[Enthalpy]

JPL's document considers that UV abstracts haloatoms selectively from alkanes. Data gathered from MPI's site, which I grouped on the joined diagrams, supports this notion for Cl, Br and I, and looks interesting for synthesis.

Alkanes absorb only the shorter waves, in constrast to alkenes.

• More carbons shift the absorption to lower energy: the cross section increases quicker below the fuzzy energy threshold.
• Primary and secondary carbons (cyclopentane, cyclohexane) show little difference, but strain (cyclopropane) reduces the energy threshold.
• Tertiary carbons (isobutane) reduce the threshold, quaternary (neopentane) more so. Crowding may act more than any influence on the tertiary C-H.
• If UV breaks a C-C rather than a C-H at quaternary carbons, it would suggest a way to synthesize bigger crowded molecules, to be extracted in real time from the reaction zone.
  
• All is over at 200nm for these alkanes, opening the way to selective haloatoms abstraction.

The reaction byproducts absorb wavelengths closer to haloalkanes, and some dilution gases are transparent:

• Dihalo molecules absorb little

at wavelengths useful for the corresponding haloalkanes.

• The hydrogen halides absorb as much as the corresponding haloalkanes and at similar wavelengths, so they should better be removed.
• N₂ and H₂ absorb very little above 100nm. I've no data for noble gases.
  
• Wavelength windows suggest to replace an I by Br at alkanes with a single photon, or a Cl or Br by H, and maybe a Br by Cl if the chain reaction with surrounding Cl₂ can be effectively quenched.
  

Fluoroalkanes absorb UV where alkanes do.

• No selective abstraction of F is expected.
• Only for fluor, more haloatoms reduce the absorption and increase the threshold.
• Maybe UV abstracts H rather then F. This would suggest a way to synthesize bigger, cyclic or polycyclic fluorocarbons.
  

Isolated Br and I on alkanes, less so Cl, absorb UV where alkanes don't.

• Br and I, less so Cl, absorb light that semiconductors can produce and silica transmit.
• Abstracting Br but not Cl (nor F) looks easy, I but not Br nor Cl also.
  
• Abstracting Br but not I looks feasible on the diagram, at AlN or KrCl wavelength, but with small selectivity. Abstracting Cl but not Br seems difficult.
• Vicinal carbons act little. They increase the energy threshold with Cl but ease it with I. UV could remove I only from tertiary C with some selectivity, to replace it with H, Br, maybe Cl.
  
• Vicinal F or OH don't change bromoalkane's absorption, but two vicinal Br or Cl more than double the absorption of a single one and keep the threshold.

Geminal haloatoms increase the UV absorption, much more than vicinals do.

• They ease the absorption threshold of the heaviest haloatom, F less so than Cl does.
• If they're equally big, the absorption can increase faster than their number, especially for big atoms.
• Geminal Cl absorb UV where alkanes don't, including in the present range of semiconductors.
• Abstracting one of two geminal Cl but not a lone Br looks better, and one of two geminal Br but not a lone I seems easy on the diagram.

Radicals absorb much UV: thin lines on the "geminal haloatoms" diagram.

• -CH₂Cl, -CH₂Br, -CH₂I absorb UV several magnitudes stronger than the CH₂X₂ if the photon energy is chosen a bit weak.
  
• Outside its 216nm peak, CH₃ absorbs much less than the CH₂X.
  
• If this means that the remaining haloatom becomes more receptive, and not the lone electron, it's the efficient way to carbenes. The radical would quickly absorb a second photon, so the population with one haloatom less would be smaller than with two less, even with light less concentrated.
  

The abstraction of two haloatoms per photon, the other path to carbenes cited previously for CHBr₃, is mentioned in JPL's document for CCl₄ as well. I don't have this information for CHI₃, CBR₄, CI₄, CCl₂Br₂ and the like, including longer molecules.

