Document Sample

    Tonio Buonassisi1, Scott Reitsma2, Robert Sweeney2, Matthew D. Pickett2*, Weidong Huang2, Jon Lesniak3, Matthew L.
              Massachusetts Institute of Technology, Cambridge, MA 02139, USA; Formerly: Evergreen Solar, Inc.
                        Evergreen Solar Inc., 257 Cedar Hill St., Suite #2, Marlborough, MA 01752, USA
                           *Currently at: Hewlett-Packard Laboratories, Palo Alto, CA 94304, USA
                               Stress Photonics Inc., 3002 Progress Road, Madison, WI 53716, USA
                                      Correspondence: Tonio Buonassisi, buonassisi@mit.edu

     We employ a combination of commercially available analytical tools to monitor the evolution of bulk and edge shear
     stresses at stages of commercial production of String Ribbon wafers and cells produced by Evergreen Solar. Through
     absolute and differential near-infrared (NIR) photoelasticity images taken on as-grown and processed wafers, shear
     stress distributions and changes thereof can be imaged with an integration time of only a few seconds. The relevance
     of NIR photoelasticity data is demonstrated through a clear correlation with four-bar flexure testing, a measure of
     breakage strength and a surrogate for yield. Additional applications of this technique, e.g., for evaluating the quality
     of laser-cut edges and crack detection, are exemplified.
     Keywords: Stress, breakage, yield.

1    INTRODUCTION                                                     gradient during growth, descending from 1400°C to room
                                                                      temperature in less than one hour. If the slope of the
     During the fabrication of a multicrystalline silicon             thermal gradient is not strictly controlled, large residual
(mc-Si) solar cell – from crystal growth to tabbing and               stresses within the as-grown material can result, as shown
stringing – both the substrate and the applied layers are             in Figure 2.
subject to a variety of thermal stresses. These stresses can
become “locked in” during cooling, resulting in local
residual stresses on the order of several tens to a few
hundreds of megapascals. While compressive stresses can
be beneficial and inhibit cracking, tensile stresses assist
crack propagation and can be disastrous for wafer
mechanical integrity. Residual stress can thus be
considered among the most important parameters for
predicting process yield.
     Unfortunately, compared to widely disseminated
techniques for measuring solar cell electrical properties,
stress measurement tools are still largely underdeveloped.
Until this year, few techniques were capable of
measuring stress in a contactless, non-destructive, and
repeatable manner, easily accommodating wafers of
different sizes and geometries, with high throughput.
Furthermore, none of these techniques were
commercially adapted to the PV industry. Meanwhile, in
other industries such as glass manufacturing, stress
measurements using automated photoelastic techniques                  Fig. 1: IR photoelasticity tool developed by Stress
were commercially developed to detect both residual                   Photonics, applied to String Ribbon wafers.
stresses and defects such as cracks and inclusions [1]. In
this contribution, we describe the application of a
commercially-developed           near-infrared        (NIR)
photoelasticity tool (Fig. 1), produced by Stress
Photonics, to monitor the evolution of bulk and edge
stresses in stages of commercial production of Evergreen
Solar String Ribbon wafers and cells. Additionally,
differential photoelasticity (DPE) can be useful for
quantifying the effect of a given processing step on bulk
stresses, edge stresses, or defect (e.g., crack) generation.
Previous investigations using similar techniques were                 Fig. 2: Stress images of 8x15 cm2 Evergreen String
performed by the groups of Danyluk [2] and Vallera [3].               Ribbon wafers with low bulk stress (left) and high bulk
                                                                      stress (right), by NIR photoelasticity. Several wafers are
                                                                      averaged for each condition, to normalize the effect of
2 COMPARISON            OF    AS-GROWN         RESIDUAL               crystal-orientation-specific (anisotropic) birefringence.
STRESSES                                                              On the right, bright orange regions along long edges
                                                                      indicate tensile residual stresses resulting from a sub-
     As-grown ribbon silicon experiences a large thermal              optimal thermal profile during cooling.
3 STRESS REDUCTION               IN   BULK      WAFERS:

    High-temperature annealing is commonly employed
in the glass industry to alleviate grown-in residual stress.
We demonstrate that annealing can have a similar
beneficial impact on residual stress within bulk mc-Si         Fig. 4: Microscopic IR photoelasticity images of two
wafers. NIR photoelasticity measurements were taken on         laser-cut edges: low-stress (left), and high-stress (right).
the same wafer before and after annealing; these two           Images are a few hundred microns across, and the stress
images were then subtracted to normalize any crystal-          field corresponds to the distance between laser pulses.
orientation-specific (anisotropic) birefringence. The
resulting differential DPE image indicates the change in
stresses directly. In Figure 3, we compare the in-plane
shear stress component of experimental DPE
measurements to the theoretically predicted shear stress
component of the residual stress, assuming the
distribution of as-grown stresses shown in Figure 2. The
excellent correlation indicates that residual stresses are
alleviated through annealing, with the magnitude being
quantifiable with DPE.

