High Speed Movement of Workpieces in Vacuum Processing

The present invention relates to a method for moving workpieces in a vacuum chamber for high processing throughput. More specifically, it involves a method which maintains a nearly continuous flow of workpieces for processing as they are introduced into a vacuum chamber, aligned to a suitable orientation, processed one at a time, and removed from the vacuum chamber.

During the processing of silicon wafers in the manufacture of microelectronic circuits, up to twenty different types of tools are employed for effecting several hundred processing steps. Most of these processing steps must be performed in a vacuum chamber at pressures less than 1.times.10.sup.-3 torr, and each requires from ten seconds to three minutes per wafer. Most of the processing tools operate on wafers one at a time in order to optimize control and reproducibility in a manufacturing environment.

In general, each operation on a wafer must be performed in a particular order, so that each operation must wait until rite completion of a preceding one, and, in turn, affects the time a wafer is available for a subsequent step. Tool productivity or throughput for vacuum processes that are relatively short, such as ion implantation, can be severely limited if the work flow to the processing location or platen is interrupted by sequential events, which may include, for example, the introduction of the wafer into the vacuum system, the transfer of a wafer in the vacuum chamber or the exchange of wafer carriers or cassettes.

In order to increase throughput, it is desirable to shorten the duration of sequential events, i.e., those events which must be performed consecutively. For this reason, the pumpdown times (to high vacuum) and the venting times (to atmospheric pressure) are usually accomplished as quickly as possible, in less than about a half minute. However, high levels of foreign material or particles carried deep into the process line may result from the turbulence of such rapid cycling. These particles become distributed on the surface of the silicon wafer and result in more defects for a higher fraction of the micro-electronic circuits being fabricated, reducing manufacturing yields significantly.

The prior art has sought to address these concerns in a number of ways. Early systems employed wafer handling stations that introduced wafers into the vacuum process chamber through a load lock one at a time by mean of a gravitation slide. Each wafer was processed, and exited the vacuum process chamber through a second load lock as shown, for example, in FIG. 1 of U.S. Pat. No. 4,282,924. These systems suffered from several problems as wafer sizes employed in micro-electronic device fabrication evolved to larger diameters of six and eight inches, including wafer breakage and high levels of defect generation resulting from both sliding and rapid vacuum cycling. Also, manufacturers tracked work in process through the factory by tracking the progress of the wafer carriers, not the individual wafers, through the line, and since wafers were not returned to their original carrier or cassette, this complicated tracking of the work in process through the factory.

Another prior art approach used a wafer handling station in which the two cassettes are introduced, one each through load locks #1 and #2, and are pumped down simultaneously. Wafers were processed in an alternating manner from both load locks; for example, the first wafer from load lock #1, the second from load lock #2 and so on until all wafers from both load locks had been processed. In this system, wafers are returned to their original cassette, however the pumpdown and venting to atmosphere are sequential to the processing. Also this scheme requires that both cassettes of wafers be introduced to or removed from the load lock chambers in essentially the same time interval while processing is halted. Since production lots may very often be only a single cassette of wafers, this is not always possible. Another problem with this system is the requirement that the operator be present immediately following processing to exchange the processed cassettes of wafers for the next cassettes waiting to be processed, in order not to lose valuable process time.

It would be desirable to arrange the handling or movement of wafers in a manner which assured a continuous flow of wafers from cassettes which have been transferred into the vacuum, while increasing time available for slowly venting the load locks, exchanging the cassettes of processed wafers for unprocessed ones, and slowly pumping the load locks containing the cassettes to a suitable vacuum.

It would also be desirable to effect wafer handling with a small number of motive mechanisms while enhancing the flow of wafers.

These and other desirable ends are achieved in accordance with the present invention by a process line in which two cassettes are introduced via separate load locks. Multiple transport mechanisms, such as robots, move successive wafers from one cassette to a processing station and back to the cassette in a substantially continuous flow. The robots then process successive wafers from the second cassette. This allows ample time for introduction or removal of the first cassette, and its attendant pumpdown or venting operations, during processing of the second cassette. Wafers of the second cassette are then processed while a new first cassette is introduced or removed, and wafers of the first cassette are processed while the second cassette is introduced or removed. This provides a nearly continuous flow of workpieces in the vacuum chamber, as they are aligned to a suitable orientation and processed, while allowing a leisurely cycle of introduction and pumpdown, or venting and removal, of the other cassette.

FIGS. 1A-1C illustrate a prior art wafer handling system;

FIG. 2 illustrates another prior art wafer handling system;

FIG. 3 illustrates a water handling system in accordance with the present invention; and

FIG. 4 illustrates exemplary transport paths in a simple embodiment of the wafer handling system of FIG. 3.

