Dynamic voltage control method for anodic bonding process

Abstract

The invention relates to the technical field of MEMS (Micro Electro Mechanical System) manufacturing, in particular to a dynamic voltage control method for an anodic bonding process, which comprises the following steps of: setting an initial voltage value to be 50-70% of a fixed voltage value, and monitoring a bonding interface resistance value, interface temperature field distribution and local stress change in real time; fuzzy PID and sliding mode variable structure control are fused, control parameters are dynamically adjusted according to current density deviation and temperature overshoot, voltage output ripples are made to be smaller than + / -0.5%, if the local temperature of an electrode exceeds 90% of a material tolerance threshold value, a reverse voltage pulse sequence is activated, and if the local temperature of the electrode exceeds 90% of the material tolerance threshold value, voltage level response is carried out; according to the invention, through multi-mode feedback monitoring and layered dynamic regulation and control, the voltage can be adjusted in real time, and uneven current distribution and local overheating are avoided, so that the risk of electrode burnout is remarkably reduced, and the bonding reliability is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of MEMS manufacturing, and more particularly to a dynamic voltage control method for an anodic bonding process. Background Art

[0002] Anodic bonding is a key microfabrication connection technology in the semiconductor and MEMS fields. It can firmly bond two or more different materials, such as silicon wafers and glass, metal and glass, etc. In the standard anodic bonding process, metal electrodes are required to apply voltage to generate an electrostatic field and heat the bonding interface at the same time. When the voltage rises, it can drive the movement of charged ions and ultimately achieve atomic-level bonding between materials.

[0003] However, when equipment like EVG bonding machines perform metal electrode anodic bonding, metal electrodes often burn out. Once the electrode burns out, the reliability of the bond is greatly reduced, and the device performance is also affected. Uneven current distribution is the primary factor for electrode burnout. Ideally, the current should be evenly distributed over the entire bonding area, but in reality, it is difficult to achieve perfection. If any link is not done well, such as electrode design, contact conditions, and wafer surface conditions, it may cause uneven current distribution. In particular, when the local current density is high, the Joule heat will increase, and the heat dissipation cannot keep up. The local temperature is likely to exceed the limit, exceeding the tolerance limit of the electrode material, and burnout is inevitable.

[0004] Anodic bonding itself is a heating process. The electric field force drives the movement of ions, and the current passes through the resistor to generate heat (Joule heat). In addition, if the voltage is too high, the time is too long, or the cooling efficiency is not good, heat will easily accumulate. Especially when the current is unevenly distributed, heat is more likely to drill into the area with high current density, and local overheating will be more serious.

[0005] There is also the traditional anodic bonding process, in which the voltage control generally uses a constant voltage mode, and the voltage does not change during the entire bonding process. Using this method, in the early stages of the process, especially when the resistance of the bonding interface is still relatively large, the current rises too quickly, causing local overheating of the electrode and burning. Summary of the invention

[0006] To this end, the purpose of the present invention is to provide a dynamic voltage control method for the anodic bonding process to solve the difficult problem that the metal electrode is easily burned out during the anodic bonding process. By optimizing the process parameter control method and the overall operation process, the risk of electrode burning out is fundamentally reduced to ensure the stability and reliability of the anodic bonding process.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A dynamic voltage control method for an anodic bonding process comprises the following steps:

[0009] S1. Setting the initial voltage: setting the initial voltage value to 50%-70% of the fixed voltage value, the initial voltage range is 200V-300V;

[0010] S2. Multi-modal feedback monitoring: real-time monitoring of bonding interface resistance, interface temperature field distribution and local stress changes;

[0011] S3. Layered dynamic control: voltage level response according to interface resistance drop rate and temperature gradient:

[0012] S31, when the interface resistance decrease rate is higher than the first threshold, start the linear voltage ramp-up mode and increase the voltage in proportion to the rate;

[0013] S32, when the resistance drops to the preset point, the step-by-step pressure increase is triggered, and each step decreases by 20%-30%;

[0014] S33, when the interface temperature field uniformity index is lower than 0.9, switch to pulse voltage control to alleviate local heat accumulation;

[0015] S4, Adaptive fuzzy sliding mode control: Fuzzy PID is integrated with sliding mode variable structure control to dynamically adjust control parameters according to current density deviation and temperature overshoot, so that the voltage output ripple is less than ±0.5%.

