Open Access
Add to BibSonomy
Abstract
The flip-chip bonding technique utilized in ultra-large array infrared detectors has a substantial impact on connectivity rates. The electrical connectivity of the flip-chip bonding process exhibits randomness due to the difficulties in the surface control of large-scale devices. This restriction hinders the development of ultra-large array devices. In this work, the surface shape matching calculation is performed based on the surface shape distributions measured from infrared detector chips and readout circuits. The multi combinations and multi rotation angles are employed to calculate the distribution of combined surface distances, and the combined PV (peak-to-valley) value is applied to describe the severity of surface mismatch. Test devices with combined PV values ranging from 7.460 µm to 4.265 µm are prepared and tested, and the connectivity rate achieves an improvement from 74.57% to 99.75% between mismatched devices and matching devices. The electrical test results of test devices indicate that disconnections tend to cluster in areas where surface distance is over 5 µm, which is determined by extracting and analyzing the surface distance correlated to electrical test results. A standard based on the combined PV value is established to select matching combinations and ensure a high connectivity rate of 99% or 97% for infrared detectors, while the connectivity rates of randomly selected devices are no higher than 91%. This work presents a systematic method to predict and improve the connectivity rate of flip-chip bonding process for ultra-large array infrared detector.
© 2024 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Infrared focal plane (IRFPA) detectors are advancing towards ultra-high resolution. Focal plane specifications have developed from 320 × 256, 640 × 512 to 2k × 2 k, 4k × 4 k, and the pitch size is decreasing from 50 µm, 30 µm to 8 µm, 5 µm [1–6]. The preparation process for IRFPA detectors is facing new challenges due to the rapid expansion of large-scale devices and the continuous reduction in pitch size. IRFPA detector is composed of detector chip and readout circuit (ROIC) that are connected using flip-chip bonding technology, and the photoelectric signals generated at the detector chip are transmitted to the ROIC through the indium bump interconnection structure [7,8]. The high specifications of IRFPA detectors presents challenges in surface control, which results in unpredictable and incidental interconnection outcomes of the flip-chip bonding process and impedes the progress of ultra-large array infrared detector.
Prior to the flip-chip bonding process, it is essential to measure the surface shape of both the detector chip and the ROIC. The PV (peak to valley) value describes the surface warpage and is used to determine the feasibility of the flip-chip bonding process for each component. The reduced pitch size imposes significant constraints on the height and uniformity of the indium bump interconnection structure [3,9,10], which increases the demands on the device's surface flatness. The large size of the ultra-large array device, variations in processes, internal stresses, and contaminants make it difficult to measure and control the surface shape of the large-scale devices, including photon detectors [11–13], MEMS devices [14,15] and other electronic devices [16–20]. Negative pressure suction, multi-layer parameter optimization [21–24] and backside compensation layer [9,25] have been developed to suppress deformation. Based on conventional screening criteria, a considerable number of detector chips and ROICs have PV values that exceed the corresponding standards, rendering them unsuitable for further preparation processes.
This work investigates a systematic method for predicting the connectivity rate of flip-chip bonding process for ultra-large array devices. Based on the surface shape measurement results, the surface shape distribution is obtained by Zernike polynomial fitting [26–28], and the PVr value (“r” represents robust) [29,30] is calculated to describe the magnitude of surface warpage. The combined surface distance distribution of the device's flip-chip bonding process is subsequently computed, considering multiple combinations of detector chip-ROICs and four different rotation angles, and the combined PV value is calculated. Based on the matching calculation results, the connectivity rate of the test devices improves from 74.57% to 99.25% depending on the combined PV values of several test devices. A correlation is observed between the disconnections in test devices and the areas with surface distances over 5 µm. The difference between matching devices and mismatched devices has been applied to set a standard to reach high connectivity rate of 99% or 97% for ultra-large array infrared detector. This method can identify the suitable combinations and rotation angles for detector chips and ROICs with high PV values, which in turn enhances the connectivity rate of flip-chip bonding process and maximizes the device preparation yield.
