Novel technology for handling very thin wafers
10/01/2003
Overview
Increasing use of very thin, flexible wafers for IC manufacturing brings with it extremely challenging automated handling problems. Individually, vacuum- or Bernoulli-based end-effectors cannot solve all problems. But the combination of these two techniques in one end-effector design provides effective automated handling for several emerging applications.
In today's wafer-processing applications requiring wafer thinning, 150mm wafers may be thinned to a 50µm limit, 200mm wafers to 100µm. At these thicknesses, wafers are flexible, fragile objects. If a wafer bows from a backside metal, standard vacuum end-effectors or electrostatic handling fail, making routine handling a significant challenge.
 Figure 1. End-effector that combines vacuum and Bernoulli wafer "gripping." |
null
The instructions to "never handle a thin wafer without tape" cannot always be followed. First, no tapes are available that withstand high-temperature backend processes. Furthermore, carriers or protection tapes are not applicable before test because a wafer has to be uncovered on both the front- and backsides for electrical contact and visual inspection.
With the forecast growing demand for thin silicon wafers, it is often predicted that conventional wafer handling will not be able to achieved targeted wafer breakage rates [1]. Currently, many thin-wafer-handling operations are carried out manually due to lack of automated handling systems; this explains high breakage rates.
With these challenges in mind, we set out to develop a novel, thin-wafer gripping method with sensor-controlled robotics that will aid fully automated thin-wafer production.
Principle of operation
Standard vacuum grippers have some major disadvantages with thin wafers. First, a wafer bends on both sides of the gripper. This can cause problems during wafer insertion into a carrier because the wafer edge is razor-sharp and cuts into the carrier material. Second, if the wafer bow exceeds the suction capacity of the vacuum pads, such grippers fail to function properly.
Pure Bernoulli-type chucks have problems with convex bows when gripping the wafer frontside and centering pins that are needed to prevent the wafer from drifting sideward, which can impede the handling task.
Our new gripping method overcomes disadvantages associated with just vacuum or just Bernoulli wafer handling by integrating both into an end-effector (Fig. 1). Vacuum action is realized through a set of small diameter holes around the perimeter of the gripper surface. Bernoulli action (Bernoulli effect [2, 3]) comes from a number of angular nozzles near the center of the gripper. The gripper is a "sandwich" of aluminum parts; its surface area is significantly larger than a standard vacuum gripper, but it is still smaller than a wafer so that wafers can be inserted into carriers.
The thin-wafer-gripping process of this new end-effector involves three consecutive phases:
- Vacuum is switched on in the first 100 msec.
- A short — 150–300 msec — pulse of nitrogen is applied through the Bernoulli nozzles, pulling the wafer onto the gripper surface.
- The wafer is fixed to the gripper by the suction force of vacuum pads.
In our test, we have shown that the combination gripper handles thin wafers essentially flat because it is capable of flattening wafer bow up to 10mm. In addition, via the directed Bernoulli nozzle, it is possible to align wafers on the gripper; during the Bernoulli pulse, the wafer is gently "blown" toward the rear border of the batch, thus pre-aligning it to a certain extent. And, the vacuum pads exert minimal stress on thin wafers. We have not measured the stress, but we have optimized the diameter of the vacuum holes. It is obvious that small bores produce less stress than bigger ones. The total thickness of the gripper is 2mm, allowing the handling of wafers in a 25-slot batch.
Gripper characterization
For our characterizations of the new gripper and its effect on wafer bow and warpage, we needed to measure the force of the Bernoulli effect because the stronger the effect, the better the wafer grip. But we also needed an idea of the tensions placed on the wafer.
We did our measurements on a special test apparatus (Fig. 2), where the gripper is mounted topside down on a linear slide and acts on a flat circular plate mounted to the top of a load cell measuring the force (Fig. 2a); in this way, force-distance data can be obtained. The warping force measurement of wafers is done on the same device in a different configuration (Fig. 2b), where a piston is pressed onto the wafer until it is in flat contact with the load cell.
 Figure 2. Measurement setup for Bernoulli effect and warping force measurement. |
null
Data gathered with the test apparatus give a full characterization of the Bernoulli-vacuum gripper and warping force measurements (Fig. 3). The grip force qualitatively obeys a square function, except for "buckling" in the vicinity of the plane. This opposing force to the Bernoulli effect, starts approximately -1mm from the plate. The Bernoulli effect is sustained by streamlines parallel to the plate. At distances close to the plate, the normal component of the streamlines impedes Bernoulli force. The floating distance of a wafer is ~100µm, where the attracting and repulsing forces are in an equilibrium.
