Impedance and power matching method and device of plasma power supply

Abstract

The invention belongs to the technical field of semiconductor manufacturing, and provides an impedance and power matching method and device for a plasma power supply, and the method comprises the steps: constructing a cost function based on the plasma power and the resonant converter input impedance at a k + 1 moment, judging whether an angle switching frequency exists to enable F to be equal to 0, and calculating a corresponding optimal switching frequency; judging whether the cost function has a minimum value point with respect to the angle switching frequency, and if yes, calculating the optimal switching frequency corresponding to the minimum value of the cost function; otherwise, updating the cost function weight based on a simulated annealing algorithm. According to the invention, adaptive adjustment of the switching frequency is realized by searching the extreme value of the cost function, the input impedance fluctuation of the plasma power supply is effectively inhibited, stable power transition is realized, the risk of wafer breakdown possibly caused by sudden rise or sudden drop of the power after switching is avoided, and the response time of the power supply from a standby state to a working state is shortened.

Description

Technical Field

[0001] This invention relates to the field of semiconductor manufacturing technology, and more specifically to an impedance and power matching method and apparatus for a plasma power supply. Background Technology

[0002] In precision processes such as semiconductor dry etching and deposition, the generation and maintenance of plasma depend on a stable RF power supply. During the process, it is often necessary to switch between different power levels according to the requirements of each step, or to employ pulsed power modes to optimize process performance. However, the plasma load exhibits a high degree of nonlinearity, and its impedance undergoes drastic and rapid changes during power startup (ignition), shutdown (extinguishing), and transient transitions between different power levels. This dynamic impedance abrupt change makes it difficult for traditional impedance matching networks and power control loops to achieve real-time and accurate power matching, leading to a series of key technical challenges: First, as a nonlinear time-varying load, plasma's impedance undergoes drastic and rapid changes during power-on (ignition), power-off (extinguishing), and switching between different power levels. Traditional impedance matching networks and power feedback control loops, due to their inherent limitations in response bandwidth, adjustment speed, and algorithm lag, struggle to achieve real-time and accurate power matching during such transient processes. This "mismatch" state leads to a significant amount of reflected power, causing the actual net power delivered to the plasma to deviate severely from the set value.

[0003] Secondly, due to the aforementioned lack of dynamic matching capability, power "overshooting" or "undershooting" phenomena are common during the transient process of power switching. Power overshooting (i.e., the actual peak power far exceeds the set value) will cause excessive energy impact on the wafer in the cavity, significantly increasing the risk of damage to device feature structures or electrical breakdown; while power undershooting (i.e., the actual power cannot quickly reach the set value) may cause the plasma to be accidentally extinguished due to interruption or instability of energy supply. Once the plasma is extinguished, not only does the current process step fail, but the entire process cavity state is reset, seriously damaging the continuity, repeatability, and consistency between wafers of the process.

[0004] Furthermore, existing systems often require lengthy adjustment and stabilization times to cope with wide variations in load impedance and re-establish a stable matching point. This forces process designs to incorporate longer waiting times, reducing equipment throughput and production efficiency. Simultaneously, to avoid the lengthy and uncertain process of re-ignition caused by a complete power shutdown during process intervals, existing solutions typically require maintaining a high base power during standby to "keep the plasma alive," resulting in continuous energy waste and additional thermal management burdens.

[0005] In summary, the core deficiency of existing technologies lies in their control architecture and matching mechanisms' inability to adapt to the extreme nonlinear impedance changes of plasma loads during transient processes. This leads to a series of interconnected problems, including insufficient power control accuracy, poor process stability, wafer damage risk, low production efficiency, and high energy consumption. This has become a key bottleneck restricting the further development of advanced plasma processes, especially pulsed processes requiring high-frequency, rapid power modulation. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an impedance and power matching method and apparatus for plasma power supplies, thereby solving the problem of insufficient power control accuracy in current plasma power supplies, which leads to poor process stability and the risk of wafer damage.

