Tag Archives: diffusion

Einstein’s theory of diffusion in liquids, and my extension.

In 1905 and 1908, Einstein developed two formulations for the diffusion of a small particle in a liquid. As a side-benefit of the first derivation, he demonstrated the visible existence of molecules, a remarkable piece of work. In the second formulation, he derived the same result using non-equilibrium thermodynamics, something he seems to have developed on the spot. I’ll give a brief version of the second derivation, and will then I’ll show off my own extension. It’s one of my proudest intellectual achievements.

But first a little background to the problem. In 1827, a plant biologist, Robert Brown examined pollen under a microscope and noticed that it moved in a jerky manner. He gave this “Brownian motion” the obvious explanation: that the pollen was alive and swimming. Later, it was observed that the pollen moved faster in acetone. The obvious explanation: pollen doesn’t like acetone, and thus swims faster. But the pollen never stopped, and it was noticed that cigar smoke also swam. Was cigar smoke alive too?

Einstein’s first version of an answer, 1905, was to consider that the liquid was composed of atoms whose energy was a Boltzmann distribution with an average of E= kT in every direction where k is the Boltzmann constant, and k = R/N. That is Boltsman’s constant equals the gas constant, R, divided by Avogadro’s number, N. He was able to show that the many interactions with the molecules should cause the pollen to take a random, jerky walk as seen, and that the velocity should be faster the less viscous the solvent, or the smaller the length-scale of observation. Einstein applied the Stokes drag equation to the solute, the drag force per particle was f = -6πrvη where r is the radius of the solute particle, v is the velocity, and η is the solution viscosity. Using some math, he was able to show that the diffusivity of the solute should be D = kT/6πrη. This is called the Stokes-Einstein equation.

In 1908 a French physicist, Jean Baptiste Perrin confirmed Einstein’s predictions, winning the Nobel prize for his work. I will now show the 1908 Einstein derivation and will hope to get to my extension by the end of this post.

Consider the molar Gibbs free energy of a solvent, water say. The molar concentration of water is x and that of a very dilute solute is y. y<<1. For this nearly pure water, you can show that µ = µ° +RT ln x= µ° +RT ln (1-y) = µ° -RTy.

Now, take a derivative with respect to some linear direction, z. Normally this is considered illegal, since thermodynamic is normally understood to apply to equilibrium systems only. Still Einstein took the derivative, and claimed it was legitimate at nearly equilibrium, pseudo-equilibrium. You can calculate the force on the solvent, the force on the water generated by a concentration gradient, Fw = dµ/dz = -RT dy/dz.

Now the force on each atom of water equals -RT/N dy/dz = -kT dy/dz.

Now, let’s call f the force on each atom of solute. For dilute solutions, this force is far higher than the above, f = -kT/y dy/dz. That is, for a given concentration gradient, dy/dz, the force on each solute atom is higher than on each solvent atom in inverse proportion to the molar concentration.

For small spheres, and low velocities, the flow is laminar and the drag force, f = 6πrvη.

Now calculate the speed of each solute atom. It is proportional to the force on the atom by the same relationship as appeared above: f = 6πrvη or v = f/6πrη. Inserting our equation for f= -kT/y dy/dz, we find that the velocity of the average solute molecule,

v = -kT/6πrηy dy/dz.

Let’s say that the molar concentration of solvent is C, so that, for water, C will equal about 1/18 mols/cc. The atomic concentration of dilute solvent will then equal Cy. We find that the molar flux of material, the diffusive flux equals Cyv, or that

Molar flux (mols/cm2/s) = Cy (-kT/6πrηy dy/dz) = -kTC/6πrη dy/dz -kT/6πrη dCy/dz.

where Cy is the molar concentration of solvent per volume.

Classical engineering comes to a similar equation with a property called diffusivity. Sp that

Molar flux of y (mols y/cm2/s) = -D dCy/dz, and D is an experimentally determined constant. We thus now have a prediction for D:

D = kT/6πrη.

This again is the Stokes Einstein Equation, the same as above but derived with far less math. I was fascinated, but felt sure there was something wrong here. Macroscopic viscosity was not the same as microscopic. I just could not think of a great case where there was much difference until I realized that, in polymer solutions there was a big difference.

Polymer solutions, I reasoned had large viscosities, but a diffusing solute probably didn’t feel the liquid as anywhere near as viscous. The viscometer measured at a larger distance, more similar to that of the polymer coil entanglement length, while a small solute might dart between the polymer chains like a rabbit among trees. I applied an equation for heat transfer in a dispersion that JK Maxwell had derived,

where κeff is the modified effective thermal conductivity (or diffusivity in my case), κl and κp are the thermal conductivity of the liquid and the particles respectively, and φ is the volume fraction of particles. 

