Ding Xiashaw@teclab.cnRuitaike (China) Co., Ltd

Summary:This article introduces various non-contact detection methods based on laser ultrasound for evaluating the quality of chip packaging. By comparing with ultrasonic microscope, X-ray and scanning electron microscope, it was found that laser ultrasonic testing technology has a relatively higher detection possibility in the quality inspection of the second layer bonding; Compared with common time-domain or amplitude imaging methods, the resonance time-domain amplitude correlation coefficient method is easier to implement, can significantly improve detection speed, and has a relatively high detection rate. Finally, several potential non-destructive testing methods for chip packaging were proposed and preliminarily validated.

keyword:Laser ultrasonic chip packaging inspection ultrasonic microscope

Background:

In the ever-changing international situation, the domestic chip industry is thriving and catching up. As the industrial chain gradually improves, the demand for quality control of chips will also increase day by day. Inverted chips are closer to the market due to their high performance and interface density. However, the reliability of detecting such chips has always been a challenge. At present, non-destructive testing of chips often uses ultrasonic microscopes and X-rays. The damaging method is to cut and polish the chip cross-section and observe it under a scanning electron microscope. Ultrasonic microscope is a cost-effective and harmless non-destructive testing method for the human body. It can sensitively detect the bonding quality of the first layer, but often the bonding quality of the second layer is not ideal. X-ray testing also faces the same problem. Laser ultrasound has great potential in chip packaging inspection due to its following advantages:

1. Non contact, no need for coupling agents, no need to immerse the chip in coupling liquid for testing, avoiding the possibility of damage to the chip by the coupling liquid and eliminating the subsequent process of removing the coupling liquid. It is a more suitable testing method for full inspection.

2. It can stimulate very broadband ultrasound signals, increasing the possibility of high-precision imaging of internal defects

3. Due to the use of laser beams for both excitation and detection, it is possible to reduce the spot size to the micrometer level through optical path focusing technology, thereby improving the spatial resolution of imaging.

Based on laser ultrasonic testing, the packaging quality of chips can be mainly layered in the following ways:

1. Laser ultrasonic microscope: This method requires a scanning system with micrometer level positioning accuracy and repeatability, a narrow pulse width excitation laser, a focusing optical path with a very small focal spot, a high-power detection laser, and a high signal-to-noise ratio mixing interference receiving system. In order to obtain high-precision imaging results, imaging algorithms may need to use SAFT aperture focusing acquisition or even FMC full matrix acquisition as appropriate. This method can perform full field scanning of chips, accurately locate defects in both plane and depth, and even perform equivalent evaluation of defects. However, the difficulty of system construction is high, and chips are often not homogeneous or isotropic in the thickness direction. When the sound velocity propagates in multi-layer structures, it will change due to its anisotropy, and appropriate sound velocity adaptation calibration is needed. Currently, this method has not been used by mature cases and literature to characterize defects in chips.

2. Resonance time-domain amplitude analysis method: This method divides the chip surface into multiple regions, excites ultrasonic signals at the center of each region with an excitation laser, and receives ultrasonic signals at different positions in each region through an interferometer for single or multiple times. Further analyze the bonding quality of the region by comparing the time-domain signals of intact chips and defective chips. This method is a non full field scanning scheme that can quickly assess whether there are quality defects in various areas. It can roughly locate the planar position of the defects, but cannot accurately determine the defect degree and size equivalent. This article will focus on the application of this technology in chip packaging bonding quality inspection.

3. Resonance spectrum analysis method: This method involves applying sweep excitation to the chip and plotting the frequency spectrum of the received signal to obtain the resonance frequency spectrum of the chip within a certain frequency range. By comparing the resonance frequency spectra of intact chips and chips with defects, it will be found that the resonance frequency peak of chips with defects will have a relative frequency shift compared to intact chips. Evaluate whether the chip has defects by measuring the frequency offset. This method can quickly evaluate whether the signal is qualified. But the reason for the non conformance cannot be determined (it may be incomplete bonding, chip size deviation, or even raw material deviation, etc.).

Chip sample:

The experimental object adopts a PWB board with a size of 180mm x 180mm as shown in the figure below. The inverted chip (18.5mm x 20mm) is packaged in the FCBGA (52.5mm x 52.5mm) in the middle position. The entire chip has 2597 bonding points distributed in a matrix of 51 x 51 with a spacing of 1mm, and each bonding point has a diameter of about 0.5-0.6 mm.

Figure 1. Physical picture and internal structure diagram of inverted chip

The experiment used three identical intact chips, with one intact chip as the reference object. The other two chips were subjected to drop tests to induce incomplete bonding defects in their internal bonding areas due to cracks.