[0007] In a first aspect, the present invention provides an impedance and power matching method for a plasma power supply, the plasma power supply comprising an inverter circuit, a resonant converter, a matching circuit, and a reaction chamber. Inverter circuit, connected to DC voltage source; A resonant converter converts the output of an inverter circuit into an AC source. A matching circuit is used to match the input impedance of a resonant converter with the actual load impedance. The method includes: S1. Based on k Plasma power at time +1 and resonant converter input impedance Construct the cost function; the cost function is... , For reference power, As the reference impedance, The weights are the cost function weights; S2. Determine if an angle switch frequency exists. If F=0, and it exists, calculate the corresponding optimal switching frequency and determine the optimal driving sequence in conjunction with the duty cycle; otherwise, proceed to S3. S3. Determine the cost function with respect to the angle switch frequency. If a minimum point exists, calculate the optimal switching frequency corresponding to the minimum value of the cost function, and determine the optimal driving sequence in conjunction with the duty cycle; otherwise, proceed to S4. S4. Update the cost function weights based on the simulated annealing algorithm. ,include: If the updated cost function weights are within the number of iterations of the simulated annealing algorithm... The optimal switching frequency corresponding to the minimum value of the cost function is output, and the optimal driving sequence is determined in conjunction with the duty cycle; otherwise, the output sequence is substituted into the corresponding cost function for rolling optimization, and the sequence that minimizes the output of the cost function is determined as the optimal sequence.

[0008] Optionally, the plasma power at time k+1: , in, for k The output current of the resonant converter at time +1 for k The output voltage of the resonant converter at time +1 For current With voltage The vector angle; , , This is the inverter output voltage. The characteristic impedance of the equivalent circuit. The quality factor of the equivalent circuit is given by j, where j is the imaginary unit. The inductance ratio of the resonant converter; .

[0009] Optionally, the input impedance at time k+1: .

[0010] Optionally, determine the cost function with respect to Does a local minimum exist, including: With respect to the cost function F Find the first and second partial derivatives if the cost function simultaneously satisfies and If , then there exists a local minimum point.

[0011] Optionally, the cost function weights are updated based on the simulated annealing method, including: Step T1: Initialize parameters and set the initial temperature. T 0. Termination temperature T final and current temperature T , d The initial state is randomly generated, representing the temperature decay parameter. The number of iterations for each temperature is L ;in Including the weights of the cost function ; Step T2: Check the state Apply perturbation Generate a new state ,but Disturbance Following a uniform distribution, the corresponding cost function is given by Become Calculation formula: ; The cost function is expressed as: ; in, It is the model prediction value Compared with the true value The losses between N It is the number of samples; Step T3: If Δ F If the value is less than 0, then the next state value is... If Δ F >0, then ; Then, it is compared with the generated random number in the range (0,1). ξ Compare, if ξ < p The next state value is Otherwise, the next state value is still 0. ; Step T4: Determine if the number of iterations has been reached. If the number of iterations has been reached, determine if the cost function has a minimum value. Otherwise, repeat steps T2 and T3. If a minimum value exists, output the optimal cost function weights. ; If no minimum value exists, then the temperature decay parameter is used. d Gradually decrease the current temperature T Repeat steps T2 and T3 until... T = T final Obtain the optimized cost function weights .

[0012] In a second aspect, the present invention provides an impedance and power matching device for a plasma power source, which, when executing a computer program, implements the steps of the method described in any possible implementation of the first aspect.

[0013] By adopting the above technical solution, this application has the following beneficial effects: This invention achieves adaptive adjustment of the switching frequency by searching for the extreme value of the cost function, effectively suppressing input impedance fluctuations and realizing smooth power transition. It solves the power matching delay problem caused by large impedance fluctuations during the start-up and shutdown switching of plasma power supplies, and can control the peak power adjustment accuracy within ±1%. This invention achieves a smooth power transition, thereby effectively avoiding the risk of wafer breakdown that may be caused by a sudden increase or decrease in power after switching, while ensuring that the plasma remains stable and does not extinguish during pulse switching; This invention shortens the response time of the RF power supply from standby to working state, can maintain plasma stability in low power mode, and accurately execute process actions in high power mode, thus taking into account both the requirements of state maintenance and efficient operation. Attached Figure Description

[0014] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0015] Figure 1 A matching flowchart of a plasma power supply provided by an embodiment of the present invention is shown; Figure 2 A control block diagram of a plasma power supply is shown; Figure 3 One of the flowcharts of an impedance and power matching method for a plasma power source provided in an embodiment of the present invention is shown; Figure 4 The second flowchart illustrates an impedance and power matching method for a plasma power source provided in an embodiment of the present invention. Figure 5 This invention provides a flowchart illustrating the process of updating the cost function weights using a simulated annealing method. Figure 6 A schematic diagram of an impedance and power matching device for a plasma power source provided in an embodiment of the present invention is shown. Detailed Implementation

