Advances in magnetic microscopy for stacked-die and package-level fault isolation
12/01/2004
Recent advances have enabled magnetic-field imaging to identify high-resistance defects in packaged ICs within a resolution of 20µm on the device. This is done in a noncontact, nondestructive approach for several packaging technologies, including stacked die. For die-level applications, submicron resolution of HR defects can be achieved. High-resolution current mapping is achieved with two types of magnetic microscopes. One is a scanning fiber/superconducting quantum interference device (SQUID) microscope, which uses a SQUID sensor coupled to a nanoscale ferromagnetic probe. The second employs a magnetoresistive sensor.
In recent years, current imaging through magnetic-field detection has become a mainstream approach for short localization in a packaged device [1]. This approach to fault isolation is also heavily utilized for die-level applications [2]. Current imaging is based on the measurement of the magnetic field generated by current carried in the structures of the device under test (DUT). The magnetic fields are typically measured with a SQUID, which is a very sensitive sensor.
A new approach to isolating high-resistance (HR) defects has been developed using SQUID-based current imaging. HR defects describe any damage in a current path that results in an increase in resistance of that path. Some of these physical defects include cracked traces, nonwet or cracked C4 solder bumps, and delaminated vias. Higher-resolution failure analysis techniques will be required as IC processes move to 90nm linewidths and below. Magnetic imaging technology can help serve this need by enabling designers and failure analysts to see current paths in the nanoscale regime.
>Two approaches have been developed for nanoscale imaging using magnetic fields. One approach couples a magnetic fiber to a SQUID. This fiber can be etched to a tip that is as small as 10nm and used as a probe to deliver high spatial resolution. The other approach involves a magnetoresistive (MR) sensor, which is intrinsically less sensitive to magnetic fields than the SQUID but is readily miniaturized to the nano-scale. If this sensor is brought within about 1µm of currents to be measured, it has the magnetic sensitivity to map out submicron current lines carrying <500µA of current.
The growth in handheld, portable systems in high-volume consumer electronics applications has moved the semiconductor industry toward complex packaging structures, including stacked-die assemblies with two, three, and four ICs in a package. The magnetic-field microscope can detect fields through all these packaging structures, including the stacked devices. Since there is no shielding by the packaging material, current levels can be imaged as low as hundreds of nanoamps.
How magnetic-field imaging works
Source currents in electronic devices can be calculated from magnetic-field images. By mapping the current in an IC, stacked die, or package, short-circuit defects can be localized and designs can be verified to determine if electricity is flowing where expected. HR defects can be found by identifying differences in the magnetic field. Unlike thermal, optical, ion, or electron-beam techniques, magnetic-field detection is not affected by materials in ICs or the package. Therefore, current imaging can be performed from both the front and backside of a device through many layers of metal, the die, or packaging materials. The only difficulty in this approach is that the strength of the magnetic field decreases with current magnitude and the increase in separation between the sensor and source currents. The rate of decrease depends upon the nature of the current source, but the magnetic-field strength for ICs is typically inversely proportional to the distance between the sensor and the source.
A magnetic-field imaging instrument, called the Magma C20, has been developed for failure analysis in ICs, stacked-die components, chip packages, and board assemblies using a high-temperature SQUID with a sensitivity of 20 picotesla (two million times lower than Earth’s magnetic field). The scanning SQUID microscope (SSM) has been designed to keep the SQUID cold and in vacuum while the DUT is at room temperature and in air. The design of the magnetic-field microscope also facilitates positioning of the SQUID as close as 70µm from the DUT. The system can run samples requiring high-resolution current images (at the die level and for wafers) as well as samples requiring high sensitivity (low current and HR). The instrument’s sensitivity is high enough to detect currents as small as 10nA at a 100µm working distance with 1-sec averaging but low enough to enable the instrument to function in an unshielded environment.
