Stretched exponential function: Difference between revisions

Revision as of 17:13, 20 December 2021

Latest revision as of 13:38, 24 July 2025

(28 intermediate revisions by 17 users not shown)

Line 1:

{{Short description|Mathematical function common in physics}}

[[Image:Pibmasterplot.png|325px|thumb|'''Figure 1'''. Illustration of a stretched exponential fit (with ''β''=0.52) to an empirical master curve. For comparison, a least squares single and a [[Double exponential function|double exponential]] fit are also shown. The data are rotational [[anisotropy]] of [[anthracene]] in [[polyisobutylene]] of several [[molecular mass]]es. The plots have been made to overlap by dividing time (''t'') by the respective characteristic [[time constant]].]]

[[File:Stretched exponential.svg|325px|thumb|'''Figure 1'''. Plot of {{math|1=''f''_''β''(''t'') = ''e''^{−''t''^''β''}}} for varying values of ''β'', with stretched exponentials ({{math|1=''β'' < 1}}) in reddish colors, compressed exponentials ({{math|1=''β'' > 1}}) in green and blue colors, and the standard exponential function in yellow. The [[degenerate case]]s {{math|1=''β'' → 0}} and {{math|1=''β'' → +∞}} are marked in dotted lines.]]

The '''stretched exponential function''' <math display="block">f_\beta (t) = e^{ -t^\beta }</math> is obtained by inserting a fractional [[power law]] into the [[exponential function]]. In most applications, it is meaningful only for arguments {{mvar|t}} between 0 and +∞. With {{math|1=''β'' = 1}}, the usual exponential function is recovered. With a ''stretching exponent'' ''β'' between 0 and 1, the graph of log&nbsp;''f'' versus ''t'' is characteristically ''stretched'', hence the name of the function. The '''compressed exponential function''' (with {{math|1=''β'' > 1}}) has less practical importance, with the notable exceptions of {{math|1=''β'' = 2}}, which gives the [[normal distribution]], and of compressed exponential relaxation in the dynamics of [[amorphous solids]].<ref>{{Cite journal |last1=Trachenko |first1=K. |last2=Zaccone |first2=A.|date=2021-06-14 |title=Slow stretched-exponential and fast compressed-exponential relaxation from local event dynamics |url=https://iopscience.iop.org/article/10.1088/1361-648X/ac04cd |journal=Journal of Physics: Condensed Matter |language=en |volume=33 |issue= |pages=315101 |doi= 10.1088/1361-648X/ac04cd|bibcode= |issn=0953-8984|arxiv=2010.10440 }}</ref>

The '''stretched exponential function'''

:<math>f_\beta (t) = e^{ -t^\beta }</math>

is obtained by inserting a fractional [[power law]] into the [[exponential function]].

In most applications, it is meaningful only for arguments ''t'' between 0 and +∞. With ''β''&nbsp;=&nbsp;1, the usual exponential function is recovered. With a ''stretching exponent'' ''β'' between 0 and 1, the graph of log&nbsp;''f'' versus ''t'' is characteristically ''stretched'', hence the name of the function. The '''compressed exponential function''' (with ''β''&nbsp;>&nbsp;1) has less practical importance, with the notable exception of ''β''&nbsp;=&nbsp;2, which gives the [[normal distribution]].

In mathematics, the stretched exponential is also known as the [[Cumulative distribution function#Complementary cumulative distribution function (tail distribution)|complementary cumulative]] [[Weibull distribution]]. The stretched exponential is also the [[characteristic function (probability theory)|characteristic function]], basically the [[Fourier transform]], of the [[stable distribution|Lévy symmetric alpha-stable distribution]].

Line 19 ⟶ 17:

| bibcode = 1854AnP...167...56K

}}.</ref> thus it is also known as the '''Kohlrausch function'''. In 1970, G. Williams and D.C. Watts used the [[Fourier transform]] of the stretched exponential to describe [[dielectric spectroscopy|dielectric spectra]] of polymers;<ref>{{cite journal

|author1=Williams, G. |author2=Watts, D. C.

|name-list-style=amp | year = 1970

| title = Non-Symmetrical Dielectric Relaxation Behavior Arising from a Simple Empirical Decay Function

Line 26 ⟶ 24:

| pages = 80–85

| doi = 10.1039/tf9706600080

|s2cid=95007734

}}.</ref> in this context, the stretched exponential or its Fourier transform are also called the '''Kohlrausch–Williams–Watts (KWW) function'''.

