The manufacture of semiconductor wafers comprises a series of steps, beginning with the pulling of a monocrystalline ingot, typically of single crystal silicon. The raw ingot is machined to a cylindrical shape and a groove is then formed along the surface of the cylindrical ingot, parallel to its longitudinal axis. The ingot is then sliced into a plurality of disc-shaped wafers. The groove formed in the ingot prior to slicing results in each wafer having a notch in its peripheral edge. This notch provides a reference or index mark to assist in correctly orienting the wafer during subsequent processing of the wafer and is commonly referred to as an"orientation

notch". The notch is normally V-shaped, with an internal angle of 90°.

The slicing of the ingot provides roughly finished wafers which then undergo a series of processes, typically including mechanical, thermal and chemical processes to provide finished wafers. The precise details of such processes vary according to the required characteristics of the finished wafer, but generally include a variety of grinding, lapping and polishing steps. Normally, the peripheral edges of the wafers, including the edges of the notch, are chamfered. Typically, at least one of the major surfaces of the wafer and the chamfered edges are polished to a mirror finish.

The slicing of the ingot is normally performed using a wire saw apparatus as illustrated schematically in Fig.

1 of the accompanying drawings. The ingot 10 is bonded securely to a mounting beam 12, which is in turn held by a clamp (ingot holder) 14. The clamp 14 is mounted on a supporting table 16. The wire saw comprises a multiplicity of cutting wires 18 arranged to form a web of parallel wires which extend between wire guide rollers 20 on either side of the table 16. The rollers extend parallel to the table 16 so that the wires 18 extend across the table at right angles to the longitudinal axis thereof. The wires 18 are spaced along the rollers 20 at intervals corresponding to the thickness of the roughly sliced wafers.

The longitudinal axis of the ingot 10 will be substantially parallel to the longitudinal axis of the table 16, but may be set at an angle thereto in order to align the crystal axis of the ingot at right angles to the cutting wires 18, as is well known in the art.

During a cut, the wires 18 are driven lengthwise at high speed, between a supply spool and a take-up spool (not shown), via the guide rollers 20. At the same time, the table 16 is driven in the direction of arrow A (the feed direction) so that the ingot contacts the wires 18, creating a bow in the wire web (this bow is exaggerated in Fig. 1). As the table 16 continues to be fed in the direction A, the wires cut through the ingot and into the mounting beam, until the ingot has been sliced completely through and the cut is complete.

During a cut, various forces are generated in various directions. As illustrated in Fig. 1, the major forces on the wires 18 are: Force F (X) along the lengths of the wires, created by the lengthwise running of the wires.

Force F (X2) is a component of F (X) in the horizontal plane.

Force F (Z) which is the reaction force between the wires and the ingot, created by the feeding of the table 16. Force F (Z2) is the component of F (Z) along the bowed wire web. The force F (Z) and the component thereof F (Z2) are related to the degree of bow of the wires.

These forces are distributed amongst all of the wires 18 in the web. It is assumed here that the forces are evenly distributed amongst all of the wires. The total number of wires is dependent on the length of the ingot and the thickness of the slices. In this discussion, the forces F (X), F (Z) etc. refer to forces applied to individual wires.

The basic factors affecting the level of these forces are the speed at which the wires are run, the speed at which the table is fed in direction A, and the composition and supply parameters (flow rate and uniformity) of the slurry applied during the cut. If these factors are all kept constant during a cut, the forces generated will vary during the progress of the cut. If the sum of the forces F (X) and F (Z2) exceeds the breaking load of the cutting wire, the wire will break. If force F (X2) exceeds the strength of the glue bonding the ingot 10 to the mounting beam 12, the ingot 10 will break loose during the cut. These cutting forces are inversely proportional to the slicing efficiency and proportional to the slicing work load/resistance. The forces also depend upon the length of the ingot and the spacing of the wires (i. e. the total number of wires for a given ingot length).

In order to avoid wire breakage or dismounting of the ingot, it is necessary to regulate the cutting forces during the cut. This is most easily accomplished by varying the table feed rate whilst maintaining the wire running speed and slurry supply constant. Fig. 2 illustrates a typical example of a table speed profile,

in which the table speed varies with the position of the table in the feed direction A.

