Difference between revisions of "Skolem-Mahler-Lech theorem"
Ulf Rehmann (talk | contribs) m (moved Skolem–Mahler–Lech theorem to Skolem-Mahler-Lech theorem: ascii title) |
Ulf Rehmann (talk | contribs) m (tex encoded by computer) |
||
Line 1: | Line 1: | ||
− | + | <!-- | |
+ | s1101601.png | ||
+ | $#A+1 = 35 n = 2 | ||
+ | $#C+1 = 35 : ~/encyclopedia/old_files/data/S110/S.1100160 Skolem\ANDMahler\ANDLech theorem | ||
+ | Automatically converted into TeX, above some diagnostics. | ||
+ | Please remove this comment and the {{TEX|auto}} line below, | ||
+ | if TeX found to be correct. | ||
+ | --> | ||
− | + | {{TEX|auto}} | |
+ | {{TEX|done}} | ||
− | + | A recurrence sequence $ ( a _ {h} ) $ | |
+ | of order $ n $ | ||
+ | is a solution to a linear homogeneous [[Recurrence relation|recurrence relation]] with constant coefficients | ||
− | + | $$ | |
+ | a _ {h + n } = s _ {1} a _ {h + n - 1 } + \dots + s _ {n} a _ {h} ( h = 0,1, \dots ) . | ||
+ | $$ | ||
− | + | Thus, the generating function $ \sum _ {h \geq 0 } a _ {h} X ^ {h} $ | |
+ | of a recurrence is a rational function $ { {r ( X ) } / {s ( X ) } } $ | ||
+ | where $ s ( X ) = 1 - s _ {1} X - \dots - s _ {n} X ^ {n} = \prod _ {i = 1 } ^ {m} ( 1 - \alpha _ {i} X ) ^ {n _ {i} } $, | ||
+ | say; the polynomial $ r $ | ||
+ | of degree less than $ n $ | ||
+ | is determined by the initial values $ a _ {0} \dots a _ {n - 1 } $. | ||
+ | If so, the distinct complex numbers $ \alpha _ {i} $ | ||
+ | are called the roots of the recurrence, and the $ n _ {i} $ | ||
+ | their multiplicities. It follows that the $ a _ {h} $ | ||
+ | are given by a generalized power sum $ a _ {h} = a ( h ) = \sum _ {i = 1 } ^ {m} A _ {i} ( h ) \alpha _ {i} ^ {h} $( | ||
+ | $ h = 0,1, \dots $); | ||
+ | the polynomial coefficients $ A _ {i} $ | ||
+ | are respectively of degree $ n _ {i} - 1 $. | ||
− | The theorem is | + | The theorem of Skolem–Mahler–Lech asserts that if a recurrence (equivalently, a generalized power sum) has infinitely many zeros, then those zeros occur periodically. That is, given a power series $ \sum _ {h \geq 0 } a _ {h} X ^ {h} $ |
+ | representing a rational function with infinitely many zero Taylor coefficients, the set $ \{ h : {a _ {h} = 0 } \} $ | ||
+ | is a finite union of complete arithmetic progressions (cf. [[Arithmetic progression|Arithmetic progression]]), plus (a pre-period of) finitely many isolated zeros. By virtue of Ritt's quotient theorem [[#References|[a6]]], it is equivalent that an infinitude of integer zeros of a complex exponential polynomial $ a ( z ) = \sum _ {i = 1 } ^ {m} A _ {i} ( z ) { \mathop{\rm exp} } ( z { \mathop{\rm log} } \alpha _ {i} ) $ | ||
+ | is accounted for by it being divisible (in the ring of exponential polynomials) by functions $ \sin { {2 \pi ( z - r ) } / d } $. | ||
+ | |||
+ | Skolem's argument was generalized by K. Mahler to algebraic number fields, and eventually to arbitrary fields of characteristic zero by C. Lech and by Mahler. The elegant argument of J.W.S. Cassels [[#References|[a1]]] bypasses the technical complications in the chain of successive generalizations. In brief, one observes that there are rational primes $ p $ | ||
+ | so that (technically, after embedding the data in the field $ \mathbf Q _ {p} $ | ||
+ | of $ p $- | ||
+ | adic rationals) one has $ \alpha _ {i} ^ {p - 1 } \equiv 1 ( { \mathop{\rm mod} } p ) $ | ||
+ | for each root. Then, for $ 0 \leq r < p - 1 $, | ||
+ | each of the $ p - 1 $ | ||
+ | different $ p $- | ||
