Flattening process of silicon wafers.

Dobrescu, Tiberiu ; Dorin, Alexandru

1. INTRODUCTION

Most integrated circuit (IC) chips are built on single crystal silicon wafers. Wafer diameter has increased steadily from less than 50 mm in the 1970s to 200 mm today and possibly to 300 mm in the near future.

Manufacturing of silicon wafers starts with growth of silicon ingots. A sequence of processes is needed to turn an ingot into wafers. This typically consists of following processes (Dobrescu, 1996):

* Slicing, silicon ingot into wafers of this disk shape;

* Edge profiling (chamfering), to chamfer the peripheral edge portion of the wafer;

* Flattening (lapping or grinding), to achieve a high degree of parallelism and flatness of the wafer;

* Etching, to chemically remove the damage induced by slicing and flattering without introducing further mechanical damage;

* Rough polishing, to obtain a mirror surface on the wafer;

* Fine polishing, to obtain final mirror surface;

* Cleaning, to remove the polishing agent or dust particles from the wafer surface.

The lapping operation is shown in figure 1. A batch of wafers (for example, 20 pieces) is manually loaded into a lapping machine. The loaded wafers are then lapped by the abrasive slurry (typically a mixture of alumina and glycerin) injected between two lapping plates rotating in opposite directions (Dorin. & Dobrescu, 2007).

The lapping operation would generate sub surface damages in silicon wafers, which need to be removed by its sub sequent processes.

Doble-side lapping is an established process in modern wafer production.

A common feature of all lapping processes is that the material removal takes place by the action of loase abrasive grains, which are dispersed, in a watery or oily lapping fluid. The workpiece shape is produced by the relative motion between the lapping wheel pieces surfaces coupled with the action of the loose lapping compound located between these two active surfaces.

The engagement of the lapping grains takes place non-directionally and stochastically so that directional abrasion grooves are generally avoided. A lusterless, homogeneous surface result.

[FIGURE 1 OMITTED]

For large volume wafer production, mestly high powered special purpose doublesided lapping machines with planetary drive systems are use. The primary features of these modern lapping machines are (Dobrescu, 1998):

* Programme and gauging controls which allow pressure-controlled operation;

* The infinitely variable speed adjustment of the drive motors (soft-start control);

* The application of grooved, temperature-controlled lapping wheels.

The necessity to gradually increase the lapping pressure in parallel with ensuring a soft-start control may be attributed to the extreme sensitivity of the wafers to brittle fracture.

Single-sided surface grinding operations can be applied to removal diffusion layers or to reduce the thickness of completely processed semiconductor wafers (thinning). The wafer thickness is reduced from 450-600 [micro]m down to approximately 150-380 [micro]m. After the thinning process, the parallelism may not exceed a total thickness value of 5 [micro]m, in certain cases of 3 [micro]m. In pure thinning surface roughness's [R.sub.max] of greater than 2 [micro][m are generally not permitted. The advantages of surface grinding compared to lapping are the high material removal rates and the lower level of crystal damage. In addition, grinding is a cleaner process and has good potential for automatisation in which fully integrated cassette-to-cassette manufacturing can be achieved.

2. COMPARE BETWEEN FACE TANGENTIAL GRINDING AND FACE PLUGE GRINDING

All specially developed grinding machines for semiconductor manufacture have vertically oriented diamond cup wheels (Figure 2).

Grinding machines differ in process kinematics and the number of simultaneously machinable wafers, which are fixed on vacuum chucks. Depending on the depth of cut [a.sub.p] and the table feed speed [v.sub.ft] face tangential grinding processes can be further classified into "reciprocation grinding" and "creep feed grinding".

[FIGURE 2 OMITTED]

In the first case, the material removal takes place in several cuts each with low depths of cut (1-10 [micro]m) and a high table feed speed (up to 8000 mm/min) whereas creep feed grinding is characterized by high depths of cut (100-200 [micro]m) and low table feed speeds (100-300 mm/min). In face tangential grinding with a rotary table it is favorable to have the process conditions: contact length, contact area and grinding forces remain constant since this directly affects the form accuracy attainable.

