Non-destructive testing method for bond strength of semiconductor device package coatings

Abstract

The application provides a nondestructive testing method for the bonding strength of a coating of a semiconductor device shell, and belongs to the technical field of semiconductor device testing. The method solves the problem that the bonding strength of a coating and a substrate cannot be nondestructively evaluated in the prior art. The method comprises the following steps: arranging an acoustic wave coupling medium and an acoustic wave sensor on the surface of the coating, selecting at least three laser impact points, focusing a laser beam generated by a pulsed laser on the impact points, collecting a longitudinal wave signal after laser impact, reading a first echo signal peak voltage and a second echo signal peak voltage, calculating an acoustic pressure reflection coefficient R , performing laser impact on the same impact points for a second time at the same energy density, obtaining a reflection coefficient change amount R , calculating an average value R of the reflection coefficient change amounts R of the multiple impact points, and determining whether the bonding strength of the coating and the substrate meets a requirement according to the average value R avg When the average value R avg is greater than 0.05, it is determined that the bonding strength of the coating and the substrate does not meet the requirement, otherwise, the bonding strength of the coating and the substrate meets the requirement. The application can be used for the rapid and nondestructive evaluation of the bonding strength of a coating of a semiconductor device shell.

Description

Technical Field

[0001] This invention belongs to the field of semiconductor device testing technology. More specifically, this invention relates to a non-destructive testing method for the bonding strength of coatings on the casing of semiconductor devices. Background Technology

[0002] In semiconductor equipment manufacturing, the outer casing is typically coated with functional coatings to provide properties such as corrosion resistance, wear resistance, or contamination protection. The bonding strength between the coating and the substrate material is a key indicator determining the coating's lifespan and reliability. Insufficient bonding strength can lead to coating peeling or cracking under subsequent thermal cycling, mechanical vibration, or chemical atmospheres. The detached coating particles can contaminate delicate components inside the equipment, resulting in decreased product yield or even equipment failure. Therefore, effective testing of coating bonding strength during the production process is of paramount importance.

[0003] Currently, common methods for evaluating coating bond strength mainly include tensile testing, scratch testing, and indentation testing. Tensile testing requires a loading head to be fixed to the coating surface using an adhesive on a specialized specimen prepared under the same processing conditions as the outer shell. A tensile force perpendicular to the surface is then applied using a tensile device until the coating peels off from the substrate. The maximum tensile force at peeling is recorded and converted into bond strength. Scratch testing uses a diamond indenter to apply a gradually increasing vertical load to the coating surface while simultaneously moving it horizontally. The critical load at which the coating peels off is recorded to evaluate bond strength. Indentation testing uses an indenter to create an indentation on the coating surface. The bond quality is semi-quantitatively judged by observing cracks or peeling around the indentation.

[0004] The methods described above are all destructive or damaging tests, sharing common inherent drawbacks. First, these methods cause irreversible damage to the test object. Tensile tests completely destroy the specimen, scratch tests leave obvious scratches and peeling areas on the coating surface, and indentation tests leave permanent indentations. Therefore, these methods cannot be directly applied to finished semiconductor device casings that have already been manufactured. In actual production, only sampling inspection of samples from the same batch can be used, meaning that the test results of a small number of samples taken from the same production batch are considered representative of the quality of the entire batch of casings. Second, coating bonding strength is affected by various process factors, including the cleanliness of the substrate surface, spraying parameters, curing temperature and time, and ambient humidity. These factors inevitably fluctuate within a certain range during mass production, leading to significant differences in bonding strength between different casings within the same batch, and even between different locations within the same casing. Sampling inspection can only reflect the quality of the sampled specimens and cannot eliminate the risk of individual casings failing to meet bonding strength standards. For high-value semiconductor equipment, the consequences of coating peeling leading to equipment contamination or damage are often very serious, and missed detection can cause significant economic losses. Third, these destructive testing methods are relatively cumbersome to operate. Tensile tests require waiting for the adhesive to fully cure, resulting in a long testing cycle; scratch and indentation tests require high equipment precision and operator skills, and the interpretation of results is somewhat subjective. These factors limit testing efficiency and make it difficult to meet the rapid, high-throughput testing needs of production lines.

[0005] In summary, existing destructive testing methods cannot meet the requirements for component-by-component, non-destructive, and rapid testing of the bonding strength of coatings on semiconductor device housings. There is an urgent need in this field for a testing method that can directly and reliably evaluate the bonding strength between the coating and the substrate without damaging the finished product housing. Summary of the Invention

[0006] The purpose of this invention is to provide a non-destructive testing method that can directly evaluate the bonding strength between the coating and the substrate of a semiconductor device housing on a piece-by-piece basis in a rapid and reliable manner without damaging the finished product housing. This overcomes the shortcomings of existing destructive testing methods, which cannot be used for full inspection of finished products and have the risk of missed inspections in random sampling.

[0007] To address the aforementioned problems and achieve the objectives and other advantages of this invention, a non-destructive testing method for the bonding strength of coatings on semiconductor device housings is provided, comprising: An acoustic coupling medium is disposed on the coating surface of the semiconductor device housing, and an acoustic sensor is coupled to the acoustic coupling medium. Select on the coating surface M There are 1 laser impact points, among which M ≥3; A laser beam generated by a pulsed laser is focused onto an arbitrary laser impact point using a focusing lens; the energy density of the laser beam is 0.1-1 mJ / mm². 2 The acoustic sensor is located on the coating surface at a distance of 5-20 mm from the laser impact point; The longitudinal wave signal generated after each laser impact is collected by an acoustic wave sensor. The first longitudinal wave pulse that arrives after the laser impact is the first echo signal, and the next longitudinal wave pulse that arrives after the first echo signal is the second echo signal. Connect the longitudinal wave signal output from the acoustic wave sensor to the input channel of the oscilloscope, set the oscilloscope's trigger mode to edge trigger, and use the oscilloscope's cursor measurement function to read the peak voltage of the first echo signal. V Peak voltage of the first and second echo signals V 2; Using formula Calculate the acoustic pressure reflection coefficient after each laser impact. R ; right M Each laser impact point is subjected to a first laser impact, and the reflection coefficient corresponding to each laser impact point is recorded to obtain a reflection coefficient sequence. R i , i =1,2,…, M ; Using the same laser energy density as the first impact, for the same M A second laser impact is performed on each laser impact point, and the reflection coefficient of each point after the second impact is recorded to obtain a reflection coefficient sequence. ; Calculate the change in reflection coefficient at each laser impact point. and calculate M △ R i The average value △ R avg ; When △ R avg If the strength is greater than 0.05, the bonding strength between the coating and the substrate is deemed insufficient; otherwise, the bonding strength between the coating and the substrate is deemed sufficient.

[0008] Preferably, in the non-destructive testing method for the bonding strength of the semiconductor device housing coating, the method for... M After each laser impact point is subjected to one laser impact, a waiting time interval Δ is required. T Then a second laser shock is performed, in which △ T The time required for the coating surface temperature to recover to the temperature before laser shock is greater than or equal to the time required for the coating surface temperature to recover. The coating surface temperature is monitored in real time by an infrared thermometer.

[0009] Preferably, in the non-destructive testing method for the bonding strength of the semiconductor device housing coating, in the calculation M △ R i The average value △ R avg Before that, let's start from M △ R i Remove the maximum and minimum values ​​from the list, and then process the remaining values... M -2 △ R i Calculate the average value and use this average value as Δ. R avg .

[0010] Preferably, in the non-destructive testing method for the bonding strength of the semiconductor device housing coating, the formula is used. Before calculating the sound pressure reflection coefficient, the coating thickness is first measured using an eddy current thickness gauge. h Then, the sound wave attenuation coefficient is measured through the following steps. α : The acoustic wave sensor was fixed sequentially at the coating surface at distances of 5 mm and 20 mm from the laser impact point. The same acoustic wave coupling medium as used in the actual test was applied to each location. After at least 10 seconds for the coupling medium to stabilize, a laser impact was generated using a pulsed laser. The peak voltage of the first echo signal at each of the two locations was measured. V 1,5 and V 1,20 Using formula calculate α , where 15 is the distance difference between the two measurement locations, in mm; Using formula The peak voltage of the second echo signal is attenuated and compensated. Replace the original V 2. Substitute into the formula for calculating the reflection coefficient .

[0011] Preferably, in the non-destructive testing method for the bonding strength of the semiconductor device housing coating, in M In addition to the laser impact point, an additional reference impact point is selected. This reference impact point is located on a reference sample prepared with the same material and coating process as the semiconductor device housing. The adhesion strength between the coating and the substrate at the reference impact point on the reference sample has been confirmed as qualified through a tensile test. The reference impact point is then subjected to the same... M Two identical laser impact steps were performed at the same laser impact point. The change in the reflection coefficient Δ at the reference impact point was measured. Rref When △ R ref When ≤0.01, use the formula The change in reflection coefficient at each laser impact point is corrected, and the corrected value is then... Replace the original △ R i Calculate the average value △ R avg When △ R ref If the value is greater than 0.01, the detection system is deemed to be in an abnormal state. After recalibration, the detection is performed again.

[0012] Preferably, in the non-destructive testing method for the bonding strength of the semiconductor device housing coating, an annular sound-absorbing layer is formed on the coating surface with the laser impact point as the center. The annular sound-absorbing layer has an inner diameter of 1.5-3 mm, an outer diameter of 4-6 mm, and a thickness of 0.5-1 mm. The annular sound-absorbing layer is formed by coating silicone rubber sound-absorbing material. The annular sound-absorbing layer completely surrounds the laser impact point. The acoustic wave sensor is located outside the annular sound-absorbing layer, and the distance between the center of the acoustic wave sensor and the laser impact point is 5-20 mm.

[0013] Preferably, in the non-destructive testing method for the bonding strength of the semiconductor device housing coating, when the coating thickness... h When the coating thickness is ≥0.3mm, no deconvolution processing is performed on the longitudinal wave signal acquired by the oscilloscope; when the coating thickness is... h When the error is <0.3mm, after acquiring the longitudinal wave signal using an oscilloscope, deconvolution processing is first performed on the acquired longitudinal wave signal to separate the overlapping first and second echo signals in the time domain into two independent pulse waveforms. Then, the peak voltage of the first echo signal is read from the separated independent pulse waveforms. V Peak voltage of the first and second echo signals V 2. The deconvolution process is performed according to the following steps: The longitudinal wave time-domain signal acquired by the oscilloscope is denoted as... y ( t The impulse response function of the acoustic wave sensor is denoted as... s ( t ),right y ( t )and s ( t The frequency domain signal was obtained by performing Fourier transforms on each of the two methods. Y ( f )and S ( f ); The frequency domain estimate of the reflection coefficient sequence is calculated using the Wiener filter deconvolution formula. R ( f ): in S* ( f )for S ( f The conjugate of the complex number, | S ( f )| 2 for S ( f The power spectrum of ) K For regularization parameters, K The value is | S ( f )| 2 1% of the maximum value; right R ( f Performing an inverse Fourier transform yields the time-domain reflection coefficient sequence. r ( t ); exist r ( t From two independent pulse waveforms that appear in chronological order, the peak voltage of the first pulse waveform is read as the peak voltage of the first echo signal. V 1. Read the peak voltage of the second pulse waveform as the peak voltage of the second echo signal. V 2; Impulse response function of acoustic sensor s ( t Obtain it in advance through the following steps: a) Prepare a metal standard test block with a thickness greater than 20 mm, wherein the lower surface of the metal standard test block is flat and parallel to the upper surface; b) Determine a laser impact point on the upper surface of the metal standard test block; c) Fix the acoustic wave sensor on the upper surface of the metal standard test block. The distance between the acoustic wave sensor and the laser impact point is 5-20mm. Apply the same acoustic wave coupling medium as the one used in the formal test between the acoustic wave sensor and the metal standard test block. d) Use a pulsed laser to generate a laser beam, focus the laser beam on the laser impact point to generate a laser impact, and the energy density of the laser beam is the same as that of the formal test; e) Acquire the echo signal directly reflected from the upper surface of the metal standard test block using the acoustic wave sensor, and use the acquired echo signal as the impulse response function of the acoustic wave sensor. s ( t ).