}}.</ref> in this context, the stretched exponential or its Fourier transform are also called the '''Kohlrausch–Williams–Watts (KWW) function'''. The Kohlrausch–Williams–Watts (KWW) function corresponds to the time ___domain charge response of the main dielectric models, such as the [[Cole–Cole equation]], the [[Cole–Davidson equation]], and the [[Havriliak–Negami relaxation]], for small time arguments.<ref>{{Cite journal |last=Holm|first=Sverre|title=Time ___domain characterization of the Cole-Cole dielectric model |journal=Journal of Electrical Bioimpedance |year=2020 |volume=11 |issue=1 |pages=101–105|doi=10.2478/joeb-2020-0015|pmid=33584910 |pmc=7851980 }}</ref>

In phenomenological applications, it is often not clear whether the stretched exponential function should be used to describe the differential or the integral distribution function—or neither. In each case, one gets the same asymptotic decay, but a different power law prefactor, which makes fits more ambiguous than for simple exponentials. In a few cases,<ref>{{cite journal

In each case, one gets the same asymptotic decay, but a different power law prefactor, which makes fits more ambiguous than for simple exponentials. In a few cases,<ref>{{cite journal

|author1=Donsker, M. D. |author2=Varadhan, S. R. S.

|name-list-style=amp | journal = Comm. Pure Appl. Math.

Line 63 ⟶ 61:

| year = 1993

| title = Stretched exponential decay at intermediate times in the one-dimensional Ising model at low temperatures

|doi = 10.1016/0378-4371(93)90015-V | bibcode = 1993PhyA..197..569B

}}</ref> it can be shown that the asymptotic decay is a stretched exponential, but the prefactor is usually an unrelated power.

== Mathematical properties ==

Line 70 ⟶ 68:

=== Moments ===

Following the usual physical interpretation, we interpret the function argument ''t'' as time, and ''f''_β(''t'') is the differential distribution. The area under the curve can thus be interpreted as a ''mean relaxation time''. One finds

:<math display="block">\langle\tau\rangle \equiv \int_0^\infty dt\, e^{-(t/\tau_K)^\beta} = {\tau_K \over \beta } \Gamma {\left({1 \overfrac 1 \beta }\right)}</math>

can thus be interpreted as a ''mean relaxation time''. One finds

where {{math|Γ}} is the [[gamma function]]. For [[exponential decay]], 〈{{math|1=⟨''τ''〉&nbsp;⟩ =&nbsp; ''τ''_''K''}} is recovered.

The higher [[moment (mathematics)|moments]] of the stretched exponential function are<ref name="Zwillinger_2014">{{cite book |author-first1=Izrail Solomonovich |author-last1=Gradshteyn |author-link1=Izrail Solomonovich Gradshteyn |author-first2=Iosif Moiseevich |author-last2=Ryzhik |author-link2=Iosif Moiseevich Ryzhik |author-first3=Yuri Veniaminovich |author-last3=Geronimus |author-link3=Yuri Veniaminovich Geronimus |author-first4=Michail Yulyevich |author-last4=Tseytlin |author-link4=Michail Yulyevich Tseytlin |author-first5=Alan |author-last5=Jeffrey |editor-first1=Daniel |editor-last1=Zwillinger |editor-first2=Victor Hugo |editor-last2=Moll |editor-link2=Victor Hugo Moll |translator=Scripta Technica, Inc. |title=Table of Integrals, Series, and Products |publisher=[[Academic Press, Inc.]] |date=2015 |orig-year=October 2014 |edition=8 |language=English |isbn=978-0-12-384933-5 |lccn=2014010276 <!-- |url=https://books.google.com/books?id=NjnLAwAAQBAJ |access-date=2016-02-21-->|title-link=Gradshteyn and Ryzhik |chapter=3.478. |page=372}}</ref>

:<math>\langle\tau\rangle \equiv \int_0^\infty dt\, e^{-(t/\tau_K)^\beta} = {\tau_K \over \beta } \Gamma \left({1 \over \beta }\right)</math>

: <math display="block">\langle\tau^n\rangle \equiv \int_0^\infty dt\, t^{n-1}\, e^{-(t/\tau_K)^\beta} = {{\tau_K}^n \over \beta }\Gamma {\left({\frac n \over \beta }\right)}.</math>

=== Distribution function ===

where Γ is the [[gamma function]]. For exponential decay, 〈''τ''〉&nbsp;=&nbsp;''τ''_''K'' is recovered.