[0016] S5, self-repair logic trigger: If the local temperature of the electrode exceeds 90% of the material tolerance threshold, the reverse voltage pulse sequence is activated to offset the inertia of ion migration and reinitialize the current control cycle.

[0017] The present invention is further configured as follows: the preset quantiles are 80%, 50% and 20% of the initial value.

[0018] The present invention is further configured as follows: each step voltage increment of the step boost in step S3 satisfies ,in For the The voltage increment of each step; is the attenuation factor of the thermal conductivity of the material; is the maximum voltage; is the ladder sequence number; is an exponential decay term.

[0019] The present invention is further configured as follows: the step-by-step boost stage also includes a degradation mechanism: if the resistance fluctuation amplitude is monitored to exceed 10% of the value corresponding to the current voltage, the voltage is returned to the previous step and the dwell time is extended by 20%-50%.

[0020] The present invention is further configured as follows: the frequency of the pulse voltage control is 10 Hz-50 Hz, and the rising edge and falling edge time of a single pulse does not exceed 1 ms.

[0021] The present invention is further configured as follows: the single pulse amplitude of the reverse voltage pulse sequence is 30%-60% of the forward voltage, the pulse width is 0.2-0.5 times of the forward period, and the sequence duration is 10ms-50ms.

[0022] The present invention is further configured as follows: the sliding mode surface function of the fuzzy sliding mode control in step S4 is defined as: ;

[0023] in is the voltage deviation, and The fuzzy rule base dynamically adjusts according to the uniformity of the temperature field and the depressurization rate.

[0024] The present invention is further configured as follows: the first threshold value is 70%-80% of the initial bonding interface resistance value.

[0025] The present invention is further configured as follows: the multimodal feedback monitoring includes:

[0026] An infrared thermal imager is used to obtain the temperature field distribution of the bonding interface at a sampling rate of 100Hz-200Hz;

[0027] The interface resistance value was measured by the four-probe method, and the measuring electrode spacing was less than 1 mm and was distributed in concentric circles with the bonding area;

[0028] The interface stress changes were monitored using a piezoelectric sensor array.

[0029] The present invention is further configured to synchronously start the electrode contact pressure adjustment when returning to the previous step voltage:

[0030] The contact pressure between the electrode and the wafer is dynamically adjusted by the pneumatic servo mechanism, and the pressure adjustment range is 5N-20N;

[0031] The pressure adjustment rate is positively correlated with the resistance fluctuation amplitude, satisfying ,in Indicates the pressure regulation rate; is the resistance fluctuation amplitude; is the initial resistance value, is the adjustment coefficient, and ;

[0032] After the pressure stabilizes, it needs to be maintained for at least 3 step cycles before the voltage can be triggered to increase again.

[0033] Compared with the shortcomings of the prior art, the beneficial effects of the present invention are:

[0034] Through multimodal feedback monitoring and hierarchical dynamic regulation, the voltage can be adjusted in real time to avoid uneven current distribution and local overheating, thereby significantly reducing the risk of electrode burnout and improving bonding reliability.

[0035] The dynamic voltage control strategy can make adaptive adjustments based on interface resistance changes and temperature gradients, shorten bonding time, and improve production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a process flow chart of the present invention;

[0037] Figure 2 Flow chart for implementing the present invention. DETAILED DESCRIPTION

[0038] Reference Figure 1 to Figure 2 The present invention further describes a dynamic voltage control method for an anodic bonding process, which includes the following steps:

[0039] The first step is to set the initial voltage: set the initial voltage value to 50%-70% of the fixed voltage value.