[0014] Preferably, in the non-destructive testing method for the bonding strength of the semiconductor device housing coating, after obtaining... MThe peak voltage of the first echo signal at each laser impact point V 1i After that, when any one V 1i When the voltage is <5mV, the laser impact point is from... M Remove from the laser impact points, and record the number of remaining laser impact points as follows: M ',use M The average reflection coefficient of ' laser impact points Δ R avg .

[0015] Preferably, in the non-destructive testing method for the bonding strength of the semiconductor device housing coating, the method for... M Before each laser impact point undergoes its first laser impact, the surface roughness of the coating at each laser impact point is measured using a non-contact optical surface roughness measuring instrument. The surface roughness at any given laser impact point is then... Ra When the value is >1.5μm, the laser shock point is moved from... M Remove from the laser impact points, and record the number of remaining laser impact points as follows: M ",use M The average reflection coefficient of each laser impact point is calculated as Δ. R avg .

[0016] The present invention has at least the following beneficial effects: This invention achieves completely non-destructive testing of the bonding strength of the coating on the casing of qualified semiconductor equipment. Unlike existing destructive methods such as tensile testing, scratch testing, or indentation testing, this invention uses pulsed laser shock to generate acoustic signals. By collecting the first and second echo signals, the acoustic pressure reflection coefficient and its change are calculated. After testing, only a laser spot with a depth of micrometers remains on the surface of the coating of the qualified finished product casing. It will not penetrate the coating or cause damage such as cracks or peeling. Therefore, the qualified finished product casing can still be used normally for semiconductor equipment assembly, thus truly achieving full inspection of each piece and completely eliminating the risk of missed inspection caused by process fluctuations in traditional sampling inspection methods.

[0017] This invention can directly evaluate the bonding strength between the coating and the substrate without damaging the outer shell, and the test results are highly consistent with the destructive tensile test results. In the examples, the test results of the six outer shells are completely consistent with the tensile test results, indicating that the present invention can accurately distinguish between qualified and unqualified coatings with a threshold of 0.05, and the detection accuracy rate reaches 100%, overcoming the technical difficulty of quantitatively evaluating the bonding strength of existing non-destructive methods.

[0018] This invention enables a rapid and efficient testing process. It eliminates the need to wait for adhesive curing or perform complex sample preparation; a single laser impact and signal acquisition can be completed within milliseconds to seconds. The total testing time for multiple impact points typically does not exceed a few minutes, far faster than tensile testing (which requires 24 hours of curing) or scratch testing (which requires precise loading and interpretation). This invention effectively meets the high-throughput, rapid testing requirements of semiconductor production lines.

[0019] This invention determines bonding strength by calculating the average value of the reflection coefficient changes at multiple laser impact points (M≥3), effectively reducing random errors caused by local non-uniformity of the coating. Furthermore, by removing a maximum and a minimum value before calculating the average, the influence of abnormal data caused by occasional interference (such as poor local coupling, instantaneous fluctuations in laser energy, electromagnetic interference, etc.) on the overall evaluation is eliminated, ensuring that Δ... R avg It more accurately reflects the overall level of coating bonding strength, improving the robustness and anti-interference ability of the testing method.

[0020] This invention uses an infrared thermometer to monitor the coating surface temperature in real time and waits for the temperature to return to the level before the second laser impact. This eliminates the influence of the local temperature rise caused by the first impact on the sound velocity and attenuation coefficient, ensuring that the change in reflectance Δ between the two impacts is accurately measured. R The acoustic parameter changes originate solely from microscopic damage at the interface, rather than thermal effects, thus improving the accuracy and repeatability of the judgment results.

[0021] This invention, by creating a ring-shaped sound-absorbing layer to completely surround the laser impact point, effectively absorbs surface waves propagating along the coating surface, suppresses surface interference with the longitudinal wave echo signal, significantly improves the signal-to-noise ratio and reading stability of the second echo signal, and reduces the reflection coefficient. R and change Δ R The calculations are more accurate, making it particularly suitable for detection scenarios involving hard coatings or sensors at a distance.

[0022] This invention uses an eddy current thickness gauge to measure the coating thickness. h The acoustic attenuation coefficient was calculated by measuring the peak voltage of the first echo at two different distances (5 mm and 20 mm). α This allows for attenuation compensation of the second echo signal, eliminating the influence of coating thickness differences and material attenuation on the reflection coefficient calculation, thus ensuring the calculated... R The value is only related to the interface reflection characteristics, ensuring the comparability of test results between coatings of different thicknesses.

[0023] This invention introduces a reference impact point correction method. The change in reference reflection coefficient Δ is measured on a reference sample prepared using the same material and process as the shell to be tested, and whose bonding strength is known to be acceptable.R ref And use it to measure the △ at each point to be measured R i The correction can effectively eliminate common system errors caused by laser energy drift, changes in sensor coupling state, and ambient temperature drift in the detection system, further improving the consistency and accuracy of the detection results. It is especially suitable for long-term continuous detection or production sites with large fluctuations in environmental conditions.

[0024] This invention adaptively selects whether to perform deconvolution processing on the longitudinal wave signal based on whether the coating thickness is less than 0.3 mm. For thin coatings ( h <0.3mm), Wiener filtering and deconvolution are used to separate the overlapping first and second echo signals in the time domain into two independent pulses, thereby accurately reading their respective peak voltages; for thick coatings ( h For thicknesses ≥0.3mm, direct reading is performed without complex calculations. This strategy ensures the feasibility of detecting thin coatings while maintaining the efficiency of detecting thick coatings, making this invention applicable to a wide range of coating thicknesses from tens of micrometers to several millimeters.

[0025] This invention examines the peak voltage of the first echo signal. V 1. When V When the laser impact voltage is less than 5 mV, the laser-impacted point is discarded to avoid the influence of low signal-to-noise ratio data caused by poor local coupling, microbubbles, or surface contaminants on the average value. Simultaneously, a non-contact optical surface roughness measuring instrument is used to measure the surface roughness at each point before testing. Ra Value, when Ra Points with a diameter >1.5 μm are discarded to ensure that all measurement points involved in the calculation have good acoustic coupling conditions. These two screening measures effectively reduce measurement errors caused by accidental operational factors or local surface roughness, and improve the reliability and repeatability of the detection.

[0026] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Detailed Implementation

[0027] The present invention will now be described in further detail so that those skilled in the art can implement it based on the description.

[0028] This invention provides a non-destructive testing method for the bonding strength of coatings on semiconductor device housings, comprising: An acoustic coupling medium is applied to the coating surface of the semiconductor device housing, and the acoustic sensor is coupled to the coupling medium. The coupling medium can be a specialized ultrasonic testing gel (e.g., acoustic impedance approximately 1.5-1.7 M ayl, viscosity 1000-5000 cP) or a high-viscosity grease (e.g., vacuum grease). The acoustic impedance of the coupling medium should be as close as possible to the acoustic impedance of the coating material (difference < 30%) to ensure efficient acoustic wave transmission. Each application should ensure a uniform thickness (approximately 0.5-1 mm) and that there are no air bubbles between the sensor and the coating. For multiple measurements on the same housing, the same batch of coupling medium should be used.

[0029] Select on the coating surface M There are 1 laser impact points, among which M ≥3; A laser beam generated by a pulsed laser is focused onto an arbitrary laser impact point using a focusing lens; the energy density of the laser beam is 0.1-1 mJ / mm². 2 The acoustic sensor is located 5-20 mm away from the laser impact point on the coating surface; for different coating materials, the safe energy density range should be determined through pre-testing. For organic coatings, 0.1-0.3 mJ / mm² is recommended. 2 For ceramic coatings, a concentration of 0.3-0.8 mJ / mm is recommended. 2 For metallic coatings, a concentration of 0.5-1.0 mJ / mm is recommended. 2 .

[0030] The longitudinal wave signal generated after each laser impact is acquired using an acoustic wave sensor. The first longitudinal wave pulse arriving after the laser impact is the first echo signal, and the next longitudinal wave pulse arriving after the first echo signal is the second echo signal. The acoustic wave sensor is a piezoelectric longitudinal wave sensor with a center frequency of 2.5-10MHz and a bandwidth of not less than 60% of the center frequency. (Regarding coating thickness...) h For thicknesses ≥0.3mm, select a 2.5-5MHz sensor; for hFor depths <0.3mm, a 5-10MHz sensor should be selected. It should be noted that the acoustic waves excited by laser shock include longitudinal waves, transverse waves, and surface waves. To avoid interference from surface wave signals on longitudinal wave signals, the distance between the acoustic sensor and the impact point should be 10-15mm. At this distance, the arrival time difference between the surface wave and longitudinal wave signals is sufficient to distinguish them on an oscilloscope. If a distinct pulse appears in the waveform before the first echo signal, this pulse is a surface wave signal and should be suppressed by adjusting the position of the acoustic sensor or using a ring-shaped sound-absorbing layer. If an unexpected additional pulse appears between the first and second echo signals, it may be a transverse wave or a mode-conversion wave. In this case, the sensor coupling status should be checked, or the distance between the impact point and the acoustic sensor should be increased to improve the time resolution. Furthermore, laser shock generates mode-conversion waves (such as reflected waves after a longitudinal wave converts to a transverse wave) at the coating-substrate interface. The arrival time of this wave at the sensor may be close to that of the second echo signal. To reduce interference from mode conversion waves, the sensor center frequency should be matched with the coating thickness (generally, the longitudinal wave two-way time corresponding to the coating thickness should be greater than twice the sensor pulse response width). The consistency of the waveform should be observed to determine the impact: if the second echo waveform is stable and the amplitude repeatability is good, it indicates that the influence of the mode conversion wave is negligible; if the second echo signal waveform is distorted or the amplitude fluctuates greatly, the sensor position should be adjusted appropriately or deconvolution processing should be used to separate the waveform.

[0031] Connect the longitudinal wave signal output from the acoustic wave sensor to the input channel of the oscilloscope, set the oscilloscope's trigger mode to edge trigger, and use the oscilloscope's cursor measurement function to read the peak voltage of the first echo signal. V Peak voltage of the first and second echo signals V 2. The oscilloscope's sampling rate should be no less than 1GS / s, analog bandwidth no less than 100MHz, and vertical resolution no less than 8 bits.

[0032] Using formula Calculate the acoustic pressure reflection coefficient after each laser impact. R It should be noted that, because the coating attenuates high-frequency sound waves more strongly, the second echo signal may lose more high-frequency components compared to the first echo signal. This is not the sound pressure reflection coefficient in the strict sense, but rather the frequency-dependent equivalent reflection coefficient. However, this method uses the change in reflection coefficient Δ between two impacts. R As a criterion for judgment, since the frequency components of the two impacts are the same, the frequency-related attenuation affects Δ. R The impact is effectively offset, therefore it does not affect the judgment result.

