i n dustrial automat ion products

Uniquely Challenging
Solar cell manufacturing faces several AOI challenges that can be
met by visible, near-infrared, or electroluminescent imaging

Xing-Fei He

Although the economic situation slowed tronics manufacturing processes, wafer-

down photovoltaic (PV) installation last year, based solar cell manufacture has some
the market continues to grow. According to unique processing steps that create both
the European Photovoltaic Industry Associ- opportunities and challenges for AOI. While
ation (EPIA), the market will reach $31 bil- there is no standardized production flow in
lion with a capex of $6 billion in 2010. With the solar-cell manufacturing industry, a typ-
oversupply of capacity, solar cell manufactur- ical fabrication process contains opportuni-
ers are under pressure to find ways to increase ties for AOI to aid in error detection and pro-
efficiency, improve quality, and reduce costs. cess control (see Fig. 2).
Automated optical inspection (AOI) sys- An incoming wafer inspection system
tems are growing rapidly as the solar indus- looks at the surface conditions and geometry
try is reaching the stage of maturity in which of the wafer, checking for distortion, chips,
machine vision undergoes mass adoption on saw marks, contamination, and obvious
the production floor. A properly designed cracks. It also looks for micro-cracks, which
AOI can help meet these goals. At some pro- are especially problematic for solar cell pro-
duction stages, however, AOI system design- duction because they can result in a wafer
ers have unique challenges to overcome. shattering during the following steps. Shat-
Silicon-wafer-based solar cells currently tering not only results in yield reduction but
represent 82% of the PV market, according to can shut down the production line for the
EPIA. The typical solar cell has a structure time needed to remove the debris.
(see Fig. 1) that needs a vari- Other AOI opportunities
ety of processing techniques Xing-Fei He is senior include inspecting textur-
to fabricate, including sur- product manager of ization, antireflective coat-
face texturization and the ings, printed metallization
linescan and TDI product
silk-screening of conductor layers, and a final operational
lines at DALSA,
layers for connection of the inspection of the finished
cell into a panel assembly. Waterloo, ON, Canada cells that takes advantage of
Compared with tradi- (www.dalsa.com). the solar cell’s electro-optical
tional semiconductor elec- characteristics.

14 Februar y 2010  Vision Systems Design www.vision-s y stems .com

Visible imaging • Texturization control • Sheet resistance measurement
Conventional AOI systems using visible light • Etching depth measurement

can handle a number of these inspection pro- Wafer inspection

• Surface and geometry 4
cesses. Either linescan or area-scan cameras con RC
• Micro-cracks Si iride A
l
can be used depending on the system design. • Warpage n i t
l
• Saw grooves and steps mova 2
In the geometry and surface inspection s re e
• Thickness las
n tsidting
stages, for instance, visible light and a mono- • TTV 3 p-g Froprin
r ) 5
chrome linescan or TDI camera will suffice. itte on ide ads
Emiffusi cks (p
1 2 d Ba ting
These types of cameras allow the inspection ng pri
n 5
chi ion de g t)
to occur in line, as the wafer moves along the a l etrizat s i
ck in c
i c
em ext
u Ba rint onta
production flow. The size and type of defects Ch nd t fi r ing pfull c 5 Print inspection (FS and BS)
a - (
6 Co Additional off-line equipment
for which the system is checking require cam- • Sheet resistance calibration
ion
eras with resolution in the range of 2k, 4k, or 6 ficat ting Cell-classification • Contact and line resistance
s s i sor
8k pixels per line with pixels 7–14 µm in size. Claand (FS and BS) • Optical microscopes
• Microbalances
The linescan or TDI rate needs to be fast • Ellipsometer
enough to avoid creating a bottleneck in the 1002VSDfea1F2
FIGURE 2. A solar cell production line provides many opportunities for placement of AOI sys-
production flow—typically in the 20–100- tems, some of which have requirements unique to solar cells. (Courtesy of ICOS, a division of
kHz range depending on the throughput. In KLA-Tencor)
addition, antiblooming or autoexposure fea-
tures are desirable to avoid image saturation. wavelengths still experience reflection. The linescan cameras allow inspection of moving
Inspection of the antireflection coatings exact hue and color saturation of the reflected wafers. Color cameras with 4k- to 8k-pixel res-
can also use visible light, but color camera light depends on the coating thickness. olution and 20–40-kHz scan rates are typical.
capability is required. Under white light, the Consequently, using the blue channel on The metallization print process is another
antireflective coating on the solar cell will a color linescan camera allows detection of opportunity for AOI to identify defects and
appear blue because longer wavelengths defects as well as process-control measure- provide process-control measurements using
transmit more efficiently while the shorter ment. As with the wafer inspection system, conventional cameras and techniques. Two-
camera systems can be used, providing front-
side and backside inspection simultaneously.
Ag
Pd Near-infrared imaging using backlight
Ti Antireflection layer
SiO2 One of the most demanding AOI challenges
Si
is during the incoming inspection stage:
checking for micro-cracks that can occur
during crystal growth and wafer sawing. This
-e n+
n is a critical checkpoint before beginning the
cell processing, both to ensure yield and to
reduce the occurrence of production down-
time due to a wafer shattering.
However, the micro-cracks can be too
small—less than 5 µm wide—to be seen
during a typical surface inspection. Further-
more, cracks are difficult to detect using sur-
P
face illumination because there is almost no
contrast between the reflections of the crack
+Hole p+ and the surrounding silicon.
Instead, this inspection requires the use of
Aluminum
backlighting. Because the wafer is only 150–
200 µm thick and silicon is semitransparent
FIGURE 1. Silicon-wafer-based solar cell structures, typically sized at 156 × 156 mm or 210 ×
1002VSDfea1F1
210 mm with a thickness of 150–200 µm, include antireflection coating, surface texture that in the near-infrared (NIR) spectrum, LED
helps recapture reflected light, and frontside and backside metallization layers as electrodes. backlighting using 850–950-nm wavelengths