To convert this to diffusion, I replaced κl by Dl, and κp by Dp where

Dl = kT/6πrηl

and Dp = kT/6πrη.

In the above ηl is the viscosity of the pure, liquid solvent.

The chair of the department, Don Anderson didn’t believe my equation, but agreed to help test it. A student named Kit Yam ran experiments on a variety of polymer solutions, and it turned out that the equation worked really well down to high polymer concentrations, and high viscosity.

As a simple, first approximation to the above, you can take Dp = 0, since it’s much smaller than Dl and you can take Dl to equal Dl = kT/6πrηl as above. The new, first order approximation is:

D = kT/6πrηl (1 – 3φ/2).

We published in Science. That is I published along with the two colleagues who tested the idea and proved the theory right, or at least useful. The reference is Yam, K., Anderson, D., Buxbaum, R. E., Science 240 (1988) p. 330 ff. “Diffusion of Small Solutes in Polymer-Containing Solutions”. This result is one of my proudest achievements.

R.E. Buxbaum, March 20, 2024

Isotopic effects in hydrogen diffusion in metals

For most people, there is a fundamental difference between solids and fluids. Solids have long-term permanence with no apparent diffusion; liquids diffuse and lack permanence. Put a penny on top of a dime, and 20 years later the two coins are as distinct as ever. Put a layer of colored water on top of plain water, and within a few minutes you’ll see that the coloring diffuse into the plain water, or (if you think the other way) you’ll see the plain water diffuse into the colored.

Now consider the transport of hydrogen in metals, the technology behind REB Research’s metallic  membranes and getters. The metals are clearly solid, keeping their shapes and properties for centuries. Still, hydrogen flows into and through the metals at a rate of a light breeze, about 40 cm/minute. Another way of saying this is we transfer 30 to 50 cc/min of hydrogen through each cm2 of membrane at 200 psi and 400°C; divide the volume by the area, and you’ll see that the hydrogen really moves through the metal at a nice clip. It’s like a normal filter, but it’s 100% selective to hydrogen. No other gas goes through.

To explain why hydrogen passes through the solid metal membrane this way, we have to start talking about quantum behavior. It was the quantum behavior of hydrogen that first interested me in hydrogen, some 42 years ago. I used it to explain why water was wet. Below, you will find something a bit more mathematical, a quantum explanation of hydrogen motion in metals. At REB we recently put these ideas towards building a membrane system for concentration of heavy hydrogen isotopes. If you like what follows, you might want to look up my thesis. This is from my 3rd appendix.

Although no-one quite understands why nature should work this way, it seems that nature works by quantum mechanics (and entropy). The basic idea of quantum mechanics you will know that confined atoms can only occupy specific, quantized energy levels as shown below. The energy difference between the lowest energy state and the next level is typically high. Thus, most of the hydrogen atoms in an atom will occupy only the lower state, the so-called zero-point-energy state.

A hydrogen atom, shown occupying an interstitial position between metal atoms (above), is also occupying quantum states (below). The lowest state, ZPE is above the bottom of the well. Higher energy states are degenerate: they appear in pairs. The rate of diffusive motion is related to ∆E* and this degeneracy.

A hydrogen atom, shown occupying an interstitial position between metal atoms (above), is also occupying quantum states (below). The lowest state, ZPE is above the bottom of the well. Higher energy states are degenerate: they appear in pairs. The rate of diffusive motion is related to ∆E* and this degeneracy.

The fraction occupying a higher energy state is calculated as c*/c = exp (-∆E*/RT). where ∆E* is the molar energy difference between the higher energy state and the ground state, R is the gas constant and T is temperature. When thinking about diffusion it is worthwhile to note that this energy is likely temperature dependent. Thus ∆E* = ∆G* = ∆H* – T∆S* where asterisk indicates the key energy level where diffusion takes place — the activated state. If ∆E* is mostly elastic strain energy, we can assume that ∆S* is related to the temperature dependence of the elastic strain.

