correction coefficient obtained from the second equation back into the first equation, the sum of the converted energies of this batch of calibration pulses is exactly equal to the reduced reference energy, thus the two equations are self-consistent.
[0053] Specifically, during the calibration process within a single wafer replacement idle period, the values and calculation chain for each quantity are as follows: The bias correction obtained during the zero-light input phase is 3.5 mV, and the baseline reading of reference detector 6 is obtained simultaneously. During the constant voltage phase, 500 calibration pulses are emitted using the current operating point voltage. The synchronous acquisition unit 7 records 500 peak voltages. The controller 8 first subtracts 3.5 mV from each peak voltage, then substitutes it into the factory nonlinear mapping to convert it into the energy value of that pulse. The 500 energy values are sequentially accumulated to obtain the total energy of the main detector, which is set to 10.00 mJ. The cumulative energy reading of reference detector 6 within this second is subtracted from the baseline reading to obtain the difference. This difference is then multiplied by the splitting ratio calibrated during assembly and reduced to the aperture of the branch where main detector 4 is located, with the reduction result set to 10.35 mJ. Dividing the two results in a gain correction coefficient of 1.035, which is 3.5% higher than the initial value of 1.000, indicating that the responsivity of main detector 4 has decreased proportionally relative to its factory state. The controller 8 then writes 1.035 into the energy domain scaling parameter of the calibration curve and 3.5 millivolts into the voltage domain shift parameter. Subsequently, the converted energy corresponding to the same peak voltage is increased by 3.5% to match the actual energy given by the reference detector 6.
[0054] The nonlinear mapping before the update can be a piecewise polynomial obtained by fitting wide-range scan data during factory calibration. This polynomial is continuously differentiable in the neighborhood of the operating point and is easy to evaluate in real time. Alternatively, a lookup table plus linear interpolation structure established at equal intervals based on peak voltage can be used. This structure requires only one lookup and one interpolation, with fixed real-time overhead and is independent of curve complexity. The splitting ratio can be obtained by measuring the transmitted light energy of the two branches sequentially with the same standard energy meter during assembly and taking the ratio. Alternatively, it can be calculated by multiplying the designed splitting ratio of the beam splitter element by the transmittance of each branch element and verifying it with measured data.
[0055] Because online updates only require two low-dimensional parameters while preserving the shape of the original nonlinear mapping, the sample conditions required for calibration are reduced from a large sample over a wide energy range to a quantitative sample near a single operating point. Parameters are identifiable and noise is not fitted into the model, and the unsampled energy range is not degraded by online updates. Multiplicative and additive errors are carried by the gain correction coefficient and bias correction amount, respectively. These two degradation mechanisms are mathematically separated, giving the gain correction coefficient a clear physical meaning. It can be directly used as a measure of the responsivity degradation of the main detector 4 and reused for health assessment, rather than just an intermediate quantity for conversion. The numerator and denominator are taken from the same batch of pulses. The inter-pulse energy fluctuations of the laser itself are summed, averaged, and further canceled out in the ratio. Therefore, the uncertainty of the gain correction coefficient obtained from a single calibration is much lower than the uncertainty of energy measurement for a single pulse. The proportionality constant is guaranteed to be stable over a long period by the configuration, so changes in the gain correction coefficient can be uniquely attributed to the main detector 4.
[0056] In some embodiments, such as Figure 4 As shown, before opening the shutter during the interval of the burst signal or within a preset calibration window, the method further includes: S1-1. When the cumulative number of pulses emitted by the excimer laser since the last calibration reaches the preset calibration interval pulse number and a calibration window appears, the current calibration is started. The calibration window is the excimer laser's wafer change interval, batch change interval, or standby period.
[0057] The current calibration is initiated when the cumulative number of pulses emitted by the excimer laser since the last calibration reaches the preset calibration interval pulse number and a calibration window appears. The calibration window is the excimer laser's wafer change interval, batch change interval, or standby period. The cumulative pulse count, rather than the time interval, is used as the trigger criterion because the degradation drive of the main detector 4 is the cumulative received deep ultraviolet photon dose. This dose is approximately proportional to the cumulative pulse count and is independent of calendar time. The degradation degree of the same laser can differ by several times under the same schedule of full-load operation and most-time standby. Time-based triggering would lead to sparse calibration under heavy load and overly dense calibration under light load. The three forms of the calibration window increase in duration sequentially: wafer change interval is the transfer and alignment time between two adjacent wafers; batch change interval is the preparation time between two adjacent batches of wafers; and standby period is the idle time when there are no production tasks. A common characteristic of all three is that the laser does not perform beam emission tasks. The two conditions are combined; reaching the cumulative pulse count threshold only sets the calibration request, and actual startup still requires waiting for the window to appear. Therefore, the calibration execution time automatically falls within the production cycle gap. The value of the calibration interval pulse count is determined by dividing the upper limit of the allowable relative change in responsivity between two calibrations by the intrinsic degradation rate of the device, ensuring that the responsivity change of the main detector 4 between two calibrations always falls within the acceptable metrological deviation of the energy closed loop; the allowable upper limit of relative change is set to 1%, and the intrinsic degradation rate is set to 100 × 10⁻⁶. -6 The calibration interval pulse count, calculated from the million-pulse level, is on the order of one hundred million pulses. The actual value is taken with a safety margin, for example, fifty million pulses.