[0033] right M Each laser impact point is subjected to a first laser impact, and the reflection coefficient corresponding to each laser impact point is recorded to obtain a reflection coefficient sequence. Ri , i =1,2,…, M ; Using the same laser energy density as the first impact, for the same M A second laser impact is performed on each laser impact point, and the reflection coefficient of each point after the second impact is recorded to obtain a reflection coefficient sequence. For samples with unacceptable bonding strength, the first laser impact may produce microscopic local separation at the interface, ranging from nanometers to micrometers. It should be noted that this separation occurs during the first impact, and the judgment result can only be obtained after the second impact. Therefore, it is impossible to predict whether the shell is qualified at the time of the first impact. However, this microscopic separation has the following characteristics: (1) the scale is much smaller than the coating thickness and does not penetrate the coating; (2) it does not expose the substrate; (3) it does not cause coating peeling or macroscopic cracks. In other words, the macroscopic integrity and protective performance of the coating are not affected. After the second impact, if the shell is judged to be unacceptable, it will be rejected and will not enter the subsequent assembly and use stages. Therefore, this microscopic separation will not cause performance damage to any product actually put into use. For samples with acceptable bonding strength, neither impact produces any interface separation. In summary, this method is completely non-destructive to qualified products that are finally put into use, which meets the requirements of non-destructive testing. Those skilled in the art should understand that non-destructive testing is to detect whether there are defects or discontinuities in the tested object without damaging or affecting its future performance. In this method, the future performance of qualified products remains completely unaffected; unqualified products, since they will not be used in the future, also have no performance impairment. Therefore, this method belongs to non-destructive testing. The change in reflection coefficient at each laser impact point is calculated. and calculate M △ R i The average value △ R avg ; When △ R avg If the strength is greater than 0.05, the bonding strength between the coating and the substrate is deemed insufficient; otherwise, the bonding strength between the coating and the substrate is deemed sufficient.

[0034] The closest existing technologies are tensile testing, scratch testing, and indentation testing. These methods all cause irreversible damage to the coating, making it unsuitable for direct use on finished product casings. They can only be sampled during production, but process fluctuations lead to the risk of missed inspections, failing to meet the requirement of full inspection for each item.

[0035] To address the aforementioned technical challenges, this invention provides a non-destructive testing method: First, an acoustic coupling medium is placed on the coating surface of the semiconductor device housing, and an acoustic sensor is coupled to this medium. The purpose of this step is to ensure efficient transmission of acoustic waves between the coating and the sensor, reduce energy loss, and ensure that the acquired signal accurately reflects the acoustic response of the coating. Then, a selection is made on the coating surface... M One laser impact point, M ≥3. Selecting multiple impact points and averaging the results can reduce random errors caused by local non-uniformity of the coating and improve the representativeness of the test results. The laser beam generated by the pulsed laser is focused onto any one laser impact point using a focusing lens; the energy density of the laser beam is 0.1-1 mJ / mm². 2 This energy density range has been experimentally verified: below 0.1 mJ / mm². 2 The generated sound wave signal is too weak to be effectively detected by the sensor; it is higher than 1 mJ / mm 2 This could potentially cause unacceptable damage to the coating surface, violating the original purpose of non-destructive testing. The acoustic wave sensor is located on the coating surface 5-20mm away from the laser impact point. Too close a distance will block the laser beam path, while too far a distance will result in excessive acoustic wave attenuation and a decrease in the signal-to-noise ratio. The acoustic wave sensor collects the longitudinal wave signal generated after each laser impact. The first arriving longitudinal wave pulse after the laser impact is the first echo signal, representing the acoustic wave directly excited by the laser on the coating surface, reflected by the surface, and received by the sensor. The next arriving longitudinal wave pulse after the first echo signal is the second echo signal, representing the acoustic wave penetrating the coating, reaching the coating-substrate interface, reflecting back, and being received by the sensor. The longitudinal wave signal output from the acoustic wave sensor is connected to the input channel of an oscilloscope. The oscilloscope's trigger mode is set to edge trigger, and the peak voltage of the first echo signal is read using the oscilloscope's cursor measurement function. V Peak voltage of the first and second echo signals V 2. In actual operation, the method for setting the trigger level of the oscilloscope is as follows: Since the amplitude of the first echo signal is unknown before the laser shock, the trigger level can be set in any of the following ways: (1) First set the trigger level to a low fixed value (e.g., 2-5mV). This value should be lower than the expected minimum amplitude of the first echo signal (usually not lower than 10mV) to ensure that the oscilloscope can be triggered by the first echo signal after the laser shock. If the trigger is unstable, the trigger level can be appropriately lowered. (2) Use the automatic setting function of the oscilloscope. The oscilloscope will automatically set the trigger level to about 50% or 10% of the signal peak value. This function will automatically adjust after the signal appears and will not affect data acquisition. (3) Perform a test shock first and read the actual peak voltage of the first echo signal. V 1test Then set the trigger level to V 1testApproximately 10% of the signal is then measured before a formal measurement is performed. The above methods are standard operating procedures in this field, and operators can choose the appropriate method based on the actual equipment and signal conditions.

[0036] Using formula Calculate the acoustic pressure reflection coefficient after each laser impact. R .right M Each laser impact point is subjected to a first laser impact, and the reflection coefficient corresponding to each laser impact point is recorded to obtain a reflection coefficient sequence. R i , i =1,2,…, M Using the same laser energy density as the first impact, for the same... M A second laser impact is performed on each laser impact point, and the reflection coefficient of each point after the second impact is recorded to obtain a reflection coefficient sequence. R i '. Calculate the change in reflection coefficient at each laser impact point. and calculate M △ R i The average value △ R avg Change in reflection coefficient The physical meaning is as follows: the first laser shock generates a microsecond-level compressive stress wave at the interface between the coating and the substrate. This stress wave propagates to the interface, partially transmitted and partially reflected. When the interfacial bonding strength is insufficient, the reflected tensile wave can cause microcracks or localized delamination (microscopic damage) at the interface, thereby altering the acoustic impedance matching state of the interface and affecting the reflection coefficient measured during the second shock. R Compared to the first impact, a significant change occurs. Conversely, if the bonding strength is sufficient, the interface remains intact after the first impact, and the reflection coefficients are essentially the same in both impacts. When the bonding strength between the coating and the substrate is insufficient, the tensile stress wave generated by the first laser impact causes microcracks or localized delamination at the interface. The acoustic impedance of these micropores is close to 0, much lower than that of the coating or substrate material, resulting in a significant increase in the acoustic pressure reflection coefficient at the interface. Therefore, the reflection coefficient measured by the second impact... R 'Greater than the value measured during the first impact' R , △ R It is a positive value.

[0037] When △ R avg When the strength is greater than 0.05, the bonding strength between the coating and the substrate is deemed insufficient; conversely, when it is less than 0.05, the bonding strength is deemed sufficient. To determine the threshold, 30 reference samples were prepared using the same materials and processes as the shell to be tested. Of these, 15 samples achieved acceptable bonding strength (≥25 MPa) by adjusting process parameters, while 15 samples failed (<25 MPa). The Δt of each sample was measured according to this method.R avg The results are as follows: Qualified sample △ R avg The maximum value is 0.047, and the non-conforming sample △ R avg The minimum value is 0.053. Therefore, setting the threshold to 0.05 ensures no overlapping intervals. Leave-one-out cross-validation showed a 0% false positive rate. The physical meaning of the threshold is: when the average change in reflection coefficient after two impacts exceeds 5%, it indicates that the first laser impact has caused irreversible microscopic damage to the interface between the coating and the substrate, such as microcracks or localized peeling. This means the original bonding strength is insufficient to resist the stress wave generated by a laser impact of that energy density; if the change does not exceed 5%, it indicates that the interface remains intact after the first impact, and the bonding strength is acceptable. For example, five laser-shock points are selected on a batch of casings. The reflection coefficients measured by the first shock are 0.35, 0.38, 0.36, 0.37, and 0.34, respectively. The reflection coefficients measured by the second shock are 0.36, 0.39, 0.52, 0.38, and 0.35, respectively. The calculated changes at each point are 0.01, 0.01, 0.16, 0.01, and 0.01, respectively, with an average change of 0.04. Since 0.04 is less than 0.05, the coating bonding strength of the casing is deemed acceptable. If the change at the third point increases to the average value exceeding 0.05, the casing is deemed unacceptable.

[0038] It should be noted that in this method, the two laser shocks are applied to the same location on the coating surface. The spot depth generated by a single shock is in the micrometer range (typically 1-5 μm), much smaller than the coating thickness (usually ≥50 μm), and will not penetrate the coating. After the two shocks are superimposed, the increase in spot depth is less than 1.5 times that of a single shock, still much smaller than the coating thickness, and will not cause coating perforation or substrate exposure. Furthermore, the tiny indentation at the impact point does not affect the coating's corrosion resistance, wear resistance, or anti-fouling properties. Therefore, this method still falls within the scope of non-destructive testing. For extremely thin coatings with a thickness of less than 50 μm, it is recommended to use a single shock or adjust the energy density to further reduce the impact.

[0039] Example 1 This method is used to detect the coatings of semiconductor device casings in multiple batches, covering both cases where the bonding strength is qualified and unqualified. The casing substrates are all made of aluminum alloy, and the coatings are alumina ceramics with a thickness of 0.2 mm. A layer of ultrasonic coupling gel is evenly applied on the coating surface as the acoustic wave coupling medium. A piezoelectric acoustic wave sensor with a center frequency of 5 MHz is gently attached to the gel, ensuring no air bubbles between the sensor and the coating. Five laser shock points are randomly selected on the coating surface, with a distance of no less than 10 mm between each point to avoid thermal influence and acoustic wave interference between adjacent points. A Nd:YAG pulsed laser is used, with a laser wavelength of 532 nm, a pulse width of 8 ns, and an energy density set to 0.5 mJ / mm 2 . After passing through the focusing lens, the spot diameter of the laser beam is focused to 1 mm. The acoustic wave sensor is placed at a position on the coating surface 10 mm away from each laser shock point, and the position of the sensor is kept unchanged during the two shocks. The oscilloscope is set to edge trigger, and the trigger level is set to 10% of the amplitude of the first echo signal. The waiting time between the two laser shocks is 30 s.

[0040] First, three qualified examples are given. The first qualified example: For the first laser shock on the first shock point, the peak voltage of the first echo signal V 1 is 120 mV, and the peak voltage of the second echo signal V 2 is 38.4 mV. Calculate the reflection coefficient R to be 0.32. Complete the first shock for the remaining 4 points in sequence. The R values of the 5 points are 0.32, 0.33, 0.31, 0.32, 0.32 respectively. After the second shock, the R 'values of the 5 points are 0.33, 0.34, 0.33, 0.32, 0.33 respectively. Calculate the △ R for each point: 0.01, 0.01, 0.02, 0.00, 0.01. The average △ R avg = 0.01, which is less than 0.05, so it is judged to be qualified. Subsequently, a destructive tensile test is carried out on this casing, and the measured bonding strength is 32 MPa, which is higher than the process requirement of 25 MPa, so it is actually qualified. The second qualified example: The first R values are 0.35, 0.36, 0.34, 0.35, 0.35 respectively, and the second R 'values are 0.36, 0.37, 0.35, 0.36, 0.36 respectively. The △ R are all 0.01, and the average is 0.01, so it is judged to be qualified. The tensile test measures a bonding strength of 30 MPa, which is qualified. The third qualified example: The first R values are 0.30, 0.31, 0.29, 0.30, 0.30 respectively, and the second RThe values ​​of ' are 0.31, 0.32, 0.30, 0.31, and 0.31, respectively. R All values ​​were 0.01, with an average of 0.01, indicating compliance. The tensile test yielded a bond strength of 34 MPa, which is also acceptable.

