This paper demonstrates quite convincingly (because the descriptions are very detailed) that the ancient Edison light bulb (i.e. with light emitted by incandescent carbon) can be revived by using modern technologies in a form that can be better than the current LED lamps.
The first incandescent lamps had 2 disadvantages, short lifetime and low efficiency.
The evaporation and oxidation of the incandescent carbon can be avoided by a structure made of carbon nanotubes that works in argon. The claimed lifetime (based on accelerated aging tests) is as good as for LEDs.
The low efficiency is solved by using a lossless optical filter (i.e. one that transmits the visible light and reflects back the infrared light) to prevent the heat loss from the lamp by radiation other than the useful light output.
This method can reach the maximum energy efficiency determined by the human vision characteristics, unlike the current LED lamps, which are unavoidably limited to a value lower than that by the losses caused by the fluorescent light conversion.
The design of the lossless optical filter is a difficult optimization problem due to the large number of parameters. Their solution, which is completely described in the paper, is claimed to have been found by a machine-learning algorithm.
To obtain the efficiency gains, the device needs to reflect most of the IR photons back onto the hot emitter, which would be a filament in a regular incandescent lamp. But the filament's cross section is small compared to lamp's. If an IR photon must reflect 1000 times before hitting the filament (very optimistic!), and the reflector reflects 99.9% of the IR photons back inside, only about 37% of the IR photons will hit the filament. And at that they only have about 5-10% chance to be re-emitted as visible light, depending on the temperature.
So, for the economy of the photons to work, a large black glowing ball would be needed; any photons inside it would be recycled, and it would readily absorb most of the photons returned to its outer surface by the miraculous IR reflector.
Thermal losses would still be pretty noticeable because argon can't have too low pressure, else it won't prevent the evaporation of the emitter. The device would run pretty hot, which is a source of inefficiency.
I don't know whether it would be more efficient than the large and very hot arc lamps [1] which illuminate stadiums or even large shopping centers, with electrical efficiency above 50%.
You are right, but the authors are not fools, so they do not use a filament.
The incandescent part has the form of a rectangular tape, with the front side made of black carbon (nanotubes deposited by CVD) and the back side made of white boron nitride, so that most of the radiation is directed towards the front, i.e. towards the exit window, to minimize the heating of the enclosure (which is made of white alumina, to minimize the absorption of the radiation that does not go out through the window).
If the alumina enclosure gets hot, the efficiency can be further increased by reducing the heat loss with an extra surrounding layer of thermal insulator (which could be provided by a plastic or ceramic socket).
Increasing the efficiency up to the ultimate limit imposed by the human vision may become too expensive to be worthwhile, but there is no other obstacle except cost, unlike for the current LED lamps with fluorescent phosphors.
I have never heard of any arc lamp that has any chance of even remotely approaching a 50% efficiency, when counting only the useful light.
Otherwise, even an archaic incandescent lamp can approach an 100% electrical-to-light efficiency, but almost all of the light output is in the form of useless infrared light.
Besides using an incandescent emitter with a lossless optical filter, the only other way to approach the ultimate efficiency for lighting (with a good color rendering) would be with a lamp including at least 4 LEDs or laser diodes of different colors.
The latter solution is much more expensive, requires electronics for compensating the change in time of the color temperature and it is currently limited by the lack of semiconductor emitters with high efficiency for yellow and green light.
What makes you think this technology is inherently capable of scaling it's efficiency better than the phosphors that shift the frequencies in LED lights?
The frequency shift that you mention reduces the efficiency by a factor equal with the ratio of the frequencies.
When making green light out of blue light the efficiency cannot exceed about 83% and when making red light out of blue light the efficiency cannot exceed about 70%.
Real efficiencies of the fluorescent conversion are much less.
Moreover, there is no hard limit for improving the efficiency of an incandescent light emitter with better infrared filters and better thermal insulation.