2. Method
Controlling the surface shape of ultra-large array device presents a significant challenge. The PV value of the components typically maintains between 3 µm and 5 µm, which exceeds the prior standards for surface shape selection. Depending solely on the PV value is inadequate to ascertain its adequacy for the flip-chip bonding process for large-scale devices. The height of the indium bump interconnection structure is only slightly larger than the component’s PV value due to pitch size limitations, resulting in a low tolerance for flip-chip bonding process. To avoid “peak to peak, valley to valley” situations, the detector chips should be selected based on the surface shape to match the ROICs. The detector chip is designed with rotational symmetry, featuring equivalent structures after 0°, 90°, 180°, and 270° rotations. During the flip-chip bonding process, the detector chip can be interconnected with the ROIC at any of the four rotation angles, providing greater flexibility.
The principle of surface shape matching calculation is illustrated in Fig. 1. The Zernike polynomials are utilized to fit the surface shape measurement data obtained from the detector chip and ROIC. This helps eliminate chance noise and calculate the PVr value to precisely describe the amplitude of the surface warpage, and the fitted data is subsequently utilized as the surface shape distribution. The surface shape distribution of detector chip undergoes mirror and rotation transformations at rotation angles of 0°, 90°, 180°, and 270°, and is utilized to subtract the surface shape distribution of the ROIC, resulting in the combined surface distance distribution. The combined PV value of surface distance distribution is also calculated.
Fig. 1. Principle of surface shape matching calculation. (a) Detector chips and ROICs prepared in the same batch; (b) Detector chip flipped and inverted on ROIC; (c) Multiple bonding rotation angles.
The concept, surface distance, is introduced to refer to the distance between the detector chip and the ROIC during the initial stages of the flip-chip bonding process as shown in Fig. 1(b). The areas with a large surface distance are considered to be severely mismatched areas in the device, as the surface warpage of both sides accumulates in these areas.
The surface shape measurement data is normalized to the unit circle marked as $H({x,y} )$, where only the data within the detector chip or ROIC is effective in the calculation. To account for slight differences in the size of the manually selected testing area during measurement, the different measurement results must be processed into a matrix of the same order before they can be used in the combination calculation.
The PV value is defined as follow:
The 37th order Zernike polynomial is defined as follows [26]:
The surface shape measurement data can be expressed as a linear combination of Zernike polynomials:
Based on the orthogonality of Zernike polynomial, the coefficient ${a_j}$ can be calculated as:
The distribution of surface shape at the corresponding positions is then calculated:
The PVr value is defined as [29]:
where the $RMS({H({x,y} )- h({x,y} )} )$ is the root mean square of residual after fitting and removing the Zernike terms.The surface shape distribution of ROIC, ${h_{ROIC}}({x,y} )$, can be obtained through Eq. (7). The surface shape distribution of the detector chip requires mirror and rotation transformations:
where θ=0°, 90°, 180°and 270° are the four rotation angles of the detector chip relative to the ROIC.The combined surface distance is defined as:
and the PV value of combined surface distance is defined as follow:Since the ${h_{chip}}({x,y,\theta } )$ and the ${h_{ROIC}}({x,y} )$ have been fitted with the Zernike polynomial, the $P{V_{com}}$ can practically evaluate the amplitude of surface mismatching.
This calculation analyzes the distribution of surface distance between two components before the flip-chip bonding process for the IRFPA device by flipping and inverting the detector chip on the ROIC as shown in Fig. 1(b). When two components are completely incompatible, only a few protruding areas can be in contact with each other, with most of the area remaining at a large surface distance. In contrast, when two components are compatible, the majority of the area is relatively close in surface distance, making them easier to interconnect.
Infrared detectors require a high connectivity rate. Aggregated disconnections in any area can easily cause clustered blind pixels that exceed the index limit, rendering the devices unsuitable for infrared detection. When preparing multiple sets of detector chips and ROICs in the same batch, it is possible to quickly conduct surface shape matching calculations for multiple detector chip–ROIC combinations and various rotation angles. The PV value of these combinations can be reduced, consequently avoiding areas with large surface distances and decreasing the impact of surface mismatch on the effectiveness of flip-chip bonding. For matching detector chip and ROIC with combined PV lower than specified values, the connectivity rate can be predicted, allowing the device to be qualified for infrared detection applications.
