View source for Newton's method

In [[numerical analysis]], '''Newton's method''' (also known as the '''Newton–[[Joseph Raphson|Raphson]] method''' or the '''Newton–Fourier method''') is an efficient [[algorithm]] for finding approximations to the zeros (or [[root (mathematics)|roots]]) of a [[Real number|real]]-valued [[function (mathematics)|function]]. As such, it is an example of a [[root-finding algorithm]]. It can also be used to find a minimum or maximum of such a function, by finding a zero in the function's first derivative,
see [[Newton's method in optimization|Newton's method as an optimization algorithm]].

==Description of the method==

The idea of the method is as follows: one starts with an initial guess which is reasonably close to the true root, then the function is approximated by its [[tangent line]] (which can be computed using the tools of [[calculus]]), and one computes the ''x''-intercept of this tangent line (which is easily done with elementary algebra). This ''x''-intercept will typically be a better approximation to the function's root than the original guess, and the method can be [[iterative method|iterated]].

[[Image:newton_iteration.png|alt Illustration of Newton's method|thumb|right|300px|An illustration of one iteration of Newton's method (the function ''f'' is shown in blue and the tangent line is in red). We see that <math>x_{n+1}</math> is a better approximation than <math>x_n</math> for the root <math>x</math> of the function <math>f</math>.]]

Suppose ''f'' : [''a'', ''b''] → '''R''' is a [[derivative|differentiable]] function defined on the [[interval (mathematics)|interval]] [''a'', ''b''] with values in the [[real number]]s '''R'''. The formula for converging on the root can be easily derived. Suppose we have some current approximation ''x''<sub>n</sub>. Then we can derive the formula for a better approximation, ''x''<sub>n+1</sub> by referring to the diagram on the right.
We know from the definition of the derivative at a given point that it is the slope of a tangent at that point.

That is
:<math>f'(x_{n}) = \frac{ \mathrm{rise} }{ \mathrm{run} } = \frac{ \mathrm{\Delta y} }{ \mathrm{\Delta x} } = \frac{ f( x_{n} ) - 0 }{ x_{n} - x_{n+1} } = \frac{0 - f(x_{n})}{(x_{n+1} - x_{n})}\,\!</math>.
Here, ''f''&nbsp;' denotes the [[derivative]] of the function ''f''. Then by simple algebra we can derive
:<math>x_{n+1} = x_n - \frac{f(x_n)}{f'(-x_n)}\,\!</math>.
We start the process off with some arbitrary initial value  ''x''<sub>0</sub>.  (The closer to the zero, the better.  But, in the absence of any intuition about where the zero might lie, a "guess and check" method might narrow the possibilities to a reasonably small interval by appealing to the [[intermediate value theorem]].)  The method will usually converge, provided this initial guess is close enough to the unknown zero, and that <math>f'(x_0) \neq 0</math>. Furthermore, for a zero of [[multiplicity]] 1, the convergence is at least quadratic (see [[rate of convergence]]) in a [[neighbourhood (mathematics)|neighbourhood]] of the zero, which intuitively means that the number of correct digits roughly at least doubles in every step. More details can be found in the [[#Analysis|analysis section]] below.

==Example==

Consider the problem of finding the positive number ''x'' with cos(''x'') = ''x''<sup>3</sup>. We can rephrase that as finding the zero of ''f''(''x'') = cos(''x'') − ''x''<sup>3</sup>. We have ''f''&nbsp;'(''x'') = −sin(''x'') − 3''x''<sup>2</sup>. Since cos(''x'') ≤ 1 for all ''x'' and ''x''<sup>3</sup> > 1 for ''x''>1, we know that our zero lies between 0 and 1. We try a starting value of ''x''<sub>0</sub> = 0.5.

:<math>\begin{matrix}
  x_1 & = & x_0 - \frac{f(x_0)}{f'(x_0)} & = & 0.5 - \frac{\cos(0.5) - 0.5^3}{-\sin(0.5) - 3 \times 0.5^2} & = & 1.112141637097 \\
  x_2 & = & x_1 - \frac{f(x_1)}{f'(x_1)} & & \vdots & = & \underline{0}.909672693736 \\
  x_3 & & \vdots & & \vdots & = & \underline{0.86}7263818209 \\
  x_4 & & \vdots & & \vdots & = & \underline{0.86547}7135298 \\
  x_5 & & \vdots & & \vdots & = & \underline{0.8654740331}11 \\
  x_6 & & \vdots & & \vdots & = & \underline{0.865474033102}
\end{matrix}
</math>

The correct digits are underlined in the above example. In particular, ''x''<sub>6</sub> is correct to the number of decimal places given. We see that the number of correct digits after the decimal point increases from 2 (for ''x''<sub>3</sub>) to 5 and 10, illustrating the quadratic convergence. 
<!--
Consider the following example in pseudocode.