Therefore, when the comparison is made for a great enough frequency bandwidth, the efficiency of an incandescent light emitter coupled with a lossless optical filter should be able to exceed even the efficiency of a LED without fluorescent phosphor.
The best colored LEDs can reach an efficiency of 70%, but only at very low output power densities. When the output power density is increased, to reduce the cost of a LED lamp by using a smaller LED, the efficiency drops a lot, to levels that should not be hard to exceed with a filtered incandescent lamp.
The LEDs and the laser diodes are unbeatable in efficiency when it is desired to produce light within a narrow frequency band, which is not the case for lighting applications.
This seems similar to HIR technology, widely available on car headlamps:
“The HIR1 (9011) and HIR2 (9012) Halogen Infrared Reflective bulbs have micro-thin layers of electro-deposited coating on a specially-designed bulb. The coating reflects the filament's infrared (heat) back onto the filament, which results in a hotter filament that produces more light without increased power consumption, shorter bulb life, or extra heat production.“
The principle is the same, but here a much more sophisticated implementation is shown, which is able to achieve better performances than the LED lamps.
As far as I know, there isn't a law banning incandescent lamps. The law stated that such lighting needs to achieve a certain level of efficiency or greater, which in effect banned existing 1% efficient incandescent lights.
> It be great to have efficient incandescent light and put the failed promises of LED longevity behind us.
In my experience, the main failure mode of LED bulbs is the power supply -- they engineer every penny out of them even at the cost of reliability. Second, nearly all bulbs are dimmable, which complicates the power supply, because customers faced with dimmable vs non-dimmable will pick the dimmable one.
Another factor is if you read those claims about 20 year lifetimes and such, often there is a footnote saying something like "assuming it is on for 3 hours a day or less".
Finally, you are assuming that this new technology won't suffer the same fate as the LED bulbs. If they take off, they will be cost reduced at the expense of longevity. I saw the same thing happen in the era of floppy disks -- over time the price was reduced by a factor of 20 but reliability went down in parallel.
There seem to be two dominant pop commercial successes, those where brand means everything and cost is increased well beyond any utility to signal wealth, and brand AND reliability mean nothing (because like your mattress it'll be a few years later anyway) which is just a race to the bottom.
My heuristic is to just buy LEDs made by Phillips or GE, since invariably any generic or store-brand LED I've tried seems to lack rectification and/or filtering and stobes at 60 or 120 hz. I haven't tried any fancy/expensive "smart" programmable bulbs, but plain ones I do get seem pretty reasonably priced.
I've so far had good success with the ones Yuji sells; they are conceptually like buying a PC directly from Intel: Yuji makes LEDs (I'm not aware of them fabbing the semiconductor wafers in-house, nor synthesizing the phosphors in-house, but they develop them) that pride themselves in optical quality (using double and triple phosphor systems for CRI 95+ and CRI 98+, respectively) to justify their ~1$/W (electrical) price tag for strips of the lower grade (about double for the higher grade).
My main ceiling light is a now-discontinued remote phosphor 10W bulb, where the LED chip is covered by a loose plastic dome that is infused with the phosphor.
The scattering gives a really wide and even light cone to partially illuminate the ceiling for extra diffusion.
IIRC it was rated 20k hours probably due to the PSU, though I operate it uncontained for best (passive) cooling (and because I don't look at the socket it's hanging from anyways, lest I get an afterimage of the dome on my retina from staring at the lamp, so the non-fancy fixture doesn't bother me).
The PSU I got for some ultra-warn-white (iirc 1900K) strip (5m of 20W/m 24V PWM-able) from them (a knob on it does PWM dimming (iirc) just past the audible range)) was like 80 or 90 bucks, on top of the iirc ~110 bucks for the strip itself.
Given they don't seem to sell in normal retail stores, I'm not worried about brand inflation/"paying for the name", given their marketing is focused on spectrum measurements and color performance with the listed prices being "affordable" (1$/W @ 20ct/kWh means it costs as much to buy as the next 5000 hours of runtime will cost in electricity; I take "twice as much total operating cost than with literally free/gifted bulbs" as "affordable").