3. Results and discussion
To ensure precise surface shape matching calculations and establish combined PV value standards, 6 test devices under identical flip-chip bonding conditions have been prepared to simulate the process for IRFPA devices. The test device consists of a simulating detector chip and a simulating ROIC that contain the connectivity test structures and share similar mechanical features with the infrared (IR) detector except for the IR detection capability. The IR detector is based on materials such as HgCdTe to achieve IR detection However, the high cost of preparation prevents the fully functional IR device from flip-chip bonding experiment. Therefore, the simulating detector chip and the simulating ROIC are designed to exclude the preparation of detection material but maintain the mechanical features related to the flip-chip bonding process. The mechanical features include the specification and pitch size of the interconnecting structure, as well as the size of the device and the selection of the substrate.
By extracting the indium bumps from the devices and connecting them to the test electrodes, the electrical connectivity and phenomena including disconnection, adhesion and misalignment can be directly measured and the connectivity rate can be calculated. During the interconnecting experiment, the disconnection is recognized as the major factor that affects the connectivity rate based on the current interconnecting structure and bonding conditions. Severe surface warpage of devices can cause large surface distance, resulting in disconnected test points in those areas. The adhesion typically occurs in areas with small surface distances only when additional heating is introduced into the current bonding conditions. The severe misalignment that causes most test points to be adhered, is usually due to the mechanical failure. The interconnecting structure fills into the gap of the opposite instead of connecting one-to-one, resulting in adhesions regardless of surface distance. Based on the intended bonding conditions for IR detectors, the test devices with varying surface distances suffers from disconnection primarily and the details are present below.
Figure 2 displays the surface shape distribution of the simulating detector chip and simulating ROIC of test device 2, along with the surface distance distributions for the four rotation angles. The orientation of the test device depends on the surface shape distribution of the ROIC, ${h_{ROIC}}({x,y} )$. The surface shape of the detector chip, ${h_{chip}}({x,y,\theta = 0} )$, is the processed data that has undergone mirror and rotation transformations. The PVr values are provided in Table 1 corresponding. The PVr values can indicate the current control capability of surface shape. When matched at 0° and 180°, the distribution of combined surface distance for test device 2 is significantly better than when matched at 90° and 270°. The latter leads to a significant mismatched area in the lower left corner of the test device, with a surface distance of 8 µm or more, resulting in a much higher probability of disconnection.
Table 1. PVr value, combined PV value and connectivity test results of 6 test devices
Table 1 presents the PVr values of surface shape measurements, combined PV values, rotation choices and connectivity test results correspondingly for 6 test devices, as the combined surface distance distribution at the chosen rotation angle and connectivity test results of the 6 test devices are displayed in Fig. 3.
Fig. 3. Combination surface distance distribution and connectivity test results of 6 test devices. (a) Test device 1, surface distance in all areas is below 4.5 µm, and there is only one disconnection in the test device; (b) Test device 2, surface distance in almost all areas is below 5 µm, and there appears several disconnections scattered in the test device; (c) Test device 3, surface distance in almost all areas except the lower left corner is below 5 µm, and several disconnections cluster at the lower left corner; (d) Test device 4, surface distance in almost all areas except the upper left corner is below 5 µm, and a series of disconnections cluster at the upper left corner; (e) Test device 5, there are several areas with surface distances greater than 5 µm, and the disconnections scatter over the large areas; (f) Test device 6, a large area in the center with surface distance greater than 5 µm, and the disconnections distribute across the entire area.
Prior to the flip-chip bonding procedure, the surface shape of each component from the six test devices is measured, and the PVr value for each component is calculated to be approximately 3 µm to 5 µm. The combined PV value is strongly affected by rotation angle, with the combined value for rotation angles of 0° and 180° being smaller than that for rotation angles of 90° and 270° for almost all test devices, with a difference up to 3 µm.
Matching simulating detector chips and simulating ROICs are selected for test device 1, 2 and 3, and most areas have a small surface distance. The test device 4, 5 and 6 are fabricated with mismatched components chosen randomly, resulting in a larger surface distance in the large area and a combined PV value greater than the former ones.