In physics, attempts have been made to explain stretched exponential behaviour as a linear superposition of simple exponential decays. This requires a nontrivial distribution of relaxation times, ''ρ''(''u)''), which is implicitly defined by

The higher [[moment (mathematics)|moments]] of the stretched exponential function are<ref name="Zwillinger_2014">{{cite book |author-first1=Izrail Solomonovich |author-last1=Gradshteyn |author-link1=Izrail Solomonovich Gradshteyn |author-first2=Iosif Moiseevich |author-last2=Ryzhik |author-link2=Iosif Moiseevich Ryzhik |author-first3=Yuri Veniaminovich |author-last3=Geronimus |author-link3=Yuri Veniaminovich Geronimus |author-first4=Michail Yulyevich |author-last4=Tseytlin |author-link4=Michail Yulyevich Tseytlin |author-first5=Alan |author-last5=Jeffrey |editor-first1=Daniel |editor-last1=Zwillinger |editor-first2=Victor Hugo |editor-last2=Moll |translator=Scripta Technica, Inc. |title=Table of Integrals, Series, and Products |publisher=[[Academic Press, Inc.]] |date=2015 |orig-year=October 2014 |edition=8 |language=English |isbn=978-0-12-384933-5 |lccn=2014010276 <!-- |url=https://books.google.com/books?id=NjnLAwAAQBAJ |access-date=2016-02-21-->|title-link=Gradshteyn and Ryzhik |chapter=3.478. |page=372}}</ref>

: <math display="block">e^{-t^\beta} = \int_0^\infty du\,\rho(u)\, e^{-t/u}.</math>

Alternatively, a distribution <math display="block">G = u \rho (u)</math> is used.

: <math>\langle\tau^n\rangle \equiv \int_0^\infty dt\, t^{n-1}\, e^{-(t/\tau_K)^\beta} = {{\tau_K}^n \over \beta }\Gamma \left({n \over \beta }\right).</math>

===Distribution function===

In physics, attempts have been made to explain stretched exponential behaviour as a linear superposition of simple exponential decays. This requires a nontrivial distribution of relaxation times, ''ρ(u)'', which is implicitly defined by

: <math>e^{-t^\beta} = \int_0^\infty du\,\rho(u)\, e^{-t/u}.</math>

Alternatively, a distribution

: <math>G=u \rho (u)\,</math>

is used.

''ρ'' can be computed from the series expansion:<ref>{{cite journal

| author1=Lindsey, C. P. | author2=Patterson, G. D.

| name-list-style=amp | year = 1980

| title = Detailed comparison of the Williams-Watts and Cole-Davidson functions

| journal = [[Journal of Chemical Physics]]

| volume = 73

| issue = 7

| pages = 3348–3357

| doi = 10.1063/1.440530 | bibcode = 1980JChPh..73.3348L }}.

For a more recent and general discussion, see {{cite journal

| author = Berberan-Santos, M.N., Bodunov, E.N. and Valeur, B.

Line 112 ⟶ 101:

| doi = 10.1016/j.chemphys.2005.04.006

|bibcode = 2005CP....315..171B }}.</ref>

: <math display="block"> \rho (u ) = -{ 1 \over \pi u} \sum\limits_sum_{k = 0}^\infty {(-1)^k \over k!} \sin (\pi \beta k)\Gamma (\beta k + 1) u^{\beta k}</math>

</math>

For rational values of ''β'', ''ρ''(''u'') can be calculated in terms of elementary functions. But the expression is in general too complex to be useful except for the case ''β''&nbsp;=&nbsp;1/2 where