With the use of a high-resolution sensor, the Magma C20 is capable of resolving 0.3µm (300nm) features with a potential to reach 0.01µm (10nm) resolution. The sensor is held stationary while the DUT is raster-scanned under the magnetic sensor to acquire the magnetic-field image. The current supplied to the DUT typically alternates at a frequency <100kHz. A lock-in technique enables the instrument to capture only the image of the applied current, while ignoring magnetic fields generated by currents at other frequencies or static background fields.
High-resistance experimental results
Typically, an HR defect is the result of a geometrical change in a circuit element, such as a delamination, crack, or void. Clearly, the current distribution will be affected by such geometric alterations and correspondingly affect the magnetic-field distribution as illustrated in Fig. 1 for a failing C4 bump.
In this situation, one expects to see a small change in the magnetic field distribution around the defect as compared with that in a good part. Typically, this would be in the range of picotesla to a few nanotesla. A detailed image comparison between the good and failing parts would then be able to detect this difference and subsequently locate the defect region.
 Figure 1. Illustration of current distribution in a good and a delaminated (failing) C4 bump. |
The localization of HR defects through current imaging is accomplished with a detailed comparison of good and failing parts [3]. The differences in a magnetic field explained earlier are small and therefore require a very careful analysis between the good and failing parts. This requires improvements over conventional technology in two areas. First, the instrumentation for current imaging requires more precise automated control of the sample setup and data acquisition. The scan conditions must be as similar as possible between the good and failing parts, so that an effective comparison can be made. Second, even with careful sample setup and data acquisition, there will still be misalignments between the two images and potential signal differences due to different working distances, or even part deformations, for example, warping. These differences need to be sorted out from those due to the HR defect. For this, advances have been made in image difference analysis (IDA) to assist in the identification of failing defects. These defects typically show up in IDA images as dipoles, where the defect location coincides with the centroid of the dipole.
C4 bump failure
One example of a real defect tested with magnetic-field microscopy is a C4 bump failure [3]. Figure 2 shows the region of interest from the IDA image overlaid on the CAD layout for the device.
The 2D density plot shown in Fig. 2 does not adequately present the relative intensity of the magnetic anomaly. A better way of doing that is by using a 3D representation as shown in the lower left corner of Fig. 2, where the z axis is the magnetic-field intensity corresponding to the 2D IDA result. For the sake of clarity, we have zoomed in on the area around the anomaly and plotted the absolute value of the magnetic field. The two dipole peaks associated with the defect are clearly visible. The defect is located at the center between the two peaks.
In the 2D image, the metal trace connecting to the failing bump is marked with the dashed yellow line. The C4 bump is connected at the end of it before the green grid. The centroid for the magnetic anomaly is the black dot, aligning very well with the position of the bump. The defect was located with the SQUID microscope to within 30µm. Although not shown here, the best localization to date is 20µm.
High-resolution experimental results
The fiber/SQUID microscope uses a ferromagnetic metal probe 100µm thick and about 10mm long. The probe is electrochemically etched to a tip with submicron diameter [4]. This probe resides entirely in air, and its end is attached to a window at the bottom of a cone housing a SQUID. The SQUID sensor is isolated at around 77K in vacuum and the distance between the probe and SQUID can be adjusted to maximize the magnetic response of the microscope. An optical shear-force feedback method is used to scan the etched tip over the sample in near contact (separation distance of <100nm). Use of this feedback method means that topography data can be acquired on the sample at the same time as the magnetic data.
For the MR microscope, the sensor has been attached to a cantilever, which allows the sensor to be scanned over the sample in soft contact. The cantilever is attached to the SQUID microscope housing by an easily mounted bracket on a piezo stage. In this way, the sensitive SQUID can be used for large-area scans to locate regions of interest, and then the MR sensor can be used to perform high-resolution scans in these regions.