  '''function''' newtonIterationFunction(x) {
     '''return'''  x - (cos(x) - x^3) / (-sin(x) - 3*x^2)     
  }  
  
  '''var''' x := 0.5
  
  '''for''' i '''from''' 0 '''to''' 99 {
      print "Iteration: " + i
      print "Guess: " + x
      xold := x
      x := newtonIterationFunction(x) 
      '''if''' x = xold {
          print "Solution found!"
          '''break'''    
      }    
  }

Here is the same using a calculator.

<math>\begin{matrix}
\mbox{What to type} & & & \\
0.5\;\mbox{(enter)} & = & x_0 & = & 0.5 \\
Ans - \frac{\cos (Ans) - Ans^3}{-\sin (Ans) - 3 * Ans^2}\;\mbox{(enter)} & = & x_1 & = & 1.1121416371 \\
\mbox{(enter)} & = & x_2 & = & 0.909672693736 \\
\mbox{(enter)} & = & x_3 & = & 0.867263818209 \\
\vdots & & \vdots & & \vdots
\end{matrix}
</math>
-->

==History==

Newton's method was described by [[Isaac Newton]] in ''De analysi per aequationes numero terminorum infinitas'' (written in [[1669]], published in [[1711]] by [[William Jones (mathematician)|William Jones]]) and in ''De metodis fluxionum et serierum infinitarum'' (written in [[1671]], translated and published as ''[[Method of Fluxions]]'' in [[1736]] by [[John Colson]]). However, his description differs substantially from the modern description given above: Newton applies the method only to polynomials. He does not compute the successive approximations ''x''<sub>''n''</sub>, but computes a sequence of polynomials and only at the end, he arrives at an approximation for the root ''x''. Finally, Newton views the method as purely algebraic and fails to notice the connection with calculus. Isaac Newton probably derived his method from a similar but less precise method by [[François Viète]]. The essence of Viète's method can be found in the work of the [[Iran|Persian]] [[mathematician]] [[Sharaf al-Din al-Tusi]].

[[Heron of Alexandria]] used essentially the same method in book 1, chapter 9, of his ''Metrica'' to determine the square root of 720.

Newton's method was first published in [[1685]] in ''A Treatise of Algebra both Historical and Practical'' by [[John Wallis]]. In [[1690]], [[Joseph Raphson]] published a simplified description in ''[[Analysis aequationum universalis]]''. Raphson again viewed Newton's method purely as an algebraic method and restricted its use to polynomials, but he describes the method in terms of the successive approximations ''x''<sub>''n''</sub> instead of the more complicated sequence of polynomials used by Newton. Finally, in [[1740]], [[Thomas Simpson]] described Newton's method as an iterative method for solving general nonlinear equations using fluxional calculus, essentially giving the description above. In the same publication, Simpson also gives the generalization to systems of two equations and notes that Newton's method can be used for solving optimization problems by setting the gradient to zero.

[[Arthur Cayley]] in [[1879]] in ''The Newton-Fourier imaginary problem'' was the first who noticed the difficulties in generalizing the Newton's method to complex roots of [[polynomials]] with degree greater than 2 and complex initial values. This opened the way to the study of the theory of iterations of rational functions.

==Practical considerations==

In general the [[rate of convergence|convergence]] is '''quadratic''': the error is essentially squared at each step (that is, the number of accurate digits doubles in each step). There are some caveats, however. First, Newton's method requires that the derivative be calculated directly. (If the derivative is approximated by the [[slope]] of a line through two points on the function, the [[secant method]] results; this can be more efficient depending on how one measures computational effort.) Second, if the initial value is too far from the true zero, Newton's method can fail to converge. Because of this, most practical implementations of Newton's method put an upper limit on the number of iterations and perhaps on the size of the iterates. Third, if the [[root (mathematics)|root]] being sought has [[multiplicity]] greater than one, the convergence rate is merely linear (errors reduced by a constant factor at each step) unless special steps are taken.

==Counter examples==
In the following examples, the desired root is at zero for simplicity &mdash; it could have been placed elsewhere.

*If the derivative is not continuous at the root, then convergence may fail to occur.

Consider the function

:<math>f(x) = \begin{cases}
0 & \mbox{if } x = 0\\
x + x^2\sin(\frac{2}{x}) & \mbox{if } x \neq 0
\end{cases}</math>

Then <math>f'(0) = 1 \!</math> and <math>f'(x) = 1 + 2x\sin(2/x) - 2\cos(2/x) \!</math> elsewhere.

Within any neighborhood of the root, this derivative keeps changing sign as ''x'' approaches 0 from the right (or from the left) while 
<math>f(x) \ge x - x^2 > 0 \!</math> for <math>0 < x < 1 \!</math>. 