The invention will be understood following a more detailed examination of several prior art handling systems which have been briefly discussed above.

FIGS. 1A-1C illustrate a prior art wafer holder 10 which is moved by a support rod 12 to each of three rotational positions in which it receives, treats, and releases a wafer 13, respectively. In a first position, the wafer 13 slides by gravity feed into a slot 11 in the holder. The holder 10 then rotates about rod 12 up to an orientation normal to an ion beam where it holds the wafer in position for implantation or other treatment, after which it rotates downwardly, releasing the wafer to slide once again by gravity feed, along an exit path. Auxiliary handling mechanisms may be present in the input and output paths for picking the wafer from a cassette and returning it to a different cassette. This rocking stage construction provides a relatively simple mechanism for bringing wafers into and out of the processing station. However, the sliding motions may generate dust, and this continual feed-though mechanism involves highly sequential operation.

FIG. 2 illustrates the operation of the second type of prior art wafer handling system 20 described above. In this system, first and second wafer carriers 30a, 30b are received by respective load locks 22a, 22b which communicate with the main vacuum chamber 40 via isolation valves 23a, 23b and which each may be independently vented and pumped down. A single process station 25, such as an ion beam implantation station, is centrally served by a pair of robots or wafer transport mechanisms 26a, 26b. Each transport is capable of axial motion along a member 28a, 28b and pivotal or articulated motion of an arm 29a, 29b. The axial motion is used to move wafers between the vacuum chamber and the carrier residing in a load lock, and the articulated arm is used for more constrained motions within the vacuum chamber, for example, to swing from an intermediate station (27a or 27b) to the implantation station 25.

Operation of this device is as follows. For a single wafer, illustratively in a carrier in load lock 22a, the corresponding transport 26a advances linearly to grip a wafer, then retracts to place it on the alignment turntable 27a. After proper alignment of the flat on the wafer edge, the articulated arm 29a lifts the wafer and rotates to deposit in at station 25. Removal entails the reverse rotation and without stopping at the alignment table, the arm advances linearly to re-deposit the wafer in the carrier, which is then raised one space to bring the next wafer into a position accessible to the transport. As the "A" wafer (from wafer carrier 30a) is being treated at the implantation station, the second robot 26b is loading and orienting a "B" wafer from the second wafer carrier 30b. It then transports this wafer into the implantation station 25 just as the "A" wafer is being unloaded therefrom. Thus, the wafers form a continuous stream through the implantation station coming alternately from each of the two cassettes 30a, 30b. However, when each cassette has been processed, which for equal size loads will occur within a few seconds of each other, then the rather time-consuming steps of venting, cassette changing and pumpdown must be performed while the process station sits idly. As noted above, this sequential operation compels one to employ fast vacuum cycling, which, in turn, results in dirtier processing.

In distinct contrast to the foregoing prior art, a processing system 100 in accordance with the present invention allows flat thin workpieces to be introduced into a vacuum system for high speed processing in nearly a continuous flow and without introducing sequential delays. The system has two load locks, and the workpieces or wafers are processed one at a time, and each is returned to the same workpiece carrier or cassette from which it came. When all of the workpieces have been processed from the first cassette, that cassette is returned to atmospheric pressure in the load lock during the time that the next cassette of wafers is being processed from the second load lock. The flow of work is such that the pumpdown and vent times are not sequential to the processing steps, but occur in background for each cassette as the other cassette of wafers is being processed. Thus, the timing of cassette exchange is not critical to throughput as in the prior art system of FIG. 2, and may occur at any time during the processing of the other cassette. Correspondingly, slow vacuum cycling may be employed, using simpler equipment and cleaner conditions, without introducing processing delay.

A simple embodiment of the invention is shown in FIG. 3. Semiconductor wafers 13 are moved into a vacuum chamber 40 and processed one at a time or serially by introducing a cassette 30a of wafers 13 into the first load lock 22a. A second cassette of wafers may be introduced into a second load lock 22b subsequently without impacting the overall productivity or throughput of the system, and the amount of time available for introducing the second cassette and pumping down its load lock is essentially the processing time required by the entire first cassette, which is dependent on process time per wafer. For process times of approximately ten seconds, this time duration can be approximately five minutes for a cassette which holds twenty-five wafers.

The basic elements of the system 100 include first and second robots or wafer handlers 62, 64 in the vacuum chamber 40, an implantation station 25, and a transfer station 50. The implantation station may comprise, for example, a flat platen or stage with suitable registration, clamping and cooling mechanisms for positioning and holding the wafer during ion beam treatment. The transfer station 50 on the other hand may comprise a common alignment turntable which orients the wafer prior to processing, or if alignment is not necessary (for example, for implantation normal to the wafer surface) may comprise a simple pedestal or stand.