[0041] Here are three more examples of non-compliance. First example of non-compliance: The first time... R The values ​​were 0.45, 0.44, 0.46, 0.45, and 0.44 respectively. (The second time...) R The values ​​of ' are 0.52, 0.53, 0.55, 0.54, and 0.52, respectively. R The values ​​were 0.07, 0.09, 0.09, 0.09, and 0.08 respectively, with an average Δ. R avg =0.084, greater than 0.05, deemed unqualified. The tensile test measured a bond strength of 18 MPa, lower than 25 MPa, therefore practically unqualified. Second example of non-compliance: First time... R The values ​​were 0.48, 0.47, 0.49, 0.48, and 0.47 respectively. (The second time...) R The values ​​of ' are 0.56, 0.55, 0.58, 0.57, and 0.56 respectively, △ R The values ​​were 0.08, 0.08, 0.09, 0.09, and 0.09 respectively, with an average of 0.086, which was deemed unqualified. The tensile test yielded a bond strength of 16 MPa, which was also unqualified. This is the third example of non-compliance: the first time... R The values ​​were 0.42, 0.43, 0.42, 0.44, and 0.43 respectively. (The second time...) R The values ​​of ' are 0.50, 0.51, 0.49, 0.52, and 0.51 respectively, △ R The values ​​were 0.08, 0.08, 0.07, 0.08, and 0.08 respectively, with an average of 0.078, which was deemed unqualified. The tensile test yielded a bond strength of 20 MPa, which was also unqualified.

[0042] In all six examples above, this method accurately determined the defects. After testing, only a laser spot with a diameter of approximately 1 mm remained on the surface of each casing. The depth of this spot was at the micrometer level, failing to penetrate the coating. Optical microscopy revealed that the coating surface was intact, without cracks or peeling. For semiconductor device casings, such minute traces do not affect the coating's corrosion resistance, wear resistance, or anti-contamination performance, nor do they affect the casing's assembly sealing and electrical insulation; therefore, the casing remains usable. This embodiment demonstrates that this method can perform non-destructive testing on every finished product, accurately identifying both qualified and unqualified items, and the testing process does not impair the casing's functionality.

[0043] Comparative Example 1 The same batch of outer shells as in Example 1 was detected using the existing destructive tensile test method. One outer shell corresponding to each of the three qualified instances in Example 1 and one outer shell corresponding to each of the three unqualified instances were taken, for a total of six outer shells. According to the tensile test standard, a loading head with a diameter of 20 mm was pasted on the coating surface of each outer shell. After the adhesive was cured for 24 h, a universal tensile testing machine was used to apply a tensile force perpendicular to the surface at a rate of 2 mm / min until the coating was peeled off from the substrate. The measured maximum tensile forces of the three qualified outer shells were 10.05 kN, 9.42 kN, and 10.68 kN respectively. Dividing by the loading area of 314 mm 2 , the obtained bonding strengths were 32 MPa, 30 MPa, and 34 MPa respectively, all higher than the process requirement of 25 MPa, and were judged to be qualified. The measured maximum tensile forces of the three unqualified outer shells were 5.65 kN, 5.02 kN, and 6.28 kN respectively. After conversion, the obtained bonding strengths were 18 MPa, 16 MPa, and 20 MPa respectively, all lower than 25 MPa, and were judged to be unqualified. This judgment result was completely consistent with the judgment result of this method in Example 1. However, after the test, the coating on all six outer shells had completely peeled off or fallen off in large areas from the substrate, and the outer shells were damaged as a whole and could no longer be used for the assembly of semiconductor devices. In actual production, if this destructive method is used, only a small number of outer shells can be sampled from each batch for detection. Suppose a batch produces 100 outer shells in total, and the unqualified rate is about 10%. If 5 outer shells are sampled, according to probability calculation, there is about a 40% possibility of missing all unqualified products, resulting in unqualified outer shells flowing into the subsequent assembly process. And the destructive detection in Comparative Example 1 itself cannot achieve 100% inspection for each piece. In contrast, the method in Example 1 can detect each outer shell. After detection, the qualified products are continued to be used, and the unqualified products are removed. All six detections are correct, and the accuracy rate is 100%, completely eliminating the risk of missed detection. Therefore, on the premise of maintaining the same accuracy as the destructive tensile test, this method overcomes the defect that destructive detection cannot perform 100% inspection and meets the requirement of non-destructive inspection for the bonding strength of the coating of the semiconductor device outer shell for each piece.

[0044] In another solution, in the non-destructive detection method for the bonding strength of the coating of the semiconductor device outer shell, for M each laser shock point, after a laser shock is performed once, wait for the time interval △ T and then perform a second laser shock, where △ T is greater than or equal to the time required for the coating surface temperature to recover to the temperature before the laser shock. The coating surface temperature is monitored in real time by an infrared thermometer.

[0045] In the process of testing the bonding strength of a coating using two laser shocks, the first laser shock causes a localized increase in temperature on the coating surface. If the second shock is performed immediately before the temperature returns to its initial value, the sound velocity and attenuation coefficient of the coating material will change due to the residual temperature. This causes an additional change in the reflection coefficient measured by the second shock, unrelated to the bonding strength, thus interfering with the change in reflection coefficient Δ. R avg The accuracy of the calculation.

[0046] To solve this problem, after the first laser shock, a time interval △ needs to be waited. T A second laser impact is then performed. △ T The specific value is determined by real-time monitoring of the coating surface temperature using an infrared thermometer, requiring △ T The time required for the coating surface temperature to recover to its pre-laser shock temperature is greater than or equal to the time required for the temperature to return to the pre-laser shock temperature. For example, under a set of typical experimental conditions (ambient temperature 25°C, coating thickness 0.2 mm, aluminum alloy substrate, laser energy density 0.5 mJ / mm²), the temperature is considered to be within the range of 25°C. 2 Using an infrared thermal imager for real-time monitoring, the coating surface temperature rose by approximately 8°C after the first impact, and returned to the pre-impact level after about 25 seconds (temperature difference < 0.5°C). To allow for a margin of safety, the waiting time △ T Set to 30 seconds. In actual testing, △ T The values ​​should be determined in real time for each casing using an infrared thermometer; the values ​​mentioned above are for illustrative purposes only.

[0047] In another embodiment, the non-destructive testing method for the bonding strength of the semiconductor device housing coating involves calculating... M △ R i The average value △ R avg Before that, let's start from M △ R i Remove the maximum and minimum values ​​from the list, and then process the remaining values... M -2 △ R i Calculate the average value and use this average value as Δ. R avg .

[0048] In determining the coating bonding strength by using two laser shocks and calculating the average change in reflection coefficient at multiple laser shock points, the change in reflection coefficient Δ at each laser shock point is... R iIndividual values ​​deviating from the normal range may occur due to occasional interference during the measurement process. For example, during a laser shock, the laser energy may momentarily deviate from the set value due to equipment fluctuations; uneven application of the coupling medium between the acoustic sensor and the coating may create localized bubbles at that point; or the oscilloscope trigger level may experience brief fluctuations due to electromagnetic interference. These factors can all cause values ​​at that point to deviate from the normal range. V 1 or V 2. Abnormal readings, thus causing the Δ at that point to be abnormal. R i These outliers are significantly larger or smaller than other points. These outliers are unrelated to the coating's bonding strength; directly including them in the average calculation would increase the average variation Δ. R avg Deviation from the true level may lead to misjudgment: an abnormally large △ R i It may make the originally qualified △ R avg A value exceeding 0.05 is considered unacceptable; an abnormally small △ R i It may cause the originally unqualified △ R avg A value below 0.05 is considered acceptable.

[0049] To solve the above problems, in the calculation M △ R i The average value △ R avg Before that, let's start from M △ R i Remove the maximum and minimum values ​​from the list, and then process the remaining values... M -2 △ R i Calculate the average value and use this average value as Δ. R avg .here M The value of must be greater than or equal to 3, because only when is 3, is 3. M When the value is ≥3, after removing one maximum and one minimum value, at least one data point remains, making the average calculation still meaningful. For example, in actual testing, a value is usually selected... M =5 laser impact points, resulting in 5 △ R iThe values ​​were 0.01, 0.01, 0.02, 0.01, and 0.16. The value of 0.16 was significantly larger than the others. Inspection revealed a tiny air bubble beneath the sensor during the measurement, causing poor coupling. Averaging these five values ​​directly yielded 0.042, still less than 0.05, thus deemed acceptable. However, if the outlier was larger, such as 0.30, a direct average might reach 0.07, exceeding 0.05 and leading to a false positive. After removing the maximum value of 0.16 and the minimum value of 0.01, the remaining three values ​​were 0.01, 0.02, and 0.01, with an average of 0.0133, far less than 0.05, correctly deemed acceptable. In another example, the five △... R i The values ​​were 0.08, 0.09, 0.07, 0.08, and 0.02. The value of 0.02 was abnormally low (inspection revealed a momentary low laser energy at that point). After removing the maximum value of 0.09 and the minimum value of 0.02, the remaining three values ​​were 0.08, 0.07, and 0.08, with an average of 0.0767. This average is still greater than 0.05, and therefore correctly judged as unqualified. If the average (including the abnormally low value) were directly taken, the result would be 0.068, which is still greater than 0.05. However, if the abnormally low value were even lower, such as 0.00, the direct average might drop below 0.05, leading to missed judgments. Therefore, averaging after removing extreme values ​​can effectively eliminate the impact of occasional interference during the measurement process on the overall evaluation, making Δ... R avg This method better represents the true state of coating bonding strength. This step requires no additional equipment or measurement; it can be achieved solely through data processing. This simple and effective method improves the robustness and anti-interference capabilities of the testing method, avoiding misjudgments or omissions due to accidental factors during the measurement process.

[0050] In another embodiment, the non-destructive testing method for the bonding strength of the semiconductor device housing coating uses the formula... Before calculating the sound pressure reflection coefficient, the coating thickness is first measured using an eddy current thickness gauge. h Then, the sound wave attenuation coefficient is measured through the following steps. α : The acoustic wave sensor was fixed sequentially at the coating surface at distances of 5 mm and 20 mm from the laser impact point. The same acoustic wave coupling medium as used in the actual test was applied to each location. After at least 10 seconds for the coupling medium to stabilize, a laser impact was generated using a pulsed laser. The peak voltage of the first echo signal at each of the two locations was measured. V 1,5 and V 1,20 Using formula calculate α , where 15 is the distance difference between the two measurement locations, in mm; Using formula The peak voltage of the second echo signal is attenuated and compensated. Replace the original V 2. Substitute into the formula for calculating the reflection coefficient .