The test device 1 has the best combined PV value (4.265 µm) and connectivity rate (99.75%) among the 6 test devices. The simulating detector chip and simulating ROIC are selected for the most matching ones as well as an optimal rotation angle to fabricate test device 1. In result, there is only 1 disconnected test point in the device.
The combined PV value for test device 2 is 5.580 µm, with a connectivity rate of 97.22%. Despite the severe surface warpage of the simulating detector and simulating ROIC, the matching surfaces result in a tolerable combined PV value and a good connectivity rate. Disconnected test points are scattered in both area and surface distance.
In comparison, it is noteworthy that test device 3 has a combined PV value of 8.860 µm, which is significantly higher than that of test device 1 and 2. However, test device 3 achieves a connectivity rate of 95.69%. The surface deformation located in the lower left corner of test device 3 has resulted in a larger combined PV value for the entire device and the clustering of several disconnected test points in that specific area. Excluding the lower left corner, the combined PV value of test device 3 is 5.425 µm and the connectivity rate is 96.87%.
The test device 4 exhibits a similar situation to test device 3. However, the area in the upper left corner has a greater impact on the bonding process, resulting in a combined PV value of 6.932 µm and a connectivity rate of 90.99%. The majority of disconnected test points are clustered around the upper left corner.
The combined surface distance distribution of test device 5 has a PV value of 7.220 µm, which is lower than that of test device 3. However, there are several areas with greater surface distances, resulting in a higher percentage of disconnected test points scattering in those areas compared to the former ones. Consequently, the connectivity rate for test device 5 was only 88.51%.
The combined PV value of test device 6 is 7.460 µm, which is higher than all other test devices except for test device 3. The device has a large area in the middle with a significant surface distance. After the flip-chip bonding procedure, numerous disconnections accumulated in that area, and the disconnected test points were connected as a cohesive unit. The connectivity rate of test device 6 is only 74.57%.
Based on the combined surface distance distribution and connectivity test results for the six test devices shown in Fig. 3, disconnections tend to occur in areas with surface distances greater than 5 µm, while most test points at other areas remain connected. The 6 test devices exhibit varying degrees of surface mismatch, ranging from almost none, to a few at the corners, and finally to a large area in the center, and the connectivity rate decreases with disconnections appearing at the mismatched areas correspondingly.
Although one component of the test device 4, 5 and 6 has a flattened surface (with PVr values of about 3 µm), the other components with high PVr values still cause severe mismatch in the test devices. The test devices exhibit severe mismatches in the lower left corner of Fig. 3(c1) and the middle area of Fig. 3(f1), with surface distances exceeding 7 µm, resulting in most test points in those areas being disconnected.
To quantitatively evaluate the connectivity rate impacted by surface distance and combined PV value, the surface distance at connected or disconnected test points are extracted and analyzed in Fig. 4.
Fig. 4. Distribution of surface distances corresponding to test points of test devices. (a)-(f) Test device 1-6.
Figure 4 displays the surface distance distribution at the locations of connected and disconnected test points of 6 testing devices. The distribution of disconnected test points is scattered for test devices with matching surface and small combined PV values, and corresponds to random surface distances ranging from 0 µm to 5 µm. The connectivity rate of these areas can reach more than 95%, and all test points at areas with surface distances lower than 2 µm remain connected. However, test devices with mismatched surfaces and high combined PV values, as well as severely mismatched areas in devices with matching surfaces, exhibit a higher proportion of disconnected test points that show a clustered distribution in areas with large surface distances (5 µm∼8 µm). The highest connectivity rate in this area is only 85%, and the lowest is 0%.
The combined PV value is the maximum of the surface distance and represents the severity of surface mismatch. Surface mismatch can cause large surface distance and disconnections in specific areas, while test points in other areas remain connected, which is the key factor that affects connectivity rate and determines whether the device can be used for IR detection. To ensure a connectivity rate of over 99%, the combined PV value should not exceed the 4.2 µm as indicated by test device 1. To achieve a connectivity rate of about 97%, the combined value should be lower than 5.5 µm as indicated by test devices 2 and 3. However, for randomly matched devices such as test devices 4, 5, and 6, the combined PV is approximately 7 µm, and the connectivity rate is only up to 90%. The lower limit cannot be guaranteed depending on the extent of mismatch.