The robots 62, 64 are shown as multi-segmented articulated arms. However, their actual mechanical implementation may be varied to include ram-like motions produced by any of several different mechanisms, so long as they efficiently perform the required movements described, further below. In general, the robots do not actively grip the wafers, but instead slide a yoke-like support under the wafer as it is raised to an accessible position by lift pins or the like. The robot is then required to execute movements without undergoing extreme accelerations that might cause the wafer to slide off. These movements consist of a combination of one or more linear and rotational motions which occur in a program-controlled sequence that moves the water from its starting position to the next desired position.

Operation of the system 100 proceeds as follows. Load lock #1 (lock 22a) is pumped down to approximately 100 millitorr after which its isolation valve 23 opens to allow the transfer of wafers from the load lock into chamber 40 and ultimately to the process station or platen. Robot #1 (robot 62) removes a water 13 from the cassette 30a in load lock #1 and transfers the wafer to the wafer transfer station 50. In the embodiment shown, the wafer transfer station 50 is a simple wafer alignment turntable, such as employed on the Model E-220 1 sold by the Varian Company of Gloucester, Mass., and the transfer station holds a wafer and also is used for alignment. Robot #2 (robot 64) then removes the wafer from the transfer or alignment station 50 and places the wafer 13 on the platen 25 at the processing station. During this time, robot #1 returns to load lock #1, gets the next wafer and places the next wafer to be processed onto the transfer station 50. After the first wafer has been processed at the implantation station 25, robot #1 removes the processed wafer from the platen. Meanwhile, as robot #1 is picking up the processed wafer from platen 25, robot #2 has already picked up the next wafer resting at the transfer station 50, and immediately places it on the implantation platen. As this second wafer is being processed, robot #1 continues to carry the processed wafer moving it to the cassette sitting in load lock #1, where it then picks up the next wafer in line to be processed from the cassette and proceeds to place it onto the transfer station 50. Robot #1 then again advances to the process station where it proceeds to unload the second processed wafer from the implantation platen 25, while robot #2 again having picked up a wafer waiting at the transfer station 50, immediately shuttles it into position on the implantation platen 25.

Thus, to process wafers from the first cassette, robot #1 handles the first and third legs of each wafer's travel, while robot #2 handles the middle step, shuttling the wafer from the transfer station to the processing station. In particular, at least one and often both robots are occupied at each instant, and two wafers are in transit through the vacuum chambers at any given time, with both robots simultaneously participating in the loading and unloading of the process station itself so that the ion beam processing is interrupted only for a brief instant. These steps are repeated until all of the wafers in the cassette 30a residing in load lock #1 have been processed.

At that time, the two robots reverse their roles, and operate in a similar and symmetric fashion to transport, process and return all the wafers in load lock #2. In this case, robot #2 operates to pick up and move wafers from the second cassette 30b to the transfer station 50, and, later, to pick up and to return each processed wafer 13 from the implantation platen 25 to the cassette 30b, while robot #1 operates simply to shuttle the untreated wafers waiting at the transfer station 50 quickly into position on the implantation platen 25. Meanwhile, while the wafers of the second cassette 30b are being processed, the first load lock 22a is vented and a new first cassette 30a' (not shown) is introduced into lock 22a, and the lock 22a is pumped down. These venting and pumpdown steps may occur over the course of several minutes, for a typical size process batch in cassette 30b, and may thus employ gentle pressure cycling that avoids turbulent conditions.

Thus, the robots both work in a bucket-brigade fashion to achieve a high-speed throughput, and although only one cassette is processed at a time, the transport mechanisms and control regimens are simple, and all wafers are moved and processed in a similar manner, with the respective robots periodically exchanging functions as each new cassette comes on line.

Notably, during the time interval that wafers from one load lock are being processed, the other load lock is vented, any processed wafers therein are removed, a new cassette of semiconductor wafers to be processed is placed in the load lock, the door is closed, the load lock is evacuated, and the valve communicating with the transfer chamber (if used) or vacuum chamber is opened. Thus, as processing of the current set of wafers is completed, the new cassette is already in position, and the robot which accesses that cassette proceeds without any delay to pick up the first of the new wafers and deposit it at the transfer station 50. It then continues the two steps of moving a fresh wafer to the transfer station, and carrying a processed wafer from the process station back to its wafer carrier, as the opposite robot shuttles wafers from the transfer station 50 to the process station 25.