[0058] S1-2. When the number of pulses that can be accommodated in the interval of a single burst signal is less than the preset number, the calibration is divided into multiple segments. Each segment emits a portion of the calibration pulses in the interval between adjacent burst signals, and the peak voltage of each pulse and the cumulative energy reading of the reference detector are collected respectively. The total energy of the main detector in each segment is summed, and the cumulative energy reading of the reference detector in each segment is summed and then compared.
[0059] When the interval between a single burst signal can accommodate fewer pulses than a preset number, the calibration is divided into multiple segments. Each segment emits partial calibration pulses within the interval between adjacent burst signals, and the peak voltage of each pulse and the cumulative energy reading of reference detector 6 are collected. The total energy of the main detectors in each segment is summed, and the cumulative energy readings of reference detector 6 in each segment are summed before comparison. The duration of a single burst interval is on the order of tens of milliseconds, and the number of pulses it can accommodate is often less than the preset number. If it were to be emitted all at once, a longer window would have to be waited for. Segmented execution is mathematically equivalent to one-time execution because the comparison quantity itself is a summation quantity. The sum of the total energy of the main detectors in each segment is equal to the result of summing all pulses at once, and the sum of the readings of reference detector 6 in each segment after deducting the baseline is equal to the result of its one-time response to all pulses. Both summation operations satisfy the additivity. This equivalence is the direct convenience of summation comparison compared to the pulse-by-pulse pairing method. If a pulse-by-pulse one-to-one correspondence comparison method is used, segmentation will destroy the pairing relationship. It should be noted that segmented execution extends the shutter 5 opening time from one second for a single execution to spanning multiple burst cycles. During this period, the zero drift of the reference detector 6 continues to accumulate. If the baseline reading is still only collected once before the first segment, the newly generated zero drift in the increment of each segment reading cannot be deducted. Therefore, the acquisition and distribution of the baseline must be arranged separately in segmented mode.
[0060] Specifically, after a calibration request is set, the calibration control module 801 of controller 8 first checks whether there is a current replacement, batch change, or standby window. If it exists and the duration is sufficient to send 500 pulses at once, it executes in window mode: opening shutter 5, measuring dark level, sending 500 pulses at constant voltage, reading the cumulative reading of reference detector 6, and closing shutter 5, all in one go. If there is no sufficiently long idle period and production is still in burst mode, it switches to fragmented execution. Controller 8 calculates that a single segment can accommodate 25 pulses based on the calibration repetition frequency and the duration of a single interval, and divides the 500 pulses into 20 segments accordingly. Shutter 5 is opened before the first segment and remains open between segments. When each burst interval arrives, controller 8 issues a constant voltage trigger to send 25 pulses. The synchronous acquisition unit 7 records the 25 peak voltages of that segment and reads the incremental cumulative reading of reference detector 6 within that segment, then waits for the next interval. After 20 segments are completed, controller 8 sums the total energy of the main detector for each segment and sums the incremental readings of the reference detector 6 for each segment. The difference between the sums and sums is used to obtain the gain correction coefficient for this operation. Finally, shutter 5 is closed. During the segmentation period, the laser's normal burst emission continues as usual, and the calibration pulse only occupies the interval that is not normally discharging.
[0061] The determination of the calibration window can be achieved by the controller 8 receiving the status words for changing plates, changing batches, and standing by from the host computer, and using the setting of the status words as the criterion for the establishment of the window. Alternatively, the controller 8 can time the arrival interval of the trigger signal and determine whether to enter the idle window when the interval exceeds a set threshold. The latter method does not rely on the host computer interface. During fragmented calibration, the timing of the shutter 5 can be achieved by keeping it open between each segment, which is simple in timing arrangement and avoids mechanical wear caused by frequent operations. Alternatively, a segmented opening and closing method can be used, which provides more thorough light shielding for the reference detector 6 and results in a lower cumulative radiation dose. During fragmented calibration, the main beam can be handled in two ways: First, the controller 8 sends a shading request to the exposure device via the host computer. The shading blades on the exposure device side then block the main beam from the light outlet 3 during each burst interval of the calibration pulse. The controller 8 then issues constant voltage triggers for each segment after receiving confirmation that the shading is in place. Alternatively, the host computer can arrange each calibration pulse segment into a non-exposure zone between scan fields. In this case, the field stop of the exposure device is closed, and the main beam does not reach the workpiece surface. Neither method changes the exposure dose received by the workpiece. The number of pulses per segment can be determined by calculating the number based on the calibration repetition frequency and the measured interval duration, with a safety margin. Alternatively, a fixed conservative segment length can be used, and the total number of segments can be dynamically adjusted based on the actual number of segments completed. In the segmented mode, the baseline processing of the reference detector 6 can be achieved by acquiring baseline readings before each segment and subtracting them segment by segment. The subtraction amount strictly corresponds to the thermal state of that segment, at the cost of adding one baseline acquisition time per segment. Alternatively, baseline readings can be acquired before the first segment and after the last segment, and then subtracted segment by segment after linear interpolation based on the duration of each segment. This method does not occupy an additional burst interval and is simpler to implement. The determination of the calibration interval pulse number can be achieved by inversely calculating from the upper limit of the allowable relative change in responsivity and the intrinsic degradation rate of the device, which directly corresponds to the metrological deviation index. Alternatively, the degradation curves of the same batch of devices can be measured during the overall system commissioning period, and the number of pulses corresponding to reaching the upper limit of the allowable change can be taken. The latter method is more in line with the actual dispersion of devices.