Undoubtedly it is not producing the same color spectrum as new, as phosphors fade over time, its blue spike is getting wider and taller, so it is always producing more and more blue light, slowly permanently blinding you while disrupting your circadian rhythms, increasing your chances of having diabetes and heart disease, and taking years off your life. But at least you saved a buck.
Only because the sun is much brighter than any LED, but when comparing the colors of the spectrum output, the sun's color spectrum is flat compared to LED, which has what is known as a "blue spike." Most of the light a high efficiency LED outputs is blue because all high efficiency LED are blue LED. Only phosphors allow it to output other colors. And the more the LED is used, the more the phosphors fade, the more the output spectrum shifts to blue. And there is more and more scientific backing showing the problems with LED becoming more concerning. Here is a summary of one scientific study.[1] There are myriad more, and more and more are being published.
LEDs last remarkably long if used correctly. You can make a light that will easily last thousands/dozens of thousands of hours. What doesn't last is a typical LED lightbulb for legacy sockets, because of the planned obsolescence/race to the bottom, and because it's a complete device (power supply, heatsink etc) which is far too powerful for its form-factor. Not sure what of that you expect to change.
> Dubai Lamps are special LED lamps designed by Philips for high efficiency and very long operational life. A simple design change reduces stress and excess heat that otherwise significantly lowers the mean time before failure of regular LED lamps
Carbon nanotubes are about as toxic as asbestos. So unless you're making that bulb for somewhere that's absurdly hard to change, it's probably a bad idea.
Asbestos isn't toxic at all unless the fibers are airborne.
Unless you disturb the nano tubes in the device in such a way that they become airborne, they're safe. Probably they're less dangerous than the mercury in the CFL tubes.
The amount of nanotubes in a light bulb filament is measured in milligrams. There is a bigger fire hazard from the lint accumulated in your bellybutton.
I bet you've just learned to associate the continuous spectrum from incandescent bulbs with the heat they also emit. Have you ever been in a space illuminated by sun tubes or skylights? Did that light feel fake and/or sterile?
Personally I’ve always disliked feeling the heat of an incandescent lamp. And I’m fairly certain that one of the advantages of electric lighting over gas was the _reduction_ in heat produced while lighting the room.
Speak for yourself, I don’t like candles for the same reason that I don’t like incandescent lights. Not to mention the smoke, the wax guttering and dripping everywhere, the fire hazards, etc, etc.
With a bulb made of alumina and having a window made of quartz on which thin films have been deposited, it would be more expensive than a traditional incandescent lamp, but that would be compensated by the high efficiency and long life.
If in mass production, I see nothing that would prevent it from being cheaper than a comparable LED lamp.
I am not familiar with the technology used to make the incandescent band made of carbon nanotubes and boron nitride ceramic, but everything else is made with standard equipment that is widely available.
I didn't mention this in my other comment because I thought the use of alumina could be avoided, but I suggest you look up the price of some optical sapphire (ie transparent alumina) before you get excited about that. Even for a tiny, flat circle you're looking at around ten bucks. A bulb would likely be harder.
The alumina used must not be transparent, it must be as white as possible for maximum efficiency, so standard cheap alumina ceramic should be OK.
The light exits the lamp through a flat window, which is made of quartz in the prototype, but in mass production it is likely that some kind of alumosilicate glass or borosilicate glass would be fine for it.
Could it be made with a plastic and not glass exterior? One minor benefit I have enjoyed about the transition to LEDs is that they are physically more sturdy and will survive being dropped.
The alumina cannot be replaced with plastic for 2 reasons.
It must be hermetic, to keep the argon inside and the air outside, for the lifetime of the bulb.
It contributes to the high efficiency, by reflecting both the visible and the infrared light and preventing thus the heat loss from the lamp.
Nonetheless, alumina is less fragile than ordinary glass.