[0051] In determining the coating adhesion strength by using two laser shocks and calculating the change in reflectance, the reflectance... R The calculation depends on the peak voltage of the second echo signal. V 2. However, sound waves attenuate as they propagate within the coating, and the degree of attenuation varies with the coating thickness. h and the acoustic attenuation coefficient of the coating α The thicker the coating, the longer the sound wave propagation path, and the more severe the attenuation, resulting in a decrease in the amplitude of the second echo signal. V 2. The calculated reflection coefficient is lower than the actual interface reflected sound pressure, thus the calculated reflection coefficient is lower. R Low. There may be slight differences in coating thickness between different casings (e.g., 0.18-0.22 mm). Even if the interfacial bonding is exactly the same, a thicker coating will result in lower performance due to greater attenuation. R The value, and thus the change in reflection coefficient Δ after two impacts. R The accuracy of the results is affected. Therefore, attenuation compensation is needed for the second echo signal to eliminate the influence of coating thickness differences on the detection results.

[0052] To address the aforementioned issues, the coating thickness was measured using an eddy current thickness gauge before calculating the sound pressure reflection coefficient. h Eddy current thickness gauges utilize the eddy current principle to non-contactly measure the thickness of non-magnetic coatings on metal substrates, achieving a measurement accuracy of ±1μm. The acoustic attenuation coefficient is then measured through the following steps. α The acoustic wave sensor was fixed sequentially at two locations on the coating surface, 5 mm and 20 mm away from the laser impact point. The same acoustic wave coupling medium as used in the actual test was applied to these two locations. At least 10 seconds were allowed for the coupling medium to stabilize (ensuring consistent acoustic wave coupling). Then, a single laser impact was generated using a pulsed laser, and the peak voltage of the first echo signal at each of the two locations was measured. V 1,5 and V 1,20 Because the first echo signal originates from reflection from the coating surface, its propagation path is extremely short (only the distance from the coating surface to the sensor), and measurements were taken at two locations: 5mm and 20mm. V The difference in the 1 value is mainly caused by the attenuation of sound waves as they propagate on the coating surface. The formula is used... Calculate the sound wave attenuation coefficient α 15 represents the distance difference between the two measurement locations, in mm. αThe value reflects the inherent attenuation characteristics of the current coating material for sound waves.

[0053] To reduce the impact of energy fluctuations from two different laser shocks α The impact of calculation, measurement V 1,5 and V 1,20 The same pulsed laser should be used, and two measurements should be completed in the shortest possible time (e.g., first fix the sensor at 5mm, perform one laser shock and record; then move the sensor to 20mm, fire the laser shock again and record). Laser energy stability should be controlled within ±3%. To mitigate the influence of coupling differences, this method requires reapplying the coupling medium before each measurement and waiting 10 seconds for stabilization; simultaneously, the pressure applied to the sensor should be consistent during both measurements (a constant force clamp can be used). It has been verified that under good operating conditions... α The measurement repeatability error is <5%.

[0054] For typical coating materials (such as alumina, with an attenuation coefficient of approximately 0.1-0.5 Np / mm), a propagation distance of 15 mm results in an attenuation of 1.5-7.5 Np (13-65 dB), indicating significant signal attenuation. V 1,5 and V 1,20 The ratio is typically between 2 and 1000, allowing for accurate measurement. For low-attenuation materials (such as polymers, attenuation < 0.05 Np / mm), the attenuation at 15 mm is < 0.75 Np (6.5 dB). In this case, the sensor distance difference can be increased (e.g., 5 mm and 30 mm) to improve measurement sensitivity.

[0055] get α Then, the formula is used. The peak voltage of the second echo signal is attenuated and compensated, whereby... h For coating thickness. The compensated thickness will be... Replace the original V 2. Substitute into the formula for calculating the reflection coefficient The attenuation-compensated reflection coefficient was obtained. This compensation eliminates the round-trip propagation of sound waves within the coating (path length 2). h The attenuation effect caused by lateral propagation between the sensor and the impact point, as well as the calculated... R The value is only related to the interface reflection properties, and is not related to the coating thickness and material attenuation.

[0056] In another embodiment, in the non-destructive testing method for the bonding strength of the semiconductor device housing coating, in MIn addition to the laser impact point, an additional reference impact point is selected. This reference impact point is located on a reference sample prepared with the same material and coating process as the semiconductor device housing. The adhesion strength between the coating and the substrate at the reference impact point on the reference sample has been confirmed as qualified through a tensile test. The reference impact point is then subjected to the same... M Two identical laser impact steps were performed at the same laser impact point. The change in the reflection coefficient Δ at the reference impact point was measured. R ref When △ R ref When ≤0.01, use the formula The change in reflection coefficient at each laser impact point is corrected, and the corrected value is then... Replace the original △ R i Calculate the average value △ R avg When △ R ref If the value is greater than 0.01, the detection system is deemed to be in an abnormal state. After recalibration, the detection is performed again.

[0057] In determining coating adhesion strength by employing two laser shocks and calculating the average change in reflectance at multiple laser shock points, the test results may be affected by systematic energy factors such as laser energy fluctuations, changes in sensor coupling state, and ambient temperature drift. These factors can cause the change in reflectance Δ at all measurement points to be significant. R i This generates a common offset, known as systematic error. For example, if the laser output energy is 5% lower than the nominal value during a test, this will cause the acoustic wave amplitudes of the first and second impacts to decrease proportionally. While the reflection coefficient is theoretically insensitive to energy, changes in laser energy can affect the degree of interface damage, thus influencing Δ... R The more significant source of systematic error is the change in the coupling state between the acoustic sensor and the coating. If there is a slight movement in the sensor position or a change in the thickness of the coupling medium between two impacts, it will cause a systematic difference in the absolute amplitude of the two measurements, thus affecting the Δ... R i The overall value is too large or too small. This systematic error is unrelated to the bonding strength of the coating itself, but it will affect Δ. R avg The calculation results may lead to misjudgment.

[0058] To solve this problem, in MIn addition to the laser shock point, an extra reference shock point is selected. This reference shock point is located on a reference specimen, which is prepared using the exact same substrate material, coating material, and coating process as the semiconductor device under test (DUT) housing. More importantly, the adhesion strength between the coating and the substrate at the reference shock point on the reference specimen has been pre-confirmed as acceptable through tensile testing. This means that the adhesion strength at the reference shock point is known to be acceptable. Then, the reference shock point is subjected to the same... M Two identical laser impact steps were performed on a reference impact point, and the change in the reflection coefficient Δ at the reference impact point was measured. R ref Since the reference sample and the shell to be tested have the same material and manufacturing process, and the bonding strength at the reference impact point is qualified, therefore △ R ref The system error of the detection system under the current conditions should be reflected (such as sensor coupling status, laser energy drift, etc.). For a reference specimen with acceptable bonding strength, theoretically, there should be no interfacial micro-damage; therefore, Δ R ref The ideal value is 0. In actual testing, if the measured Δ R ref A value greater than 0 indicates a systematic positive drift under the current detection conditions.

[0059] To ensure the reliability of the detection results and avoid missed detections due to over-correction, this method adopts the following correction strategy: First, determine △ R ref The value of △. R ref If the error is ≤0.01, the detection system is considered to be under control, and the system error is within an acceptable small range. In this case, the formula is used. The change in reflection coefficient at each laser impact point is corrected, and the corrected value is then... Replace the original △ R i Calculate the average value △ R avg If △ R ref If the error is greater than 0.01, the system is considered abnormal, indicating a significant, non-negligible system error. In this case, mechanical correction should not be performed; instead, the detection system should be checked and recalibrated (e.g., cleaning the sensor surface, reapplying the coupling medium, and checking the laser energy output). Once the system returns to normal, the system should be recalibrated. M The system detects both a laser impact point and a reference impact point. This strategy ensures that correction operations are performed only within an acceptable range of system error, fundamentally avoiding the risk of missed detections due to large corrections.

[0060] For example, in a routine test, selectM = 5 laser shock points, and the measured △ R i are 0.04, 0.05, 0.06, 0.04, and 0.05 respectively. The direct average is 0.048, which is less than 0.05, so it is judged to be qualified. The △ of the reference shock point R ref is measured to be 0.008 (≤0.01), indicating that there is a small positive drift in the system. After correction, each is 0.032, 0.042, 0.052, 0.032, and 0.042 respectively. The average value is 0.04, which is still less than 0.05, so it is judged to be qualified. If in another test, the △ of the reference shock point R ref is measured to be 0.03 (>0.01), it indicates that the system state is abnormal and no correction should be made. The operator needs to check and recalibrate the system. After re-testing, when △ R ref drops below 0.01, then judge with the newly measured data.

[0061] Through the above correction strategy, it can effectively eliminate the influence of small system errors on the judgment result under the controlled state of the system, and avoid the risk of missed judgment caused by improper correction when the system is abnormal, making △ R avg more accurately reflect the true situation of the coating bonding strength. This method does not require complex equipment, only adding a reference point measurement and combining simple numerical judgment can improve the accuracy and consistency of detection, especially suitable for long-term continuous detection or production sites with large fluctuations in environmental conditions.

[0062] Effect description: Compared with the method without using reference point correction, this method can effectively suppress the influence of system errors on the judgment result. At the same time, by setting the correction prerequisite conditions, it prevents the risk of missed judgment caused by overcorrection and improves the robustness and repeatability of the detection method.

[0063] In another embodiment, the non-destructive testing method for the bonding strength of the semiconductor device housing coating involves creating an annular sound-absorbing layer on the coating surface centered on the laser impact point. The annular sound-absorbing layer has an inner diameter of 1.5-3 mm, an outer diameter of 4-6 mm, and a thickness of 0.5-1 mm. It is formed by coating a silicone rubber sound-absorbing material and completely surrounds the laser impact point. An acoustic sensor is located outside the annular sound-absorbing layer, with the center of the sensor 5-20 mm from the laser impact point. The outer diameter of the annular sound-absorbing layer is 4-6 mm, and the acoustic sensor is located outside the layer; that is, the distance between the center of the sensor and the laser impact point should be greater than the outer radius of the annular sound-absorbing layer (>3 mm), while still meeting the 5-20 mm distance requirement. The coating surface below the acoustic sensor should not be covered by the annular sound-absorbing layer to ensure that the sensor directly contacts the original coating surface through the coupling medium.

[0064] In the process of generating longitudinal waves using laser shock and calculating the reflection coefficient using the first and second echo signals, in addition to generating volume waves (longitudinal and transverse waves) propagating deep within the coating, the laser shock also excites surface waves (also known as Rayleigh waves) propagating along the coating surface. The energy of the surface waves is concentrated in a depth range of about one wavelength below the coating surface, and their propagation speed is slightly lower than that of the longitudinal waves. When the acoustic wave sensor is placed 5-20 mm away from the laser shock point, the surface waves propagate directly along the coating surface to the acoustic wave sensor. Their arrival time overlaps with or is close to that of the reflected longitudinal waves from the coating surface and interface, thus interfering with the accurate reception of the first echo signal (coating surface reflected longitudinal wave) and the second echo signal (interface reflected longitudinal wave) by the acoustic wave sensor. The amplitude of the surface waves is often large, which may drown out the smaller amplitude of the second echo signal, leading to… V 2. Inaccurate readings, which in turn affect the reflection coefficient. R and change Δ R The computational accuracy needs to be improved. Therefore, it is necessary to suppress the propagation of surface waves so that the acoustic wave sensor mainly receives longitudinal wave reflection signals from inside the coating.