[0062] Using the cumulative pulse count as the trigger criterion, the calibration execution density automatically follows the actual load of the laser, increasing density under heavy load and decreasing density under light load. The updates of the two correction parameters always keep up with the actual degradation process of the main detector 4, preventing prolonged reference loss during full-load operation. Execution is limited to wafer change, batch change, or standby windows, ensuring calibration falls entirely within the existing gaps in the production cycle, preventing any unplanned downtime and eliminating the need to adjust the schedule for calibration. Segmented execution further reduces the required continuous calibration time from the order of one second to the order of tens of milliseconds, enabling calibration to be performed even in continuous production conditions with no long idle periods. The calibration timing no longer depends on the randomness of the schedule.
[0063] In some embodiments, such as Figure 5 As shown, the calibration curve is updated, including: S4-1. The product of the previously calibrated gain correction coefficient and the total energy of the main detector is used as the predicted value. The cumulative energy reading of the reference detector is reduced to the first sampling branch aperture and used as the measured value. The ratio of the absolute value of the difference between the predicted value and the measured value to the measured value is used as the comparison residual.
[0064] The aforementioned comparison residuals are used to verify the reliability of the current solution results before writing, and are expressed as follows: In the formula, To compare residuals, This is the ratio of the dispersive ratios of the first sampling branch to the second sampling branch. The difference between the cumulative energy reading of reference detector 6 and the baseline reading is the product of the two, which is the measured value reduced to the aperture of the first sampling branch. Total energy of the main detector This is the gain correction factor from the previous calibration.
[0065] This formula means that the total energy that the reference detector 6 should read this time is first predicted using the gain correction coefficient of the previous calibration, and then the ratio of the absolute value of the difference between the predicted value and the measured value to the measured value is taken. Under the premise that the degradation range between the two calibrations is limited, this ratio should be a small amount. If it is abnormally large, it indicates that there is a deviation between the two readings that cannot be explained by the normal degradation of the device.
[0066] S4-2. When the comparison residual is not greater than the preset residual upper limit, the update is performed using the gain correction coefficient and bias correction amount determined in this instance.
[0067] At this point, the two readings corroborate each other, confirming that all the prerequisites for this calibration are met: shutter 5 is fully open, the baseline of reference detector 6 is stable, and both readings are taken from the same batch of pulses. The two calculated parameters are deemed reliable, and controller 8 updates the energy domain scaling parameters and voltage domain shift parameters of the calibration curve accordingly, before proceeding to the health assessment stage. The upper limit of the residual is determined to cover the sum of the normal degradation range between two calibrations and the measurement uncertainty of a single calibration. Too loose a value will allow abnormal calibrations, while too tight a value will misjudge normal calibrations as invalid. In addition, the upper limit of the residual must be significantly greater than the product of the first threshold of the response drift rate and the number of calibration interval pulses. Otherwise, any calibration whose drift rate reaches the window contamination judgment threshold will be judged as invalid in the validity judgment, and the number of consecutive exceedances required for the health assessment will not be accumulated.
[0068] S4-3. When the comparison residual is greater than the upper limit of the residual, update the calibration curve by using the gain correction coefficient and the bias correction amount of the previous calibration to keep the calibration curve unchanged and output an alarm.
[0069] The most common cause of residual exceeding limits is not degradation of the main detector 4, but rather perturbation of the calibration process itself. Possible sources include: shutter 5 not fully opening, resulting in low light flux in the reference branch; stray light entering the reference detector 6; starting acquisition before the baseline of the reference detector 6 has stabilized; pulse counts and cumulative readings belonging to different batches; and failure of constant voltage triggering to maintain a constant voltage. Furthermore, when the responsivity of the main detector 4 degrades abnormally rapidly, parameter jumps between two calibrations can also push the residual beyond the upper limit. The way to distinguish between these two is to review the drift rate sequence obtained from previous calibrations: if the current jump is an isolated one and the drift rates before and after it all fall within the normal range, then the calibration process has been perturbed; if the drift rate has been continuously and monotonically increasing, then the actual degradation of the main detector 4 has accelerated. Under perturbed conditions, the two parameters obtained in this calculation do not reflect the true state of the device; writing them in would introduce an incorrect reference into the energy closed loop. Keeping the old parameters unchanged means the system reverts to a previously verified reliable reference, at the cost of the current degradation not being compensated. The alarm output records the anomaly and prompts for troubleshooting. It can be retried when the next calibration window arrives. In the case of actual accelerated degradation, the controller 8 also maintains the old parameters and outputs an alarm, but the alarm also shows the upward trend of the drift rate, prompting the maintenance side to prioritize troubleshooting the contamination of the main detector 4 window and to immediately re-perform calibration after maintenance to restore the baseline.