It might be possible to replace the quartz window with some kind of glass window, for a lower cost, but not with any kind of plastic, because it must also be hermetic and it must be usable as a substrate for thin film deposition in vacuum and it should have a coefficient of thermal expansion matched to those of the multiple layers of the infrared-reflecting filter.
The key innovation here is the IR-selective-reflection bulb coating. As such:
>Fabrication of the VTIRF window
>The quartz substrate with a size of 25 mm by 50 mm is cleaned by bath ultrasonication using acetone, ethanol, and deionizing water for 10 min. The Al2O3 and HfO2 layers are deposited by atomic layer deposition (R200, PICOSUN) at 255°C using trimethylaluminum (Strem Chemicals) and tetrakis(dimethylamido)hafnium(IV) (Strem Chemicals) as precursors, respectively. The oxidizing agent is H2O. The precursor carrier gas is N2 (99.999% purity, Shanghai Weichuang Gas). The Ag-Ge alloy and SiO2 layers are deposited by electron beam evaporation (PVD 75, Kurt J. Lesker) in vacuum.
I'm no expert on how LEDs are made, but this coating certainly doesn't assemble itself.
All the layers are deposited successively on a larger glass panel using standard vapor deposition equipment, then the substrate is cut into the individual glass windows for the lamps.
The article mentions explicitly all the equipment that has been used in the fabrication of the prototype.
The fabrication of any semiconductor device, such as LEDs, includes many more such vapor deposition steps done in vacuum, using the same machines, but with much more stringent requirements for cleanliness and defect control.
Any defects in the deposition steps for such a lamp window will just reduce the efficiency of some lamps by a small fraction of a percent, while during LED fabrication they would have made unusable LEDs that would have been dumped.
The optical filter is critical to the design, and big curiosity to me. Normal dielectric filters have a large angle dependence. Of course and It’s solved by machine-learning, I am hopeful but skeptical. Need to dig further into this.
Is the efficiency of an incandescent lamp really a problem though? If your power comes from renewable sources then over its entire lifespan an incandescent bulb has to be way cleaner than an LED.
In general something that's a coil of wire in a glass bottle with the air sucked out that you could reasonably make in a blacksmith's forge is going to be less environmentally horrible than a hundred grams of plastic, fibreglass, epoxy, copper, tin, bismuth, gallium arsenide, phosphorus and so on that has to be put together in a multi-billion-dollar factory.
The difference in efficiency is too high to not matter.
In the past, I had used for the lighting of my house various combinations of 100-W incandescent lamps.
When better technologies have appeared, I have substituted them with other lamps having an 1500-lumen light output, which is slightly more than provided by the 100-W incandescent lamps.
With compact fluorescent lamps, I have used 22-W lamps, while during the last decade I have been using 13-W LED lamps (with 1500 lumen at a pleasant 4000 K color temperature, i.e. only slightly warm, not noticeably yellow, like the lamps with a color temperature of 3000 K or less).
Even the early prototype described in the research paper can achieve 1500 lumen at 8.7 W.
The difference between 100 W and 8.7 W is too large to ignore. In my case, both transitions, to CFL, then to LED, have been extremely visible in my monthly bills.
> Consider that electricity mostly comes from renewables now, it just doesn't matter what the efficiency is.
As far as I know we're not yet in a post-scarcity energy economy.
Think of your favorite topic of research. Now consider if you'd rather spend some of that renewable energy furthering that research, or on generating unwanted heat from a bulb and then moving that heat out of your living space.
> Consider that electricity mostly comes from renewables now
That isn't remotely true yet. And while renewables are much better for the environment than other energy sources, they're still a lot worse than not using that energy in the first place.
Considering that until 20 years ago (before switching to CFL) I had to change a few bulbs every year and now, after almost 10 years after switching to LED lamps I did not have to replace any of them yet, I believe that their effect on the environment is less than that of previous technologies, even when not counting the energy savings.
