Anda di halaman 1dari 4

Industrial Challenges For Thin Wafer Manufacturing

Per Arne Wang REC Wafer Maximum deflection, at the centre of the wafer, is given by:

ABSTRACT. Because of limited supply of Silicon, there has been an industry-driven requirement for reduction of material consumption per Wp of solar power. As a consequence, the thickness of wafers has been reduced, which in turn has introduced new challenges for the industry. During the wafer manufacturing cycle, the wafers are exposed to mechanical loads caused by sawing, manual handling, liquid jets, transport systems and pick and place equipment. The introduction of thinner wafers has reinforced the requirements for reducing these mechanical loads, in order not to increase the breakage and defect rates accordingly. The research has aimed at modeling and measuring the stresses on the wafers in the various process steps. This provides a tool for obtaining an optimal tuning of the process-parameters. In addition, this work is essential for providing specifications of better and more gentle wafer-handling technologies in the future. MECHANICAL LOADS IN WAFERLINES. During the production cycle of solar cells, from wafer-sawing and downwards, the wafers and cells are exposed to various types and levels of mechanical loads. In our waferlines, for instance, the sawn wafers are washed, stacked, separated, flipped, rotated, repositioned, measured and packed into units of about 150 pieces. Thinner wafers will require less mechanical loads in order not to increase the breakage. In addition, there are some challenges regarding cracks. Cracks might be generated in the sawing process and/or in one or more of the downstream steps. Also cracks might cause breakage in the same process steps. In this context, it has been necessary to categorize the mechanical loads in our waferlines. The mechanical loads are represented by certain forces or a certain deflections, or combination of those. Also the mechanical loads might be dynamic or static. In the following chapters, the effect of the different mechanical loads is discussed. The discussion is based on simple beam-bending models and crack propagation models. THE CHALLENGE OF WAFER THICKNESS. The following equations are valid for a simple threepoint bending process, which will be used for further analysis:

d max =

Fb 3 4 Eat 3

1)

Maximum stress, at the centre of the wafer, is given by:

s max =

3Fb 2at 2

2)

2) and 3) gives the relation between maximum stress and maximum deflection:

s max =

6 Etd max b2

3)

Where a is the width of the wafer, b is the distance between the two supporting rolls, F is the applied centered force, t is the wafer thickness and E is the elastic modulus. Equations 1) and 2) show that for a fixed force, the deflection and the stress increase when thickness is reduced. In other words, the risk of wafer breakage will increase when the thickness is reduced. On the other hand, Equation 3) shows that the stress is reduced with thickness-reduction, as long as the deflection is constant. Figures 1 and 2 shows the deflections and stresses as functions of wafer thickness, for the two situations.

1-4244-0016-3/06/$20.00 2006 IEEE

1179

Deflection as a function of wafer thickness. 50 Deflection (mm) 40 30 20 10 0 0,00 0,05 0,10 0,15 0,20 0,25

The reduced thickness of the wafer implies that it is more flexible, which means that the deflection is higher for the same applied force. On the other hand, if our process requires a certain force, the stress in the wafer is increasing with reduced thickness. This means that we have to be careful about what kind of mechanical loads we have during the process, being deflective or applied by force. As will be seen, the mechanical loads in our waferlines are of both categories. THE CHALLENGE OF WAFER MICRO-CRACKS.

Wafer thickness (mm)


0,3 1,0 2,0 4,0 Limit

Stress as a function of wafer thickness 150 Stress (MPa) 100 50 0 0,00

In the above examples, only stress concentrations as functions of thickness are presented. In order to get a complete picture, it is, for Si as with other brittle materials, important to be aware of the cracks, defects, and inherent stress concentrations in the wafer. The wafers typically have bulk defects such as dislocations, particles, voids and grain boundaries, and surface defects such as cracks, chips, saw marks and other edge defects. All these defects and anisotropies causes stress concentrations in the wafer, which in turn breaks the wafers at significantly lower stress levels than described above. A model that has been used for crack propagation in linear elastic materials, is the surface creation criterion (Griffith). In this case, there has to be an energy balance between the stress energy that is released during fracture, and the surface energy that is created during crack propagation.

0,05

0,10

0,15

0,20

0,25

Wafer Thickness (mm)


0,3 1,0 2,0 4,0 Limit

Figure 1. The above graphs show the deflection and maximum stress as functions of wafer thickness for the case of fixed forces of 0.3, 1.0, 2.0, and 4.0 Newtons. The purple line represents a stress limit of 120 MPa. As an example, when we have an applied force of 2 Newtons, wafers thinner than about 160 microns will be broken. In that case, the deflection is about 20 mm. For these calculations, the dimensions of the wafers are 156x156 mm and the elastic modulus is E=150 GPa.

s2 =

2gE pl

4)

Stress as a function of wafer thickness 140 120 100 80 60 40 20 0 0,00

On the basis of this, a stress intensity factor is 1 defined.