[0040] Among them, the initial voltage value is dynamically calculated according to the characteristics of the bonding material, and the formula is ;

[0041] in =50%-70%, is a fixed voltage value (standard bonding voltage of the material), generally 400-600V, so the initial voltage range is 200V-300V. The system loads a material database containing 30 common MEMS materials. Recommended values ​​and thermal expansion coefficient matching parameters.

[0042] The second step is multimodal feedback monitoring: a FLIR A65 infrared camera (thermal sensitivity <0.05°C) is used to capture the 50mm diameter bonding area at a sampling rate of 100-200Hz to generate a 128×128 pixel temperature matrix. The noise is eliminated by Gaussian filtering, and the temperature field uniformity index U is calculated. T , when U T <0.9 trigger pulse control.

[0043] A tungsten probe group with a concentric circle layout (inner circle diameter 2mm, outer circle diameter 4mm) was used, a 10mA constant current source was applied, and the interface resistance value was extracted through a phase-locked amplifier, with a measurement accuracy of ±0.1Ω.

[0044] 16 piezoelectric sensors (PZT-5H) are embedded in the bonding stage and distributed in a 4×4 matrix to detect local stress changes in real time and generate stress gradient vector diagrams.

[0045] The third step is layered dynamic regulation: voltage level response is performed according to the interface resistance decrease rate and temperature gradient: when the interface resistance decrease rate is higher than the first threshold, which is 70%-80% of the initial bonding interface resistance value, the linear voltage ramp-up mode is started, and the voltage is increased in proportion to the rate; the maximum boost rate is limited to 50V / s to prevent thermal shock.

[0046] Linear voltage ramp-up mode: ,in, is the rate of change of resistance with time, is the initial resistance, is the threshold resistance, is the target bonding time, which is preset to 120s, and the voltage increment formula is: (k=0.15V / Ω / s), if the resistance decrease rate is 10Ω / s, then the voltage is increased linearly at 1.5V / s until Threshold or U T <0.9.

[0047] When the resistance drops to the preset quantiles, which are 80%, 50% and 20% of the initial value, the step-by-step pressure increase is triggered, and each step decreases by 20%-30%;

[0048] Each step voltage increment of the step boost meets ,in For the The voltage increment of each step; is the attenuation factor of the thermal conductivity of the material; is the maximum voltage; is the ladder sequence number; is an exponential decay term.

[0049] The step-up voltage stage also includes a degradation mechanism: if the resistance fluctuation amplitude is detected to exceed 10% of the current voltage corresponding value, it will return to the previous step voltage and extend the dwell time by 20%-50%, such as dropping from step V3 to V2.

[0050] When returning to the previous step voltage, the electrode contact pressure adjustment is started synchronously: specifically, the contact pressure between the electrode and the wafer is dynamically adjusted through the air pressure servo mechanism, and the pressure adjustment range is 5N-20N;

[0051] The pressure adjustment rate is positively correlated with the resistance fluctuation amplitude, satisfying ,in Indicates the pressure regulation rate; is the resistance fluctuation amplitude; is the initial resistance value, is the adjustment coefficient, and ;

[0052] After the pressure stabilizes, it needs to be maintained for at least 3 step cycles before the voltage can be triggered to increase again.

[0053] When the interface temperature field uniformity index U T When it is lower than 0.9, it switches to pulse voltage control to alleviate local heat accumulation; the frequency of pulse voltage control is 10Hz-50Hz, and the rising and falling edge time of a single pulse does not exceed 1ms.

[0054] The fourth step is adaptive fuzzy sliding mode control: integrating fuzzy PID with sliding mode variable structure control, dynamically adjusting control parameters according to current density deviation and temperature overshoot, so that the voltage output ripple is less than ±0.5%.

[0055] The sliding surface function of fuzzy sliding mode control is defined as: ;

[0056] in is the voltage deviation, and The fuzzy rule base dynamically adjusts according to the uniformity of the temperature field and the depressurization rate.