Thus,

∆S* = -∆E*/Y dY/dT

where Y is the Young’s modulus of elasticity of the metal. For hydrogen diffusion in metals, I find that ∆S* is typically small, while it is often typically significant for the diffusion of other atoms: carbon, nitrogen, oxygen, sulfur…

The rate of diffusion is now calculated assuming a three-dimensional drunkards walk where the step lengths are constant = a. Rayleigh showed that, for a simple cubic lattice, this becomes:

D = a2/6τ

a is the distance between interstitial sites and t is the average time for crossing. For hydrogen in a BCC metal like niobium or iron, D=

a2/9τ; for a FCC metal, like palladium or copper, it’s

a2/3τ. A nice way to think about τ, is to note that it is only at high-energy can a hydrogen atom cross from one interstitial site to another, and as we noted most hydrogen atoms will be at lower energies. Thus,

τ = ω c*/c = ω exp (-∆E*/RT)

where ω is the approach frequency, or the amount of time it takes to go from the left interstitial position to the right one. When I was doing my PhD (and still likely today) the standard approach of physics writers was to use a classical formulation for this time-scale based on the average speed of the interstitial. Thus, ω = 1/2a√(kT/m), and

τ = 1/2a√(kT/m) exp (-∆E*/RT).

In the above, m is the mass of the hydrogen atom, 1.66 x 10-24 g for protium, and twice that for deuterium, etc., a is the distance between interstitial sites, measured in cm, T is temperature, Kelvin, and k is the Boltzmann constant, 1.38 x 10-16 erg/°K. This formulation correctly predicts that heavier isotopes will diffuse slower than light isotopes, but it predicts incorrectly that, at all temperatures, the diffusivity of deuterium is 1/√2 that for protium, and that the diffusivity of tritium is 1/√3 that of protium. It also suggests that the activation energy of diffusion will not depend on isotope mass. I noticed that neither of these predictions is borne out by experiment, and came to wonder if it would not be more correct to assume ω represent the motion of the lattice, breathing, and not the motion of a highly activated hydrogen atom breaking through an immobile lattice. This thought is borne out by experimental diffusion data where you describe hydrogen diffusion as D = D° exp (-∆E*/RT).

Screen Shot 2018-06-21 at 12.08.20 AM

You’ll notice from the above that D° hardly changes with isotope mass, in complete contradiction to the above classical model. Also note that ∆E* is very isotope dependent. This too is in contradiction to the classical formulation above. Further, to the extent that D° does change with isotope mass, D° gets larger for heavier mass hydrogen isotopes. I assume that small difference is the entropy effect of ∆E* mentioned above. There is no simple square-root of mass behavior in contrast to most of the books we had in grad school.

As for why ∆E* varies with isotope mass, I found that I could get a decent explanation of my observations if I assumed that the isotope dependence arose from the zero point energy. Heavier isotopes of hydrogen will have lower zero-point energies, and thus ∆E* will be higher for heavier isotopes of hydrogen. This seems like a far better approach than the semi-classical one, where ∆E* is isotope independent.

I will now go a bit further than I did in my PhD thesis. I’ll make the general assumption that the energy well is sinusoidal, or rather that it consists of two parabolas one opposite the other. The ZPE is easily calculated for parabolic energy surfaces (harmonic oscillators). I find that ZPE = h/aπ √(∆E/m) where m is the mass of the particular hydrogen atom, h is Plank’s constant, 6.63 x 10-27 erg-sec,  and ∆E is ∆E* + ZPE, the zero point energy. For my PhD thesis, I didn’t think to calculate ZPE and thus the isotope effect on the activation energy. I now see how I could have done it relatively easily e.g. by trial and error, and a quick estimate shows it would have worked nicely. Instead, for my PhD, Appendix 3, I only looked at D°, and found that the values of D° were consistent with the idea that ω is about 0.55 times the Debye frequency, ω ≈ .55 ωD. The slight tendency for D° to be larger for heavier isotopes was explained by the temperature dependence of the metal’s elasticity.

Two more comments based on the diagram I presented above. First, notice that there is middle split level of energies. This was an explanation I’d put forward for quantum tunneling atomic migration that some people had seen at energies below the activation energy. I don’t know if this observation was a reality or an optical illusion, but present I the energy picture so that you’ll have the beginnings of a description. The other thing I’d like to address is the question you may have had — why is there no zero-energy effect at the activated energy state. Such a zero energy difference would cancel the one at the ground state and leave you with no isotope effect on activation energy. The simple answer is that all the data showing the isotope effect on activation energy, table A3-2, was for BCC metals. BCC metals have an activation energy barrier, but it is not caused by physical squeezing between atoms, as for a FCC metal, but by a lack of electrons. In a BCC metal there is no physical squeezing, at the activated state so you’d expect to have no ZPE there. This is not be the case for FCC metals, like palladium, copper, or most stainless steels. For these metals there is a much smaller, on non-existent isotope effect on ∆E*.