Conventionally, the table speed profile is determined empirically, by trial and error, and is then applied in subsequent production cuts employing the same apparatus, type of ingot and operating parameters.

This requires the cutting of a number of test ingots, with subsequent inspection of the sliced wafers to examine factors such as warp and thickness variation.

This is expensive in terms of time and resources.

If the operational parameters are altered, e. g. by the use of new materials or the like in order to improve quality or reduce costs, then the results obtained using an existing speed profile may be adversely affected. In this event, further tests must be carried out to determine an appropriate new table speed profile.

It is an object of the present invention to provide improved methods and apparatus for slicing semiconductor ingots using wire saws, by actively monitoring the forces generated during cutting operations.

In accordance with the present invention, there is provided apparatus for slicing monocrystalline semiconductor ingots of the type including an ingot mounting assembly, a web of wire saws and means for moving the mounting assembly relative to the wire saws, and further including load sensing means for measuring

forces applied to the ingot by the interaction of the wire saws with the ingot.

Preferably, the ingot mounting assembly comprises a mounting table and an ingot holding clamp mounted on said mounting table, and said load measuring means comprises a load cell interposed between said table and said clamp.

Preferably, the load sensing means is adapted to measure forces in three orthogonal axes.

Preferably also, a signal output of the load sensing means is connected to data processing means.

Preferably also, the apparatus includes a motor for driving the table and a table speed controller for controlling the motor, and means interconnecting the load sensing means, data processing means, table speed controller and motor, such that the operation of the motor may be controlled on the basis of signals output from the load sensing means.

In accordance with a second aspect of the invention, there is provided a method of slicing a monocrystalline semiconductor ingot, in which an ingot is mounted on an ingot mounting assembly and the ingot is cut by a web of wire saws by moving said mounting assembly relative to the wire saws, including the step of measuring forces applied to the ingot by the interaction of the wire saws with the ingot.

Preferably, the ingot mounting assembly comprises a mounting table and an ingot holding clamp mounted on said mounting table, and said forces are measured by load measuring means comprising a load cell interposed between said table and said clamp.

Preferably, said forces are measured in three orthogonal axes.

Preferably, the method includes controlling the speed of movement of the ingot mounting assembly relative to the wire saws on the basis of the measured forces such that a predetermined, substantially constant force is applied to the wire saws during the progress of a cutting operation.

Preferably, the method further includes calculating at least one force component from said measured forces.

Preferably, the calculated force components include a first force F (X) generated along the lengths of the wires by the lengthwise running of the wires, and a second force F (Z2) which is that component of the reaction force between the wires and the ingot created by the feeding of the table in the direction along the wire web.

Preferably, the speed of movement of the ingot mounting assembly is controlled such that the sum of F (X) and F (Z2) does not exceed a predetermined proportion of the breaking load of the wire saws.

Embodiments of the invention will now be described, by way of example only, with reference to the accompanying drawings in which: Fig. 1 illustrates known wire saw cutting apparatus for slicing monocrystalline semiconductor ingots; Fig. 2 shows an example of a typical table speed profile for apparatus such as that of Fig. 1; Fig. 3 is a schematic perspective view of apparatus similar to that of Fig. 1, modified in accordance with the present invention; Fig. 4 is a block diagram illustrating an ingot cutting system employing the apparatus of Fig. 3; Fig. 5 is a diagram illustrating forces which are monitored by the system of Fig. 4; Fig. 6 is a further block diagram illustrating additional details of the system of Fig. 4; and Fig. 7 illustrates an example of the application of the invention.

Referring now to the drawings, Fig. 3 shows wire saw apparatus similar to that of Fig. 1, again with an ingot 10 bonded to a mounting beam 12, held by a clamp (ingot holder) 14 mounted on a supporting table 16.

The wire saw again comprises a multiplicity of cutting wires 18 guided by wire guide rollers 20.

In accordance with the present invention, the basic apparatus is modified by the inclusion of a load cell device 22, interposed between the clamp 14 and table 16. The load cell 22 may be of known type, adapted to measure forces in three orthogonal axes X, Y and Z.