For rational values of ''β'', ''ρ''(''u'') can be calculated in terms of elementary functions. But the expression is in general too complex to be useful except for the case {{math|1=''β''&nbsp; =&nbsp; 1/2}} where

: <math>

<math display="block">G(u) = u \rho(u) = { 1 \over 2\sqrt{\pi}} \sqrt{u} \exp(e^{-u/4)}

</math>

Figure 2 shows the same results plotted in both a [[linear]] and a [[Logarithm|log]] representation. The curves converge to a [[Dirac delta function]] peaked at {{math|1=''u''&nbsp; =&nbsp; 1}} as ''β'' approaches 1, corresponding to the simple exponential function.

{| class="wikitable" style="margin: 1em auto 1em auto"

|

{|

|-

| [[Image:KWW dist. function linear.png|300px]] || [[Image:KWW dist. funct. log.png|300px]]

|}

|-

| colspan=2 | '''Figure 2'''. Linear and log-log plots of the stretched exponential distribution function <math>G</math> vs <math>t/\tau</math>

for values of the stretching parameter ''β'' between 0.1 and 0.9.

|}

The moments of the original function can be expressed as

: <math display="block">\langle\tau^n\rangle = \Gamma(n) \int_0^\infty d\tau\, t^n \, \rho(\tau).</math>

The first logarithmic moment of the distribution of simple-exponential relaxation times is

: <math display="block">\langle\ln\tau\rangle = \left( 1 - {1 \over \beta} \right) {\rm Eu} + \ln \tau_K </math>

where Eu is the [[Euler constant]].<ref>{{cite journal

| doi = 10.1063/1.1446035

Line 161 ⟶ 143:

| issue = 6

| pages = 061510

| doi = 10.1103/physreve.65.061510

| pmid = 12188735

| bibcode = 2002PhRvE..65f1510H

| s2cid = 16276298

| url = https://semanticscholar.org/paper/05653287a5f7e7408582d81b63c5be2594c17512

}}</ref> For practical purposes, the Fourier transform may be approximated by the [[Havriliak–Negami relaxation|Havriliak–Negami function]],<ref>{{cite journal

| author = Alvarez, F., Alegría, A. and Colmenero, J.

Line 176 ⟶ 157:

| doi = 10.1103/PhysRevB.44.7306

| pmid = 9998642

| bibcode = 1991PhRvB..44.7306A }}</ref> though nowadays the numeric computation can be done so efficiently<ref>{{cite journal

| author = Wuttke, J.

| year = 2012

Line 184 ⟶ 165:

| issue = 4

| pages = 604–628

| doi = 10.3390/a5040604 | arxiv = 0911.4796

| s2cid = 15030084

| doi-access = free

}}</ref> that there is no longer any reason not to use the Kohlrausch–Williams–Watts function in the frequency ___domain.

== History and further applications ==

[[Image:Pibmasterplot.png|325px|thumb|'''Figure 13'''. Illustration of a stretched exponential fit (with ''&beta;β''=0.52) to an empirical master curve. For comparison, a least squares single and a [[Double exponential function|double exponential]] fit are also shown. The data are rotational [[anisotropy]] of [[anthracene]] in [[polyisobutylene]] of several [[molecular mass]]es.<ref>{{cite journal |last1=Sluch |first1=Mikhail I. |last2=Somoza |first2=Mark M. |last3=Berg |first3=Mark A. |title=Friction on Small Objects and the Breakdown of Hydrodynamics in Solution: Rotation of Anthracene in Poly(isobutylene) from the Small-Molecule to Polymer Limits |journal=The Journal of Physical Chemistry B |date=1 July 2002 |volume=106 |issue=29 |pages=7385–7397 |doi=10.1021/jp025549u}}</ref> The plots have been made to overlap by dividing time (''t'') by the respective characteristic [[time constant]].]]

As said in the introduction, the stretched exponential was introduced by the [[Germans|German]] [[physicist]] [[Rudolf Kohlrausch]] in 1854 to describe the discharge of a capacitor ([[Leyden jar]]) that used glass as dielectric medium. The next documented usage is by [[Friedrich Kohlrausch (physicist)|Friedrich Kohlrausch]], son of Rudolf, to describe torsional relaxation. [[A. Werner]] used it in 1907 to describe complex luminescence decays; [[Theodor Förster]] in 1949 as the fluorescence decay law of electronic energy donors.

