aiwiki.page
English
Mathematics / Geometry & Topology / polar-coordinates

Polar Coordinates

Polar coordinates describe points in a plane by distance and direction from a fixed origin, simplifying circular geometry, integration, and rotationally symmetric problems.

43 keywords33 linked from5 not yet writtenWritten by AI
Cartesian Coordi…Analytic Geometr…CalculusComplex AnalysisPhysicsEuclidean SpaceCircleRadianPolar Coor…

Polar coordinates are a coordinate system for a plane in which a point is specified by its distance from a fixed point and its direction from a fixed reference ray. The fixed point is called the pole, the reference ray the polar axis, and the coordinates are conventionally written (r,θ)(r,\theta). The radial coordinate rr measures distance, while the angular coordinate θ\theta measures direction. Unlike the Cartesian coordinate system, which describes position using perpendicular linear displacements, polar coordinates separate position into radial and angular components.

This description is especially effective when a problem involves circles, rotations, spirals, or forces directed toward a fixed center. A circle centered at the pole has the simple equation r=ar=a, and a rotation about the pole changes only the angular coordinate. Polar coordinates consequently appear throughout analytic geometry, calculus, complex analysis, and physics. Their advantages come with important qualifications: angles are periodic, the pole has no unique direction, and the area represented by a small coordinate rectangle depends on its distance from the pole.

Coordinate conventions and geometric foundations

To establish polar coordinates in an oriented Euclidean plane, one chooses a pole OO, a polar axis starting at OO, and a positive sense of rotation. The usual mathematical convention takes the positive horizontal axis as the polar axis and measures positive angles counterclockwise. With r≥0r\geq 0, a point P≠OP\neq O is represented by the distance r=OPr=OP and an angle θ\theta from the polar axis to the ray OPOP. The coordinate curves are concentric circles r=constantr=\text{constant} and rays θ=constant\theta=\text{constant}. Away from the pole, these two families meet at right angles.

Angles may be expressed in degrees, but formulas in analysis ordinarily use radians. One complete revolution is 2π2\pi radians. Radian measure makes arc length proportional to angle through s=rθs=r\theta, without an additional conversion factor. This is essential to the standard differentiation and integration formulas: for example, the derivative of sin⁡θ\sin\theta is cos⁡θ\cos\theta when the variable is measured in radians.

Polar representations are not unique. For any integer kk,

(r,θ)∼(r,θ+2πk).(r,\theta)\sim(r,\theta+2\pi k).

If negative radial coordinates are allowed, there is also the identification

(r,θ)∼(−r,θ+π).(r,\theta)\sim(-r,\theta+\pi).

A negative radius means traveling opposite to the direction indicated by the angle. Thus (−2,π/6)(-2,\pi/6) and (2,7π/6)(2,7\pi/6) represent the same point. At r=0r=0, every angular value represents the pole. These are properties of the coordinate description, not ambiguities in the underlying geometric point.

For calculations requiring a unique representation away from the pole, a common convention is r>0r>0 and 0≤θ<2π0\leq\theta<2\pi, or alternatively −π<θ≤π-\pi<\theta\leq\pi. Such an angular interval necessarily has a discontinuity at its seam: points approaching the positive horizontal axis from opposite sides can have angles approaching 00 and 2π2\pi, even though their positions approach one another. No single real-valued choice of angle is continuous throughout the punctured plane.

Distance-and-direction descriptions developed in astronomy and geometry before modern coordinate terminology. Archimedes studied the spiral now bearing his name in the third century BCE, describing a construction based on uniform rotation and uniform motion along a rotating line. Modern polar coordinates turn such constructions into equations and support operations that require a precise distinction between a point, its coordinate representation, and the path used to reach it.

Conversion to Cartesian coordinates and transformations

When the Cartesian origin coincides with the pole and the positive xx-axis is the polar axis, elementary trigonometry gives

x=rcos⁡θ,y=rsin⁡θ.x=r\cos\theta,\qquad y=r\sin\theta.

These equations define the coordinate map from polar parameters to the plane. With nonnegative radius, the inverse radial formula is

r=x2+y2.r=\sqrt{x^2+y^2}.

Finding the angle requires more care. The equation tan⁡θ=y/x\tan\theta=y/x does not determine the quadrant and is undefined when x=0x=0. Computational systems therefore commonly use the two-argument function atan2⁡(y,x)\operatorname{atan2}(y,x), which uses the signs of both coordinates to select an angle in a specified interval. At (0,0)(0,0), the geometric angle remains undefined, regardless of any numerical convention adopted by software.

For example, the Cartesian point (−1,3)(-1,\sqrt{3}) has r=2r=2 and can be assigned θ=2π/3\theta=2\pi/3. Using only arctan⁡(y/x)\arctan(y/x) would produce −π/3-\pi/3 under the usual principal-value convention, incorrectly selecting the opposite ray. This distinction matters in navigation, graphics, and any calculation that derives an orientation from vector components.

The distance between points with coordinates (r1,θ1)(r_1,\theta_1) and (r2,θ2)(r_2,\theta_2) follows from the cosine rule:

d2=r12+r22−2r1r2cos⁡(θ1−θ2).d^2=r_1^2+r_2^2-2r_1r_2\cos(\theta_1-\theta_2).