                                                               Fig. 5: Evolution of stresses along a laser-cut edge is
                                                               shown before and after annealing, using IR
                                                               photoelasticity (right) and four-bar flexure testing (left).

Fig. 3: An as-grown (15x8 cm2) Evergreen String
Ribbon wafer is imaged by IR photoelasticity before and
after annealing. The shear 0 (shear stress) component of
the difference image (left) agrees with finite element
analysis (FEA, right), indicating a relaxation of residual
stresses. (The vertical line on the right of the DPE image
is a registration artifact.)


    During the laser cutting process, silicon recast may
form along the laser-scribed edges. This recast material
has potential to generate alternating compressive and          Fig. 6: Evolution of wafer breakage strength with
tensile stresses on the sub-mm scale, thus reducing the        phosphorus diffusion.
mechanical integrity of the cut edge and increasing the
propensity for cracking [4]. These edge stresses are
clearly visible using NIR photoelasticity imaging, as
shown in Fig. 4. A comparison of different laser cut
qualities can be performed by comparing the magnitude          5 STRESS EVOLUTION DURING CELL
of edge stresses.                                              PROCESSING
    The stresses at a laser-cut edge can be further reduced
by altering the nature of the recast material, e.g., via           Wafer strength can be significantly increased through
etching, annealing, or polishing. Shown in Fig. 5 is the       optimized cell processing. Figure 6 demonstrates an
effect of annealing on edge stress, measured by NIR            increase in wafer breakage strength with phosphorus
photoelasticity (right) and four-bar flexure testing (left).   diffusion, as measured by 4-bar flexure testing.
A drastic reduction in edge stress is noted by both                Conversely, stresses may be induced during solar cell
techniques, with little corresponding change of recast         processing, e.g., via handling or tabbing and stringing.
morphology (measured by scanning electron microscopy,          DPE permits one to visualize these stresses in real time.
not shown).

    Localized stresses at a crack tip generate a telltale
signature in NIR photoelasticity images. This signal is
generally distinguishable from the natural crystal
birefringence of a single image, and it is exceptionally
easy to identify using DPE. By taking the difference
between two NIR photoelasticity images before and after
a given processing step, one can easily determine if a
partial crack has propagated into the wafer. In the DPE
image shown in Fig. 7, one can see that two cracks have
propagated into the wafer, one from either side.

Fig. 7: Crack tips are clearly visible in this DPE image of
an 8x15 cm2 wafer, as evidenced by the localized stress


This work would not have been possible without the
advice, inputs, and assistance of current and former
members of the R&D team at Evergreen Solar, Inc.,
including Voy Anuszkiewicz, Sarah Decourcy, Larry
Felton, Tom Ford, Eric Gabaree, Andrew M. Gabor,
David Harvey, Colan Jones, Dick Krauchune, Luey
Nyugen, Minh Le, Adam M. Lorenz, Alan Rolke, Gary J.
Tarnowski, and Richard Wallace. In addition, we
recognize the software development efforts and technical
support of Brad Boyce, Stress Photonics, Inc.


[1] G. Horn, J. Lesniak, T. Mackin, B. Boyce, Rev. Sci.
Instr. 76, 045108 (2005).

[2] Shijiang He, Ph.D. dissertation. Georgia Institute of
Technology, 2005.

[3] M. C. Brito, J. P. Pereira, J. Maia Alves, J. M. Serra,
and A. M. Vallera. Rev. Sci. Instr. 76, 013901 (2005).

[4] S. Schoenfelder, J. Bagdahn, S. Baumann, D. Kray,
K. Mayer, G. Willeke, M. Becker, S. Christiansen, 21st
European Photovoltaic Solar Energy Conference,
Dresden, Germany (2006), p. 588.