Marc Schaefer, aka Enthalpy

[Enthalpy]

Here are some examples of synthesis steps which I naively imagine could be accomplished by two absorbed photons that abstract each a haloatom from the same molecule.

Small fused cycles (sketch) seem interesting candidates, as the atoms to bind are already near to an other, and the remaining hydrogens are not in the way. Especially so on a flat drawing that conceals some nasty details...

Closing cycles (sketch) meets more hurdles. The example seeks small cycles, where carbons are not too wide apart, but the CH₂ may need to re-orient. Here the central quaternary carbon is to help avoid hydrogen jumps and the formation of double bonds.

Two successive wavelengths abstract all iodines, then all bromines. It must be possible to have untouched chlorines, make the couplings at iodines and bromines, then convert the chlorines to iodines for further couplings - or use the chlorines a classical way.

Marc Schaefer, aka Enthalpy

[discodermolide]

You really need to do a SciFinder search on some of these molecules. I don't have access currently, but may have in a couple of months.

[Enthalpy]

Existing lasers, including research setups and adaptations, permit to experiment the idea early - while pulse energy, mean power, wavelength for industrial production will need special lasers. Ask an optician for better advice.

I gave examples for exciplex lasers. They exist and deliver power at interesting wavelengths directly, but the ones that process semiconductors take a room and M$ for 60W output in long pulses. Other exciplex designs may fit better.

Solid state lasers seem more adequate. They are compact, can provide short pulses in modelock operation, but their output wavelength needs a frequency multiplier that wastes much power, and the oscillators appear to give few watts at most, so several amplifier crystals (typically 6mm*3mm*3mm plus cooling) would follow to bring the desired pulse energy and mean power.

Among them, Modelocked Ti:sapphire lasers offer pulses as short as desired: <10fs to >1ps.
http://www.rp-photonics.com/titanium_sapphire_lasers.html
They typically provide mean 1W, and 800nm needs a tripler (to target iodine) or two doublers (bromine). The pulse energy, like 10nJ, isn't brilliant neither. Worse, their pump is an already inefficient pulsed laser.

The best fit among today's zoo seems to be modelocked vanadate lasers pumped by laser diodes at 808nm.
http://www.rp-photonics.com/vanadate_lasers.html
Diode modules exist for 800W and more, about 1/2 efficient, and the vanadates can convert 1/3 of it. Nd:Vanadate outputs 1060nm, so F*3 serves little, F*4 targets iodine well, and F*5 bromine. Nd:YVO₄ is more usual, but Nd:GdVO₄ is more powerful, and the uncommon Nd:LuVO₄ possibly better.

Nd:YVO₄ pumped by a laser diode and followed by a KTP crystal doubler make green laser pointers. The power doesn't correspond, but their compact rugged design resembles what we need - just bigger... The attached diagram is gratefully pinched at Wiki
http://en.wikipedia.org/wiki/Neodymium-doped_yttrium_orthovanadate

This Nd:GdVO₄ produces 5W at 1064nm as 107MHz 47nJ 6.5ps from 50W at 808nm

"High-power picosecond Nd:GdVO4 laser mode locked by SHG in periodically poled stoichiometric lithium tantalate"
by Iliev H1, Buchvarov I, Kurimura S, Petrov V.

while that Nd:YVO₄ produces 2.5W at 1064nm as 4MHz 640nJ 17ps from 27W at 808nm

"Stable mode-locked operation of a low repetition rate diode-pumped Nd : GdVO4 laser
by combining quadratic polarisation switching and a semiconductor saturable absorber mirror"
by Christoph Gerhard, Frédéric Druon, Patrick Georges, Vincent Couderc, Philippe Leproux

----------

Here is a numerical example for two-photons coupling experiments using vanadate lasers.

With 17ps pulses, a molecule free flight time of 40ps limits unwanted reactions. This corresponds to 300K and 3.6 bar for air, say 2.5 bar for bigger radicals. -CH₂Br absorbs F*5 = 213nm with 1.5*10^-18cm² section (and BrC₂H₄Br 2*10^-18cm²), or 63% after 110µm.