Robert Buxbaum, June 21, 2018. I should probably try to answer the original question about solids and fluids, too: why solids appear solid, and fluids not. My answer has to do with quantum mechanics: Energies are quantized, and always have a ∆E* for motion. Solid materials are those where ω exp (-∆E*/RT) has unit of centuries. Thus, our ability to understand the world is based on the least understandable bit of physics.

Hydrogen permeation rates in Inconel, Hastelloy and stainless steels.

Some 20 years ago, I published a graph of the permeation rate for hydrogen in several metals at low pressure, See the graph here, but I didn’t include stainless steel in the graph.

Hydrogen permeation in clean SS-304; four research groups’ data.

One reason I did not include stainless steel was there were many stainless steels and the hydrogen permeation rates were different, especially so between austenitic (FCC) steels and ferritic steels (BCC). Another issue was oxidation. All stainless steels are oxidized, and it affect H2 permeation a lot. You can decrease the hydrogen permeation rate significantly by oxidation, or by surface nitriding, etc (my company will even provide this service). Yet another issue is cold work. When  an austenitic stainless steel is worked — rolled or drawn — some Austinite (FCC) material transforms to Martisite (a sort of stretched BCC). Even a small amount of martisite causes an order of magnitude difference in the permeation rate, as shown below. For better or worse, after 20 years, I’m now ready to address H2 in stainless steel, or as ready as I’m likely to be.

Hydrogen permeation data for SS 340 and SS 321.

Hydrogen permeation in SS 340 and SS 321. Cold work affects H2 permeation more than the difference between 304 and 321; Sun Xiukui, Xu Jian, and Li Yiyi, 1989

The first graph I’d like to present, above, is a combination of four research groups’ data for hydrogen transport in clean SS 304, the most common stainless steel in use today. SS 304 is a ductile, austenitic (FCC), work hardening, steel of classic 18-8 composition (18% Cr, 8% Ni). It shares the same basic composition with SS 316, SS 321 and 304L only differing in minor components. The data from four research groups shows a lot of scatter: a factor of 5 variation at high temperature, 1000 K (727 °C), and almost two orders of magnitude variation (factor of 50) at room temperature, 13°C. Pressure is not a factor in creating the scatter, as all of these studies were done with 1 atm, 100 kPa hydrogen transporting to vacuum.

The two likely reasons for the variation are differences in the oxide coat, and differences in the amount of cold work. It is possible these are the same explanation, as a martensitic phase might increase H2 permeation by introducing flaws into the oxide coat. As the graph at left shows, working these alloys causes more differences in H2 permeation than any difference between alloys, or at least between SS 304 and SS 321. A good equation for the permeation behavior of SS 304 is:

P (mol/m.s.Pa1/2) = 1.1 x10-6 exp (-8200/T).      (H2 in SS-304)

Because of the song influence of cold work and oxidation, I’m of the opinion that I get a slightly different, and better equation if I add in permeation data from three other 18-8 stainless steels:

P (mol/m.s.Pa1/2) = 4.75 x10-7 exp (-7880/T).     (H2 in annealed SS-304, SS-316, SS-321)

Screen Shot 2017-12-16 at 10.37.37 PM

Hydrogen permeation through several common stainless steels, as well as Inocnel and Hastelloy

Though this result is about half of the previous at high temperature, I would trust it better, at least for annealed SS-304, and also for any annealed austenitic stainless steel. Just as an experiment, I decided to add a few nickel and cobalt alloys to the mix, and chose to add data for inconel 600, 625, and 718; for kovar; for Hastelloy, and for Fe-5%Si-5%Ge, and SS4130. At left, I pilot all of these on one graph along with data for the common stainless steels. To my eyes the scatter in the H2 permeation rates is indistinguishable from that SS 304 above or in the mixed 18-8 steels (data not shown). Including these materials to the plot decreases the standard deviation a bit to a factor of 2 at 1000°K and a factor of 4 at 13°C. Making a least-square analysis of the data, I find the following equation for permeation in all common FCC stainless steels, plus Inconels, Hastelloys and Kovar:

P (mol/m.s.Pa1/2) = 4.3 x10-7 exp (-7850/T).