[0016] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of the present invention and are therefore merely examples, and should not be construed as limiting the scope of protection of the present invention. It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0017] Figure 1 A plasma power supply is shown, including an inverter circuit, a resonant converter, a matching circuit, and a reaction chamber; the inverter circuit is connected to a DC voltage source. The inverter circuit can be a half-bridge or full-bridge inverter circuit, and the inverter output voltage is... The resonant converter converts the output of the inverter circuit into an AC source, and the inductor... , and capacitor Constructing a resonant converter, the resonant current is The matching network is used to match the input impedance of the resonant converter with the actual load impedance. When the matching is successful, the impedance presented by the matching network is equal to the complex conjugate of the load impedance. At this time, the entire system behaves as a purely resistive load, and energy transfer reaches its optimal state. The reaction chamber is used to ionize gas molecules to form a high-density plasma.

[0018] like Figure 2 As shown, after the plasma power supply is powered on, the system first sets the gas current excitation threshold. i ref The excitation gas is then introduced, and the phase adjustment module adjusts the resonant output current reference value 50ms before ignition. i ref Instantly dropped to the current reference value i pmin This can effectively prevent surge current and protect switching power devices; The adjusted resonant output current reference value i ref The current value sampled by the power circuit i pri After comparison, the output current error e 1. Output duty cycle after current PI modulation d To reduce current error e 1. The switching frequency is obtained by passing the PI modulator and inverting the PI. f s Finally, the inverter switching transistors are output via PWM modulation and the drive module. D 1~ D 4.

[0019] To address the power matching delay issue caused by significant impedance fluctuations during the start-up and shutdown switching of plasma power supplies, an impedance and power matching method for plasma power supplies is proposed. This method adaptively adjusts the switching frequency to output an optimal drive sequence, effectively suppressing input impedance fluctuations and achieving smooth power transition. Figure 3-4 As shown, the method includes: S1. Based on k Plasma power at time +1 and resonant converter input impedance Construct the cost function; the cost function is... , For reference power, As the reference impedance, These are the weights of the cost function.

[0020] For example Figure 1 For the LCL resonant converter shown, due to the resonant effect of the converter, the resonant angular frequency of the converter is defined. and switching angular frequency Normalized value of switching angular frequency They are respectively: ; (1) The equivalent resistance and the root mean square value of the fundamental component of the input square wave voltage are expressed as follows: ; (2) Due to the resonant effect of the converter, the resonant frequency of the converter is defined. and switching frequency normalized value They are respectively: ; (3) Characteristic impedance of equivalent circuit And quality factor Q ; (4) The inductance ratio is: (5) The input impedance of the LCL resonant converter Z i The expression is: (6) It can be seen that the primary winding current i pri and voltage V loop They are respectively: (7) (8) From equation (8), the voltage gain can be obtained. M and current gain H They are respectively: (9) (10) Switching frequency at time k+1 It can be obtained by fitting the arctangent function: (11) in, t k+1 The cumulative sampling time at time k+1 t 0 represents the initial sampling time. f 0 represents the resonant frequency during measurement; A represents the time-varying frequency factor; and B represents the influence factor of the switching frequency change rate. Both A and B are obtained directly through fitting.

[0021] Furthermore, the first k Switching frequency at time +1 Available k Switching frequency at any time To indicate: (12) Similarly, the first k Switching frequency at any time Available k Switching frequency at time -1 To represent, that is, using the first k -1 hour to the k Predicting the trend of switching frequency at any given time to predict the first k Switching frequency at time +1 f s ( k +1); then, the first k Angle switching frequency at time +1 .

[0022] It can be predicted from equation (7) k Primary winding current at +1 time i pri ( k +1): (13) Then in k Actual power at time +1 P s ( k+ 1) Can be sampled using plasma voltage / current i pri ( k+ 1) and V loop ( k+ 1) to represent: (14) in, For current With voltage The vector angle.

[0023] From equations (9) and (10) above, it can be seen that the input impedance Z i The stability of the voltage and current gains of the LCL resonant converter is affected. Considering the impact of power stability on the etching uniformity of the RF power supply, the following cost function is designed: (15) in, The weights are the cost function weights. P ref , Z ref These are the reference plasma power and the input impedance, respectively.