4. Conclusion
This paper presents a systematic theoretical analysis and experimental validation method of flip-chip bonding connectivity prediction for ultra-large array infrared detector. The combined surface distance at the initial stages of flip-chip bonding process is obtained by calculating multiple combinations and rotation angles to suppress surface mismatch, based on surface shape measurement results. Test devices are prepared using components with PVr values ranging from approximately 3 µm to 5 µm, resulting in combined PV value varying from 7.460 µm to 4.265 µm and connectivity rate improving from 74.57% to 99.75%. The correspondence between the surface shape distribution, combined PV value, and connectivity rate is determined through surface matching calculations and electrical tests. The technique allows for the establishment of standards for the flip-chip bonding process of IR detectors to achieve a connectivity rate of 97% with combined PV value of 5.5 µm or 99% with combined PV value of 4.2 µm.
This paper presents a technique for calculating surface shape matching and identifying correspondence relationships to address the needs of flip-chip bonding for ultra-large array devices. The method can predict and improve the connectivity rate for the flip-chip bonding process, making it highly valuable for the advancement of ultra-large array infrared detector.
Funding
Innovative Project of Shanghai Institute of Technical Physics, Chinese Academy of Sciences (CX-457, CX-456).
Disclosures
The authors declare no conflicts of interest.
Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.
References
1. A. Rogalski, “Scaling infrared detectors—status and outlook,” Rep. Prog. Phys. 85(12), 126501 (2022). [CrossRef]
2. Z. Ye, H. Li, J. Wang, et al., “Recent hotspots and innovative trends of infrared photon detectors,” J. Infrared Millim. Waves 41, 15–39 (2022). [CrossRef]
3. A. Rogalski, P. Martyniuk, and M. Kopytko, “Challenges of small-pixel infrared detectors: a review,” Rep. Prog. Phys. 79(4), 046501 (2016). [CrossRef]
4. A. Rogalski, J. Antoszewski, and L. Faraone, “Third-generation infrared photodetector arrays,” J. Appl. Phys. 105(9), 91101 (2009). [CrossRef]
5. J. Abergel, F. Rochette, S. Gout, et al., “Development of sub-10 µm pitch HgCdTe infrared detectors,” Proc. SPIE 1100222, 100 (2019). [CrossRef]
6. S. Bisotto, J. Abergel, B. Dupont, et al., “7.5 µm and 5 µm pitch IRFPA developments in MWIR at CEA-LETI,” Proc. SPIE 110021C, 101 (2019). [CrossRef]
7. J. John, L. Zimmermann, P. De Moor, et al., “High-density hybrid interconnect methodologies,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip. 531(1-2), 202–208 (2004). [CrossRef]
8. R. K. Bhan and V. Dhar, “Recent infrared detector technologies, applications, trends and development of HgCdTe based cooled infrared focal plane arrays and their characterization,” Opto-Electronics Rev. 27(2), 174–193 (2019). [CrossRef]
9. W. Wang, Y. Fan, J. Si, et al., “Influence of material morphology on fabrication of large format IRFPA,” Proc.SPIE 10624, 84191M (2012). [CrossRef]
10. M. Liu, C. Wang, and L. Zhou, “Development of small pixel HgCdTe infrared detectors,” Chinese Phys. B 28(3), 037804 (2019). [CrossRef]
11. B. Safia, V. Pierre, C. Nicolas, et al., “Addressing sub-micron thermal warpage: Industrial application on semiconductor devices,” in 2023 IEEE 73rd Electronic Components and Technology Conference (ECTC) (2023), pp. 169–174.