This equation is near-identical to the equation above for mixed, 18-8 stainless steel. I would trust it for annealed or low carbon metal (SS-304L) to a factor of 2 accuracy at high temperatures, or a factor of 4 at low temperatures. Low carbon reduces the tendency to form Martinsite. You can not use any of these equations for hydrogen in ferritic (BCC) alloys as the rates are different, but this is as good as you’re likely to get for basic austenitc stainless and related materials. If you are interested in the effect of cold work, here is a good reference. If you are bothered by the square-root of pressure driving force, it’s a result of entropy: hydrogen travels in stainless steel as dislocated H atoms and the dissociation H2 –> 2 H leads to the square root.

Robert Buxbaum, December 17, 2017. My business, REB Research, makes hydrogen generators and purifiers; we sell getters; we consult on hydrogen-related issues, and will (if you like) provide oxide (and similar) permeation barriers.

A rotating disk bio-reactor for sewage treatment

One of the most effective designs for sewage treatment is the rotating disk bio-reactor, shown below. It is typically used in small-throughput sewage plants, but it performs quite well in larger plants too. I’d like to present an analysis of the reactor, and an explanation of why it works so well.

A rotating disc sewage reactor.

A rotating disk sewage reactor; ∂ is the thickness of the biofilm. It’s related to W the rotation rate in radians per sec, and to D the limiting diffusivity.

As shown, the reactor is fairly simple-looking, nothing more than a train of troughs filled with sewage-water, typically 3-6 feet deep, with a stack of discs rotating within. The discs are typically 7 to 14 feet in diameter (2-4 meters) and 1 cm apart. The shaft is typically close to the water level, but slightly above, and the rotation speed is adjustable. The device works because appropriate bio-organisms attach themselves to the disk, and the rotation insures that they are fully (or reasonably) oxygenated.

How do we know the cells on the disc will be oxygenated? The key is the solubility of oxygen in water, and the fact that these discs are only used on the low biological oxygen demand part of the sewage treatment process, only where the sewage contains 40 ppm of soluble organics or less. The main reaction on the rotating disc is bio oxidation of soluble carbohydrate (sugar) in a layer of wet slime attached to the disc.

H-O-C-H + O2 –> CO2 + H2O.

As it happens, the solubility of pure oxygen in water is about 40 ppm at 1 atm. As air contains 21% oxygen, we expect an 8 ppm concentration of oxygen on the slime surface: 21% of 40 ppm = 8 ppm. Given the reaction above and the fact that oxygen will diffuse five times more readily than sugar at least, we expect that one disc rotation will easily provide enough oxygen to remove 40 ppm sugar in the slime at every speed of rotation so long as the wheel is in the air at least half of the time, as shown above.

Let’s now pick a rotation speed of 1/3 rpm (3 minutes per rotation) and see where that gets us in terms of speed of organic removal. Since the disc is always in an area of low organic concentration, it becomes covered mostly with “rotifers”, a fungus that does well in low nutrient (low BOD) sewage. Let’s now assume that mass transfer (diffusion) of sugar in the rotifer slime determines the thickness of the rotifera layer, and thus the rate of organic removal. We can calculate the diffusion depth of sugar, ∂ via the following equation, derived in my PhD thesis.

∂ = √πDt

Here, D is the diffusivity (cm2/s) for sugar in the rotifera slime attached to the disk, π = 3.1415.. and t is the contact time, 90 seconds in the above assumption. My expectation is that D in the rotifer slime will be similar to the diffusivity sugar in water, about 3 x 10-6 cm2/s. Based on the above, we find the rotifer thickness will be ∂ = √.00085 cm2 = .03 cm, and the oxygen depth will be about 2.5 times that, 0.07 cm. If the discs are 1 cm apart, we find that, about 14% of the fluid volume of the reactor will be filled with slime, with 2/5 of this rotifer-filled. This is as much as 1000 times more rotifers than you would get in an ordinary constantly stirred tank reactor, a CSTR to use a common acronym. We might now imagine that the volume of this sewage-treatment reactor can be as small as 1000 gallons, 1/1000 the size of a CSTR. Unfortunately it is not so; we’ll have to consider another limiting effect, diffusion of nutrients.

Consider the diffusive mass transfer of sugar from a 1,000,000 gal/day flow (43 liters/sec). Assume that at some point in the extraction you have a concentration C(g/cc) of sugar in the liquid where C is between 40 ppm and 1 ppm. Let’s pick a volume of the reactor that is 1/20 the normal size for this flow (not 1/1000 the size, you’ll see why). That is to say a trough whose volume is 50,000 gallons (200,000 liters, 200 m3). If the discs are 1 cm apart, this trough (or section of a trough) will have about  4×10^8 cm2 of submerged surface, and about 9×10^8 total surface including wetted disc in the air. The mass of organic that enters this section of trough is 44,000 C g/second, but this mass of sugar can only reach the rotifers by diffusion through a water-like diffusion layer of about .06 cm thickness, twice the thickness calculated above. The thickness is twice that calculated above because it includes the supernatant liquid beyond the slime layer. We now calculate the rate of mass diffusing into the disc: AxDxc/z = 8×10^8 x 3×10-6 x C/.06 cm = 40,000 C g/sec, and find that, for this tank size and rotation speed, the transfer rate of organic to the discs is 2/3 as much as needed to absorb the incoming sugar. This is to say that a 50,000 gallon section is too small to reduce the concentration to ln (1) at this speed of disc rotation.