The fiber/SQUID microscope has been used to image a serpentine structure of micron-scale dimensions. This proof-of-principle scan has demonstrated the feasibility of the approach, but more work is necessary to demonstrate submicron capability. The approach is discussed here because of the potential for this technique to achieve 10nm resolution and to have a form factor capable of scanning in very small cavities.
 Figure 3. Optical image of the scanned region of a die with 300nm-wide lines, each separated by 300nm. The yellow line current image is overlaid on the optical image. |
The MR microscope has been fully developed for submicron imaging. The scan shown in Fig. 3 was done using a commercial magnetic-field microscope with a MR high-resolution current scan option. The image shows a die-level test structure with 300nm-wide lines 300nm apart. A current of 500µA is clearly resolved in this image and overlaid on the optical image of the part.
3D view of the current
 Figure 4. Same current as displayed in Fig. 3 in a 3D view. The currents are flowing in the same layer. |
The magnetic sensor detects a stronger field when the current path is closer to the sensor. This means that the different metal layers can be displayed in a 3D view as long as the current-carrying metal linewidths are the same for each layer and that no two metal lines from the different layers are directly on top of each other. Figure 4 shows the current image displayed in Fig. 3 in a 3D view. It is clear from the image that all currents in this chip are in the same layer.
The 3D view is beneficial when currents are in multiple levels of metal or different die in a stacked-die structure. The failure analyst can see the different layers in the chip or stacked-die structure and its related physical coordinates.
Conclusion
Magnetic image difference analysis is a new noncontact, nondestructive technique that has been demonstrated to localize high-resistance defects in package substrates and interconnections. It has been shown to localize these defects to within 30µm, which is an order-of-magnitude improvement over time domain reflectometry and without destruction of the sample.
The scans with a MR microscope demonstrate that current lines separated by 300nm can be resolved. The flux-guide scanning SQUID microscope and the MR microscope have excellent potential as tools for locating die-level defects and could play a complementary role to standard SSM with a bare SQUID. The standard SSM has unsurpassed current sensitivity and the ability to examine circuits buried hundreds of microns but is limited to applications where spatial features are larger than a micron and for global imaging of current densities. The HR microscopes described here can detect submicron anomalies in magnetic fields created by current defects. These microscopes are ideal for wafer-level studies or examination of de-processed samples.
Acknowledgments
The authors would like to thank Wang Zhiyong at Intel Corp. for providing samples for resistive-opens defect research and Dave Vallett at IBM Corp. for providing samples for high-resolution scanning.
References
- R. Dias, L. Skoglund, Z. Wang, D. Smith, “Integration of SQUID Microscopy into FA Flow,” ISTFA 2001.
- D. Vallett, “Scanning SQUID Microscopy for Die-Level Fault Isolation,” ISTFA 2002.
- A. Orozco, “Fault Isolation of High Resistance Defects using Comparative Magnetic Field Imaging,” ISTFA 2003.
- S.I. Woods, “High Resolution Current Imaging by Direct Magnetic Field Sensing,” ISTFA 2003.
J.O. Gaudestad received his MS in physics from the U. of Maryland and his MS equivalent (Sivil Ingeniør) from the physics department at the Norwegian U. of Science and Technology. He is an applications engineer focused on development of scanning SQUID microscopy for fault isolation in ICs and IC packages at Neocera Inc., 10000 Virginia Manor Rd., Beltsville, MD 20705; ph 240/441-7410, e-mail [email protected].
Solomon I. Woods received his AB in physics and mathematics from Harvard U. and his PhD in physics from the U. of California at San Diego. He is a scientist working on R&D in Neocera’s Magnetic Microscope Division.
Antonio Orozco received his MSc in theoretical physics from the Universidad Autonoma of Madrid, Spain, and his PhD in condensed matter physics from the Universidad de Cadiz, Spain. He is a scientist in the R&D Department of Neocera’s Magma Advanced Technology unit.
L.A. (Lee) Knauss received his PhD in physics from Lehigh U. He is director of advanced technology for Neocera’s scanning SQUID microscopes.