K = s pl

5)

Stress (Mpa)

0,05

0,10

0,15

0,20

0,25

Where is the remote stress, g is the surface energy, E is Youngs modulus, and l is the crack length. If the Stress Intensity Factor K is larger than a critical value Kc, the fracture will propagate through the material. For Silicon, Kc values in the range of 0.7 to 0.95 are reported. If the crack length is shorter than the critical length, the material will not break. On the other hand, when the crack length is higher than the critical length, a fracture will occur at a significantly lower stress than the nominal breakage stress level, see figure 3.
1

Wafer Thickness (mm)


1,0 5,0 10,0 20,0 Limit

Figure 2. Maximum stress as a function of wafer thickness for the case of fixed deflections of 1, 5, 10 and 20 mm. The purple line represents a stress limit of 120 MPa. In this case, the stress will be lower, when the wafer thickness is reduced.

In equation 5, a geometry factor of Y=p0.5, and crack opening mode, is assumed, which makes it strictly valid only for a straight internal crack in an infinite specimen. For an edge crack, a 12% increase is assumed.

1180

Critical Crack Length as a function of Remote Stress


1000 900

Crack Length (m)

800 700 600 500 400 300 200 100 0 0 50 100 150 Critical Crack Length Griffith Critical Crack Length Inglis

Stress (MPa)

Figure 3. Critical Crack length as a function of applied stress is shown. Griffith theory for crack propagation is used, and Inglis criteria is also included for comparison. As an example, with an applied stress of 20 MPa, the critical crack length is in the order of 500 m. Kc=0,7 is used in this example.

Figure 4. Sample crack pictures. The corner of a wafer, with transmitted NIR light (top), and reflected light (bottom).

The above discussion shows that we have to be careful about taking the right actions with respect to wafer handling equipment. Cracks are getting more important when the wafers are getting thinner, which means that our process window is getting smaller, and thereby requires that we have a better control and stability of our process equipment. MEASUREMENT OF MECHANICAL PROPERTIES OF CRACKED WAFERS. Since cracks has been identified to be one of the most critical factors regarding wafer breakage, we have performed a test of a crack detection system. The system was set up with near-infrared transmitted light for imaging, reflected visible light as a reference, and image processing algorithms for automatic detection and sorting of cracked wafers (patent pending). Since NIR transmitted light were used, the system was also able to detect internal cracks. The test showed that, with a given pixel resolution of 50 m, we were able to detect cracks on our wafers down to about 5 m width, and about 500 m length. Figure 4 shows a sample transmission and reflection image of a crack near the corner of the wafer. Figure 5 shows micrograph of a detected micro-crack on the wafer surface, taken with a normal light-microscope.

Figure 5. Sample light microscope picture of a crack, located at the centre of the wafer. This is an internal crack with length approx. 500 microns, and width of 48 microns.

With the crack detection system, wafers with microcracks were sorted out from normal A and B batches of wafers. The cracked and the un-cracked wafers from the two qualities (4 series) were tested mechanically in a 4-point bending test. The wafers were oriented randomly (0, 90, 180, 270) on the rolls. The cracks were also in most cases not perpendicular to the stress direction which means that the crack tip was often not exposed to the maximum stress. This means that all crack load modes (opening, sliding and tearing) were applied on the same crack at the same time, and that the critical crack length will be slightly higher than estimated from Equation 5 and Figure 3. The breakage force for each of the wafers was plotted in a conventional Weibull diagram, as shown in Figure 6.

1181

One can see that the median strength of the A and B batches and the cracked A-wafers is around 8.0 N, while it is around 3.0 N for cracked B-wafers. Since many of the B-wafers actually have very good mechanical properties, it might make sense to change the criteria for the B-quality.

The calculated median strength of the A and B materials is 120 MPa, which is in consistence with the previous assumptions. The above test shows that with a micro-crack detection tool, it will be easier for us to identify process equipment that introduces cracks. With an on-line system we can, at an early stage sort out bad wafers, and reduce the breakage costs along the value chain. CONCLUSION. In order to avoid increased fracture rates in our process and our customers process, we will have to use equipment that reduces the mechanical loads, according to the above mentioned criteria. We will also have to run the process in a state that the amount of weak wafers is reduced. For this purpose, we will investigate further the possibility of:

Weibull plot - accumulated broken wafers as function of strength


1,2

Broken proportion

1,0 0,8 0,6 0,4 0,2 0,0 0 2 4 6 8 10 12 14

Force [N]
B-wafers Cracked B-wafers A-wafers Cracked A-wafers

Online detection and rejection of bad wafers. Online detection and feedback for process improvement.

Figure 6. Weibull plot of the different series of wafers. The wafers were oriented randomly on the 4-point bending tester, and the cracks were therefore not systematically exposed to tension. This explains some of the unreasonably high figures for the cracked wafers.

Table of mechanical properties for Si (111)


Youngs Modulus, E, (GPa) 0.5 Toughness, Kc, (MPa m ) -2 Crack Resistance Energy, Gc, (Jm ) -2 Surface Energy, W=2g, (Jm ) 170 0.7 3.0 2.4

In order to calculate the corresponding stress in the material, the following is valid for the relation between stress and applied force for 4-point bending:

References [1] Brian Lawn : Fracture of Brittle Solids. Second Edition [2] T.L. Anderson : Fracture Mechanics [3] J.E. Gordon : The New Science of Strong materials

s max =

3Fb at 2

6)

Where a is the width of the wafer, b is the horizontal distance between the upper and the lower roll, F is the applied force, t is the wafer thickness.

1182

Anda mungkin juga menyukai