[0057] Fuzzy rule library table:

[0058]

[0059] The fifth step is the self-repair logic trigger: if the local temperature of the electrode exceeds 90% of the material tolerance threshold (assuming 90% of the material tolerance threshold of 330°C), the reverse voltage pulse sequence is activated to offset the inertia of ion migration and reinitialize the current control cycle. The single pulse amplitude of the reverse voltage pulse sequence is 30%-60% of the forward voltage, the pulse width is 0.2-0.5 times the forward cycle, and the sequence duration is 10ms-50ms. After the reverse pulse ends, the voltage is reset to 80% of the value before the trigger, and the step-by-step boost is performed again according to step three.

[0060] For example: Reverse pulse amplitude V rev is the positive pulse amplitude V forward 50%. When V forward = 300V, V rev =-150V. The reverse pulse width is the forward pulse frequency f forward The reciprocal 30%, i.e. 1 / f forward ×0.3 = 30ms (assuming the original frequency f forward = 10Hz). The total duration of the reverse pulse train is 30ms, including 3 reverse pulses.

[0061] The silicon-glass anodic bonding (20×20mm² bonding area) was tested with the present method, conventional constant voltage mode, conventional step-up voltage mode and conventional PID control at an ambient temperature of 25±1℃ and humidity <30%RH.

[0062] Equipment configuration:

[0063] External thermal imager: FLIR A65, temperature resolution 0.05°C;

[0064] Voltage source: Keysight B2900A, accuracy ±0.1%;

[0065] Pressure sensor: Kistler 9317B, range 0-50N, error ±0.5%.

[0066] Key parameters: Initial voltage: 300V, =0.25.

[0067] Traditional step-up mode detection method:

[0068] Step 1: Set a fixed voltage of 300V without any voltage adjustment; start the bonding process, monitor the interface resistance value, interface temperature field distribution and local stress changes in real time; record the bonding time, temperature uniformity index (UT), residual stress and whether the electrode is burned out.

[0069] Traditional step-boost mode detection method: set the initial voltage to 300V; start the bonding process, and monitor the interface resistance value, interface temperature field distribution and local stress changes in real time; when the resistance drops to the preset quantile (80%, 50%, 20% of the initial value), increase the voltage in step-boost mode, and each step decreases by 20%-30%; record the bonding time, temperature uniformity index (UT), residual stress, and whether the electrode is burned out.

[0070] Conventional PID control detection method: set the initial voltage to 300V; start the bonding process, and monitor the interface resistance value, interface temperature field distribution and local stress changes in real time; dynamically adjust the voltage according to the PID control algorithm to make the voltage output ripple less than ±0.5%; record the bonding time, temperature uniformity index (UT), residual stress, and whether the electrode is burned out.

[0071] The comparison table of this method and the traditional constant pressure mode, traditional step-up pressure mode and conventional PID control is as follows:

[0072]

[0073] Based on the above experimental data:

[0074] Bonding time: This method is 41.7% shorter than the traditional constant pressure mode (180s→105s);

[0075] Improved temperature uniformity: The uniformity index of this method is 0.9, which is 15.79 higher than that of conventional PID (0.85); Residual stress control: The threshold of <18MPa is much lower than that of the traditional constant pressure mode (>50MPa);

[0076] Breakthrough in scrap rate: 0.9% scrap rate is a new record in the industry (8.7% in traditional mode);

[0077] Improved voltage stability: 6% ripple factor is 50% better than conventional PID (1.2%);

[0078] Energy economy: Energy consumption is 3kW·h / batch, which is 21% lower than the step-up mode (3.8).

[0079] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Common changes and substitutions made by those skilled in the art within the scope of the technical solution of the present invention should be included in the protection scope of the present invention.