Thus, the entire sequence of events, including the repetitive alternating steps of operation of the two robots proceeds cyclically without interruption, and the changeover from one cassette to another does not delay even temporarily the next step taken by a robot, or change the rate of wafer traffic along the process line.

The advantages of this method of moving wafers into and out of a vacuum system include:

1) the continuous flow of work to the process station and the achievement of throughputs of over 200 workpieces or wafers per hour for serial process systems without interruption for operator actions, even when process times are short; and

2) pumpdown and vent to atmosphere functions which are not sequential to the processing and transport events, but rather which occur simultaneously with processing, in background, so a greater time can be allotted for these functions, reducing turbulence during vacuum cycling and resulting in reduced generation or deposition of particles, and potentially higher yields for micro-electronic device fabrication.

The invention has been described with reference to a basic embodiment wherein a plurality of robots operate to hand along wafers from a first load lock or cassette in a continuous flow through a processing station. Illustratively a transfer or orienting/transfer station 50 allows a wafer to be temporarily "parked" so that while each wafer is alternately handled by one, then the other, robot, this "bucket brigade" transport schema does not require direct robot-to-robot transfers, and thus does not tie up both robots with a single wafer at any time. This allows multiple wafers from a single cassette to be in transit at any instant. The same plurality of robots also similarly transports the wafers from an entirely separate load lock or cassette, along separate transport paths, through the same process station.

FIG. 4 is a wafer path diagram of the transport steps just described for operation of the system of FIG. 3. The adjacent load locks 22a, 22b are illustrated as positioning their cassettes 30a, 30b on left and right sides of the system, marked "L" and "R". As shown, the wafers from the first cassette follow successive path segments a,b,c, and the wafers from the second cassette follow three entirely disjoint successive path segments d,e,f, each of which curves back from one side to the other as a wafer shuttles through the process station, so that the next entering wafer is not impeded as the two robots swing the two processed and unprocessed wafers in and out of the treatment station 25. The system employs a stark minimum number of elements--two robots and one transfer station--to provide essentially uninterrupted work flow through the processing station while completely eliminating the trade-off of sequential loading and fast vacuum cycling.

As noted above, for typical fast ion beam process times of about ten seconds per wafer, the illustrated transport paths are readily implemented without resort to robot movement accelerations of a magnitude that would cause wafer sliding or transport mishaps. For example, robot #1 may return a wafer along path "c", and pick up and deposit a new wafer along path "a", and then, empty, quickly position itself to pick up the next wafer at the processing station for the next "c" movement, taking only one to three seconds for each wafer movement step.

However, it is possible that special process considerations may render the transport steps rate-limiting. For example, if the ion beam treatment time were under several seconds, then it could become difficult to operate a single robot to perform three path motions as described without risking wafer dropping. In that case, the invention further contemplates the provision of additional parking or process stations to be located along the transport line. In the event the processing station 25 operates with an extraordinarily short process time, the invention further contemplates systems including additional robots, or a combination of additional robots and stations located along the transport path, in order to maintain continuous transport of two or more wafers without requiring excessively abrupt robot movements or an acceleration consistent with the low-inertia requirements of the delicate wafer-handling process, and without introducing delays or dead time at the process station. Furthermore, the invention is not limited to a system having two load locks and two cassettes, but is readily adapted to systems with a greater number of load locks, cassettes, or both.

In addition, while the above disclosure has assumed a conventional form of cassette loading (for example, wherein a cassette sits within a load lock and rides an elevator post up and down to position successive ones of its wafers in the plane of the robot), and has assumed a pair of robots with only partially overlapping ranges, the invention is readily adapted to other forms of cassettes loading. For example, in a system wherein the entire cassette is advanced from the lock into the vacuum chamber, the system may be arranged so that the cassettes are advanced to closely contiguous positions, each of which is accessible to both robots. In that case, if no orienting station is necessary, (e.g. for a bulk doping treatment) the robots may simply alternate picking up wafers and taking them the entire distance from a cassette to the process station in a single movement, and then back.

However, rather than custom-fabricated long-reach robots, the illustrated embodiments utilize commercially available articulated robots that can have a mechanical reach made up of a series of short leg rotatable segments, wherein neither robot reaches two cassettes, but both can overlap in an intermediate portion of their range of motion, at the transfer station 50. Further, equipment costs are further reduced in that by processing one cassette of wafers as the other is being slowly pumped down, a small vacuum pump may be used, or even a single small pump which is selectively valved to evacuate both of the load locks.

The invention being thus disclosed, variations and modifications will occur to those skilled, in the art, and such variations are understood to be a part of the invention, as set forth in the claims appended hereto.