[0070] Specifically, in a calibration calculation process, after the calibration parameter calculation module 802 of controller 8 obtains the gain correction coefficient for this calibration, it does not immediately write it down. Instead, it first performs a validity determination. The calculation module 802 retrieves the gain correction coefficient written in the previous calibration, multiplies it by the total energy of the main detector for this calibration to obtain the predicted value of the reference detector 6 reading, subtracts it from the measured and recalculated reading, takes the absolute value, and divides it by the measured and recalculated reading to obtain the comparison residual. Assuming the upper limit of the residual is 2%, if the residual for this calibration is approximately 0.68% based on the total energy of the main detector for this calibration (10.00 mJ), the measured and recalculated reading (10.35 mJ), and the gain correction coefficient written in the previous calibration (1.028), which is less than the upper limit, then this calibration is accepted. The calculation module 802 writes the gain correction coefficient and the bias correction amount for this calibration into the calibration curve and submits the result for this calibration, along with the cumulative pulse count at that time, to the health assessment module 803. If the residual calculated this time is 3.5%, which is greater than the upper limit, the calibration is deemed invalid. The calculation module 802 does not write any new parameters, and the calibration curve continues to use the two parameters from the previous time. At the same time, it outputs a calibration abnormality alarm to the host computer and records the residual value of this time for troubleshooting of shutter 5 action, baseline and synchronization time base, etc. The controller 8 then closes shutter 5 and waits for the next calibration window to be executed again.
[0071] The construction of the comparison residuals can be achieved by using the relative deviation form with the measured reading of the reference detector 6 as the denominator, which is dimensionless. Alternatively, the ratio of the difference between the two calibration gain correction coefficients to the previous value can be used as the criterion, which directly reflects the parameter jump amplitude. The upper limit of the residuals can be determined by combining the upper limit of the normal degradation amplitude between two calibrations with the root sum of squares of the single calibration measurement uncertainty. Alternatively, the distribution of the residuals obtained from multiple normal calibrations can be statistically analyzed during the whole machine commissioning period, and a number of standard deviations can be taken.
[0072] This decision-making process places the writing of calibration results after a gating mechanism, preventing a disturbed calibration from writing an incorrect reference into the energy closed loop. In the metrology chain, the consequences of an incorrect reference update and a failure to update are not asymmetrical. The former causes the closed loop to lock the emitted light energy at the incorrect energy point, continuously affecting the exposure dose until the next calibration, while the latter only temporarily leaves the degradation uncompensated. Therefore, choosing a conservative gating mechanism is reasonable. Both quantities used for the decision are taken from intermediate results already generated in this calibration, without adding any measurement actions or hardware, and without consuming additional calibration window time. Alarm outputs make the calibration chain's own faults observable events; problems such as shutter 5 not operating properly, unstable baseline, and misaligned time base, which would otherwise be silently absorbed into the gain correction coefficient, are explicitly exposed.
[0073] In some embodiments, after determining the gain correction factor, the method further includes: S5. Normalize the relative change of the gain correction coefficient with respect to the gain correction coefficient of the previous calibration to the cumulative number of pulses emitted by the excimer laser between the current calibration and the previous calibration to obtain the responsivity drift rate per unit pulse: In the formula, For response drift rate, This is the gain correction factor. This is the gain correction factor from the last calibration. This represents the cumulative number of pulses. The response drift rate is measured in units of pulses, but when compared with the first threshold, it is converted to a caliber of pulses per million.
[0074] The formula first takes the relative change of the gain correction coefficient to eliminate the influence of dimensions and initial values, and then divides it by the cumulative pulse number to distribute the change to each pulse. Normalization must be performed on the pulse number rather than the time because the time interval between two calibrations is not proportional to the pulse number. If normalized by time, the same detector will give completely different readings under different schedules, making it impossible to set a uniform threshold. The driving factors for window contamination and responsivity decay are the cumulative photon dose and are approximately proportional to the cumulative pulse number. The drift rate obtained after normalizing the pulse number becomes the intrinsic degradation rate of the device, which is independent of the schedule.
[0075] S6. When the responsivity drift rate is greater than the preset first threshold and the preset number of consecutive occurrences are met, the window of the main detector is determined to be contaminated, a cleaning request or replacement request for the main detector is output, and the output of the excimer laser is kept uninterrupted.
[0076] The purpose of requiring a preset number of consecutive calibrations is to suppress false alarms caused by noise from a single calibration. When the cumulative number of pulses between two calibrations is too small, the smaller denominator will amplify the measurement uncertainty in the drift rate. A single instance exceeding the limit is insufficient to indicate accelerated degradation. The criterion is to use a drift rate greater than the first threshold rather than its absolute value being greater than the threshold. This is because window contamination reduces the incident light flux and decreases the peak voltage for the same energy. To maintain the calculated energy equal to the actual energy, the gain correction coefficient must monotonically increase. Therefore, the drift rate is positive under contamination conditions, while a significantly negative value indicates another type of anomaly and should be alerted separately. The basis for maintaining uninterrupted light output is that contamination is a gradual process, and its conversion effect has been compensated for by this calibration. The purpose of the request is to indicate the maintenance window arrangement, not to demand immediate shutdown.