So <math>f(x)/f'(x) \!</math> is unbounded near the root and Newton's method will not converge, even though: the function is differentiable everywhere; the derivative is not zero at the root; <math>f \!</math> is infinitely differentiable except at the root; and the derivative is bounded in a neighborhood of the root (unlike its reciprocal).

*If there is no second derivative at the root, then convergence may fail to be quadratic.

Indeed, let 
:<math>f(x) = x + x^{4/3} \!</math>
Then 
:<math>f'(x) = 1 + (4/3)x^{1/3} \!</math> 
And 
:<math>f''(x) = (4/9)x^{-2/3} \!</math> 
except when <math>x = 0 \!</math> where it is undefined. Given <math>x_n \!</math>,

:<math>x_{n+1} = x_n - \frac{f(x_n)}{f '(x_n)} = \frac{(1/3)x_n^{4/3}}{(1 + (4/3)x_n^{1/3})} \!</math>

which has approximately 4/3 times as many bits of precision as <math>x_n \!</math> has. This is less than the 2 times as many which would be required for quadratic convergence. So the convergence of Newton's method (in this case) is not quadratic, even though: the function is continuously differentiable everywhere; the derivative is not zero at the root; and <math>f \!</math> is infinitely differentiable except at the desired root.

*If the first derivative is zero at the root, then convergence will not be quadratic.
Indeed, let 
:<math>f(x) = x^2 \!</math>
Then <math>f'(x) = 2x \!</math> and consequently <math>x - f(x)/f'(x) = x/2 \!</math>. So convergence is not quadratic, even though the function is infinitely differentiable everywhere.

==Analysis==

Suppose that the function ''f'' has a zero at α, i.e., ''f''(α) = 0.

If ''f''&nbsp; is continuously differentiable and its derivative does not vanish at α, then there exists a [[topological neighborhood|neighborhood]] of α such that for all starting values ''x''<sub>0</sub> in that neighborhood, the [[sequence]] {''x''<sub>''n''</sub>} will [[limit of a sequence|converge]] to α.

If the function is continuously differentiable and its derivative does not vanish at α and it has a second derivative at α then the convergence is quadratic or faster. If the second derivative does not vanish at α then the convergence is merely quadratic.

If the derivative does vanish at α, then the convergence is usually only linear. Specifically, if ''f'' is twice continuously differentiable, <math>f'(\alpha) = 0 \!</math> and <math>f''(\alpha) \ne 0 \!</math>, then there exists a neighborhood of α such that for all starting values ''x''<sub>0</sub> in that neighborhood, the sequence of iterates converges linearly, with [[rate of convergence|rate]] log<sub>10</sub>&nbsp;2 (Süli & Mayers, Exercise 1.6). Alternatively if <math>f'(\alpha) = 0 \!</math> and <math>f'(x) \ne 0 \!</math> elsewhere, in a [[topological neighborhood|neighborhood]] ''U'' of α, α being a zero of [[multiplicity]] ''r'' and if <math>f \in C^r(U)</math> then there exists a neighborhood of α such that for all starting values ''x''<sub>0</sub> in that neighborhood, the sequence of iterates converges linearly.

However, even linear convergence is not guaranteed in pathological situations.

In practice these results are local and the neighborhood of convergence are not known ''a priori'', but there are also some results on global convergence, for instance, given a right neighborhood ''U<sub>+</sub>'' of α, if ''f'' is twice differentiable in ''U<sub>+</sub>'' and if <math>f' \ne 0 \!</math>, <math>f \cdot f'' > 0 \!</math> in ''U<sub>+</sub>'', then, for each ''x<sub>0</sub>'' in ''U<sub>+</sub>'' the sequence ''x<sub>k</sub>'' is monotonically decreasing to α.

==Generalizations==
===Nonlinear systems of equations===
One may use Newton's method also to solve systems of ''k'' (non-linear) equations, which amounts to finding the zeros of continuously differentiable functions ''F'' : '''R'''<sup>''k''</sup> <tt>&rarr;</tt> '''R'''<sup>''k''</sup>. In the formulation given above, one then has to left multiply with the inverse of the ''k''-by-''k'' [[Jacobian]] [[matrix (mathematics)|matrix]] ''J''<sub>''F''</sub>(''x''<sub>''n''</sub>) instead of dividing by ''f''&nbsp;'(''x''<sub>''n''</sub>). Rather than actually computing the inverse of this matrix, one can save time by solving the [[system of linear equations]] 

:<math>J_F(x_n) (x_{n+1} - x_n) = -F(x_n)\,\!</math>

for the unknown ''x''<sub>''n''+1</sub> − ''x''<sub>''n''</sub>.
Again, this method only works if the initial value ''x''<sub>''0''</sub> is close enough to the true zero. Typically,  a region which is [[well-behaved]] is located first with some other method and Newton's method is then used to "polish" a root which is already known approximately.