[0065] To address the aforementioned issues, a ring-shaped sound-absorbing layer is fabricated on the coating surface, centered on the laser impact point. This ring-shaped sound-absorbing layer is formed by coating with silicone rubber sound-absorbing material (peelable silicone BR-5808B). Silicone rubber possesses excellent acoustic absorption properties, effectively absorbing sound wave energy propagating along the surface. The inner diameter of the ring-shaped sound-absorbing layer is 1.5-3 mm, the outer diameter is 4-6 mm, and the thickness is 0.5-1 mm. The inner diameter must be larger than the laser spot diameter (typically around 1 mm) to avoid the sound-absorbing material covering the laser impact point and affecting the normal incidence of laser energy; however, the inner diameter cannot be too large, otherwise the exposed surface between the impact point and the sound-absorbing layer will allow surface waves to propagate outward. The outer diameter determines the coverage area of ​​the sound-absorbing layer; an outer diameter of 4-6 mm ensures that within the sensor distance (5-20 mm), surface waves must pass through the sound-absorbing layer before reaching the sound wave sensor, thus being fully absorbed. A thickness of 0.5-1mm is sufficient to allow surface waves to be reflected and attenuated multiple times within the sound-absorbing layer, while avoiding excessive thickness that could affect the installation of the acoustic sensor or cause unnecessary stress to the coating. The annular sound-absorbing layer completely surrounds the laser impact point, forming a complete ring to ensure that surface waves propagating in all directions are absorbed. The acoustic sensor is located outside the annular sound-absorbing layer, with the distance between the center of the acoustic sensor and the laser impact point remaining 5-20mm, but the original acoustic coupling medium is still maintained between the acoustic sensor and the coating surface. In this way, the signal received by the acoustic sensor mainly comes from longitudinal waves propagating inside the coating (including longitudinal waves reflected back from the surface and interface), while surface waves propagating directly along the surface are significantly attenuated when passing through the annular sound-absorbing layer, thereby reducing the interference of surface waves on the echo signal.

[0066] For example, without using a ring-shaped sound-absorbing layer, the peak voltage of the first echo signal in a certain detection... V 1 is 120mV, the peak voltage of the second echo signal. V The voltage is 38mV, but there is a large surface wave pulse with an amplitude of 50mV before or between the first echo signal in the signal waveform. This pulse overlaps with the second echo signal, causing... V 2. Reading error ±5mV. After fabricating a ring-shaped silicone rubber sound-absorbing layer with an inner diameter of 2mm, an outer diameter of 5mm, and a thickness of 0.8mm, the surface wave amplitude attenuated to below 5mV, and the second echo signal waveform was clear. V 2. The reading stabilizes at 38mV, and the repeatability error is less than 1mV. This is how the calculated... R Value and △ RMore accurate and reliable. The dimensions of the annular sound-absorbing layer need to be optimized based on the laser spot size and sensor position: if the sound sensor is close (e.g., 5mm), a smaller outer diameter (e.g., 4mm) can be used; if the sound sensor is far away (e.g., 20mm), a larger outer diameter (e.g., 6mm) can be used to ensure that the surface waves are fully absorbed before reaching the sound sensor. The inner diameter should ensure that the edge of the sound-absorbing layer is at least 0.5mm away from the edge of the impact point to avoid direct laser ablation of the sound-absorbing material.

[0067] Effect Description: Compared to methods without annular sound-absorbing layers, this method significantly reduces the interference of surface waves on the longitudinal wave echo signal by absorbing surface-propagating sound waves, thereby improving the signal-to-noise ratio and reading stability of the second echo signal, and thus reducing the reflection coefficient. R and change Δ R The calculation is more accurate, reducing measurement errors caused by surface wave interference, and is especially suitable for detection scenarios where the surface acoustic wave propagation characteristics of the coating are strong (such as hard coatings) or the sensor is far away.

[0068] In another embodiment, in the non-destructive testing method for the bonding strength of the semiconductor device housing coating, when the coating thickness... h When the coating thickness is ≥0.3mm, no deconvolution processing is performed on the longitudinal wave signal acquired by the oscilloscope; when the coating thickness is... h When the error is <0.3mm, after acquiring the longitudinal wave signal using an oscilloscope, deconvolution processing is first performed on the acquired longitudinal wave signal to separate the overlapping first and second echo signals in the time domain into two independent pulse waveforms. Then, the peak voltage of the first echo signal is read from the separated independent pulse waveforms. V Peak voltage of the first and second echo signals V 2. The deconvolution process is performed according to the following steps: The longitudinal wave time-domain signal acquired by the oscilloscope is denoted as... y ( t The impulse response function of the acoustic wave sensor is denoted as... s ( t ),right y ( t )and s ( t The frequency domain signal was obtained by performing Fourier transforms on each of the two methods. Y ( f )and S ( f ); The frequency domain estimate of the reflection coefficient sequence is calculated using the Wiener filter deconvolution formula. R ( f ): in S* (f )for S ( f The conjugate of the complex number, | S ( f )| 2 for S ( f The power spectrum of ) K For regularization parameters, K The value is | S ( f )| 2 1% of the maximum value; when the coating thickness h When the pulse width is ≥0.3mm, the two echo pulses are basically separated, and deconvolution processing is usually not performed on the longitudinal wave signal acquired by the oscilloscope. However, if the pulse edges stick together due to sensor bandwidth limitations or high coating sound velocity, the operator can still choose to perform deconvolution processing to obtain a more accurate peak voltage.

[0069] right R ( f Performing an inverse Fourier transform yields the time-domain reflection coefficient sequence. r ( t ); exist r ( t From two independent pulse waveforms that appear in chronological order, the peak voltage of the first pulse waveform is read as the peak voltage of the first echo signal. V 1. Read the peak voltage of the second pulse waveform as the peak voltage of the second echo signal. V 2; the result obtained after deconvolution processing r ( t The peak voltage of the pulse in the signal has the same physical dimensions and proportional relationship as the peak voltage of the original longitudinal wave signal, and can be directly used for subsequent reflection coefficient calculation.

[0070] Impulse response function of acoustic sensor s ( t Obtain it in advance through the following steps: a) Prepare a metal standard test block with a thickness greater than 20 mm. The lower surface of the metal standard test block is flat and parallel to the upper surface. Given a longitudinal wave velocity of 6000 m / s, the arrival time of the bottom echo signal is approximately 2 × 20 / 6000 ≈ 6.67 μs. The arrival time of the surface echo signal, corresponding to a distance of 5-20 mm between the sensor and the impact point, is approximately (5-20) / 6000 ≈ 0.83-3.33 μs. Therefore, the bottom echo signal arrives at least 3.3 μs later than the surface echo signal. Since the pulse response width of the sensor is typically less than 0.5 μs, the bottom echo signal will not interfere with the measurement of the surface echo signal. If a material with a higher sound velocity (such as copper, with a sound velocity of 4700 m / s) is used, the arrival time of the bottom echo signal will be even later, while still maintaining safety.

[0071] b) Determine a laser impact point on the upper surface of the metal standard test block; c) Fix the acoustic wave sensor on the upper surface of the metal standard test block. The distance between the acoustic wave sensor and the laser impact point is 5-20 mm. Apply the same acoustic wave coupling medium as the one used in the formal test between the acoustic wave sensor and the metal standard test block. d) Use a pulsed laser to generate a laser beam, focus the laser beam on the laser impact point to generate a laser impact, and the energy density of the laser beam is the same as that of the formal test; e) Acquire the echo signal directly reflected from the upper surface of the metal standard test block using the acoustic wave sensor, and use the acquired echo signal as the impulse response function of the acoustic wave sensor. s ( t ).

[0072] In the process of calculating the reflection coefficient using two laser shocks and the first and second echo signals, the time interval between the first echo signal (reflection from the coating surface) and the second echo signal (reflection from the coating-substrate interface) depends on the two-way propagation time of the sound wave within the coating, i.e., Δ. t =2 h / v ,in h For coating thickness, v This refers to the longitudinal wave velocity in the coating. When the coating thickness is relatively thick (e.g., ...), h When ≥0.3mm), △ t The two echo pulses are relatively large, and they are clearly separated on the oscilloscope, allowing direct reading of their respective peak voltages. V 1 and V 2. However, when the coating thickness is thin (e.g.) h When <0.3mm), △ tThe two echo pulses are very small, and they partially or completely overlap in the time domain. The oscilloscope cannot distinguish which pulse is the first echo signal and which is the second echo signal in the acquired longitudinal wave signal, resulting in inaccurate readings. V 1 and V 2. Consequently, the reflection coefficient cannot be calculated. R and change Δ R For example, if the coating thickness is 0.15 mm and the longitudinal wave velocity is 5000 m / s, the two-way time is only 60 ns, while the pulse response width of a typical acoustic sensor is about 100 ns. The two echo signals completely overlap into a single composite waveform. Therefore, signal processing techniques are needed to separate the overlapping echo signals.

[0073] To solve the above problems, based on the coating thickness h Whether or not deconvolution processing is performed depends on whether the thickness is less than 0.3 mm. The 0.3 mm threshold is calculated based on the lower limit (3000 m / s) of the typical coating's longitudinal wave velocity range (3000-7000 m / s): two-way travel time Δ t =2×0.3mm / 3000m / s=200ns. Typically, the pulse response width of an acoustic sensor is approximately 100-150ns. When Δt ≥ 150ns, the two echo signals can be partially separated. t A separation is considered essentially complete when the sound velocity of the coating exceeds 200 ns. If the sound velocity of the coating is greater than 7000 m / s, the two-way time is shorter, and a thicker threshold may be required. In this method, 0.3 mm is a recommended value; in practical applications, the actual sound velocity of the coating material can be determined using the formula... h min Adjust by multiplying the pulse width by the velocity of sound by 2. When the coating thickness... h When the difference is ≥0.3mm, the two echo signals are sufficiently separated, and the original signal acquired by the oscilloscope can be directly read. V 1 and V 2 is sufficient; no additional processing is required to save calculation time. (Regarding coating thickness...) h When the pulse width is <0.3mm, after the oscilloscope acquires the longitudinal wave signal, deconvolution processing is first performed on the acquired signal to separate the first and second echo signals that overlap in the time domain into two independent pulse waveforms. Then, the pulse waveforms are read from the separated independent waveforms. V 1 and V 2. Deconvolution processing is performed according to the following steps: The longitudinal wave time-domain signal acquired by the oscilloscope is denoted as... y ( t The impulse response function of the acoustic wave sensor is denoted as... s ( t ).right y ( t )and s ( tThe frequency domain signal was obtained by performing Fourier transforms on each of the two methods. Y ( f )and S ( f The frequency domain estimate of the reflection coefficient sequence is calculated using the Wiener filter deconvolution formula. R ( f The formula is: ,in S *( f )for S ( f The conjugate of the complex number, | S ( f )| 2 for S ( f The power spectrum of ) K These are the regularization parameters. K The value can be | S ( f )| 2 The maximum value is 1%, which serves to prevent numerical instability caused by a denominator of zero and to suppress the amplification of high-frequency noise. Then, for... R ( f Performing an inverse Fourier transform yields the time-domain reflection coefficient sequence. r ( t ).exist r ( t In the sequence, two independent pulse waveforms will appear: the first pulse corresponds to the first echo signal (reflection from the coating surface), and the second pulse corresponds to the second echo signal (reflection from the interface). The peak voltage of the first pulse waveform is then read as... V 1. The peak voltage of the second pulse waveform is used as... V 2. Separate these two... V 1 and V Substituting 2 into the subsequent formula for calculating the reflection coefficient, an accurate result can be obtained. R value.