[0077] Specifically, in one execution of the health assessment module 803, the flow of each quantity is as follows: After the residual judgment is passed, the calibration parameter calculation module 802 sends the current gain correction coefficient and the current cumulative pulse count to the health assessment module 803. The assessment module 803 retrieves the previous calibration gain correction coefficient and the current cumulative pulse count from the historical sequence, subtracts the two counts to obtain the cumulative pulse count between the current and previous times, and sets it to 5 × 10. 7 (That is, 50 million pulses). Assuming the previous gain correction factor was 1.028 and this one is 1.035, the relative change is 0.007 ÷ 1.028 ≈ 0.68%, which is divided by the cumulative pulse count of 5 × 10⁻⁶. 7 And converted to per million pulses, that is, 0.68% ÷ (5 × 10) 7 )×10 6 ≈1.36×10 -4 The response drift rate was approximately 136 × 10⁻⁶. -6 / Million pulses. Assume the first threshold is 100 × 10 -6 If a million pulses are emitted consecutively for three consecutive times, this is the first instance of exceeding the limit. Evaluation module 803 records the result but does not output a request. If the drift rates obtained from the subsequent two calibrations are still 142 × 10⁻⁶, then... -6 / Million pulses and 151×10 -6 If one million pulses are emitted and the limit is exceeded three times consecutively, the evaluation module 803 immediately determines that the incident window of the main detector 4 is contaminated and sends a cleaning request to the host computer, while providing the drift rate values and corresponding cumulative pulse counts for the three pulses. During this period, the energy closed-loop module 804 continues to execute pulse-by-pulse energy closed-loop, and the light output task is not interrupted. The maintenance actions are arranged by the process side and performed in the subsequent batch change or shutdown maintenance window.
[0078] The normalization benchmark for the drift rate can be the cumulative number of pulses between two calibrations, which is approximately proportional to the photon dose. Alternatively, the cumulative emitted light energy between two calibrations can be used, with the latter providing a more accurate dose representation when the pulse energy setting varies significantly with the process layer. The first threshold can be determined by measuring the intrinsic degradation rate distribution of devices in the same batch during the overall system commissioning period and taking its upper quantile. It can also be determined by working backwards from the target replacement cycle of the main detector 4 and the allowable cumulative response decrease. The purpose of each bias correction can be to characterize the misalignment and temperature drift of the main detector 4's amplification and integration circuit using its changing trend, and to output a maintenance prompt in the electronics channel when its deviation exceeds a preset range. This distinguishes it from the incident window contamination characterized by the gain correction coefficient in fault location. Alternatively, its short-term fluctuation amplitude can be used as a criterion for whether the number of dark level samples is sufficient, and the number of samples can be automatically increased when the fluctuation exceeds the limit. The latter can adaptively maintain the uncertainty of the bias correction when the noise environment changes.
[0079] Health assessment adds no measurement hardware or actions; all inputs come from intermediate quantities already generated during calibration. Therefore, degradation diagnosis becomes a zero-cost byproduct of calibration. A single comparison for benchmark correction simultaneously produces the observations needed for predictive maintenance. Normalizing the cumulative pulse count makes the drift rate an intrinsic device degradation rate independent of production scheduling. The same threshold can be applied to machines with different loads, and results can be compared across machines and used for lifetime statistics of the same batch of devices. Continuous exceedances of limits distinguish random noise from single calibrations from actual degradation acceleration, significantly reducing the false alarm rate of cleaning requests and preventing maintenance resources from being consumed by invalid requests. Uninterrupted light output detection transforms detector degradation—previously only detectable after degradation in critical dimensions during the exposure process—into a level that provides early warning before process parameters deteriorate.
[0080] In some embodiments, the method further includes: S7. When the deviation of the gain correction coefficient from the initial gain correction coefficient is greater than the preset second threshold, the main detector is determined to be faulty, a replacement request for the main detector is output, and a degraded operation flag is set.
[0081] The criterion is expressed as follows: In the formula, This is the gain correction factor used in this calibration. This is the initial gain correction factor. This is a preset second threshold. The second threshold and the first threshold serve different functions. The drift rate used for the first threshold only reflects the current rate of degradation, not the cumulative degradation. The deviation used for the second threshold only reflects how far it has deviated from the initial state, not the current rate of change. They are not interchangeable; slow but continuous degradation may not trigger the first threshold for a long time but will eventually exceed the second threshold. The criterion for determining failure is that when the deviation exceeds the second threshold, the signal-to-noise ratio and linearity of the main detector 4 have significantly deteriorated. Even if the calibration curve can still compensate for its mean, the random error of the single-pulse measurement has exceeded the energy closed-loop requirements.
[0082] S8. When the degraded operation flag is set, keep the shutter open, switch the reference of the pulse-by-pulse energy closed loop that uses single-pulse energy as feedback to adjust the discharge voltage of the excimer laser from the main detector to the reference detector, and switch the pulse-by-pulse energy closed loop to a dose sliding window closed loop that uses the cumulative energy of a preset number of pulses as feedback.