www.vision-s y stems .com Vision Systems Design   Februar y 2010 15

 industrial automation products

will provide enough illumination to obtain

useful images.
Under these backlighting conditions, a
crack will scatter light and create a dark line
against a light background that is readily
detectable. Sensors need a resolution of 2–8k
pixels with 7–14-µm pixel size to be able to
detect these defects.
The challenge is that CCD image sensors
lose quantum efficiency (QE) in NIR wave-
lengths, resulting in a relatively weak signal. FIGURE 4. Micro-cracks in polysilicon wafers can be captured using NIR backlighting and TDI
Moreover, camera systems vary consider- cameras with sufficient sensitivity at those wavelengths. In the IS360 Wafer Inspector, pre- (left)
ably in their NIR sensitivity. Some cameras and post-processed (right) images of micro-cracks are analyzed and the wafers sorted based on
their dimensions using proprietary algorithms. (Photo courtesy of Meikle Automation)
exhibit as much as 30–40% QE at 900 nm
while others can be much lower. Therefore,
sensor and camera selection are critical in Increasing the NIR backlight intensity is line sensor with 100x of sensitivity. For exam-
developing an AOI system for backlit micro- one option for reducing exposure time but is ple, DALSA’s Eclipse TDI has a QE of 32%
crack inspection. limited by the availability of appropriate LED at 900 nm and 96 TDI stages, which provide
System developers should also consider the light sources. A more practical approach is to a combined broadband responsivity of 1950
exposure time needed to obtain a usable image utilize a time-delay integration (TDI) sensor. DN/nj/cm2 at 0-dB gain.
under NIR backlight because this directly These sensors utilize multiple TDI stages (as Systems for performing backlight inspec-
affects the system’s production throughput. many as 100x) synchronized to the scan rate tions have begun to appear. For example,
The shorter the exposure time, the faster the to accumulate multiple exposures at each Meikle Automation developed the IS360
production line can run at this step. stage. The result is the equivalent of a single Wafer Inspector (see Fig. 3). As one of its
inspection steps, the IS360 detects and ana-
FIGURE 3. The Meikle Automation IS360
lyzes micro-cracks using proprietary software
Wafer Inspector (top) uses NIR backlight- and sorts wafers based on the dimension and
ing to detect and measure micro-cracks. area information of the defects (see Fig. 4).
The system incorporates DALSA Pira- These data also provide operators with feed-
nha and Eclipse linescan cameras, which back to improve process control on the man-
interface to National Instruments Camera
ufacturing floor.
Link and LVDS frame grabbers that are
installed in high-speed image-process-
ing PCs. The PCs run Meikle’s MASS Electroluminescence
material-handling and image-process- Another AOI opportunity unique to solar cell
ing software. Standard encoder-based manufacturing, along with backlit micro-
image acquisition along with continu-
crack wafer inspection, is final quality inspec-
ous light and speed control ensures con-
sistent image quality. Parallel processing
tion. This step takes advantage of the solar cell’s
is required to achieve one-wafer-per-sec- electro-optical characteristics, which allow the
ond throughput (bottom). cell to generate luminescence for imaging.
This step can pinpoint micro-cracks and
Geometry/ Micro-crack/ Resistivity/TTV Surface
chip-out saw cut tester(s) contamination other production-induced defects in the fin-
DALSA DALSA
ished cell/panel that can cause early failure
Linear DALSA
Piranha 3 Eclipse
red but might not be detectable via conventional
Piranha 2
LEDs electrical testing. Eliminating defective
cells at final inspection can ensure that solar
panels fabricated from the remaining cells
have a product lifetime exceeding 20 years.
Linear red Linear NIR Linear red The application of a current density of about
LEDs LEDs LEDs DALSA 40 mA/cm2 to a cell causes electrolumines-
Piranha 2
Wafer travel (one per second) cence (EL) and results in the emission of light