===Nonlinear equations in a Banach space===
Another generalization is the Newton's method to find a zero of a function ''F'' defined in a [[Banach space]]. In this case the formulation is
:<math>X_{n+1}=X_n-(F'_{X_n})^{-1}[F(X_n)]</math>,
where <math>F'_{X_n}</math> is the [[Fréchet derivative]] applied at the point <math>X_n</math>. One needs the Fréchet derivative to be invertible at each <math>X_n</math> in order for the method to be applicable.

===Complex functions===
[[image:newtroot_1_0_0_0_0_m1.png|thumb|Basins of attraction for x<sup>5</sup> &minus; 1 = 0; darker means more iterations to converge.]]
{{main|Newton fractal}}
When dealing with [[complex analysis|complex functions]], however, Newton's method can be directly applied to find their zeros. For many complex functions, the boundary of the set (also known as the [[basin of attraction]]) of all starting values that cause the method to converge to a particular zero is a [[fractal]].

==References==

* Tjalling J. Ypma, Historical development of the Newton-Raphson method, ''SIAM Review'' '''37''' (4), 531&ndash;551, 1995.
* P. Deuflhard, ''Newton Methods for Nonlinear Problems. Affine Invariance and Adaptive Algorithms.'' Springer Series in Computational Mathematics, Vol. 35. Springer, Berlin, 2004. ISBN 3-540-21099-7.
* C. T. Kelley, ''Solving Nonlinear Equations with Newton's Method'', no 1 in Fundamentals of Algorithms, SIAM, 2003. ISBN 0-89871-546-6.
* J. M. Ortega, W. C. Rheinboldt, ''Iterative Solution of Nonlinear Equations in Several Variables.'' Classics in Applied Mathematics, SIAM, 2000. ISBN 0-89871-461-3.
* W. H. Press, B. P. Flannery, S. A. Teukolsky, W. T. Vetterling, ''[[Numerical Recipes|Numerical Recipes in C: The Art of Scientific Computing]]'', Cambridge University Press, 1992. ISBN 0-521-43108-5 (online, with code samples: [http://www.nrbook.com/a/bookcpdf.html]), sections 9.4 [http://www.nrbook.com/a/bookcpdf/c9-4.pdf] and 9.6 [http://www.nrbook.com/a/bookcpdf/c9-6.pdf].
* W. H. Press, B. P. Flannery, S. A. Teukolsky, W. T. Vetterling, ''[[Numerical Recipes|Numerical Recipes in  Fortran]]'', Cambridge University Press, 1992. ISBN 0-521-43064-X (online, with code samples: [http://library.lanl.gov/numerical/bookfpdf/f9-4.pdf])
* Endre Süli and David Mayers, ''An Introduction to Numerical Analysis'', Cambridge University Press, 2003. ISBN 0-521-00794-1.

==External links==

* [http://mathworld.wolfram.com/NewtonsMethod.html Newton's method on Wolfram.com]
* [http://math.fullerton.edu/mathews/a2001/Animations/RootFinding/NewtonMethod/NewtonMethod.html Animations for Newton's method]
* [http://twt.mpei.ac.ru/mas/worksheets/newton.mcd Newton's method on the Mathcad Application Server] (with animation)
* [http://numericalmethods.eng.usf.edu/topics/newton_raphson.html Newton-Raphson Method Notes, PPT, Mathcad, Maple, Matlab, Mathematica] at [http://numericalmethods.eng.usf.edu Holistic Numerical Methods Institute]
*[http://math.fullerton.edu/mathews/n2003/Newton'sMethodMod.html Module for Newton’s Method by John H. Mathews]
* [http://plus.maths.org/issue9/puzzle/solution.html worked example]

==See also==

* [[Euler method]]
* [[Integer square root]]
* [[Methods of computing square roots]]

[[Category:Optimization algorithms]]
[[Category:Root-finding algorithms]]

[[ar:طريقة نيوتن]]
[[bg:Метод на Нютон]]
[[ca:Mètode de Newton]]
[[cs:Metoda tečen]]
[[da:Newtons metode]]
[[de:Newton-Verfahren]]
[[es:Método de Newton]]
[[fr:Méthode de Newton]]
[[it:Metodo delle tangenti]]
[[he:שיטת ניוטון-רפסון]]
[[nl:Newton-Raphson]]
[[ja:ニュートン法]]
[[pl:Metoda Newtona]]
[[pt:Método de Newton]]
[[fi:Newtonin menetelmä]]
[[sv:Newtons metod]]
[[zh:牛顿法]]