[0083] Shutter 5 is changed to remain open to ensure that reference detector 6 is continuously connected to the second sampling branch, providing continuous feedback. The trade-off is that reference detector 6 is subjected to continuous irradiation, and its responsivity is no longer stable over long periods. Therefore, degraded operation is a finite-term transitional state, not a steady state. Switching the closed-loop reference means that reference detector 6 changes from a metrological standard used only during calibration to a feedback source used daily. The switch in closed-loop configuration is a necessary result of reference switching. Reference detector 6 cannot distinguish individual pulses, and pulse-by-pulse energy closed-loop cannot be maintained. The controlled object changes from the energy of each pulse to the total dose of a pulse segment. The sliding window means that each time a new pulse is emitted, the earliest pulse is moved out of the window, and the accumulated energy within the window is recalculated. It should also be noted that in normal operating mode, reference detector 6 only participates in the ratio comparison of two readings; its output does not establish a correspondence with absolute energy. Therefore, before switching to the closed-loop reference, its energy scale is determined according to a predetermined method so that the accumulated output within the window can be converted into accumulated energy and compared with the target dose.
[0084] Specifically, in one judgment by the health assessment module 803, the initial gain correction coefficient is set to 1.000. The gain correction coefficient obtained from each calibration increases monotonically with the cumulative pulse count, and the value obtained in this calibration is 1.21, with a deviation of 21% from the initial value. Assuming the second threshold is 20%, the deviation exceeds the threshold, and the assessment module 803 immediately determines that the main detector 4 has failed, outputs a replacement request for the main detector 4 to the host computer, and sets the degraded operation flag. Upon receiving the degraded operation flag, the calibration control module 801 of the controller 8 no longer closes the shutter 5, allowing the reference detector 6 to remain connected to the optical path. The energy closed-loop module 804 synchronously switches the feedback channel, no longer using the single-pulse energy calculated from the main detector 4, but instead reading the continuous cumulative output of the reference detector 6, and using the cumulative energy of the most recent 200 pulses as the feedback quantity, adjusting the discharge voltage of subsequent pulses according to its deviation from the target dose. The controller 8 simultaneously reports the current degraded operation mode and its dose stability index to the host computer, and outputs the cumulative number of pulses during the degraded operation. The process side decides whether to continue production based on this, and the maintenance side arranges the replacement schedule of the main detector 4 accordingly.
[0085] The failure determination metric can be the relative deviation of the gain correction coefficient from its initial value, which can be compared across different machines. Alternatively, the criterion can be whether the absolute value of the gain correction coefficient exceeds the normal range set during assembly; the latter is simpler to implement. During the degradation period, the shutter 5 can be kept open throughout, providing continuous feedback and minimizing closed-loop delay. Alternatively, it can be opened with a higher duty cycle, which reduces the cumulative radiation dose to the reference detector 6 while maintaining the feedback update rate, thereby extending the duration of degradation operation. The feedback quantity of the dose sliding window can be the difference between the accumulated energy within the window and the target dose, used directly as the adjustment quantity. Alternatively, it can be calculated using a proportional-integral method before outputting the discharge voltage increment; the latter provides better suppression of slowly varying disturbances. The determination of the energy scale of the reference detector 6 before degraded operation can be achieved by reverse calculation from the result of the last valid calibration before failure, that is, by fixing its responsivity calibration value by the ratio of the total energy of the main detector known in that calibration to the reading of the reference detector 6 after reduction, without the need to add any measurement hardware; or by separately calibrating the second sampling branch with a standard energy meter during the assembly of the whole machine and storing the result in the controller 8, the scale is independent of the state of the main detector 4, which is more reliable when the main detector 4 has been severely degraded.
[0086] The dual-threshold design divides detector degradation into two levels, corresponding to two different treatments. The first threshold corresponds to the rate of change of the threshold tube and the cleanable window contamination, while the second threshold corresponds to the cumulative deviation of the threshold tube and the failure of the component requiring replacement. This avoids the dilemma of threshold value selection that inevitably arises when dealing with both progressive contamination and final failure using a single criterion. Failure determination does not end with a shutdown. Even after the main detector 4 is determined to have failed, the laser can still emit light. The reference detector 6 in the same hardware thus gains a third purpose: it serves as an online metrology reference during normal operation, a source of health observations during the evaluation period, and a redundant closed-loop reference after failure, without requiring any additional components for redundancy. The closed-loop configuration is reduced from pulse-by-pulse to a dose sliding window, which is the maximum control capability that can be retained under the constraint of the limited speed of the reference detector 6. The exposure process is truly concerned with the cumulative dose rather than the energy of each pulse, so this configuration can still cover a considerable portion of the process layer requirements.
[0087] The following are system embodiments of the present invention, which can be used to execute the online calibration method for excimer laser pulse energy detectors involved in the present invention. For details not disclosed in the system embodiments of the present invention, please refer to the method embodiments of the online calibration method for excimer laser pulse energy detectors involved in the present invention.