[0074] Impulse response function of acoustic sensor s ( tThis requires prior acquisition; theoretical values ​​cannot be used directly because the actual response of each sensor varies. The acquisition steps are as follows: Prepare a metal standard test block (e.g., aluminum alloy or stainless steel) with a thickness greater than 20mm. The lower surface of the test block should be flat and parallel to the upper surface. The thickness of more than 20mm ensures that the echo signal from the lower surface is much later than the echo signal from the upper surface, thus not interfering with the acquisition of the upper surface echo signal. Determine a laser impact point on the upper surface of the metal standard test block. Fix the acoustic wave sensor to the upper surface of the metal standard test block, with the distance between the sensor and the laser impact point the same as in the actual test (e.g., 5-20mm). Apply the same acoustic coupling medium as in the actual test between the sensor and the metal standard test block. Use a pulsed laser to generate a laser beam, focus it on the laser impact point, and generate a laser impact. The energy density of the laser beam is the same as in the actual test. At this time, the laser impact excites acoustic waves on the upper surface of the metal standard test block. Since the acoustic impedance of the metal material is much greater than that of air, almost all the acoustic waves are reflected back to the sensor from the upper surface, while the acoustic waves that penetrate into the metal will not return in a short time because the lower surface is far away. Therefore, the signal acquired by the sensor is the waveform of the acoustic wave excited by the laser shock after reflection from the upper surface. This waveform includes the sensor's own impulse response as well as the responses of the entire measurement chain, including the coupling medium and the preamplifier. This waveform is then used as... s ( t ).Should s ( t This reflects the sensor output waveform corresponding to an ideal unit reflection event under the current detection conditions. (Obtained) s ( t Then it can be used for subsequent deconvolution processing.

[0075] For example, regarding thickness h =0.15mm alumina coating (sound velocity approximately 6000m / s, two-way time 50ns), directly acquired by oscilloscope. y ( t The two echo signals in the image completely overlapped and could not be read. V 1 and V 2. Pre-acquire sensor impulse response s ( t (Its main pulse width is approximately 80 ns). For y ( t )and s ( t Perform deconvolution to obtain r ( t Two clearly separated pulses in the image, the first pulse peak V 1 = 118mV, the peak value of the second pulse V 2 = 35mV, calculate R=0.297. Without deconvolution, the peak values ​​forcibly read from the overlapping waveforms might be 100mV and 50mV, leading to... R =0.5, which deviates significantly from the true value. By using deconvolution, the accurate value was obtained... V 1 and V 2, thus ensuring △ R The accuracy of the calculation.

[0076] Results: Compared to methods that directly read signals regardless of coating thickness, this method uses a 0.3mm threshold to determine whether deconvolution is needed. This ensures processing efficiency for thick coatings (eliminating the need for complex calculations) while resolving the issue of overlapping echo signals that cannot be read in thin coatings. This allows the detection method to be applied to a wider range of coating thicknesses (from tens of micrometers to millimeters). Regularization parameters in deconvolution. K Taking 1% of the maximum power spectrum value is an empirical value that achieves a good balance between noise suppression and maintaining signal resolution. The sensor impulse response function is obtained in advance. s ( t The method is simple and reliable, requires no expensive equipment, and can be re-acquired before each test to adapt to the current testing conditions. Therefore, this method significantly improves the feasibility and accuracy of thin coating bonding strength testing.

[0077] In another embodiment, in the non-destructive testing method for the bonding strength of the semiconductor device housing coating, after obtaining... M The peak voltage of the first echo signal at each laser impact point V 1i After that, when any one V 1i When the voltage is <5mV, the laser impact point is from... M Remove from the laser impact points, and record the number of remaining laser impact points as follows: M ',use M The average reflection coefficient of ' laser impact points Δ R avg .

[0078] In determining the coating adhesion strength by employing two laser shocks and calculating the average change in reflection coefficient at multiple laser shock points, the peak voltage of the first echo signal at each laser shock point is... V 1i It reflects the intensity of the longitudinal wave reflected from the coating surface and received by the acoustic sensor. V 1iThe size is affected by factors such as the coupling state between the sensor and the coating, the local smoothness of the coating surface, and the presence of contaminants or microbubbles. When the coupling conditions at a certain laser impact point are poor (e.g., uneven application of the coupling medium leading to local defects, microbubbles under the sensor, oil or particulate matter on the coating surface), the size of that point will be affected. V 1i It will be significantly lower than the normal value. If V 1i If the value is too small, the signal-to-noise ratio will decrease, and the second echo signal will be affected. V 2i The reflection coefficient measured at that point may be affected by noise. R i and change Δ R i Unreliable. Including such unreliable data points in the average calculation will reduce Δ. R avg This could compromise the representativeness of the data and even lead to misjudgments. Therefore, it is necessary to identify and eliminate those results caused by abnormal measurement conditions. V 1i Too low a point.

[0079] To solve the above problems, after obtaining M The peak voltage of the first echo signal at each laser impact point V 1i After that, check each one. V 1i The value. When any one V 1i When the voltage is less than 5mV, the laser impact point is moved from... M Removed from the specified points. It should be noted that 5mV is the minimum acceptable threshold determined based on a typical noise floor of 0.5mV and a signal-to-noise ratio of not less than 10dB. In practical testing, if the system noise floor is lower or higher measurement reliability is desired, it is recommended to increase the rejection threshold to 10-30mV. For example, under normal coupling... V When the typical value is 120mV, using a 20mV threshold ensures a signal-to-noise ratio of no less than 32dB, resulting in a more stable second echo signal reading. The aforementioned increase in the threshold is a conventional optimization by those skilled in the art and does not exceed the scope of protection of this invention. Therefore, [the following is a continuation of the previous sentence, but the context is unclear]. V 1i Points with a voltage <5mV are considered invalid and discarded. The number of remaining points after discarding is recorded as follows: M '. Then adopt M Calculate the average value of the reflection coefficient at each point Δ R avg The average value is then compared with a threshold of 0.05 to determine whether the bonding strength is acceptable. It should be noted that after rejection... MThe value should not be less than 3; otherwise, the test should be repeated.

[0080] For example, in a certain test, select M =5 laser impact points, measured V 1i The values ​​were 120mV, 115mV, 3mV, 118mV, and 122mV, respectively. The third point... V 1 = 3mV, below 5mV. Inspection revealed a tiny air bubble beneath the sensor at this point, causing poor coupling. After removing this point, the remaining... M = 4 points. The first reflection coefficient of these four points. R i The values ​​are 0.32, 0.33, 0.32, and 0.33 respectively, representing the second reflection coefficient. R i Given ' = 0.33, 0.34, 0.33, and 0.34 respectively, calculate the Δ for each point. R i The values ​​are 0.01, 0.01, 0.01, and 0.01 respectively, with an average Δ. R avg =0.01, less than 0.05, is considered acceptable. If this outlier is not removed, the value of this point... V When 1=3mV V 2 might only be around 1mV, with an extremely low signal-to-noise ratio. The measured reflection coefficient might be 0.33 or 0.30, but it fluctuates greatly, leading to Δ R The calculation is unreliable. Even worse, if the point... V 2. If the reading is 0 due to noise, then R =0, △ R The values ​​may be abnormally small or large, interfering with the average. Therefore, they are removed. V Points with 1 < 5mV can avoid the influence of these unreliable data on the final judgment result.

[0081] Effect description: Compared to not performing V Compared to the method of directly calculating the average value using all points, this method, by setting a 5mV threshold, can automatically identify and eliminate invalid measurement points caused by poor local coupling or surface contamination, thereby improving the accuracy of ΔV measurement. R avg This method improves reliability and repeatability, reducing misjudgments caused by accidental operational factors. Furthermore, it is simple and easy to implement, requiring no additional equipment and only requiring readings from an oscilloscope. V A single value is sufficient for judgment, making it suitable for rapid testing on the production site.

[0082] In another embodiment, the non-destructive testing method for the bonding strength of the semiconductor device housing coating includes... MBefore each laser impact point undergoes its first laser impact, the surface roughness of the coating at each laser impact point is measured using a non-contact optical surface roughness measuring instrument. The surface roughness at any given laser impact point is then... Ra When the value is >1.5 μm, the laser impact point is moved from... M Remove from the laser impact points, and record the number of remaining laser impact points as follows: M ",use M The average reflection coefficient of each laser impact point is calculated as Δ. R avg .

[0083] In determining coating adhesion strength using two laser shocks and calculating the average change in reflection coefficient at multiple laser shock points, the local roughness of the coating surface directly affects the coupling state between the acoustic sensor and the coating, as well as the reflection characteristics of acoustic waves on the coating surface. When the surface roughness of the coating at a certain point... Ra When the value is large, the coating surface is uneven, and the acoustic coupling medium cannot completely fill the micro-depressions on the surface, resulting in local air gaps between the sensor and the coating. These air gaps significantly increase the reflection loss of acoustic waves at the interface, causing the peak voltage of the first echo signal received by the sensor to decrease. V Peak voltage of 1st and second echo signals V Both decreased proportionally, but the ratio between the two... V 2 / V The reflectance coefficient (1) may also vary due to scattering at different frequencies. More importantly, rough surfaces enhance diffuse scattering of sound waves, converting some longitudinal wave energy into transverse or surface waves, thereby reducing the signal-to-noise ratio and stability of the second echo signal. Furthermore, the energy distribution of the laser beam on a rough surface may be uneven, affecting the reproducibility of interface damage. Therefore, the reflectance coefficient measured at points with excessive roughness will vary. R i and change Δ R i The value may deviate from the true value; including these unreliable data points in the average calculation will reduce Δ. R avg The representativeness of the impact could be compromised, potentially leading to misjudgment. Therefore, it is necessary to evaluate the surface roughness of each impact point before testing and eliminate points with excessive roughness.

[0084] To solve the above problems, in the context of MBefore each laser impact point undergoes its first laser impact, the surface roughness of the coating at each impact point is measured using a non-contact optical surface roughness measuring instrument. Based on the principles of white light interference or laser confocalization, this non-contact optical measuring instrument can quickly measure the roughness parameters of a local area without contacting the coating surface, achieving a measurement accuracy of up to 0.01 μm. When measuring the surface roughness of the coating at each laser impact point using the non-contact optical surface roughness measuring instrument, the measurement area should be a circular region centered on that impact point with a diameter not less than the diameter of the laser spot (typically 1 mm). Measurements should be taken along at least three contour lines in different directions, and their values ​​should be calculated. Ra The arithmetic mean of the values ​​at that point Ra Value. When any laser strikes a point Ra When the value is >1.5μm, the laser shock point is moved from... M Removed from individual points. The 1.5μm threshold is calculated based on ultrasonic coupling theory: when the surface roughness... Ra When the wavelength exceeds approximately 1 / 20 of the acoustic wave (this method uses a 5MHz longitudinal wave, with a typical coating wavelength of approximately 1-1.2mm; 1 / 20 of the wavelength is approximately 50-60μm, much larger than 1.5μm), the effect of surface roughness on coupling can be ignored. However, in actual experiments, when... Ra When the reflectance is greater than 1.5 μm, the filling capacity of the coupling medium decreases, and the standard deviation of the reflectance coefficient measured multiple times at the same location exceeds 10%. Therefore, the threshold is set to 1.5 μm. This value is suitable for most ceramic and metal coatings, and can be adjusted appropriately for softer or harder coatings. Ra When the surface micro-peak and valley heights are greater than 1.5 μm, significant differences occur, making it difficult for the coupling medium to completely fill the gaps. This leads to an unstable coupling state, and the reflection coefficient fluctuation at the same location may exceed 10% in multiple measurements, severely impacting the reliability of the detection results. Therefore, [the following is missing from the original text: "will..."] Ra Points larger than 1.5 μm were deemed unsuitable for detection and were discarded. The number of remaining points after discarding was recorded as follows: M Then adopt M "Calculate the average value of the reflection coefficient at each point △" R avg The average value is then compared with a threshold of 0.05 to determine whether the bonding strength is acceptable. It should be noted that after rejection... M "The number should not be less than 3; otherwise, the testing area should be reselected or the outer casing surface should be properly cleaned. If the number of remaining points after rejection..." M If the value is less than 3, at least 3 new laser impact points should be selected for testing. Before retesting, the coating surface should be inspected and cleaned to ensure that the coupling medium is applied evenly.