 Figure 3. Characterization of Bernoulli effect and wafer-warping force. |
null
Our data clearly show that the Bernoulli effect can overcome the warping force of the wafers we tested (Fig. 3). This is a prerequisite for the combined vacuum-Bernoulli grip method. Otherwise, the wafer would not be sufficiently flat to grip. (The discontinuities in the wafer curves are due to switches of the warpage orientation of the wafer during the measurement procedure.)
Application
Our analyses have shown that applications for the Bernoulli-vacuum principle are plentiful. We have successfully completed feasibility studies for:
- the loading and unloading of thin wafers in carriers for metal-coating processes;
- a thin-wafer flat-aligner station; and
- various equipment retrofits.
The new gripper has also been applied to the problem of automated loading and unloading of quartz carriers. In contrast to standard wafer transfer equipment, the transport of a whole batch of thin wafers is not applicable for wafer thicknesses <200µm. The most advantageous method is a single-wafer transfer by means of sensor-controlled robotics [4].
Figure 4 shows the optical paths of a custom design fiber optical sensor head. Here, the objective is to detect slots on 150mm and 200mm quartz boats, the presence of wafers, the position of the wafer edge, and the bow of inserted wafers.
 Figure 4. Design and optical paths of a quartz boat sensor head. |
null
Before wafers are inserted, the sensor head scans the whole quartz boat. For the loading procedure, the end-effector moves to a position vertical to the slot. Then, the wafer edge position is measured and, if necessary, aligned to the center position of the slot by the robot. This measurement procedure is especially necessary for safe and exact insertion of thin wafers with fragile wafer edges.
When there are heavily bowed wafers, the sensor head detects a slot overlap and communicates with the handling robot to omit one or more slots for the insertion of the next wafer, avoiding wafers touching.
For unloading, after slot and wafer scans, the end-effector aligns to the wafer and the Bernoulli effect is applied to pull the wafer toward the end-effector. Subsequently, vacuum pads fix the wafer on the end-effector and it can be vertically moved out of the slots.
Our feasibility study with 150mm wafer thickness >70µm and 200mm wafer thickness >100µm resulted in loading times of 18 sec/wafer and unloading times of 10 sec/wafer. The loading procedure includes the transfer from batch to a wafer alignment station and from there to the quartz boat. The unloading procedure takes roughly half the time, because the wafer alignment can be omitted.
Outlook
The proposed handling technique can easily be adapted to custom problems. This has been demonstrated in the past and will be a part of future work on the application side. Further work has to be done in the field of fluid dynamics simulation, where a reliable prediction model for forces for specific gripper designs is under development. With such a model, the development time for a custom design could be significantly reduced, because time-consuming prototyping could be abandoned.
Future applications are not only foreseen in the semiconductor industry, but also in solar cell production, where thin cells are likely to go into series production in 2–3 years.
Alfred Binder, Gerhard Kroupa, Carinthian Tech Research AG, Villach St. Magdalen, Austria
Acknowledgments
This work was co-funded by the K-plus program of the Federal Ministry of Transport, Innovation, and Technology, and by Infineon Technology in Austria. We appreciate collaboration with our colleagues at CTR, in particular, Josef Sulzer and Roland Moser.
References
- F.A. Schraub, Solid State Technology, pp. 59–64, September 2002.
- F. Ziegler, Mechanics of Solids and Fluids, Springer Verlag, 2nd ed., January 1995.
- Bernoulli and vacuum combined gripper, EP1091389.
- Method and apparatus for loading and unloading wafers, EP1207548.
Alfred Binder received his diploma in mechatronics from Johannes Kepler University, Linz. He is an R&D engineer at Carinthian Tech Research AG, Europastrasse 4/1, A-9524 Villach, St. Magdalen, Austria; ph 43/0 4242 56300 - 210, fax 43/0 4242 56300 - 400, e-mail [email protected].
Gerhard Kroupa received his diploma in technical physics from the Technical University of Vienna. He is an R&D engineer at Carinthian Tech Research AG.