+ | adic exponential polynomials $ a ( r + t ( p - 1 ) ) = \sum _ {i = 1 } ^ {m} A _ {i} ( r + t ( p - 1 ) ) \alpha _ {i} ^ {r} { \mathop{\rm exp} } ( t { \mathop{\rm log} } \alpha _ {i} ^ {p - 1 } ) $ | ||
+ | is a $ p $- | ||
+ | adic analytic function in the disc $ \{ {t \in \mathbf Q _ {p} } : {| t | _ {p} < p ^ {1 - {1 / {( p - 1 ) } } } } \} $. | ||
+ | It follows that if any one of these functions has infinitely many zeros (it turns out, as few as $ n $ | ||
+ | zeros [[#References|[a3]]]) in the unit disc, then it must vanish identically, yielding the theorem with arithmetic progressions with common difference $ d = p - 1 $. | ||
+ | It follows that a recurrence sequence can have infinitely many zeros only if it is degenerate, that is, some quotient $ { {\alpha _ {i} } / {\alpha _ {j} } } $ | ||
+ | of its distinct roots is a root of unity. | ||
+ | |||
+ | The theorem is provable without visible appeal to $ p $- | ||
+ | adic analysis [[#References|[a2]]]. But a generalization, whereby if two recurrence sequences have infinitely many elements in common then they coincide along certain of their arithmetic subprogressions (see [[#References|[a4]]]), as yet, (1996) relies on a $ p $- | ||
+ | adic generalization of Schmidt's subspace theorem. A different generalization, Shapiro's conjecture, according to which two exponential polynomials with infinitely many complex common zeros must have a common exponential polynomial factor, is still (1996) mostly conjecture [[#References|[a5]]]. | ||
A general reference surveying this and other relevant material is [[#References|[a7]]]. | A general reference surveying this and other relevant material is [[#References|[a7]]]. |
Revision as of 08:14, 6 June 2020
A recurrence sequence $ ( a _ {h} ) $
of order $ n $
is a solution to a linear homogeneous recurrence relation with constant coefficients
$$ a _ {h + n } = s _ {1} a _ {h + n - 1 } + \dots + s _ {n} a _ {h} ( h = 0,1, \dots ) . $$
Thus, the generating function $ \sum _ {h \geq 0 } a _ {h} X ^ {h} $ of a recurrence is a rational function $ { {r ( X ) } / {s ( X ) } } $ where $ s ( X ) = 1 - s _ {1} X - \dots - s _ {n} X ^ {n} = \prod _ {i = 1 } ^ {m} ( 1 - \alpha _ {i} X ) ^ {n _ {i} } $, say; the polynomial $ r $ of degree less than $ n $ is determined by the initial values $ a _ {0} \dots a _ {n - 1 } $. If so, the distinct complex numbers $ \alpha _ {i} $ are called the roots of the recurrence, and the $ n _ {i} $ their multiplicities. It follows that the $ a _ {h} $ are given by a generalized power sum $ a _ {h} = a ( h ) = \sum _ {i = 1 } ^ {m} A _ {i} ( h ) \alpha _ {i} ^ {h} $( $ h = 0,1, \dots $); the polynomial coefficients $ A _ {i} $ are respectively of degree $ n _ {i} - 1 $.
The theorem of Skolem–Mahler–Lech asserts that if a recurrence (equivalently, a generalized power sum) has infinitely many zeros, then those zeros occur periodically. That is, given a power series $ \sum _ {h \geq 0 } a _ {h} X ^ {h} $ representing a rational function with infinitely many zero Taylor coefficients, the set $ \{ h : {a _ {h} = 0 } \} $ is a finite union of complete arithmetic progressions (cf. Arithmetic progression), plus (a pre-period of) finitely many isolated zeros. By virtue of Ritt's quotient theorem [a6], it is equivalent that an infinitude of integer zeros of a complex exponential polynomial $ a ( z ) = \sum _ {i = 1 } ^ {m} A _ {i} ( z ) { \mathop{\rm exp} } ( z { \mathop{\rm log} } \alpha _ {i} ) $ is accounted for by it being divisible (in the ring of exponential polynomials) by functions $ \sin { {2 \pi ( z - r ) } / d } $.