Table 1. Chip Drop Test

Laser ultrasonic chip detection system:

Figure 2. Laser ultrasonic chip packaging quality testing device and schematic diagram

The excitation laser power needs to be set high enough to generate strong ultrasonic signals entering the chip, while also being low enough to avoid ablation on the chip surface. This article uses a 5 ns pulse width and 80 mW excitation laser to generate ultrasonic signals on the chip surface at a 45 ° angle. The surface temperature of the chip incident point quickly reaches its highest value after being excited by laser, and rapidly decreases within 40-50 ns.

The temperature curve T (t) of the silicon wafer surface can be expressed as follows:

I₀Representing the surface laser absorption flux density

K represents thermal conductivity

K represents thermal diffusivity

T represents time

In this article, the excitation laser spot area A is approximately 6.14 mm²The laser absorbs energy E₀=(1-R) E, where E represents the incident laser intensity and R represents the reflectivity. When a 1064 nm wavelength laser is incident on a silicon wafer at a 45 ° angle, the photon reflectivity is approximately 0.43.

In order to ensure the thermal elastic effect generated by the excitation laser on the surface of the chip, so that ultrasound can propagate inside and on the surface of the chip without ablation, based on the above formula, the temperature curves of the excitation laser with a pulse width of 5 ns and an excitation laser of 80 mW at different depths on the surface of the silicon wafer were analyzed.

Figure 3. Temperature curves at different depths on the surface of silicon wafers

From the curve, it can be seen that the maximum temperature on the surface of the silicon wafer is about 640 K, far below the melting point of the silicon wafer, and drops to 350 K in a very short time of about 50 ns, so the excitation laser will not cause ablation damage on the surface of the silicon wafer.

Figure 4. Schematic diagram of laser excited ultrasound and transient out of plane displacement interference reception

Detection method:

The farther away from the excitation laser point, the greater the ultrasonic attenuation. At the same time, soft materials, multi-layer structures, and uneven test surfaces can also cause rapid attenuation of ultrasonic signals over short distances. In this experiment, there were a large number of soft materials filled in the chip and an 11 layer substrate was used. In order to ensure that the laser interferometer can receive a sufficiently strong signal, two methods can be considered: 1) using a very high-power laser to incident on the central surface of the chip, and 2) layering the chip into 9 sub regions, as shown on the left in Figure 5, and using a relatively low-power laser to excite the center position of each sub region surface, while the laser interferometer receives ultrasonic signals in that sub region. Considering the balance of thermoelastic and ablation effects, this article adopts the second method for detection.

Figure 5. Chip detection area division 1-9 (left), area 5 signal acquisition method corresponding to X-ray imaging (right)

If laser ultrasound microscopy imaging method is used, full field scanning is required to collect ultrasound signals for each bonding point. The chip has 2597 bonding points, which takes a long time. In order to achieve rapid detection, signal acquisition is performed every 3x3 bonding point matrix, and it is assumed that defects or abnormal bonding within the bonding matrix will affect the interferometer's received signal. On the right side of Figure 5, black represents the bonding solder joint, and blue represents the laser interferometer signal acquisition point.

Table 2. Testing Parameters Table

In order to quantitatively analyze the signals received by the laser interferometer and evaluate the bonding quality of the signals, the following time-domain amplitude correlation coefficient correction method was designed:

In the formula, Rn represents the received signal of the intact chip reference board, and R represents the average value of Rn

An represents the received signal at the tested location, and Ā represents the average value of An

N represents the number of signal sampling points

According to the formula, when the calculation result is 0, the test signal and reference signal match perfectly, indicating that the chip has no defects or abnormalities. Similarly, when the result is greater than a certain threshold, it indicates that there is an abnormality in the chip.

Due to the non full field scanning method, the time required to complete the entire chip inspection will be significantly shorter than the time required for point by point scanning imaging with laser ultrasound microscopy.

Test results:

Figure 6. # 2 chip edge 51 detection result (left) # 3 chip edge 1 detection result (right)

Figure 6. # 2 chip edge 51 detection result (left) # 3 chip edge 1 detection result (right)

From the analysis of the test results, there is a possibility of bonding failure in the four corners of the two test chips. At the same time, abnormal values were found in the 2 and 6 areas of the # 3 test chip, which were analyzed to be caused by the unevenness and roughness of the chip surface.

Result verification:

To verify the accuracy and effectiveness of the detection method, finite element simulation was first performed on the chip after the drop experiment, and the results are shown in Figure 8. The simulation results are highly consistent with the laser ultrasonic detection results.

Figure 8. Finite Element Simulation Results of Chip

2. Through high-precision scanning imaging of chip corners using a 200 MHz immersion high-frequency ultrasonic microscope, it was found that the ultrasonic microscope can detect defects in the first layer, but cannot detect and identify defects in the second layer. The same X-ray 2D imaging results indicate that microcracks are hidden in bonding points or other structural projections and cannot be accurately identified and extracted.