Based on the above calculation, I’m inclined to increase the speed of rotation to .75 rpm. At this speed, the rotifer-slime layer will be 2/3 as thin 0.2 mm, and we expect an equally thinner diffusion barrier in the supernatant. At this faster speed, there is now 3/2 as much diffusion transfer per area because the thinner diffusion barrier, and we can expect a 50,000 liter reactor section to reduce the initial concentration by a fraction of 1/2.718 or C/e. Note that the mass transfer rate to the discs is always proportional to C. If we find that 50,000 gallons of tank reduces the concentration to 1/e, we find that we need 150,000 gallons of reactor to reduce the concentration of sugar from 40 ppm to 2 ppm, the legal target, ln (40/2) = 3. This 150,000 gallons is a remarkably small volume to reduce the sBOD concentration from 40 ppm to 2 ppm (sBOD = soluble biological oxygen demand), and the energy use is small too if the disc bearings are good.

The Holly sewage treatment plant is the only one in Oakland county, MI using the rotating disc contacted technology. It has a flow of 1,000,000 gallons per day, and has a contactor trough that is 215,000 gallons, about what we’d expect though their speed is somewhat higher, over 1 rpm and their input concentration is likely lower than 40 ppm. For the first stage of sewage treatment, the Holly plant use a vertical-draft, trickle-bed reactor. That is they drizzle the sewage-liquids over a slime-coated packing to reduce the sBOD concentration from 200 ppm to before feeding the flow to the rotating discs. My sense of the reason they don’t do the entire extraction with a trickle bed is that the discs use far less energy.

I should also add that the back-part of the disc isn’t totally useless oxygen storage, as it seems from my analysis. Some non-sugar reactions take place in the relatively anoxic environment there and in the liquid at the bottom of the trough. In these regions, iron reacts with phosphate, and nitrate removal takes place. These are other important requirements of sewage treatment.

Robert E. Buxbaum, July 18, 2017. As an exercise, find the volume necessary for a plug flow reactor or a stirred tank reactor (CSTR) to reduce the concentration of sugar from 40 ppm to 2 ppm. Assume 1,000,000 gal per day, an excess of oxygen in these reactors, and a first order reaction with a rate constant of dC/dt = -(0.4/hr)C. At some point in the future I plan to analyze these options, and the trickle bed reactor, too.

Dr. Who’s Quantum reality viewed as diffusion

It’s very hard to get the meaning of life from science because reality is very strange, Further, science is mathematical, and the math relations for reality can be re-arranged. One arrangement of the terms will suggest a version of causality, while another will suggest a different causality. As Dr. Who points out, in non-linear, non-objective terms, there’s no causality, but rather a wibbly-wobbely ball of timey-wimey stuff.

Time as a ball of wibblely wobbly timey wimey stuff.

Reality is a ball of  timey wimpy stuff, Dr. Who.

To this end, I’ll provide my favorite way of looking at the timey-wimey way of the world by rearranging the equations of quantum mechanics into a sort of diffusion. It’s not the diffusion of something you’re quite familiar with, but rather a timey-wimey wave-stuff referred to as Ψ. It’s part real and part imaginary, and the only relationship between ψ and life is that the chance of finding something somewhere is proportional Ψ*|Ψ. The diffusion of this half-imaginary stuff is the underpinning of reality — if viewed in a certain way.

First let’s consider the steady diffusion of a normal (un-quantum) material. If there is a lot of it, like when there’s perfume off of a prima donna, you can say that N = -D dc/dx where N is the flux of perfume (molecules per minute per area), dc/dx is a concentration gradient (there’s more perfume near her than near you), and D is a diffusivity, a number related to the mobility of those perfume molecules. 

