LED lighting’s ‘best years’ will be 2013 to 2017
The value of the market for high-brightness LEDs used in general
lighting applications will likely peak in 2017 before dropping away as
the cost of the technology falls sharply.
That’s according to IMS Research market analyst Jamie Fox, who told
delegates at a euroLED “market kicker” conference hosted by the
Birmingham Science Park - Aston, UK, that the “best years” for LED vendors selling into the general lighting market would be 2013-2017.
Over that period, market penetration is expected to grow massively
from its current unit level of only a couple of per cent, to 10% in 2014
and 25% by 2017. And although the cost of LEDs is likely to fall
quickly over the next five years, that won’t be enough to hold back the
value of the market for LED chips in general lighting until after 2017,
Fox indicated.
But from 2018 onwards, saturation will take over and the falling
prices of LEDs, combined with reduced rates of lamp replacement thanks
to the long lifetime of the technology, will see the market for LEDs
used in general lighting start to shrink.
The impact of that wave of adoption will also be seen in the wider
lighting market. The emergence of LEDs will initially spark a period of
significant growth, from around $85 billion now to $120 billion by 2017.
But as lower costs and saturation kick in, that figure may then shrink
back to $105 billion by the end of the decade, Fox thinks.
Multiple pathways to LED cost reduction
One of the key drivers of that cost reduction could be a move to gallium
nitride LED chip production on silicon wafers instead of either
sapphire or silicon carbide. The novel approach is being pursued by a
number of the major players in the industry today – albeit with varying
degrees of success.
Speaking at the same event, European marketing director at chip
producer Bridgelux Tom van den Bussche said that the switch to a silicon
platform would cut costs by 75% - partly through the lower price of
silicon, but also through enabling a larger (8-inch) platform that can
take advantage of obsolete silicon wafer manufacturing fabs whose owners
have now largely moved on to 12-inch silicon.
Van den Bussche added that Bridgelux, fresh from another
$20 million injection of venture funding
that brings its total capital raised thus far to more than
$200 million, could be in production with GaN-on-silicon LEDs by the end
of this year.
The company’s latest investment came from China’s Kaistar Lighting, a
joint venture between Taiwan-based LED maker Epistar and Shenzhen’s
Kaifa Technology. Bridgelux is expected to take advantage of an 8-inch
silicon facility in Japan for back-end production of the chips
fabricated on the larger wafer format.
According to van den Bussche, Bridgelux has recently shown
silicon-based LEDs delivering an efficacy of 160 lm/W at the cool-white
color temperature of 4700 K, as well as 125 lm/W at the much warmer
3000 K color – indicating a performance that can rival
today’s state-of-the-art devices.
One of the key questions now to be answered is whether the company can
replicate that performance level in volume and with a high-yielding
process.
If that can be done, then a silicon-like manufacturing approach ought
to deliver serious cost reduction, because the current method – largely
restricted to 4-inch wafers – requires a number of inefficient and
time-consuming stages, for example the laser lift-off step that is used
to separate the GaN wafer from its sapphire host.
Lattice mismatch
Skeptics of the GaN-on-silicon-based approach point to the inherent
lattice mismatch between the two crystals, and the need to produce a
complex buffer layer between them to overcome the stresses and strains
encountered in the high-temperature MOCVD deposition process, as a key
weakness. They suggest that sapphire will remain the material platform
of choice once scaling difficulties are overcome and the current
overcapacity is absorbed. That overcapacity is currently
holding back a planned transition to 6-inch sapphire and SiC LED wafer production at some of the industry’s leading manufacturers.
But other factors are at work, too. Fox says that the vast numbers of
MOCVD reactors, used to deposit GaN layers onto the wafers, shipped to
Chinese LED makers in recent years are now “mostly” switched on and in
production, even if the industry’s overall capacity utilization remains
low at the moment.
The analyst says that this shift of the LED manufacturing center of
gravity will continue to drive some future cost reduction, even though
leading-edge manufacturers do retain a significant performance advantage
at this point.
But with some lower-priced LED lamps already being marketed at a
price of around $10 in the US and China, the signs are that this may be
happening faster than Fox had predicted in some regions. Average prices
do remain much higher, however, and are not expected to break through
the $10 barrier until 2014, before dropping to $5 in 2017.
Quality of light and ‘useful lumens’
While cost is perhaps the leading factor in driving adoption, the
quality of the light produced by LED lamps is also critical, and at
present only the very best chips are really suitable for general
lighting applications. Worryingly, a survey carried out last year by the
lighting firm Havells-Sylvania found that 85% of consumers who had
bought LED lamps were unhappy with the quality of light that they
produced, and would not purchase them again.
Speaking at the same euroLED event, James Hooker, R&D manager for
LED and special lamps at the company’s Belgian operation, said that
those findings “shocked us to the core”, with the entire industry
desperate to avoid a repeat of the light quality issues that have
plagued compact fluorescent (CFL) lamps.
Predictably, it is not the technology itself that is causing this
problem. Hooker said that the market had become flooded with inferior
LED light sources, with some replacements for 50 W halogens found to be
delivering only the equivalent amount of light as a 25 W bulb.
In fact, says Hooker, precisely zero LED lamps have so far emerged as
genuine replacements the 50 W halogen lamp – aside from a recent
introduction from his own company (the Sylvania RefLED ES50 GU10), that
is. One problem is that standards and regulations have yet to really
catch up with the fast-moving industry. “LEDs have to match the lumen
flux of the things that are being replaced,” he said, while pointing out
that simply measuring the total lumen output of a product is
insufficient because it does not consider the optical quality of a lamp
design – in other words the “useful lumens” that it emits.
Encouragingly, the European Commission (EC) is now planning to
recognize a new “useful lumen” metric officially, something that should
help to ensure that LED lamps do meet key quality criteria and stop
disappointing consumers. The metric is based on the lumens emitted in a
directional cone of space, or beam angle, rather than the total lumens
figure that it is customary to publish for lamps sold in the US.
By specifying the lumens emitted within a 90° cone, says Hooker, it is possible to calculate an efficacy figure in “
useful
lumens per watt” for reflector lamps, which will be employed for
judging whether LED-based lamps really are suitable replacements for
halogens - and in which applications they should be used. The new
Havells-Sylvania lamp delivers 350 lumens in total, of which 300 lumens
are contained within the 90° cone, and is said to be the only such
replacement to meet IEC size requirements.
But Hooker also warned that close market surveillance would be
essential to ensure compliance with future regulations, and to stop
rogue lamps from being sold within the EU. “Without adequate controls,
inferior products may remain on sale and jeopardize consumer
satisfaction and rate of LED growth,” he warned. Finding and testing LED
lamps looks like being critical to the success of this major emerging
market, and a real growth area for photonics specialists.
• The full euroLED 2012 conference will take place 13-14 June at
Birmingham's National Exhibition Centre (NEC). To register, visit the event web site.