12. T. Yang, D. Cheng, and Y. Wang, “Aberration analysis for freeform surface terms overlay on general decentered and tilted optical surfaces,” Opt. Express 26(6), 7751–7770 (2018). [CrossRef]
13. A. N. Danilewsky, P. J. McNally, and B. K. Tanner, “X-ray imaging of silicon die within fully packaged semiconductor devices,” Powder Diffr. 36(2), 78–84 (2021). [CrossRef]
14. K. N. Tu, “Recent advances on electromigration in very-large-scale-integration of interconnects,” J. Appl. Phys. 94(9), 5451–5473 (2003). [CrossRef]
15. Y. Ku, P.-Y. Chang, and C. Shen, “Experimental investigation of three-dimensional interconnect processing wafers,” J. Micro/Nanolithography, MEMS, MOEMS 11(4), 043002-1 (2012). [CrossRef]
16. S. Xue, Y. Dai, S. Zeng, et al., “Interferometric stitching method for testing cylindrical surfaces with large apertures,” Opt. Express 29(13), 19767–19789 (2021). [CrossRef]
17. W. Li and D. Yu, “Experimental and numerical investigations of the warpage of fan-out wafer-level packaging with SAW filters,” IEEE Trans. Components, Packag. Manuf. Technol. 12(5), 869–877 (2022). [CrossRef]
18. P.-Y. Chang and Y.-S. Ku, “Wafer warpage characterization measurement with modified fringe reflection method,” Proc.SPIE 9525, 952528 (2015). [CrossRef]
19. T.-C. Chiu and E.-Y. Yeh, “Warpage simulation for the reconstituted wafer used in fan-out wafer level packaging,” Microelectron. Reliab. 80, 14–23 (2018). [CrossRef]
20. K. Peng, W. Xu, Z. Qin, et al., “Reflow warpage induced interconnect gaps between package/PCB and PoP top/bottom packages,” in 2017 IEEE 67th Electronic Components and Technology Conference (ECTC) (2017), pp. 1378–1383.
21. A. A. Mani, N. Caicedo, N. Miloud-Ali, et al., “Cu pillar based advanced packaging, for large area & fine pitch heterogeneous devices,” in 2020 IEEE 8th Electronics System-Integration Technology Conference (ESTC) (2020), pp. 1–6.
22. Y. Duan, G. Liu, W. Wang, et al., “Finite element modeling and analysis method for predicting and optimizing the warpage of construction before flip chip bonding in System-on-Wafer process flow,” Microelectron. Reliab. 151, 115260 (2023). [CrossRef]
23. P. Rajmane, K. Dhandapani, M. Schwarz, et al., “Investigation of the factors affecting the warpage prediction of multi-chip package,” in 2021 IEEE 71st Electronic Components and Technology Conference (ECTC) (2021), pp. 1514–1520.
24. H.J. Lee, S.M. Park, and S.J. Park, “Minimization of warpage for wafer level package using response surface method,” Int. J. Precis. Eng. Manuf. 17(9), 1201–1207 (2016). [CrossRef]
25. W. Zhang, X. Chen, and Z. Ye, “A study on the surface correction of large format infrared detectors,” Semicond. Sci. Technol. 35(12), 125007 (2020). [CrossRef]
26. J. Loomis, “A computer program for analysis of interferometric data,” in Reduction and Interpretation of Interferograms, J. Loomis and A. S. for T. and Materials, eds. (ASTM International, 1978), Vol. STP 666, pp. 71.
27. J. Y. Wang and D. E. Silva, “Wave-front interpretation with Zernike polynomials,” Appl. Opt. 19(9), 1510–1518 (1980). [CrossRef]
28. K. Niu and C. Tian, “Zernike polynomials and their applications,” J. Opt. 24(12), 123001 (2022). [CrossRef]
29. C. J. Evans, “PVr—a robust amplitude parameter for optical surface specification,” Opt. Eng. 48(4), 043605 (2009). [CrossRef]
30. C. J. Evans, “Robust estimation of pv for optical surface specification and testing,” in Frontiers in Optics 2008/Laser Science XXIV/Plasmonics and Metamaterials/Optical Fabrication and Testing, OSA Technical Digest (CD) (Optica Publishing Group, 2008), pp. OWA4.
31. E. P. Goodwin and J. C. Wyant, Field Guide to Interferometric Optical Testing (SPIE PRESS, 2006), pp. 48–49.
32. M. Born and E. Wolf, Principles of Optics (Pergamon, 1980), pp. 459–490.