Skolem's argument was generalized by K. Mahler to algebraic number fields, and eventually to arbitrary fields of characteristic zero by C. Lech and by Mahler. The elegant argument of J.W.S. Cassels [a1] bypasses the technical complications in the chain of successive generalizations. In brief, one observes that there are rational primes $ p $ so that (technically, after embedding the data in the field $ \mathbf Q _ {p} $ of $ p $- adic rationals) one has $ \alpha _ {i} ^ {p - 1 } \equiv 1 ( { \mathop{\rm mod} } p ) $ for each root. Then, for $ 0 \leq r < p - 1 $, each of the $ p - 1 $ different $ p $- adic exponential polynomials $ a ( r + t ( p - 1 ) ) = \sum _ {i = 1 } ^ {m} A _ {i} ( r + t ( p - 1 ) ) \alpha _ {i} ^ {r} { \mathop{\rm exp} } ( t { \mathop{\rm log} } \alpha _ {i} ^ {p - 1 } ) $ is a $ p $- adic analytic function in the disc $ \{ {t \in \mathbf Q _ {p} } : {| t | _ {p} < p ^ {1 - {1 / {( p - 1 ) } } } } \} $. It follows that if any one of these functions has infinitely many zeros (it turns out, as few as $ n $ zeros [a3]) in the unit disc, then it must vanish identically, yielding the theorem with arithmetic progressions with common difference $ d = p - 1 $. It follows that a recurrence sequence can have infinitely many zeros only if it is degenerate, that is, some quotient $ { {\alpha _ {i} } / {\alpha _ {j} } } $ of its distinct roots is a root of unity.
The theorem is provable without visible appeal to $ p $- adic analysis [a2]. But a generalization, whereby if two recurrence sequences have infinitely many elements in common then they coincide along certain of their arithmetic subprogressions (see [a4]), as yet, (1996) relies on a $ p $- adic generalization of Schmidt's subspace theorem. A different generalization, Shapiro's conjecture, according to which two exponential polynomials with infinitely many complex common zeros must have a common exponential polynomial factor, is still (1996) mostly conjecture [a5].
A general reference surveying this and other relevant material is [a7].
References
[a1] | J.W.S. Cassels, "An embedding theorem for fields" Bull. Austral. Math. Soc. , 14 (1976) pp. 193–198 (Addendum: 14 (1976), 479–480) |
[a2] | G. Hansel, "Une démonstration simple du théorème de Skolem–Mahler–Lech" Theor. Comput. Sci. , 43 (1986) pp. 91–98 |
[a3] | A.J. van der Poorten, R.S. Rumely, "Zeros of -adic exponential polynomials II" J. London Math. Soc. (2) , 36 (1987) pp. 1–15 |
[a4] | A.J. van der Poorten, H.-P. Schlickewei, "Additive relations in fields" J. Austral. Math. Soc. , 51 (1991) pp. 154–170 |
[a5] | A.J. van der Poorten, R. Tijdeman, "On common zeros of exponential polynomials" L'Enseign. Math. Série , 21 (1975) pp. 57–67 |
[a6] | J.F. Ritt, "On the zeros of exponential polynomials" Trans. Amer. Math. Soc. , 31 (1929) pp. 680–686 |
[a7] | A.J. van der Poorten, "Some facts that should be better known; especially about rational functions" R.A. Mollin (ed.) , Number Theory and Applications , NATO ASI , Kluwer Acad. Publ. (1989) pp. 497–528 |
Skolem-Mahler-Lech theorem. Encyclopedia of Mathematics. URL: http://encyclopediaofmath.org/index.php?title=Skolem-Mahler-Lech_theorem&oldid=23027