As said in the introduction, the stretched exponential was introduced by the [[Germans|German]] [[physicist]] [[Rudolf Kohlrausch]] in 1854 to describe the discharge of a capacitor ([[Leyden jar]]) that used glass as dielectric medium. The next documented usage is by [[Friedrich Kohlrausch (physicist)|Friedrich Kohlrausch]], son of Rudolf, to describe torsional relaxation. [[A. Werner]] used it in 1907 to describe complex luminescence decays; [[Theodor Förster]] in 1949 as the fluorescence decay law of electronic energy donors.{{Citation needed|date=May 2023}}

Outside condensed matter physics, the stretched exponential has been used to describe the removal rates of small, stray bodies in the solar system,<ref>{{cite journal

Outside [[condensed matter physics]], the stretched exponential has been used to describe the removal rates of small, stray bodies in the solar system,<ref>{{cite journal

| author = Dobrovolskis, A., Alvarellos, J. and Lissauer, J.

| year = 2007

Line 200 ⟶ 184:

| issue = 2

| pages = 481–505

| doi = 10.1016/j.icarus.2006.11.024 | bibcode = 2007Icar..188..481D }}</ref> the diffusion-weighted MRI signal in the brain,<ref>{{cite journal

| author = Bennett, K. | year = 2003

| title = Characterization of Continuously Distributed Water Diffusion Rates in Cerebral Cortex with a Stretched Exponential Model

| journal = Magn. Reson. Med.

Line 207 ⟶ 191:

| issue = 4

| pages = 727–734

| doi = 10.1002/mrm.10581 | pmid = 14523958

| display-authors=etal | doi-access = free}}</ref> and the production from unconventional gas wells.<ref>{{Cite journal |last1=Valko|first1=Peter P. |last2=Lee|first2=W. John |date=2010-01-01 |title=A Better Way To Forecast Production From Unconventional Gas Wells|journal=SPE Annual Technical Conference and Exhibition |language=english | publisher=Society of Petroleum Engineers | doi=10.2118/134231-ms | isbn=9781555633004}}</ref>

=== In probability, ===

If the integrated distribution is a stretched exponential, the normalized [[Probabilityprobability distribution|probability density function]] is given by{{Citation needed|date=May 2023}}

:<math display="block"> p(\tau \mid \lambda, \beta)~d\tau = \frac{\lambda}{\Gamma(1 + \beta^{-1})} ~ e^{-(\tau \lambda)^\beta} ~ d\tau</math>

Note that confusingly some authors have been known to use the name "stretched exponential" to refer to the [[Weibull distribution]].<ref>{{cite book

:<math> p(\tau \mid \lambda, \beta)~d\tau = \frac{\lambda}{\Gamma(1 + \beta^{-1})}~e^{-(\tau \lambda)^\beta}~d\tau</math>

Note that confusingly some authors<ref>{{cite book

| author = Sornette, D.

| year = 2004

| title = Critical Phenomena in Natural Science: Chaos, Fractals, Self-organization, and Disorder}}.</ref> have been known to use the name "stretched exponential" to refer to the [[Weibull distribution]].

=== Modified functions ===

A modified stretched exponential function

:<math display="block">f_\beta (t) = e^{ -t^{\beta(t)} }</math>

with a slowly ''t''-dependent exponent ''&beta;β'' has been used for biological survival curves.<ref>{{cite journal

|author1=B. M. Weon |author2=J. H. Je

|name-list-style=amp | year = 2009

Line 231 ⟶ 214:

| journal = Biogerontology

| volume = 10

| issue = 1

| pages = 65–71

| doi = 10.1007/s10522-008-9156-4 | pmid=18560989

|s2cid=8554128

}}</ref><ref>

Line 248 ⟶ 231:

}}</ref>

=== Wireless Communicationscommunications ===

In wireless communications, a scaled version of the stretched exponential function has been shown to appear in the Laplace Transform for the interference power <math>I</math> when the transmitters' locations are modeled as a 2D [[Poisson point process|Poisson Point Process]] with no exclusion region around the receiver.<ref>{{cite journalbook

| author = Ammar, H. A., Nasser, Y. and Artail, H.