In face plunge grinding, the rotating compound grinding, wheel, which covers the entire wafer surface, is fed continuously in axial direction (Figure 2). Shortly before reaching the final thickness, the axial feed is disengaged. Subsequently a tangential finishing process takes place in which the fine-grained diamond outer ring is engaged as the wafer leaves the grinding zone. In face pluge grinding, the engagement conditions concerning cutting speed [v.sub.c] and cutting length [1.sub.ck] vary within the wafer surface and the grinding layer. Furthermore, the grinding forces in pluge rough grinding are approximately an order of magnitude higher than those for tangential grinding due to the large wheel / wafer contact area [A.sub.k] (Li et al., 2006)

Therefore, the final tangential finishing process diminishes form inaccuracies, which result from deflections of the machine tool. In this final operational step, very smooth surfaces can be obtained by using a fine grained outer ring. In tangential grinding, multi-step process as described above can only be realized on a multi-spindle machine tool system.

Face plunge grinding results in smaller depths of damage than that for tangential grinding. The normal component [F".sub.n] ranges between 2 and 36 N/[cm.sup.2] depending on the "process kinematics and the material removal rate. The highest values of area-related grinding forces have been obtained in tangential reciprocation grinding. The influence of the feed speed on the depth of damage and the forces per unit area is higher than that of the depth of cut for the same material removal rate (Sun et al., 2004).

Even at material removal rates of up to 1000 [micro]m /min, considerably low depths of damage were found with face plunge grinding. This can be explained by the large number of diamond grains, which are engaged simultaneously. This causes low level process-induced stresses and damage in the crystal material although the material removal rate is remarkably high. The material removal process is governed by micro-plastic abrasion. Brittle chipping and cracking at the wafer surface usually associated with coarse abrasion occurred less often in face plunge grinding. Face plunge grinding is thus a very efficient abrasion process for roughing. In tangential grinding at material removal rate per unit area =1000 [micro]m /min, more than 90 % of the wafers fractured during grinding or in subsequent processing for damage measurement. It was therefore not possible to evaluate depths of damage.

3. CONCLUSION

The following conclusions can be drowning from this experiment:

* When the two grinding wheels rotated in different directions the surface on one side of the wafer was different from the other side. This was a limitation for simultaneous double side grinding if the identical wafer surface on both sides were required;

* Lower wheel speed and higher feed rate produce rougher surface;

* In silicon wafer manufacturing, the removal amount of the subsequent polishing process to be large enough to eliminate all grinding marks generated in the simultaneous double side grinding or single side grinding operation. Further reduction of polishing amount necessitates optimization of the grinding process so that the grinding marks can be eliminated with minimum amount of polishing;

* The topography of the abrasives, in particular the larger ones, is reflected on the wafer surface in the form of arc-formed grinding grooves. Thus, a final tangential grinding process as a finishing operation is indispensable for high form accuracies and surface qualities.

4. REFERENCES

Dobrescu, T. (1996), Surface Grinding on a Rotary Table of Silicon Wafers, Research Reports, LAPT, University of Naples "FedericoII", pp. 35-38, Italy

Dobrescu, T. (1998), Cercetari privind optimizarea masinilor de superfinisat materiale fragile, PhD Thesess, Universitatea "Politehnica" din Bucuresti

Dorin, A & Dobrescu, T. (2007). The Manufacturing Process Flow for Silicon Wafers. Proceedings of the 5th International Conference of Advanced Manufacturing Technologies, AGIR (Ed), pp. 29-34, Sibiu, july 2007, Academy of Technical Sciences of Romania

Li. Z., Pei, Z. & Fisher, G. (2006). Simultaneous double side grinding of silicon wafers: a literature review. International Journal of Machine Tools & Manufacturing, no. 46, pp. 1449-1458

Sun, W., Pei, Z. & Fisher, G. (2004). Fine grinding of silicon wafers: a mathematical model for the wafer shape. International Journal of Machine Tools & Manufacturing, no. 44, pp. 707-716