213nm light concentrated on r=5µm diverge by r+1.4µm at 55µm before and after the focal plane. The irradiated 12400µm³ contain 7.5*10¹¹ molecules, and two 5.8eV photons for each need 63% of 2.2µJ, the output of 35 amplifier crystals after 10% upconversion. Cumulated 9W treat 0.44mol in 24h, that's 30g housane.

Marc Schaefer, aka Enthalpy

[Enthalpy]

The spectra of reply #15 want light around 260nm (I), 210nm (Br), ideally 170nm (Cl) - as short pulses when the goal is cross or internal coupling. Existing lasers fit imperfectly:

• Exciplexes make directly very short wavelengths. Chip fabrication emphasized optical quality and smooth operation, chemical synthesis would optimize pulses and efficiency, more like LANL and NRL (4kJ 4ns, supposedly huge) did with KrF http://en.wikipedia.org/wiki/Krypton_fluoride_laser
• Vanadate and other colour-centre lasers need a lossy frequency multiplier, lossy optical pumping, and they get huge over at few 10W. Fine as a proof-of-concept only.

It seems that semiconductor lasers would bring the efficiency and power, but chemical production wants characteristics that must be developed.

• >80W per laser diode exists, >50% efficient, assembled in 10kW arrays, but the GaAs based diodes emit around 800nm as a continuous wave with poorly coherent superradiant mode and independently from an other.
• 405nm GaN based laser diodes are produced industrially with 0.3W output, and 245nm or 265nm Al_xGa_1-xN based Led (not laser) emitting 100µW can be bought. 210nm AlN-based Led and 245nm GaAlN-based laser diodes have been demonstrated.
  
• Modelocking a GaN-based laser diode was demonstrated, with 4ps pulses of mean 72mW at 422nm repeating at 28GHz, or with 15ps at 406nm repeating at 0.84GHz to produce mean 260mW in 600µm*5µm active area, and more.
• Short pulses amplified on-chip up to mean >10W were demonstrated on a usual material, probably at 1550nm.

So 10ps pulses of mean 10kW at 210nm will still take a few years until mass production... Fortunately, vanadate lasers permit to experiment meanwhile.

----------

What could be the development directions?

The smallest semiconductor wavelength is 210nm with AlN (fits Br), but lasers need an additional material with a wider gap and smaller index. Al_xGa_1-xN makes 245nm - or 260nm (for I) - with additional layers of AlN. Pure BN has the wrong band structure with a smaller gap. Though, common speculation backed by other compounds wants to grow thin layers of B_xAl_1-xN, where <20% B keep a direct gap, wider than AlN itself which could then lase.

No semiconductor produces 160-180nm for Cl, so Al_xGa_1-xN's frequency must be doubled or Ga_xIn_1-xN's tripled. This can be an evolution path since In_xGa_1-xN is present technology and it also targets Br and I once doubled, the fraction x adjusting the wavelength.

If the chip's facets fail under pulse power, could light exit through the big face over a long length first to a thicker material? Or the material's ends have a long thickness taper like the cigar antenna? This would also linder reflections at an amplifier.

Strong coherent light pulse are usually created by a small oscillator and then amplified. One paper describes a flare just 5µm wide for mean 0.26W; several paths fed commonly and reconnected later must increase the power. Chemical production would need several amplifier chips; those illuminating the same spot must share the oscillator.

Marc Schaefer, aka Enthalpy

[Enthalpy]

Devery et al published in 2016 a de-bromination by light
http://pubs.acs.org/doi/abs/10.1021/acscatal.6b01914
similar to what I had described here
http://www.chemicalforums.com/index.php?topic=77307.msg283221#msg283221
but using visible light (blue LED) and a catalyst. Good yields even for less easy targets like cyclopropane and benzene.

[wildfyr]

Why is it that every time you see some really fancy chemistry going on it involves something ridiculous like iridium or ruthenium?