[0088] like Figure 2 As shown in the figure, an online calibration system for an excimer laser pulse energy detector provided in this embodiment of the invention includes a beam-splitting sampling unit 2 disposed on the output optical path of the excimer laser, and further includes: The main detector 4 is located on the first sampling branch branch of the beam splitting sampling unit 2, and its peak voltage output is converted into single pulse energy by the calibration curve. Reference detector 6 is located on the second sampling branch branch of beam splitting sampling unit 2. Its response speed is lower than that of main detector 4, and it outputs the cumulative energy reading of a batch of calibration pulses. Shutter 5 is located on the second sampling branch, and is opened during the calibration period and closed outside the calibration period, thus blocking the second sampling branch. The synchronous acquisition unit 7 is used to acquire the peak voltage of each pulse of the main detector 4 and the cumulative energy reading of the reference detector 6 under the same trigger time base; The controller 8 is used to open the shutter 5 during the interval of the burst signal or within a preset calibration window, measure the dark level of the main detector 4 in the zero light input state and use it as the bias correction amount, control the excimer laser to emit a preset number of calibration pulses with a constant discharge voltage, subtract the bias correction amount from the peak voltage of each pulse, convert and sum them through the calibration curve to obtain the total energy of the main detector 4, determine the gain correction coefficient by the ratio of the splitting ratio of the first sampling branch to the second sampling branch based on the cumulative energy reading of the reference detector 6, and then compare it with the total energy of the main detector 4. The calibration curve is updated with the gain correction coefficient and the bias correction amount while maintaining the nonlinear shape of the calibration curve, and then the shutter 5 is closed.
[0089] In some embodiments, continue reading Figure 2 As shown, the controller 8 includes a calibration control module 801, a calibration parameter calculation module 802, a health assessment module 803, and an energy closed-loop module 804. The calibration control module 801 is used to determine the calibration window and arrange the shutter opening and closing timing and dark level acquisition timing. The calibration parameter calculation module 802 is used to perform summation and comparison operations and solve for the gain correction coefficient and the bias correction amount; The health assessment module 803 is used to record the gain correction coefficient, the bias correction amount, the relative change of the gain correction coefficient relative to the previous calibration gain correction coefficient, and the responsivity drift rate obtained by normalizing the number of pulses emitted by the excimer laser between the current calibration and the previous calibration. The energy closed-loop module 804 is used to calculate the peak voltage output of the main detector 4 according to the updated calibration curve, and to adjust the discharge voltage of the excimer laser according to the calculated single-pulse energy.
[0090] It should be noted that the online calibration system for excimer laser pulse energy detectors provided in the above embodiments is only illustrated by the division of the above functional modules when performing online calibration. In actual applications, the above functions can be assigned to different functional modules as needed. That is, the internal structure of the online calibration system for excimer laser pulse energy detectors will be divided into different functional modules to complete all or part of the functions described above.
[0091] Furthermore, the online calibration system for excimer laser pulse energy detectors and the online calibration method for excimer laser pulse energy detectors provided in the above embodiments belong to the same concept. The specific way each module performs its operation has been described in detail in the method embodiments, and will not be repeated here.
[0092] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of the present invention.
Claims
1. An online calibration method for an excimer laser pulse energy detector, wherein the excimer laser output optical path is provided with a beam-splitting sampling unit, a main detector, and a reference detector, the main detector is located on a first sampling branch branched by the beam-splitting sampling unit, the peak voltage output of the main detector is converted into single-pulse energy via a calibration curve, and the reference detector has a lower response speed than the main detector and is located on a second sampling branch branched by the beam-splitting sampling unit, which is equipped with a shutter, characterized in that... include: The shutter is opened during the interval of the burst signal or within a preset calibration window, and the dark level of the main detector is measured in a zero light input state and used as the bias correction amount. The excimer laser is instructed to emit a preset number of calibration pulses at a constant discharge voltage, and the peak voltage of each pulse of the main detector and the cumulative energy reading of the reference detector are collected simultaneously. The peak voltage of each pulse is subtracted from the bias correction amount and then converted using the calibration curve. The energy of each pulse obtained from the conversion is summed to obtain the total energy of the main detector. The gain correction coefficient is determined by the ratio of the splitting ratio of the first sampling branch to the second sampling branch, which is then reduced to the aperture of the first sampling branch, to the total energy of the main detector based on the cumulative energy reading of the reference detector. The calibration curve is updated with the gain correction coefficient and the offset correction amount. The shutter is then closed to end the current calibration. The nonlinear shape of the calibration curve remains unchanged.
2. The online calibration method for excimer laser pulse energy detector according to claim 1, characterized in that, The main detector includes a photodiode and an amplification and integration circuit. The reference detector is a pyroelectric energy detector. The response stability of the reference detector is higher than that of the main detector. The shutter is closed outside the calibration period and blocks the second sampling branch.
3. The online calibration method for an excimer laser pulse energy detector according to claim 1, characterized in that, Measuring the dark level of the main detector includes: In the zero-light-input state, the output of the main detector is continuously sampled a preset number of times and the arithmetic average is taken. The baseline reading of the reference detector is acquired simultaneously under the zero light input state, and the baseline reading is subtracted from the cumulative energy reading of the reference detector.
4. The online calibration method for the excimer laser pulse energy detector according to claim 3, characterized in that, The updated calibration curve and the gain correction coefficient are as follows: In the formula, The pulse energy is calculated based on the updated calibration curve. The gain correction coefficient is... This is the bias correction amount. The peak voltage output by the main detector. The calibration curve represents the nonlinear mapping before the update. For the first The peak voltage of the calibration pulse. The preset number of calibration pulses. The difference between the cumulative energy reading of the reference detector and the baseline reading is calculated by subtracting the baseline reading from the cumulative energy reading of the reference detector. It is the ratio of the splitting ratio of the first sampling branch to that of the second sampling branch.
