Semiconductor Measurement Technology: THE DESTRUCTIVE BOND PULL TEST John Albers, Editor ABSTRACT This report summarizes the work done at NBS on the destructive bond pull test as applied to small-diameter (approximately 1 mil or 25 μm) ultrasonically bonded aluminum wire. This work was performed during the period from 1969 to 1974. The report begins with a brief summary of the calculation of the resolution-of-forces operative in the bond system during the application of the pulling force. Next, comparisons of the theoretical and experimental dependencies of the pull strength on the variables involved in the resolution-of-forces calculation are given. Some of the variables which are not directly involved in this calculation are then considered and their effects on the measured pull strength are presented. The report ends with a sensitivity calculation as to how well the variables must be controlled to maintain the variability of the pull strength within given limits. Bond pull specifications for large-diameter wire as well as recommended force levels to be used in the application of the nondestructive bond pull test, both of which have resulted from the pull test work, are considered in the appendices. Key Words: Bond angle; bonding; bond pull test; bond-to-bond spacing; large wire; loop height; microelectronics; non-destructive bond pull test; position of hook, pull rate; pull strength; resolution-of-forces; semiconductor devices; ultrasonic bonding; wire bond. 1. INTRODUCTION This report is an edited summary of the NBS effort on the destructive bond pull test. The rationale for this work lies in the widespread use of this test method in the electronics industry to evaluate the mechanical strength of wire bonds in semiconductor devices and in the large gap between the use of the pull test and the acquisition of reproducible calculable quantities which could be used to quantify the test results. Hence, at the request of several agencies, and with the guidance of standards organizations, an extensive survey of test methods for wire bonds in general [1] and an evaluation of the pull test in particular were begun. Initial in-house efforts to produce reproducible wire bonds to be used in the evaluation showed that the currently practiced industrial procedures were inadequate. This led to a study of ultrasonic wire bonding techniques. The results of this study pointed to the lack of shock and vibration isolation and the subsequent bond variability due to environmental and operator induced stresses as being primarily responsible for this inadequacy. An offshoot of this investigation was a better understanding of ultrasonic bonding through the use of magnetic pickup, capacitor microphone, and low-power laser detectors of tool motion during the bonding process. Optimized bond quality through bonding schedule stud ies resulted from this study. An exposition of a mechanism for the formation of ultrasonic bonds also followed from this study. A detailed treatment of these facets of the NBS work appears elsewhere [2]. Once bonder and related problems were solved, work was begun to quantify the results of the pull test. In order to carry out this program, it was necessary to correlate the measured pull strength as determined in the pull test with the stress in the wire, which depends on the geometry of the bond system. This was done through the resolution-of-forces calculation. An extensive series of experiments was undertaken in order to relate observed pull strengths to the results of the resolution-of-forces calculation. To further enhance the understanding of the pull test, certain variables which could not be introduced in the two-dimensional resolution-of-forces calculation were experimentally studied to ascertain their effects on the pull test results. This report is concerned with the results and conclusions of this investigation. The results of the pull test work have been used by ASTM Committee F-1 on Electronics in the preparation of a document on the pull test method. At the time of this writing, the document is well on its way to becoming an accepted standard. To support this evolution, a round robin is presently being run using test vehicles prepared at NBS. On each specimen, three groups of bonds of different deformation were prepared. One-half of the bonds of each group were pulled to destruction at NBS and the remaining half were pulled at the individual laboratories participating in the round robin. The round robin is not complete; however, preliminary results indicate that the mean pull strengths as determined by NBS and by each of the other participating laboratories differ by about ten percent. It should be noted that these conclusions are preliminary and are based upon raw data. A more complete treatment of the round robin and the data analysis will appear in the future when the round robin is complete. The better understanding of the pull test has led to other developments. These include a bond pull specification for large diameter wire as well as statistical and metallurgical rationales for a nondestructive wire bond pull test. These are discussed in appendices C and D