Perhaps some of the constants of Nature are non-computable numbers:
G = 6.673 x 10^-11 N m^2 kg^-2
c = 2.99 x 10^8 m s^-1
h = 6.626 x 10^-34 J s
It would be difficult to test experimentally whether they are non-computable. Quantities can only be measured to a small number of significant figures. QED has been tested to 10 sig fig, whereas general relativity has been tested to 14 sig fig. But we don't know what happens in the infinite number of decimal places beyond the known value.
There are solutions to the Schrodinger equation which take the form of eigenvalues. They equal the energy level of (say) an electron in a hydrogen atom. The Schrodinger eqn is:
H psi = (i * h / 2 * pi) * (d psi / dt)
pi is computable because it can be found from an infinite series:
pi = 4 arctan(1)
= sum of k = 0 to infinity ((-1)^k / (2k+1))
Also, e occurs frequently in physics equations, including some of the solutions to the Schrodinger equation, and
e = sum of k = 0 to infinity (1 / k factorial)
I am not sure whether we can really say that i (the imaginary unit) is computable! Of course, it is not a real number, so I am not sure how you would program a Turing machine to calculate it. On the other hand, its absolute value is 1, and it is the solution of the quadratic x^2 + 1 = 0.
This leaves h. Now, h is a physical constant, rather than a mathematical constant. It is measured rather than derived from first principles. Thus, h is truly fundamental. It cannot be derived from anything else, or defined. I think it is very probable that h is non-computable, because in a measure-theoretic sense, most real numbers are not computable. There are is only a countably infinite set of Turing machines, and hence only countably many computable numbers. Therefore, it you were to pick a number x completely at random from R (the set of all real numbers), it would almost certainly be non-computable. The set of computable numbers has measure zero.
Returning to the Schrodinger equation, let us assume that h is non-computable. Then we have this non-computable number as a kind of "constant of proportionality". Some solutions take the form of eigenvectors (wavefunctions) and eigenvalues (energy levels). Once again, it is very likely that each eignevalue is non-computable.
Versions of the S equation for systems with a large number of particles, eg, a large biomolecule (such as DNA, RNA, microtubule, neurotransmitter) could not be solved analytically. Either no analytical solution exists or else we humans are not clever enough to solve it.
In such situations, people resort to numerical approximations (perhaps using computers), but Nature itself makes no such approximation. The various physical values are real numbers which are probably non-computable.
In Nature, there are many physical systems that evolve dynamically, based on the interplay of mostly non-computable numbers, with a few computable numbers like e, pi, square root of 2, a/b thrown in for good measure! All physical quantities (eg, mass, energy, length, time, velocity, etc) may be described by numbers (scalars) or combinations of numbers (tensors). I would suggest the following hypothesis:
Newtonian physics is non-computable and deterministic
Quantum mechanics is non-computable and non-deterministic (the latter property making free will possible)
Thus, a system composed of conventional matter, such as the human brain operates and goes about its daily business in a non-computable way. However, this means that even a conventional computer (such as an ordinary desktop PC) is (strictly speaking) non-computable. This is true, but a PC has built-in features to "iron out" the non-computability. Each bit is either 0 or 1, obviously computable numbers. How could voltages subject to random noise be kept at strictly computable values? This could be achieved using tolerance bands:
Let x be a voltage.
If -0.1 <= x <= 0.1, then x should be flattened out to 0
If 0.9 <= x <= 1.1, then x should be flattened out to 1
Occasionally, there are random errors, due to non-computable physics, but these are interpreted as a flip between 0 and 1. This is the kind of thing that happens when a data file gets corrupted.
By contrast, the human brain (and perhaps even the eternal soul) harnesses the richness of the real number system. Furthermore, there is a hierarchy of transfinite numbers. But I shall leave it there for the moment, and say more about these transfinite numbers on subsequent blog postings.
This comment has been removed by the author.
ReplyDeleteA comment from Dr John Allsup:
ReplyDeleteComputability of reals and actual physical constants have little in common given that infinite precision is impossible.
If there is a limit to precision in measuring distance, say, and my understanding of Physics is that our usual idea of distance
becomes meaningless at scales below the Planck length, then any physical quantity in a given system of units is experientally
indistinguishable from some close physical quantity that can be expressed as a rational number in that system of units.
For example, consider the distance of 1 metre. Now take some non-computable real, say epsilon, between 10^{-100} and 10^{-101}
and consider the distance (1+epsilon) metres. Within what is possible with physical experiments, these two distances are
indistinguishable. In particular, if we defined a new system based on the SI units, such that, in the new system, the metre
was (1+epsilon) times the present-day metre, all practical physics would look the same. But quantities that are exact in
the current system, and use metres in their units, would have non-computable expressions in the new system. On the other hand
these non-computable values would be arbitarily close to rational values.
The other issue with computability and foundational mathematics to ponder is the existence of infinity and its consequences.
There is a finite bound on the amount of information in the observable universe (the Bekenstein Bound). This places a limit
on how high you can count. Similarly, on a computer, there is a limit to how much memory you can allocate before an allocation
failure. The axiom of infinity effectively assumes not only that you can count higher and higher without bound (as opposed to
physical reality where the ability to count is bounded, albeit by some unimaginably high bound), but also that a collection
of counting numbers exists as a whole. These are convenient for mathematical reasoning, but are physically unrealistic. I do
wonder if this has issues when it comes to the mathematical foundations of physics, but cannot see how to concretely investigate.