Figure 9. Ultrasonic microscope chip corner scanning imaging (left) X-ray 2D imaging (right)

3. In order to obtain concrete evidence to prove the feasibility of this laser ultrasonic detection chip, the tested chip was cut and polished, and then the cross-section was observed by scanning electron microscopy, as shown in Figures 10 and 11. It can be seen that there are obvious crack defects with incomplete bonding in all four corners of the chip.

Figure 10. Scanning electron microscopy imaging of the corner bonding position of chip # 2 shows obvious microcracks emerging from the bonding area

Figure 11 # 3 Scanning electron microscopy imaging of the edge and corner bonding positions of the chip shows obvious cracks on the substrate surface below the solder joints

Other potential non-destructive testing technologies for chips:

1. Resonance spectrum analysis method

The resonance spectrum analysis method can be implemented in various forms, as shown in Figure 12. Air coupled ultrasound is used as the excitation signal source to excite the chip sample in ultrasonic resonance, and a laser vibrometer is used to measure the vibration velocity (frequency) of the chip surface. It is found that there is a significant frequency shift phenomenon in the chip position with defects under high-order resonance modes.

Figure 12. Experimental setup for air coupling excitation and laser vibration detection chip packaging quality

Figure 13. Resonance frequency vs. resonance frequency order curve (left), frequency shift distribution at different frequencies for single and multiple bond failure points (right)

The contact method sweep resonance is achieved by gently clamping the chip between two resonant ultrasound transducers and exciting one transducer with a sweep signal. When the sweep signal is at the same frequency as a certain resonant mode of the chip, the chip generates frequency resonance, and the receiving transducer will receive a stronger frequency peak. By scanning a frequency range, the resonant frequency spectrum of the chip within that frequency range can be obtained, with each frequency peak corresponding to a resonant mode of the chip. When there are defects inside the chip, the resonance frequency peak shifts relative to the resonance frequency peak of the intact chip. This phenomenon can be used to evaluate whether the chip quality is qualified, but it cannot determine the type, location, and size of defects.

Figure 14. Schematic diagram of contact sweep resonance experimental device and connection

Figure 15. Comparison of Peak Shifts of Resonance Frequencies for Different Defects

2. Thickness resonance analysis method

The thickness resonance method is to excite an ultrasonic signal on the chip and receive the ultrasonic signal passing through the chip through echo or transmission. When the wavelength of the ultrasound signal reflected back and forth inside the chip is an integer multiple of half the thickness of the chip, thickness resonance occurs. The thickness resonance frequency can be obtained by performing FFT on the collected ultrasound time-domain signal. When there is a crack in the internal wiring of the chip, the resonance thickness changes, and the thickness resonance frequency will also change accordingly.

Figure 16.2mm thick silicone rubber plate (sound speed 967 m/s) thickness resonance frequency analysis (left), C-scan imaging of resonance frequencies of samples with different thicknesses

3. Laser excitation photoacoustic reception detection method

The laser excitation photoacoustic reception detection chip eliminates the trouble of immersion in water compared to conventional water immersion ultrasonic microscopes. And the detection speed is relatively fast. Due to the limitation of the bandwidth of the photoacoustic receiving module, which can only reach a maximum of 2 MHz, the imaging resolution is relatively low compared to ultrasound microscopes, but from the detection results, it can still reach around 200 μ m.

Figure 17. Laser excitation photoacoustic receiving chip quality detection system and schematic diagram

Figure 18. On the left of the tested chip (red box), in the water immersion ultrasonic microscope detection result a, the laser photoacoustic detection result b is on the right

Summary:

Although there are many non-destructive testing techniques for chip packaging quality, there are currently few technical means that can achieve full field scanning and multi-layer bonding quality testing. Laser ultrasound, as a non-contact and coupling free non-destructive testing technology, has great potential in the chip inspection industry. Although this article demonstrates that the laser ultrasonic resonance time-domain amplitude analysis method can achieve fast, non-destructive, and coupling free detection of multi-layer bonding quality in chips. But its detection scanning method is limited to transient out of plane displacement with limited local field. Unable to accurately locate and assess the size equivalent of defects. Although laser ultrasound microscopy can achieve more accurate detection, there are still many technical challenges that need to be overcome.

reference:

Assessment of 2nd level interconnect quality in flip chip ball grid array (FCBGA) package using laser ultrasonic inspection technique

Detection of solder bump defects on a flip chip using vibration analysis, DOI 10.1007/s11465-012-0314-7

Enhanced non-contact ultrasonic testing using an air-coupled optical microphone

Air-Coupled and Resonant Pulse-Echo Ultrasonic Technique