[0024] k The input impedance at time +1 is as follows: .

[0025] S2. Determine if an angle switch frequency exists. If F=0, and it exists, calculate the corresponding optimal switching frequency and determine the optimal driving sequence in conjunction with the duty cycle; otherwise, proceed to S3.

[0026] To ensure that the cost function outputs a minimum value, we first determine if it exists. Make F =0, where, Set the cost function weights to default values ​​if they exist. If the cost function is made to be 0, then the corresponding switching frequency is calculated, which is the optimal frequency corresponding to minimizing the cost function; if no such frequency exists... To make the cost function equal to 0, it is necessary to determine whether the cost function has a minimum value that minimizes the cost function value.

[0027] S3. Determine the cost function with respect to the angle switch frequency. If a minimum point exists, calculate the optimal switching frequency corresponding to the minimum value of the cost function, and determine the optimal driving sequence in conjunction with the duty cycle; otherwise, proceed to S4.

[0028] In step S3, the cost function F is adjusted with respect to... Find the first and second partial derivatives if the cost function simultaneously satisfies and Then there exists a local minimum point; including: To determine whether the cost function has an extreme point, for the cost function... F about Find the first-order partial derivative; (16) make The solution involves determining whether an extreme point exists, and if so, whether it is a minimum point, based on the cost function. F about Taking the second-order partial derivative, we get: (17) Determine the cost function F about Check if the second-order partial derivative is greater than 0 to confirm whether the extreme value is a local minimum; if a local minimum exists, then use... and Given two conditions, calculate the optimal switching frequency corresponding to the minimum value of the cost function.

[0029] S4. Update the cost function weights based on the simulated annealing algorithm. ,include: If the updated cost function weights are within the number of iterations of the simulated annealing algorithm... The optimal switching frequency corresponding to the minimum value of the cost function is output, and the optimal driving sequence is determined in conjunction with the duty cycle; otherwise, the output sequence is substituted into the corresponding cost function for rolling optimization, and the sequence that minimizes the output of the cost function is determined as the optimal sequence.

[0030] In S4, the cost function weights are updated based on the simulated annealing method. Utilizing the Metropolis criterion within the annealing algorithm, the global optimum can be effectively found, improving recognition accuracy and avoiding local optima. The idea of ​​simulated annealing (SA) was first proposed by Metropolis et al., such as... Figure 5 As shown, the steps are as follows: Step T1: Initialize parameters and set the initial temperature. T 0. Termination temperature T final and current temperature T , d The initial state is randomly generated, representing the temperature decay parameter. The number of iterations for each temperature is L ;in Including the weights of the cost function ; Step T2: Check the state Apply perturbation Generate a new state ,but Disturbance Following a uniform distribution, the corresponding cost function is given by Become Calculation formula: ; The cost function is expressed as: ; in, It is the model prediction value Compared with the true value The losses between N It is the number of samples; Step T3: If Δ F If the value is less than 0, then the next state value is... If Δ F >0, then ; Then, it is compared with the generated random number in the range (0,1). ξ Compare, if ξ < p The next state value is Otherwise, the next state value is still 0. ; Step T4: Determine if the number of iterations has been reached. If the number of iterations has been reached, determine if the cost function has a minimum value; otherwise, repeat steps T2 and T3. The condition for determining the minimum value remains the same. and ; If a minimum value exists, output the optimal weight. ; If no minimum value exists, then the temperature decay parameter is used. d Gradually decrease the current temperature T Repeat steps T2 and T3 until... T = T final Obtain optimized weights .

[0031] If the cost function weight update count reaches its maximum, it indicates that there is neither a switching frequency in the current switching cycle that makes the cost function zero, nor a switching frequency that makes the cost function close to a minimum value. The model prediction algorithm is then used to iterate the driving signal, i.e., the output sequence... , , , Substituting these values ​​into the corresponding cost function for rolling optimization, then, by k Time to k The switching frequency can be calculated from the driving signal at time +1, which is the optimal switching frequency corresponding to the minimum value of the cost function.