16
1002VSDfea1F3b
Februar y 2010  Vision Systems Design www.vision-s y stems .com
 industrial automation products

centered at 1.15 µm, corresponding to silicon’s use indium gallium arsenide (InGaAs) sensor
bandgap energy (1.1 eV). Any defects in the sil- technology, which has greater sensitivity at 1.1
icon will appear by inspecting the uniformity µm. Unfortunately, it is also expensive. Cam-
of the EL (see Fig. 5). Photoluminescence (PL) eras using InGaAs sensors cost from $20,000–
uses a similar method but with laser excitation $40,000, depending on the type of camera.
rather than an electrical current. Also, the resolutions of currently available
The EL process has the additional attri- sensors—1k for linescan or 640 × 512 pixels
bute that the amount of light a cell generates for area-scan—are not high enough to meet
for a given applied current can serve as a mea- the needs of these inspection applications.
sure of the solar cell’s conversion efficiency. Imaging an entire wafer with sufficient detail
would require the use of multiple cameras,
multiplying the already high system cost.
Automated optical inspection An alternative is to use a fluorescent pig-
ment coating on the silicon sensor. Infrared
(AOI) systems are growing
photons striking the coating would generate
rapidly as the solar industry is shorter-wavelength secondary photons, effec-
tively converting the IR to visible light that the
reaching the stage of maturity
silicon sensor can more readily detect. This
in which machine vision conversion is known as an anti-Stokes process
and it has a low conversion efficiency (0.2~2%),
undergoes mass adoption which results in limited improvement in sen-
on the production floor. A sitivity requirements. The coating also could
reduce the sensor’s modulation transfer func-
properly designed AOI can tion, thereby degrading sensor performance.
help meet these goals. While the industry seeks to find a better
solution to EL/PL inspection challenges, the
opportunities for AOI in solar cell manufac-
This means the final inspection step can not turing remain. Many of the inspection needs
only detect defects, it can help sort and grade can be met with fairly conventional AOI
finished cells by their output characteristics. system designs and allow a tradeoff among
The results can also assist in process control performance and cost to match different pro-
as well as in matching cells for compatibility duction line needs. To address solar cell pro-
in a solar panel assembly. duction’s unique AOI opportunities, however,
developers should seek effective ways of han-
Handling IR insensitivity dling NIR/IR imaging.
While AOI in the final inspection has the
potential to provide significant manufacturing
benefits, it also represents a significant chal- Company Info
lenge for which there is as yet no ideal solu-
tion. The problem is that conventional silicon DALSA,Waterloo, ON, Canada
imaging sensors have extremely poor sensitiv- www.dalsa.com

ity beyond 1 µm, with a QE of almost zero. European Photovoltaic

Industry Association
As a result, the inspection requires long Brussels, Belgium
exposure times—on the order of seconds— www.epia.org
to obtain usable images. Right now, area-scan ICOS, a division of KLA-Tencor
cameras are being used in EL because of the Leuven, Belgium
www.icos.be
long exposure time required and are creating
Meikle Automation
a bottleneck in production. Kitchener, ON, Canada
The image sensor industry is actively seek- www.meiklesolar.com
ing a better alternative. One approach is to

www.vision-s y stems .com Vision Systems Design   Februar y 2010 17