Consider that if we take the subset of the reals, consisting of those of the form
x = Sum_{-1000<i<1000}a_i10^i
that is, numbers which only have up to 1000 nonzero digits either side of the decimal point, we can still do practical
experimental physics to any desired degree of accuracy -- experimental error will become a problem before we exhaust this
system of numbers.
As for computability of i, bear in mind with computation that you must define a system of notation (such as the binary
floating point format our computers use). If you define one (perhaps by interleaving bits) so that two real-valued
floating point numbers (of arbitrary precision) can be stored in a single number, then use this to store numbers
of the form x+iy, then i is trivially computable. Bear in mind that by saying Pi is computable, we really mean something
like that there is an algorithm which will output an arbitrarily high number of correct digits (and no incorrect digits).
Clearly a machine intended to output every digit of pi would never halt.
Finally, consider what happens if you replace the Turing machine model of computability with physically plausible computation devices, so that
time and space and other resources are actual bounded resources that limit what computations are possible.
John has made some very interesting points here, which I very much appreciate.
ReplyDeleteHere are some thoughts about all of this:
(1) The fact we can only measure physical values to a finite number of sig fig means that we never quite capture the full complexity of Nature, which I believe is utilizing the real number system. Solutions to the Schrodinger equation map from R^3 -> C. Both R and C are non-countable infinite sets.
(2) An actual computer (eg, a desktop PC) is less powerful than an idealized Turing machine, because it only has finite memory and finite time to carry out the algorithm. This any computer model of a physical system even less accurate and realistic. Once again, the physical system transcends the computer. (Although, I have to tread carefully here, because an actual PC is part of the physical world.)
(3) From what John has said, it is clear that irrational numbers like pi and e are not really computable on an actual PC. Thus solutions to differential equations that contain these transcend computability. I find this a very liberating thing.
(4) Thanks to John for pointing out that i is computable. You just have to represent the complex numbers on your Turing machine. In the Schrodinger equation, I believe that h is the non-computable constant.
(5) The real mystery is what happens at and below the Planck scale. We won't know for sure until we have a correct theory of Quantum Gravity.
Any more comments are most welcome!
This comment has been removed by the author.
ReplyDeleteA comment from Dr John Allsup:
ReplyDeleteComputability of reals and actual physical constants have little in common given that infinite precision is impossible. If there is a limit to precision in measuring distance, say, and my understanding of Physics is that our usual idea of distance becomes meaningless at scales below the Planck length, then any physical quantity in a given system of units is experientally indistinguishable from some close physical quantity that can be expressed as a rational number in that system of units. For example, consider the distance of 1 metre. Now take some non-computable real, say epsilon, between 10^{-100} and 10^{-101} and consider the distance (1+epsilon) metres. Within what is possible with physical experiments, these two distances are indistinguishable. In particular, if we defined a new system based on the SI units, such that, in the new system, the metre was (1+epsilon) times the present-day metre, all practical physics would look the same. But quantities that are exact in the current system, and use metres in their units, would have non-computable expressions in the new system. On the other hand these non-computable values would be arbitarily close to rational values.
The other issue with computability and foundational mathematics to ponder is the existence of infinity and its consequences. There is a finite bound on the amount of information in the observable universe (the Bekenstein Bound). This places a limit on how high you can count. Similarly, on a computer, there is a limit to how much memory you can allocate before an allocation failure. The axiom of infinity effectively assumes not only that you can count higher and higher without bound (as opposed to physical reality where the ability to count is bounded, albeit by some unimaginably high bound), but also that a collection of counting numbers exists as a whole. These are convenient for mathematical reasoning, but are physically unrealistic. I do wonder if this has issues when it comes to the mathematical foundations of physics, but cannot see how to concretely investigate.
Consider that if we take the subset of the reals, consisting of those of the form
x = Sum_{-1000<i<1000}a_i10^i
that is, numbers which only have up to 1000 nonzero digits either side of the decimal point, we can still do practical experimental physics to any desired degree of accuracy -- experimental error will become a problem before we exhaust this system of numbers.
As for computability of i, bear in mind with computation that you must define a system of notation (such as the binary floating point format our computers use). If you define one (perhaps by interleaving bits) so that two real-valued floating point numbers (of arbitrary precision) can be stored in a single number, then use this to store numbers of the form x+iy, then i is trivially computable. Bear in mind that by saying Pi is computable, we really mean something like that there is an algorithm which will output an arbitrarily high number of correct digits (and no incorrect digits). Clearly a machine intended to output every digit of pi would never halt.
Finally, consider what happens if you replace the Turing machine model of computability with physically plausible computation devices, so that time and space and other resources are actual bounded resources that limit what computations are possible.
The above comment from Dr John Allsup has already been posted, but the formatting of the text was very poor, so I have re-edited it, and posted it again.
ReplyDeleteI'm interested if there is a mathematical proof / physical experiment that proves the quantum vacuum is a continuum. Thanks.
ReplyDeleteI'm interested if there is a mathematical proof / physical experiment that proves the quantum vacuum is a continuum. Thanks.
ReplyDelete