| journaltitle = 2018 IEEE International Conference on Communications (ICC)

| titlechapter = Closed Form Expressions for the Probability Density Function of the Interference Power in PPP Networks

| year = 2018

| pages = 1–6

| title = Closed Form Expressions for the Probability Density Function of the Interference Power in PPP Networks

| doi = 10.1109/ICC.2018.8422214 | arxiv = 1803.10440 }}</ref>| isbn = 978-1-5386-3180-5

| journal = 2018 IEEE International Conference on Communications (ICC)

| pagess2cid = 1-64374550

}}</ref>

| doi = 10.1109/ICC.2018.8422214 | arxiv = 1803.10440 }}</ref>

The [[Laplace Transformtransform]] can be written for arbitrary [[fading]] distribution as follows:

:<math display="block"> L_I(s) = \exp\left(-\pi \lambda \mathbb{E}{\left[g^\frac{2}{\eta} \right]} \Gamma{\left(1 - \frac{2}{\eta} \right)} s^\frac{2}{\eta}\right) = \exp\left(- t s^\beta \right)</math>

where <math>g</math> is the power of the fading, <math>\eta</math> is the [[Path loss#Loss_exponentLoss exponent|path loss exponent]], <math>\lambda</math> is the density of the 2D Poisson Point Process, <math>\Gamma(\cdot)</math> is the Gamma function, and <math>\mathbb{E}[x]</math> is the expectation of the variable <math>x</math>.{{Citation needed|date=May 2023}}

The same reference also shows how to obtain the inverse Laplace Transform for the stretched exponential <math>\exp\left(-s^\beta \right)</math> for higher order integer <math>\beta = \beta_q \beta_b </math> from lower order integers <math>\beta_a</math> and <math>\beta_b</math>.{{Citation needed|date=May 2023}}

=== Internet streaming ===

The [[Laplace Transform]] can be written for arbitrary [[fading]] distribution as follows:

:<math> L_I(s) = \exp\left(-\pi \lambda \mathbb{E}\left[g^\frac{2}{\eta} \right] \Gamma\left(1 - \frac{2}{\eta} \right) s^\frac{2}{\eta}\right) = \exp\left(- t s^\beta \right)</math>

where <math>g</math> is the power of the fading, <math>\eta</math> is the [[Path loss#Loss_exponent|path loss exponent]], <math>\lambda</math> is the density of the 2D Poisson Point Process, <math>\Gamma(\cdot)</math> is the Gamma function, and <math>\mathbb{E}[x]</math> is the expectation of the variable <math>x</math>.

The stretched exponential has been used to characterize Internet media accessing patterns, such as YouTube and other stable [[streaming media]] sites.<ref>{{Cite conference |author= Lei Guo, Enhua Tan, Songqing Chen, Zhen Xiao, and Xiaodong Zhang| title="The Stretched Exponential Distribution of Internet Media Access Patterns" |conference= PODC' 08| pages=283–294|year=2008 | doi=10.1145/1400751.1400789 }}</ref> The commonly agreed power-law accessing patterns of Web workloads mainly reflect text-based content Web workloads, such as daily updated news sites.<ref>{{cite journal |last1=Adamic|first1=Lada A. |last2=Bernardo A. |first2=Huberman |year=2000 |title=Power-Law Distribution of the World Wide Web |url= |journal=Science |volume=287 |issue=5461 |pages=2115-2115 |doi=10.1126/science.287.5461.2115a}}</ref>

The same reference also shows how to obtain the inverse Laplace Transform for the stretched exponential <math>\exp\left(-s^\beta \right)</math> for higher order integer <math>\beta = \beta_q \beta_b </math> from lower order integers <math>\beta_a</math> and <math>\beta_b</math>.

== References ==

<references/>

== External links ==

* J. Wuttke: [http://apps.jcns.fz-juelich.de/kww libkww] C library to compute the Fourier transform of the stretched exponential function

{{DEFAULTSORT:Stretched Exponential Function}}

[[Category:Exponentials]]