5. The online calibration method for an excimer laser pulse energy detector according to claim 1, characterized in that, Before opening the shutter during the interval of the burst signal or within a preset calibration window, the method further includes: The current calibration is initiated when the cumulative number of pulses emitted by the excimer laser since the last calibration reaches the preset calibration interval pulse number and the calibration window appears. The calibration window is the excimer laser's wafer change interval, batch change interval, or standby period. When the number of pulses that can be accommodated in the interval of a single burst signal is less than the preset number, the calibration is divided into multiple segments. Each segment emits a portion of the calibration pulses in the interval between adjacent burst signals, and the peak voltage of each pulse and the cumulative energy reading of the reference detector are collected respectively. The total energy of the main detector in each segment is summed, and the cumulative energy reading of the reference detector in each segment is summed and then compared.
6. The online calibration method for an excimer laser pulse energy detector according to claim 1, characterized in that, Updating the calibration curve includes: The product of the previously calibrated gain correction factor and the total energy of the main detector is used as the predicted value. The cumulative energy reading of the reference detector is reduced to the first sampling branch aperture and used as the measured value. The ratio of the absolute value of the difference between the predicted value and the measured value to the measured value is used as the comparison residual. When the comparison residual is not greater than the preset residual upper limit, the update is performed using the gain correction coefficient and the bias correction amount determined in this instance. When the comparison residual is greater than the upper limit of the residual, the update is performed using the gain correction coefficient and the bias correction amount of the previous calibration to keep the calibration curve unchanged and output an alarm.
7. The online calibration method for an excimer laser pulse energy detector according to claim 1, characterized in that, After determining the gain correction coefficient, the method further includes: The relative change of the gain correction coefficient relative to the previous calibration is normalized to the cumulative number of pulses emitted by the excimer laser between the current calibration and the previous calibration to obtain the responsivity drift per unit pulse: In the formula, The responsivity drift rate, The gain correction coefficient is... This is the gain correction factor from the last calibration. The cumulative number of pulses; When the responsivity drift rate is greater than a preset first threshold and this condition is met for a preset number of consecutive times, the window of the main detector is determined to be contaminated, a cleaning request or replacement request for the main detector is output, and the light output of the excimer laser is kept uninterrupted.
8. The online calibration method for an excimer laser pulse energy detector according to claim 7, characterized in that, Also includes: When the deviation of the gain correction coefficient from the initial gain correction coefficient is greater than a preset second threshold, the main detector is determined to be faulty, a replacement request for the main detector is output, and a degraded operation flag is set, wherein the initial gain correction coefficient is the gain correction coefficient corresponding to the calibration curve at the factory calibration time. When the degraded operation flag is set, the shutter is kept open, and the reference for adjusting the discharge voltage of the excimer laser using the single pulse energy as feedback is switched from the main detector to the reference detector. The pulse-by-pulse energy closed loop is then switched to a dose sliding window closed loop using the cumulative energy of a preset number of pulses as feedback.
9. An online calibration system for an excimer laser pulse energy detector, comprising a beam-splitting sampling unit disposed in the output optical path of the excimer laser, characterized in that, Also includes: The main detector is located on the first sampling branch of the beam-splitting sampling unit, and its peak voltage output is converted into single-pulse energy by the calibration curve. The reference detector, located on the second sampling branch of the beam-splitting sampling unit, has a lower response speed than the main detector and outputs the cumulative energy reading of a batch of calibration pulses. The shutter is located on the second sampling branch, and it is opened during the calibration period and closed and blocks the second sampling branch outside the calibration period. The synchronous acquisition unit is used to acquire the peak voltage of each pulse of the main detector and the cumulative energy reading of the reference detector under the same trigger time base; The controller is used to open the shutter during the interval of the burst signal or within a preset calibration window, measure the dark level of the main detector in a zero-light input state and use it as a bias correction, control the excimer laser to emit a preset number of calibration pulses with a constant discharge voltage, subtract the bias correction from the peak voltage of each pulse, convert and sum them through the calibration curve to obtain the total energy of the main detector, determine the gain correction coefficient by the ratio of the splitting ratio of the first sampling branch to the second sampling branch, and then calculate the ratio of the first sampling branch aperture to the total energy of the main detector based on the cumulative energy reading of the reference detector, and update the calibration curve with the gain correction coefficient and the bias correction while maintaining the nonlinear shape of the calibration curve unchanged, and then close the shutter.
10. The online calibration system for an excimer laser pulse energy detector according to claim 9, characterized in that, The controller includes a calibration control module, a calibration parameter calculation module, a health assessment module, and an energy closed-loop module. The calibration control module is used to determine the calibration window and arrange the opening and closing timing of the shutter and the dark level acquisition timing. The calibration parameter calculation module is used to perform summation and comparison operations and solve for the gain correction coefficient and the bias correction amount; The health assessment module is used to record the gain correction coefficient, the bias correction amount, the relative change of the gain correction coefficient relative to the previous calibrated gain correction coefficient, and the responsivity drift rate obtained by normalizing the number of pulses emitted by the excimer laser between the current calibration and the previous calibration. The energy closed-loop module is used to calculate the peak voltage output of the main detector according to the updated calibration curve, and to adjust the discharge voltage of the excimer laser according to the calculated single-pulse energy.
Citation Information
Patent Citations
Excimer laser device
CN108352674B
Control method and system for energy of double-cavity laser
CN113783097A