I have worked in semiconductor device manufacturing, so I am much more aware of the quantities of materials and energy that goes into them than most.
Except for special applications like solar panels, most semiconductor devices are very small (much smaller than their visible packages) and they form a very small fraction of a complete electronic device. In many cases a low-tech device and a high-tech device of similar size may have similar environmental impacts, determined more by the amount of copper and other metals and of plastic and glass, than by the types of semiconductor chips that might be included in them.
The only solution for the environmental impact of industrial manufacturing would be legislation that would make a list with chemical elements that are less abundant and/or more difficult to extract or to dump without environmental damage and which would mandate that they must be used in closed circle.
That means that for any industrial product that would contain some of the listed elements, the manufacturing company or another company contracted by them would have to design a process for recovering those elements from the finished product, and not only, like it is done now, the manufacturing process that converts the raw materials into the finished product.
To receive the license to sell any such product, the manufacturer should have to demonstrate that the recycling process exceeds some recovery efficiency mandated by law, e.g. 90%.
The selling price of all affected products should include the cost of the recycling process, not like the fake prices of today, when many people enjoy what look like cheap products, but that is only because someone else will have to pay in the future for the consequences of how they have been made.
Until such laws are enacted everywhere, any other measures just delay the undesirable outcome.
That's remarkable. I was afraid this was going to be a pure theory/simulation paper, but they built an incandescent light source that selectively recycles infrared photons back to the hot emitter, thereby dramatically improving the luminous efficacy. It's a white light source that produces a continuous spectrum in the visible range, like other incandescent light bulbs, but has efficiency comparable to LED lighting.
This uses carbon nanotubes, which can have asbestos-like effects on the lungs. Nanotubes are great for tons of applications, but especially those where safe disposal procedures are possible.
If these are a replacement for LEDs with better properties, these would be widely deployed in everything from lighting to devices. I wonder whether at that scale this presents a safety challenge, given the probability of breaking and the way people dispose things. The glass substrate does hold them in place, but what happens when the glass breaks in all the myriad ways it can in real life?
When there's articles about Gallium arsenide (GaAs) semiconductors, you get the same comments. "Arsenic is a poison!"
Yes, we know. Are you cracking the chips open an licking the semiconductor wafers inside for their flavour? No? Then you have nothing to worry about.
This is the same category of thing. There's micrograms of carbon nanotubes in these lights as a thin-film inside a very strong ceramic casing. Unless you're smashing the thing open and then scraping the surface so you can snort it with a straw like it's cocaine, you'll be fine.
I am just excited as others when reading the TLDR comment.
But this warning made me think of do I still want it?
Is there anything in the paper explaining about the safety of nano carbon they use?
The MIT article is cited in this research paper and it is likely that it has been the one which inspired the start of this research.
However the progress made from the very primitive device described by MIT and the high-performance prototype made by the Chinese team has been huge.
The Chinese have used completely other materials, which enabled also a long lifetime, not only a better efficiency. There are many new constructive details which contribute to an increased efficiency. The design methodology used by the Chinese for the lossless optical filter is far more advanced and it enabled a much better filter with much less layers.
The MIT article showed an interesting idea, for which it was not clear if it can be successful in practice, while this shows a real device that needs only minor changes in order to become suitable for mass production (e.g. the replacement of HfO2 with TiO2 in the window coating and the replacement of the O-ring used in the prototype with hermetic glass-alumina bonding).
This device could be easily converted into a commercial product in a couple of years.
A yardstick that I like to use for the exponentially increasing pace of technological advancement is the transition from one lighting technology to the next.
Wood fires were the only lighting we could muster for a few hundred thousand years.
Wick-based lighting like candles and oil lamps for a few thousand.
Incandescent electric lighting started about 150 years ago.
Fluorescent lighting about 70 years ago.
CFL lighting about 30 years ago.
LED lighting about 15 years ago.
OLED lighting about 7 years ago.