[0085] For example, in a certain test, select M =5 laser impact points, and the surface roughness of each point was measured. Ra The values are 0.8μm, 0.9μm, 1.8μm, 0.7μm, and 1.6μm respectively. Among them, the third point Ra =1.8μm is greater than 1.5μm, and the fifth point Ra =1.6μm is also greater than 1.5μm. These two points are removed. The remaining M "=3 points, and their Ra values are all below 0.9μm. The first and second laser shocks are performed on these 3 points, and the measured △ R i are 0.01, 0.01, and 0.02 respectively. The average △ R avg =0.0133, which is less than 0.05, and it is determined to be qualified. If the points with excessive roughness are not removed, due to the rough surface of the third point and poor coupling, the value measured in the first shock R may be abnormally high (such as 0.45). The change amount △ R after the second shock may be 0.20, resulting in an average △ R avg reaching above 0.05 and being misjudged as unqualified. Or due to roughness, the signal of the fifth point is unstable, and the △R of the two shocks fluctuates greatly, which will also interfere with the average value. Therefore, removing Ra points >1.5μm can avoid the interference of surface roughness on the acoustic wave coupling and the measurement of the reflection coefficient, and improve the reliability and consistency of the detection.

[0086] Effect description: Compared with the method of directly detecting all points without surface roughness screening, this method measures the Ra values of each point in advance and removes the points with excessive roughness, ensuring that the measurement points participating in the calculation have good acoustic coupling conditions, thereby improving the accuracy and repeatability of △ R avg and reducing misjudgments caused by local roughness on the coating surface. The use of a non-contact optical measuring instrument does not damage the coating and meets the requirements of non-destructive testing. This method is especially suitable for the outer shell surfaces that have been pretreated by sandblasting, grinding, etc., and these surfaces may have uneven local roughness. By removing the rough points, the detection results can more truly reflect the quality of the coating bonding strength itself.

[0087] Although the embodiments of the present invention have been disclosed above, it is not limited to only the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those familiar with the field, additional modifications can be easily achieved. Therefore, without departing from the general concept defined by the claims and the equivalent scope, the present invention is not limited to specific details.

Claims

1. A non-destructive testing method for the bonding strength of coatings on semiconductor device housings, characterized in that, include: An acoustic coupling medium is disposed on the coating surface of the semiconductor device housing, and an acoustic sensor is coupled to the acoustic coupling medium. Select on the coating surface M There are 1 laser impact points, among which M ≥3; A laser beam generated by a pulsed laser is focused onto an arbitrary laser impact point using a focusing lens; the energy density of the laser beam is 0.1-1 mJ / mm². 2 The acoustic sensor is located on the coating surface at a distance of 5-20 mm from the laser impact point; The longitudinal wave signal generated after each laser impact is collected by an acoustic wave sensor. The first longitudinal wave pulse that arrives after the laser impact is the first echo signal, and the next longitudinal wave pulse that arrives after the first echo signal is the second echo signal. Connect the longitudinal wave signal output from the acoustic wave sensor to the input channel of the oscilloscope, set the oscilloscope's trigger mode to edge trigger, and use the oscilloscope's cursor measurement function to read the peak voltage of the first echo signal. V Peak voltage of the first and second echo signals V 2; Using formula Calculate the acoustic pressure reflection coefficient after each laser impact. R ; right M Each laser impact point is subjected to a first laser impact, and the reflection coefficient corresponding to each laser impact point is recorded to obtain a reflection coefficient sequence. R i , i =1,2,…, M ; Using the same laser energy density as the first impact, for the same M A second laser impact is performed on each laser impact point, and the reflection coefficient of each point after the second impact is recorded to obtain a reflection coefficient sequence. ; Calculate the change in reflection coefficient at each laser impact point. and calculate M △ R i The average value △ R avg ; When △ R avg If the strength is greater than 0.05, the bonding strength between the coating and the substrate is deemed insufficient; otherwise, the bonding strength between the coating and the substrate is deemed sufficient.

2. The non-destructive testing method for the bonding strength of the semiconductor device housing coating as described in claim 1, characterized in that, right M After each laser impact point is subjected to one laser impact, a waiting time interval Δ is required. T Then a second laser shock is performed, in which △ T The time required for the coating surface temperature to recover to the temperature before laser shock is greater than or equal to the time required for the coating surface temperature to recover. The coating surface temperature is monitored in real time by an infrared thermometer.

3. The non-destructive testing method for the bonding strength of the semiconductor device housing coating as described in claim 1, characterized in that, In calculation M △ R i The average value △ R avg Before that, let's start from M △ R i Remove the maximum and minimum values ​​from the list, and then process the remaining values... M -2 △ R i Calculate the average value and use this average value as Δ. R avg .

4. The non-destructive testing method for the bonding strength of the semiconductor device housing coating as described in claim 1, characterized in that, When using the formula Before calculating the sound pressure reflection coefficient, the coating thickness is first measured using an eddy current thickness gauge. h Then, the sound wave attenuation coefficient is measured through the following steps. α : The acoustic wave sensor was fixed sequentially at the coating surface at distances of 5 mm and 20 mm from the laser impact point. The same acoustic wave coupling medium as used in the actual test was applied to each location. After at least 10 seconds for the coupling medium to stabilize, a laser impact was generated using a pulsed laser. The peak voltage of the first echo signal at each of the two locations was measured. V 1,5 and V 1,20 Using formula calculate α , where 15 is the distance difference between the two measurement locations, in mm; Using formula The peak voltage of the second echo signal is attenuated and compensated. Replace the original V 2. Substitute into the formula for calculating the reflection coefficient .

5. The non-destructive testing method for the bonding strength of the semiconductor device housing coating as described in claim 1, characterized in that, exist M In addition to the laser impact point, an additional reference impact point is selected. This reference impact point is located on a reference sample prepared with the same material and coating process as the semiconductor device housing. The adhesion strength between the coating and the substrate at the reference impact point on the reference sample has been confirmed as qualified through a tensile test. The reference impact point is then subjected to the same... M Two identical laser impact steps were performed at the same laser impact point. The change in the reflection coefficient Δ at the reference impact point was measured. R ref When △ R ref When ≤0.01, use the formula The change in reflection coefficient at each laser impact point is corrected, and the corrected value is then... Replace the original △ R i Calculate the average value △ R avg When △ R ref If the value is greater than 0.01, the detection system is deemed to be in an abnormal state. After recalibration, the detection is performed again.

6. The non-destructive testing method for the bonding strength of the semiconductor device housing coating as described in claim 1, characterized in that, An annular sound-absorbing layer is formed on the coating surface with the laser impact point as the center. The inner diameter of the annular sound-absorbing layer is 1.5-3mm, the outer diameter is 4-6mm, and the thickness is 0.5-1mm. The annular sound-absorbing layer is formed by coating silicone rubber sound-absorbing material. The annular sound-absorbing layer completely surrounds the laser impact point. The sound wave sensor is located outside the annular sound-absorbing layer, and the distance between the center of the sound wave sensor and the laser impact point is 5-20mm.

7. The non-destructive testing method for the bonding strength of the semiconductor device housing coating as described in claim 1, characterized in that, When coating thickness h When the coating thickness is ≥0.3mm, no deconvolution processing is performed on the longitudinal wave signal acquired by the oscilloscope; when the coating thickness is... h When the error is <0.3mm, after acquiring the longitudinal wave signal using an oscilloscope, deconvolution processing is first performed on the acquired longitudinal wave signal to separate the overlapping first and second echo signals in the time domain into two independent pulse waveforms. Then, the peak voltage of the first echo signal is read from the separated independent pulse waveforms. V Peak voltage of the first and second echo signals V 2. The deconvolution process is performed according to the following steps: The longitudinal wave time-domain signal acquired by the oscilloscope is denoted as... y ( t The impulse response function of the acoustic wave sensor is denoted as... s ( t ),right y ( t )and s ( t The frequency domain signal was obtained by performing Fourier transforms on each of the two methods. Y ( f )and S ( f ); The frequency domain estimate of the reflection coefficient sequence is calculated using the Wiener filter deconvolution formula. R ( f ): in S* ( f )for S ( f The conjugate of the complex number, | S ( f )| 2 for S ( f The power spectrum of ) K For regularization parameters, K The value is | S ( f )| 2 1% of the maximum value; right R ( f Performing an inverse Fourier transform yields the time-domain reflection coefficient sequence. r ( t ); exist r ( t From two independent pulse waveforms that appear in chronological order, the peak voltage of the first pulse waveform is read as the peak voltage of the first echo signal. V 1. Read the peak voltage of the second pulse waveform as the peak voltage of the second echo signal. V 2; Impulse response function of acoustic sensor s ( t Obtain it in advance through the following steps: a) Prepare a metal standard test block with a thickness greater than 20 mm, wherein the lower surface of the metal standard test block is flat and parallel to the upper surface; b) Determine a laser impact point on the upper surface of the metal standard test block; c) Fix the acoustic wave sensor on the upper surface of the metal standard test block. The distance between the acoustic wave sensor and the laser impact point is 5-20mm. Apply the same acoustic wave coupling medium as the one used in the formal test between the acoustic wave sensor and the metal standard test block. d) Use a pulsed laser to generate a laser beam, focus the laser beam on the laser impact point to generate a laser impact, and the energy density of the laser beam is the same as that of the formal test; e) Acquire the echo signal directly reflected from the upper surface of the metal standard test block using the acoustic wave sensor, and use the acquired echo signal as the impulse response function of the acoustic wave sensor. s ( t ).

8. The non-destructive testing method for the bonding strength of the coating on a semiconductor device housing as described in claim 1, characterized in that, In obtaining M The peak voltage of the first echo signal at each laser impact point V 1i After that, when any one V 1i When the voltage is <5mV, the laser impact point is from... M Remove from the laser impact points, and record the number of remaining laser impact points as follows: M ',use M The average reflection coefficient of ' laser impact points Δ R avg .

9. The non-destructive testing method for the bonding strength of the coating on a semiconductor device housing as described in claim 1, characterized in that, right M Before each laser impact point undergoes its first laser impact, the surface roughness of the coating at each laser impact point is measured using a non-contact optical surface roughness measuring instrument. The surface roughness at any given laser impact point is then... Ra When the value is >1.5μm, the laser shock point is moved from... M Remove from the laser impact points, and record the number of remaining laser impact points as follows: M ",use M The average reflection coefficient of each laser impact point is calculated as Δ. R avg .

Images