Claims

1. A dynamic voltage control method for an anodic bonding process, characterized in that: The following steps are involved: S1. Setting the initial voltage: setting the initial voltage value to 50%-70% of the fixed voltage value, the initial voltage range is 200V-300V; S2. Multi-modal feedback monitoring: real-time monitoring of bonding interface resistance, interface temperature field distribution and local stress changes; S3. Layered dynamic control: voltage level response according to interface resistance drop rate and temperature gradient: S31, when the interface resistance decrease rate is higher than the first threshold, start the linear voltage ramp-up mode and increase the voltage in proportion to the rate; S32, when the resistance drops to the preset point, the step-by-step pressure increase is triggered, and each step decreases by 20%-30%; S33, when the interface temperature field uniformity index is lower than 0.9, switch to pulse voltage control to alleviate local heat accumulation; S4, Adaptive fuzzy sliding mode control: Fuzzy PID is integrated with sliding mode variable structure control to dynamically adjust control parameters according to current density deviation and temperature overshoot, so that the voltage output ripple is less than ±0.5%; S5, self-repair logic trigger: If the local temperature of the electrode exceeds 90% of the material tolerance threshold, the reverse voltage pulse sequence is activated to offset the inertia of ion migration and reinitialize the current control cycle.

2. A dynamic voltage control method for anodic bonding process according to claim 1, characterized in that: The preset quantiles are 80%, 50% and 20% of the initial value.

3. A dynamic voltage control method for anodic bonding process according to claim 2, characterized in that: Each step voltage increment of the step boost in step S3 satisfies ,in For the The voltage increment of each step; is the attenuation factor of the thermal conductivity of the material; is the maximum voltage; is the ladder sequence number; is an exponential decay term.

4. The dynamic voltage control method for anodic bonding process according to claim 1, characterized in that: The step-up voltage stage also includes a degradation mechanism: if the resistance fluctuation amplitude is detected to exceed 10% of the value corresponding to the current voltage, the voltage will be returned to the previous step and the dwell time will be extended by 20%-50%.

5. A dynamic voltage control method for anodic bonding process according to claim 4, characterized in that: The frequency of the pulse voltage control is 10 Hz-50 Hz, and the rising edge and falling edge time of a single pulse does not exceed 1 ms.

6. The method for dynamic voltage control in an anodic bonding process according to claim 5, characterized in that: The single pulse amplitude of the reverse voltage pulse sequence is 30%-60% of the forward voltage, the pulse width is 0.2-0.5 times of the forward cycle, and the sequence duration is 10ms-50ms.

7. A dynamic voltage control method for anodic bonding process according to claim 6, characterized in that: The sliding mode surface function of the fuzzy sliding mode control in step S4 is defined as: ; in is the voltage deviation; and The fuzzy rule base dynamically adjusts according to the uniformity of the temperature field and the depressurization rate.

8. The dynamic voltage control method for anodic bonding process according to claim 3, characterized in that: The first threshold value is 70%-80% of the initial bonding interface resistance value.

9. The dynamic voltage control method for anodic bonding process according to claim 1, characterized in that: The multimodal feedback monitoring includes: An infrared thermal imager is used to obtain the temperature field distribution of the bonding interface at a sampling rate of 100Hz-200Hz; The interface resistance value was measured by the four-probe method, and the measuring electrode spacing was less than 1 mm and was distributed in concentric circles with the bonding area; The interface stress changes were monitored using a piezoelectric sensor array.

10. The method for dynamic voltage control in an anodic bonding process according to claim 4, characterized in that: When returning to the previous step voltage, the electrode contact pressure adjustment is started synchronously: The contact pressure between the electrode and the wafer is dynamically adjusted by the pneumatic servo mechanism, and the pressure adjustment range is 5N-20N; The pressure adjustment rate is positively correlated with the resistance fluctuation amplitude, satisfying ,in Indicates the pressure regulation rate; is the resistance fluctuation amplitude; is the initial resistance value, is the adjustment coefficient, and ; After the pressure stabilizes, it needs to be maintained for at least 3 step cycles before the voltage can be triggered to increase again.

Images