[0032] In summary, when the cost function can be determined to have a minimum / maximum value, the optimal angle switching frequency corresponding to the minimum / maximum value should be determined. ,pass Convert the corresponding switching frequency By coordinating the duty cycle with the optimal drive sequence, the switching transistors of the full-bridge inverter circuit are controlled to turn on and off; when the cost function cannot determine the minimum / maximum value, the output sequence is... , , , Substitute the corresponding cost function and determine the sequence that minimizes the output of the cost function as the optimal driving sequence, thereby controlling the on / off state of the switching transistors in the full-bridge inverter circuit.

[0033] Among them, the output sequence , , , They represent: This indicates that switches S1 and S3 are on; This indicates that switches S1 and S4 are on. This indicates that switches S2 and S3 are conducting. This indicates that switches S2 and S4 are conducting.

[0034] In one embodiment, an impedance and power matching device for a plasma power supply is provided, such as... Figure 6 As shown, the steps for performing the aforementioned impedance and power matching method for plasma power sources are described to achieve the same technical effect as the aforementioned method.

[0035] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for impedance and power matching of a plasma power source, characterized in that, A plasma power source includes an inverter circuit, a resonant converter, a matching circuit, and a reaction chamber. Inverter circuit, connected to DC voltage source; A resonant converter converts the output of an inverter circuit into an AC source. A matching circuit is used to match the input impedance of a resonant converter with the actual load impedance. The method includes: S1. Based on k Plasma power at time +1 and resonant converter input impedance Construct the cost function; the cost function is... , For reference power, As the reference impedance, The weights are the cost function weights; S2. Determine if an angle switch frequency exists. If F=0, and it exists, calculate the corresponding optimal switching frequency and determine the optimal driving sequence in conjunction with the duty cycle; otherwise, proceed to S3. S3. Determine the cost function with respect to the angle switch frequency. If a minimum point exists, calculate the optimal switching frequency corresponding to the minimum value of the cost function, and determine the optimal driving sequence in conjunction with the duty cycle; otherwise, proceed to S4. S4. Update the cost function weights based on the simulated annealing algorithm. ,include: If the updated cost function weights are within the number of iterations of the simulated annealing algorithm... The optimal driving sequence is determined by finding the minimum value of the cost function and outputting the optimal switching frequency corresponding to the minimum value, along with the duty cycle. Otherwise, the output sequences are substituted into the corresponding cost functions for rolling optimization, and the sequence that minimizes the output of the cost function is determined as the optimal driving sequence.

2. The method according to claim 1, characterized in that, Plasma power at time k+1: ， in, for k The output current of the resonant converter at time +1 for k The output voltage of the resonant converter at time +1 For current With voltage The vector angle; ， ， This is the inverter output voltage. The characteristic impedance of the equivalent circuit. The quality factor of the equivalent circuit is given by j, where j is the imaginary unit. The inductance ratio of the resonant converter; 。 3. The method according to claim 2, characterized in that, Input impedance at time k+1: 。 4. The method according to claim 2, characterized in that, Determine the cost function with respect to Does a local minimum exist, including: With respect to the cost function F Find the first and second partial derivatives if the cost function simultaneously satisfies and If , then there exists a local minimum point.

5. The method according to claim 2, characterized in that, The cost function weights are updated based on the simulated annealing method, including: Step T1: Initialize parameters and set the initial temperature. T 0. Termination temperature T final and current temperature T , d The initial state is randomly generated, representing the temperature decay parameter. The number of iterations for each temperature is L ;in Including the weights of the cost function ; Step T2: Check the state Apply perturbation Generate a new state ,but Disturbance Following a uniform distribution, the corresponding cost function is given by Become Calculation formula: ； The cost function is expressed as: ； in, It is the model prediction value Compared with the true value The losses between N It is the number of samples; Step T3: If Δ F If the value is less than 0, then the next state value is... If Δ F >0, then ; Then, it is compared with the generated random number in the range (0,1). ξ Compare, if ξ < p The next state value is Otherwise, the next state value is still 0. ; Step T4: Determine if the number of iterations has been reached. If the number of iterations has been reached, determine if the cost function has a minimum value. Otherwise, repeat steps T2 and T3. If a minimum value exists, output the optimal cost function weights. ; If no minimum value exists, then the temperature decay parameter is used. d Gradually decrease the current temperature T Repeat steps T2 and T3 until... T = T final Obtain the optimized cost function weights .

6. An impedance and power matching device for a plasma power source, characterized in that, When used to execute a computer program to implement the steps of the method according to any one of claims 1 to 5.

Images