respectively. The work on the nondestructive pull test is also being used by ASTM Committee F-1 on Electronics in the development of a recommended practice for carrying out this type of test. The following section contains a brief summary of a calculation of the resolution of the forces operative in the bond system during the application of the pulling force. A more complete derivation is presented in Appendix A and programs (in several different languages) which may be used for numerical calculation of the resolution-of-forces equations are given in Appendix B. The discussion of the resolution-of-forces calculation is followed in Section 3 by a description of the fabrication of the specimens used for the single-level and two-level bond studies. In Section 4, comparisons of the theoretical and experimental dependencies of the pull strength on the variables which are involved in the resolution-of-forces calculation are given. Of particular importance are the loop height, the position of the hook, and the angle between the direction of the pulling force and the normal to the line joining the bond terminals. In Section 5, consideration is given to some of the variables which are not directly involved in the resolution of forces calculation. The variables considered are the rate of pull and the angle between the direction of pull and the normal to the substrate in the plane perpendicular to both the substrate and the plane of the bond loop. Section 6 contains a sensitivity computation from which it can be estimated how well the variables must be controlled in order to maintain the accuracy of the results to within a given interval. A summary of the results comprises the final portion of the report. 1.1 UNITS The American semiconductor industry has traditionally used mixed English and metric units, but presently there is a trend in the direction of the International System (SI) units. For the purposes of conversion, it should be noted that 1 mil = 0.001 in. = 25.4 μm and that 1 gram force (gf) = 9.8 millinewtons (mN). 2. RESOLUTION-OF-FORCES CALCULATION The destructive bond pull test, more commonly called simply the pull test, is one of the most widely used test methods employed in the evaluation of the quality of wire bond systems. In its elemental form, the test consists of pulling the wire span between the bond on the terminal and the bond on the semiconductor die with a hook assembly until rupture takes place. The pulling force at which rupture occurs is usually referred to as the pull strength of the bond. The pull force at rupture is most frequently presented in units of grams force. An analysis of the resolution-of-forces provides relationships between the pull strength and the forces in the bond system at rupture. In figure 1, the geometrical variables are presented for a typical two-level bond. The angle is that between the direction of the applied force F and the normal to the substrate. At the bond on the semiconductor die, the angle between the wire and the surface of the die is denoted by ¿• At the terminal, the angle between the wire and the plane of the terminal surface is denoted by 0 The forces in the wire at the terminal and die are designated by F and F respectively. These forces are related to the applied pull force as follows: t' wt wd' For the simple case of a single-level bond pulled normal to the substrate and at the middle of the wire span, 4-0 and 0-0, In this instance, the above equations reduce to may be related to the variables h, H, d, and ɛ which are more easily accessible to measurement. It may be seen from figure 1 that h is the height of the wire span (at the point of pull) above the terminal contact surface. The vertical distance between the terminal contact surface and the semiconductor die surface is denoted by H. The horizontal distance between the bond on the terminal and the bond on the die is denoted by d and is usually referred to as the bond-to-bond spacing. Finally, ε, where Osc≤1, is the dimensionless fractional horizontal distance between the terminal bond and the point of application of the pulling force. In terms of these distance variables, F and F are related to F for wd wt If the bond is pulled normal to the substrate (=0), eqs (4) and (5) reduce to For a single-level bond which is pulled at mid-span, in a direction normal to the substrate, both and H are zero and ɛ is equal to one-half. Under these conditions, eqs (4) and (5) reduce to It should be noted that eqs (1) through (8) may be used for calculating pull strengths only when heel breakage (tensile failure) is observed. For the convenience of readers interested in numerical computations, Appendix B contains programs in several different languages which make use of eqs (4) and (5) to calculate the pull strength. For purposes of graphical illustration of the predictions of eqs (1) through (8), plots of the force in the wire on the terminal or die divided by the pulling force as functions of the angular variables or the length variables are presented in figures 2 to 4. Also, table 1 contains typical results from the resolution-offorces calculation for single-level and two-level bonds [3]. |