Output signals from the load cell are transmitted by any suitable means to a computer 24 (Fig. 6), via a suitable amplifier 26 and signal filter 28 as may be required. Fig. 6 also illustrates a motor 30 for driving the table 16, a table speed controller 32 for controlling the motor 30, and an encoder 34 providing feedback between the motor 30 and the speed controller 32. In preferred embodiments of the invention, the system further includes programmable logic control means 36, allowing the load cell 22 and computer 24 to be interfaced with table speed controller 32, as shall be discussed further below.

In effect, the load cell 22 measures the total forces, in three dimensions, generated by the interaction between the ingot 10 and the wire web. The signals generated by the load cell are processed by the computer 24 to calculate those force components which are of interest.

The present invention can be applied in one of two basic modes of operation, as illustrated in Fig. 4.

In its simplest form, the system simply monitors the forces generated during cutting operations. The measured forces can be displayed during the cutting

operation and/or recorded for future reference. A predetermined table speed profile can still be employed, which controls the table motor 30 in a conventional manner and which may be switched on or off as desired.

Fig. 4 also shows a further example of a cutting speed profile 40, illustrating how, in an ideal case, the variation of the table speed during the cut results in a constant cutting force.

In the second mode, the system operates to actively control the table speed so as to maintain a pre-set cutting force.

With reference to Fig. 1 and Fig. 5, the pre-set cutting force is determined by the breaking load (BL) of the cutting wire and might typically be set as 70% of the maximum breaking load (0.7 x BL), so as to provide a safety margin. Accordingly, the system may monitor the forces F (X) and F (Z2) and control the table speed so as to ensure that the sum of F (X) and F (Z2) is maintained at a substantially constant value substantially equal to 0.7 x BL. That is, during the progress of a cut, the relevant force components are computed from the output of the load cell 22 and an appropriate speed control signal is applied to the table speed controller 32.

As shown in Fig. 4, the system further employs a number of parameters as follow: pre-set maximum table speed;

pre-set acceleration/deceleration factor; wire groove pitch; i. e. the spacing between the wires; pre-set cutting force, calculated as noted above, dependent on the strength of the wires; target web bow; ingot length; pre-set base table speed; i. e. a"safe"table speed to be employed at particular times during the cutting operation as discussed below.

The target web bow is selected to have a value for which the resultant force FZ (2) on the wire is well within the wire breaking load, and the target web bow is input into the system. The force generated by a selected web bow can be measured manually in advance; e. g. by using a spring weight scale to hold one of the wires at the target bow value. The actual force measured in this way is F (Z) (see Figs. 1,5 and 7), from which the component F (Z2) may be calculated.

Typically, the web bow may be selected such that F (Z2) is about 10% of the wire breaking load. The remainder of the breaking load (minus a suitable safety factor) is then available for use by F (X), generated by the running of the wire.

The forces on the load cell also increase with the length of the ingot, so that the forces on the individual wires depend on the length of the ingot and the spacing of the wires. Accordingly, these parameters must also be taken into account when controlling the table speed on the basis of the total

forces detected by the load cell. The system can then calculate the number of wires being used. In the accompanying diagrams, F (Z') is the reaction force corresponding to F (Z) as measured by the load cell (theoretically, F (Z) = F (Z')). F (Z') is equal to the total force measured by the load cell in the Z direction divided by the number of wires.

At the beginning of a cutting operation, the table 16 is driven in the direction A at the pre-set maximum speed until the wire web contacts the ingot. The speed is then reduced to zero and subsequently increased in accordance with the pre-set acceleration/deceleration factor until the pre-set cutting force is reached, whereafter the speed is controlled as described to maintain a constant cutting force of the desired value.

The load cell 22 can also be employed to detect when the ingot contacts the wire web for this purpose (i. e. while the measured force N = 0). The acceleration may also be controlled on the basis of the load cell output, so that the measured force N divided by the distance travelled is maintained at a constant value until the desired target force is reached.

Fig. 4 also illustrates (42) how the table speed and cutting force varies during the progress of the cut under automatic control based on the pre-set cutting force, maximum table speed and acceleration factor.