We can further generalize the diffusion of an ordinary material for a case where concentration varies with time because of reaction or a difference between the in-rate and the out rate, with reaction added as a secondary accumulator, we can write: dc/dt = reaction + dN/dx = reaction + D d2c/dx2. For a first order reaction, for example radioactive decay, reaction = -ßc, and 

dc/dt = -ßc + D d2c/dx2               (1)

where ß is the radioactive decay constant of the material whose concentration is c.

Viewed in a certain way, the most relevant equation for reality, the time-dependent Schrödinger wave equation (semi-derived here), fits into the same diffusion-reaction form:

dΨ/dt = – 2iπV/h Ψ + hi/4πm d2Ψ/dx               (2)

Instead of reality involving the motion of a real material (perfume, radioactive radon, etc.) with a real concentration, c, in this relation, the material can not be sensed directly, and the concentration, Ψ, is semi -imaginary. Here, h is plank’s constant, i is the imaginary number, √-1, m is the mass of the real material, and V is potential energy. When dealing with reactions or charged materials, it’s relevant that V will vary with position (e.g. electrons’ energy is lower when they are near protons). The diffusivity term here is imaginary, hi/4πm, but that’s OK, Ψ is part imaginary, and we’d expect that potential energy is something of a destroyer of Ψ: the likelihood of finding something at a spot goes down where the energy is high.

The form of this diffusion is linear, a mathematical term that refers to equations where solution that works for Ψ will also work for 2Ψ. Generally speaking linear solutions have exp() terms in them, and that’s especially likely here as the only place where you see a time term is on the left. For most cases we can say that

Ψ = ψ exp(-2iπE/h)t               (3)

where ψ is not a function of anything but x (space) and E is the energy of the thing whose behavior is described by Ψ. If you take the derivative of equation 3 this with respect to time, t, you get

dΨ/dt = ψ (-2iπE/h) exp(-2iπE/h)t = (-2iπE/h)Ψ.               (4)

If you insert this into equation 2, you’ll notice that the form of the first term is now identical to the second, with energy appearing identically in both terms. Divide now by exp(-2iπE/h)t, and you get the following equation:

(E-V) ψ =  -h2/8π2m d2ψ/dx2                      (5)

where ψ can be thought of as the physical concentration in space of the timey-wimey stuff. ψ is still wibbly-wobbley, but no longer timey-wimey. Now ψ- squared is the likelihood of finding the stuff somewhere at any time, and E, the energy of the thing. For most things in normal conditions, E is quantized and equals approximately kT. That is E of the thing equals, typically, a quantized energy state that’s nearly Boltzmann’s constant times temperature.

You now want to check that the approximation in equation 3-5 was legitimate. You do this by checking if the length-scale implicit in exp(-2iπE/h)t is small relative to the length-scales of the action. If it is (and it usually is), You are free to solve for ψ at any E and V using normal mathematics, by analytic or digital means, for example this way. ψ will be wibbely-wobbely but won’t be timey-wimey. That is, the space behavior of the thing will be peculiar with the item in forbidden locations, but there won’t be time reversal. For time reversal, you need small space features (like here) or entanglement.

Equation 5 can be considered a simple steady state diffusion equation. The stuff whose concentration is ψ is created wherever E is greater than V, and is destroyed wherever V is greater than E. The stuff then continuously diffuses from the former area to the latter establishing a time-independent concentration profile. E is quantized (can only be some specific values) since matter can never be created of destroyed, and it is only at specific values of E that this happens in Equation 5. For a particle in a flat box, E and ψ are found, typically, by realizing that the format of ψ must be a sin function (and ignoring an infinity). For more complex potential energy surfaces, it’s best to use a matrix solution for ψ along with non-continuous calculous. This avoids the infinity, and is a lot more flexible besides.

When you detect a material in some spot, you can imagine that the space- function ψ collapses, but even that isn’t clear as you can never know the position and velocity of a thing simultaneously, so doesn’t collapse all that much. And as for what the stuff is that diffuses and has concentration ψ, no-one knows, but it behaves like a stuff. And as to why it diffuses, perhaps it’s jiggled by unseen photons. I don’t know if this is what happens, but it’s a way I often choose to imagine reality — a moving, unseen material with real and imaginary (spiritual ?) parts, whose concentration, ψ, is related to experience, but not directly experienced.

