A. Test Structures
The element described above has been implemented in a 0.25um CMOS process with a poly thickness of about 0.2um [3]. Ti-silicide films resulting in sheet resistance ranging from 3 to 4 ohms per square have been studied. Initial and post program electrical characteristics of a variety of element designs have been investigated. This includes the effects of poly doping (n, p, undoped), fuse length and width, fuse shape, programming and sensing voltage and current, and programming time.

B. Programming dynamics
In order to program an element, a certain amount of current is needed. The voltage needed for injecting this current must obviously be smaller than the available power supply voltage. Under constant voltage stress, as the element gets hot enough, agglomeration starts to occur, thereby, increasing the element resistance. As a result, the current through the element drops to a low value consistent with the elements final resistance and the element cools down. This mechanism is one with negative feedback. Therefore, a given fuse may be stressed only once and it's post program resistance will not increase with additional voltage stress.

Figure 3 shows the I-V characteristics of a typical fuse element. As the voltage is increased, current increases in a nonlinear fashion due to resistance change caused by self heating. When the dissipated power reaches a critical value, fusing occurs and element goes to a much higher resistance.

C. Response parameters
Initial and post program resistance of the element are the two key parameters affecting any circuit meant to sense the state of the element. A maximum value of initial resistance and a minimum value of post program resistance are needed to guarantee proper circuit function (about 100W and 1kW respectively in our circuit).

Initial fuse resistance depends on element geometry and silicide thickness and quality. Silicide quality in turn depends on process conditions, poly line width, and doping [2,4]. Silicide imperfections are more likely for long narrow elements and best silicide lines were found to be the ones made from p-doped poly. Imperfections in the silicide layer (cracks, high resistance Ti-Si phase) result in a resistive element Figure 4 shows cumulative distribution of the resistance of a typical fuse structure made with two processes with different thermal cycles and Ti thickness. A high resistance tail corresponding to silicide imperfections is evident in the distribution of the resistance of the unoptimized process.

Post program resistance varies greatly from device to device and depends on the shape and size of the discontinuity in the link. Due to this variation, any aspect of this resistance must be studied statistically. Many factors affect the level of fusing and therefore, post program resistance. They include:

Programming voltage, current, and time: Even though fusing can occur quickly and at fairly low currents and voltages (in the order of 1V, 8mA, 1mS), post program resistance is significantly enhanced if more energy is dumped into the element Therefore, increased voltage and current levels are needed for a longer time to guarantee a sufficiently large resistance. In this work, minimum programming conditions which resulted in statistically adequate post program resistance were a current of 20mA injected for 100ms with a voltage compliance of 2.5V.

Initial fuse integrity: Measured data shows that fuses that are more robust initially (by process or geometry) result in more successfully programmed elements. This is due to the fact that for a given voltage compliance and a given value of fusing current, a smaller resistance results in a larger amount of energy transferred to the device. The fortuitous result is that process conditions which result in good silicide formation and robust unprogrammed fuses also produce elements which program successfully.

Fuse shape: In addition to the relation between fuse size and it's initial resistance, the shape of the fuse has a marked effect on the distribution of its post program resistance. This has been found to be due to the fact that in addition to the high temperature necessary for agglomeration, the level of temperature gradient (and therefore stress) in the element plays a key role in the fusing event. Fusing has been found to occur near the sides of the element close to the point which has the highest temperature gradient (see Figure 2,5). Additionally, line width plays a significant role in fusing success with narrower lines having the advantage of better fusing. Figure 6 shows four different fuse shapes of the same length. Figures 7,8 show the distribution of initial and post program resistance for these elements. The difference between post program resistance of elements a,b corresponds to the effect of element width while differences between structures c,d show the effect of temperature gradient.

 

    Page 4 of 9