In this second mode of operation, it is preferred that the table speed is displayed and/or recorded. The

cutting force could also be displayed/recorded as in the first mode.

As indicated in Fig. 6, the computer 24 may provide a variety of control, monitoring and management functions, including: "Recipe Management"This refers to the operational parameters ("recipe") of a particular cutting process, including the wire speed, slurry supply, table speed profile, automatic control parameters, ingot length, wire pitch, pre-set cutting force, etc.

"Diary Management"This refers to the process of logging data relating to cutting operations, including all of the parameters defined by the recipe and any other relevant data.

"Display"This refers to the data displayed during cutting operations, which may include the table speed, cutting force (i. e. the actual measured forces and/or computed force components), diary data, etc.

A typical example of a cutting operation managed by the system will now be described, with reference to Figs. 4, 5 and 7.

Firstly, the target bow is selected and tested as described above, and the bow value is input into the system along with the other pre-set parameters previously described, including the designed cutting force (e. g. 0.7 x wire breaking load).

A suitable recipe is selected based on the ingot length.

The exact ingot length is entered and the number of wires calculated.

The cutting operation is started, beginning with a warm- up period (typically 20 minutes) before the table begins to translate the ingot towards the wires.

The ingot table travels at the pre-set maximum speed until the load cell detects contact between the ingot and the wire web. The system recognises this as the beginning of the table stroke proper (typically 205 mm).

The ingot table then travels at the pre-set base speed for a short distance (e. g. 2-3 mm), and F (T) and F (X2) are monitored: F (T) = IF (Z') | + IF (ZX) |. The forces generated while the table is running are illustrated in Fig. 7A.

The translation of the ingot table is paused at this position, so that the load cell now detects only the force generated by the web bow (Fig. 7B), with no force being generated by the running of the wires. BY comparing the forces measured while the table is moving and while the table is paused, the system can calculate the actual bow at this position, taking account of other parameters such as the wire guide span (Fig. 7C).

The system then begins feeding the table forward once more at the maximum pre-set speed until it creates the target bow on the web and re-starts the cut. The system then monitors F (T) and F (X2), calculates F (Tw) and

determines whether F (Tw) is still within the pre-set cutting force. If not, the cut stops and an alarm is activated. At this point, the table travels at the pre- set base speed again.

During the cut, the direction of running the wires is reversed periodically. Whenever the direction is reversed, the above described process of pausing the cut and calculating the actual bow is performed, whereafter the cut resumes and F (Tw) is re-calculated to determine whether the load on the wire is still within acceptable limits.

Beyond this point, the system may increase or decrease the table speed, based on the acceleration/deceleration factor, in order to optimise the table speed without exceeding the pre-set cutting force and the target web bow.

The invention may be usefully employed during trials of new or modified cutting processes and/or during actual production operations. For example, (and with reference to Fig. 5) in a situation where a new, thinner type of cutting wire is being evaluated, there would be a risk that established process parameters might result in breakage of the wire. In determining an appropriate cutting force, the desired amount of bow on the cutting wires would first be decided (e. g. 30 mm, with a 440 mm span between the guide rollers 20) and the cutting force calculated as 70 % of the breaking load of the new wire. These parameters may be input into the system computer and all of the relevant

force components may then be computed from the output of the load cell 22.

After completion of the cut, the data logged may be reviewed alongside the actual wafers produced to determine whether the control parameters are satisfactory. If not, the parameters may be adjusted or the data obtained may be used to assist in determining a table speed profile which may then be employed in an otherwise conventional manner.

In this example, the breaking load of the wire is 70 N, so that the pre-set cutting force is 49 N. This value is much lower than the strength of the glue securing the ingot to the mounting beam (as discussed previously above), so that the latter may be ignored.

The invention may be usefully employed in a number of different ways. Ideally, production cutting processes may be conducted entirely under the control of the system, based on a small number of input parameters.

Alternatively, the system might be employed in combination with previously established table speed profiles etc. to improve process control and to provide additional data for future reference. At least, the invention may be employed in test operations to facilitate the process of determining appropriate table speed profiles for new and modified cutting processes.

Improvements and modifications may be incorporated without departing from the scope of the invention.