This is not the only way the equations can be rearranged. Another way of thinking of things is as the sum of path integrals — an approach that appears to me as a many-world version, with fixed-points in time (another Dr Who feature). In this view, every object takes every path possible between these points, and reality as the sum of all the versions, including some that have time reversals. Richard Feynman explains this path integral approach here. If it doesn’t make more sense than my version, that’s OK. There is no version of the quantum equations that will make total, rational sense. All the true ones are mathematically equivalent — totally equal, but differ in the “meaning”. That is, if you were to impose meaning on the math terms, the meaning would be totally different. That’s not to say that all explanations are equally valid — most versions are totally wrong, but there are many, equally valid math version to fit many, equally valid religious or philosophic world views. The various religions, I think, are uncomfortable with having so many completely different views being totally equal because (as I understand it) each wants exclusive ownership of truth. Since this is never so for math, I claim religion is the opposite of science. Religion is trying to find The Meaning of life, and science is trying to match experiential truth — and ideally useful truth; knowing the meaning of life isn’t that useful in a knife fight.

Dr. Robert E. Buxbaum, July 9, 2014. If nothing else, you now perhaps understand Dr. Who more than you did previously. If you liked this, see here for a view of political happiness in terms of the thermodynamics of free-energy minimization.

Another Quantum Joke, and Schrödinger’s waves derived

Quantum mechanics joke. from xkcd.

Quantum mechanics joke. from xkcd.

Is funny because … it’s is a double entente on the words grain (as in grainy) and waves, as in Schrödinger waves or “amber waves of grain” in the song America (Oh Beautiful). In Schrödinger’s view of the quantum world everything seems to exist or move as a wave until you observe it, and then it always becomes a particle. The math to solve for the energy of things is simple, and thus the equation is useful, but it’s hard to understand why,  e.g. when you solve for the behavior of a particle (atom) in a double slit experiment you have to imagine that the particle behaves as an insubstantial wave traveling though both slits until it’s observed. And only then behaves as a completely solid particle.

Math equations can always be rewritten, though, and science works in the language of math. The different forms appear to have different meaning but they don’t since they have the same practical predictions. Because of this freedom of meaning (and some other things) science is the opposite of religion. Other mathematical formalisms for quantum mechanics may be more comforting, or less, but most avoid the wave-particle duality.

The first formalism was Heisenberg’s uncertainty. At the end of this post, I show that it is identical mathematically to Schrödinger’s wave view. Heisenberg’s version showed up in two quantum jokes that I explained (beat into the ground), one about a lightbulb  and one about Heisenberg in a car (also explains why water is wet or why hydrogen diffuses through metals so quickly).

Yet another quantum formalism involves Feynman’s little diagrams. One assumes that matter follows every possible path (the multiple universe view) and that time should go backwards. As a result, we expect that antimatter apples should fall up. Experiments are underway at CERN to test if they do fall up, and by next year we should finally know if they do. Even if anti-apples don’t fall up, that won’t mean this formalism is wrong, BTW: all identical math forms are identical, and we don’t understand gravity well in any of them.

Yet another identical formalism (my favorite) involves imagining that matter has a real and an imaginary part. In this formalism, the components move independently by diffusion, and as a result look like waves: exp (-it) = cost t + i sin t. You can’t observe the two parts independently though, only the following product of the real and imaginary part: (the real + imaginary part) x (the real – imaginary part). Slightly different math, same results, different ways of thinking of things.

Because of quantum mechanics, hydrogen diffuses very quickly in metals: in some metals quicker than most anything in water. This is the basis of REB Research metal membrane hydrogen purifiers and also causes hydrogen embrittlement (explained, perhaps in some later post). All other elements go through metals much slower than hydrogen allowing us to make hydrogen purifiers that are effectively 100% selective. Our membranes also separate different hydrogen isotopes from each other by quantum effects (big things tunnel slower). Among the uses for our hydrogen filters is for gas chromatography, dynamo cooling, and to reduce the likelihood of nuclear accidents.

Dr. Robert E. Buxbaum, June 18, 2013.

To see Schrödinger’s wave equation derived from Heisenberg for non-changing (time independent) items, go here and note that, for a standing wave there is a vibration in time, though no net change. Start with a version of Heisenberg uncertainty: h =  λp where the uncertainty in length = wavelength = λ and the uncertainty in momentum = momentum = p. The kinetic energy, KE = 1/2 p2/m, and KE+U(x) =E where E is the total energy of the particle or atom, and U(x) is the potential energy, some function of position only. Thus, p = √2m(E-PE). Assume that the particle can be described by a standing wave with a physical description, ψ, and an imaginary vibration you can’t ever see, exp(-iωt). And assume this time and space are completely separable — an OK assumption if you ignore gravity and if your potential fields move slowly relative to the speed of light. Now read the section, follow the derivation, and go through the worked problems. Most useful applications of QM can be derived using this time-independent version of Schrödinger’s wave equation.