There's talk now of direct EL nano-rod lighting, this photon recycling thing, and more.
We'll reach a point where new lighting technology is developed faster than the previous lights reach their end-of-life!
I think you’re looking for a lux meter. Precision is far rougher than single photons, but good enough for any human-scale use. It’s what you’d use for things like calibrating indoor lighting, quantifying the difference between a sunny day and moonlit night, or testing flashlights. Last I checked a good basic one was under $40.
I suppose there would be a rough conversion between that value and photon count (at some wavelength range). I just imagine the error bars to be very wide.
Not really, it depends from the application.
Even something like 20 years ago the CCDs available at the time were capable of astrometric measure of star magnitudes with 1/100 of magnitude precision using some control stars.
Everything in our universe is made up of waves. Does it really matter whether it's an oscillation in one field or another? Photons are discrete systems, can you clearly differentiate why anything else in the universe is "physical"?
How so? My understand is that a photon is a discrete packet of energy. Why can't I count those? If there is an excitation in the electromagnetic field, and then there is not, and then there is one - why are those not separate, countable photons?
Would you be so gracious as to answer my questions, instead of just repeating what you previously said? I'll copy and paste:
> Why can't I count those? If there is an excitation in the electromagnetic field, and then there is not, and then there is one - why are those not separate, countable photons?
You can call and count things however you want. If you have a steady light source I'll go one better and count it for you - one, because electromagnetic radiation is not somehow switching on and off automatically.
That would be like you turning your stereo on and off and declaring the sound to be sound particles. You can do it, but you're ignoring physical reality.
Computers work with alternating voltages, but that doesn't make them 'electricity packets' or 'electricity particles'.
Electromagnetic radiation is going to be a magnitude at a certain frequency over time.
It's really more "by physics." You can pick how you want to run a tungsten filament and end up with either a long life bulb, or a high(er) efficiency bulb, but not both. The ~1000 hour goal was, at the time, an attempt to optimize the total cost of lighting - bulb and power combined. As your bulb lasts longer, you get rather fewer lumens per watt.
A typical 1000 hour rated 60W incandescent (an old one, not the new halogen "efficient" ones - 28% less power for 28% less light) will put out around 850 lumens. I have a 20k hour rated bulb with the same output - but it pulls 75W. Over the lifetime of the bulb, that increased power use is quite a bit more expensive than replacing bulbs would be - but they're designed for places that are harder to access, or that are high vibration (typically, long life incandescents also have more supports, so they work better in higher vibration environments).
There's no magical tungsten filament, 60W, 850 lumen bulb that will last 10k+ hours. It's simply not possible within the normal constraints.
Running the filaments hotter makes the bulb more efficient but also shortens the lifespan due to faster filament evaporation. You could get very long life by under-powering the bulb but then the low efficiency of incandescent lighting would go lower. That's the other remarkable thing about this paper. They use a ceramic instead of tungsten as the hot emitter, and calculate that it should have a very long lifetime.
This paper demonstrates quite convincingly (because the descriptions are very detailed) that the ancient Edison light bulb (i.e. with light emitted by incandescent carbon) can be revived by using modern technologies in a form that can be better than the current LED lamps.
The first incandescent lamps had 2 disadvantages, short lifetime and low efficiency.
The evaporation and oxidation of the incandescent carbon can be avoided by a structure made of carbon nanotubes that works in argon. The claimed lifetime (based on accelerated aging tests) is as good as for LEDs.
The low efficiency is solved by using a lossless optical filter (i.e. one that transmits the visible light and reflects back the infrared light) to prevent the heat loss from the lamp by radiation other than the useful light output.
This method can reach the maximum energy efficiency determined by the human vision characteristics, unlike the current LED lamps, which are unavoidably limited to a value lower than that by the losses caused by the fluorescent light conversion.
The design of the lossless optical filter is a difficult optimization problem due to the large number of parameters. Their solution, which is completely described in the paper, is claimed to have been found by a machine-learning algorithm.