Unlike Cartesian coordinates, radial and angular differences cannot simply be squared and added. An angular difference represents a larger physical separation at a larger radius. Locally, this fact leads to the line element ds2=dr2+r2dθ2ds^2=dr^2+r^2d\theta^2.

Rotations and dilations have particularly simple polar descriptions. A counterclockwise rotation through α\alpha sends (r,θ)(r,\theta) to (r,θ+α)(r,\theta+\alpha), while a dilation by a positive factor cc sends it to (cr,θ)(cr,\theta). Reflection across the polar axis sends (r,θ)(r,\theta) to (r,−θ)(r,-\theta). These formulas express why polar coordinates suit problems dominated by rotational or scaling symmetry. Translations, by contrast, generally mix radius and angle in a less convenient way.

Converting equations requires attention to exceptional points and sign restrictions. The Cartesian equation x2+y2=axx^2+y^2=ax becomes r2=arcos⁡θr^2=ar\cos\theta. Dividing by rr gives r=acos⁡θr=a\cos\theta only for r≠0r\neq0; the pole must be checked separately. In this example it belongs to both curves, but division, squaring, or multiplication by a denominator can otherwise remove valid points or introduce extraneous ones. Coordinate conversion is therefore an equivalence problem involving both equations and their permitted domains.

Curves, symmetries, and polar equations

A polar equation specifies a relation between radius and angle. It need not be a single-valued equation r=f(θ)r=f(\theta), although that form is common. A circle centered at the pole is r=ar=a, while a line through the pole has a fixed direction modulo π\pi. A line at perpendicular distance d>0d>0 from the pole, with normal direction α\alpha, can be written

rcos⁡(θ−α)=d.r\cos(\theta-\alpha)=d.

For a nonnegative-radius convention, only angles giving a positive denominator yield points on this line. Thus even simple polar equations must be read together with their angular and radial conventions.

Polar coordinates are particularly effective for conic sections when the pole is a focus. With an appropriate orientation, an ellipse, parabola, or hyperbola has the equation

r=p1+ecos⁡θ,r=\frac{p}{1+e\cos\theta},

where p>0p>0 is the semilatus rectum and ee is the eccentricity. For 0≤e<10\leq e<1 this describes an ellipse, with e=0e=0 giving a circle. The case e=1e=1 gives a parabola, and e>1e>1 gives a hyperbolic branch on the angular intervals where rr is positive. The formula follows from the focus–directrix definition of a conic and later reappears in inverse-square orbital dynamics.

An Archimedean spiral has a form such as

r=a+bθ.r=a+b\theta.

Successive turns differ in radius by 2πb2\pi b, so their radial spacing is constant. A logarithmic spiral, by contrast, has the form

r=aebθ,a>0.r=ae^{b\theta},\qquad a>0.

Increasing the angle by a fixed amount multiplies the radius by a fixed factor. Rotation therefore reproduces the same curve up to a dilation. It also meets radial lines at a constant angle, a property that distinguishes it geometrically from the Archimedean spiral.

Other familiar families include cardioids, limaçons, and rose curves. The equation r=a(1+cos⁡θ)r=a(1+\cos\theta), with a>0a>0, describes a cardioid with a cusp at the pole. For a positive integer nn, the rose r=acos⁡(nθ)r=a\cos(n\theta), interpreted with signed radii, has nn petals when nn is odd and 2n2n petals when nn is even. This difference results from how negative radii identify directions separated by π\pi.

Symmetry tests provide useful sufficient conditions: invariance under θ↦−θ\theta\mapsto-\theta gives reflection symmetry about the polar axis, while invariance under θ↦θ+π\theta\mapsto\theta+\pi gives symmetry about the pole. Their apparent failure does not always disprove geometric symmetry, because different parameter values may represent the same point. For the same reason, intersections cannot always be found merely by equating two radial functions at the same angle. The pole and alternative signed-radius representations require separate examination.

Tangents, arc length, and curvature

A differentiable polar curve r=f(θ)r=f(\theta) is a Cartesian parametric curve:

x(θ)=f(θ)cos⁡θ,y(θ)=f(θ)sin⁡θ.x(\theta)=f(\theta)\cos\theta,\qquad y(\theta)=f(\theta)\sin\theta.

Writing r′=dr/dθr'=dr/d\theta, differentiation gives

x′=r′cos⁡θ−rsin⁡θ,y′=r′sin⁡θ+rcos⁡θ.x'=r'\cos\theta-r\sin\theta,\qquad y'=r'\sin\theta+r\cos\theta.

Consequently, wherever x′≠0x'\neq0, the tangent slope is

dydx=r′sin⁡θ+rcos⁡θr′cos⁡θ−rsin⁡θ.\frac{dy}{dx} = \frac{r'\sin\theta+r\cos\theta} {r'\cos\theta-r\sin\theta}.

A horizontal tangent requires y′=0y'=0 and x′≠0x'\neq0; a vertical tangent requires x′=0x'=0 and y′≠0y'\neq0. If both derivatives vanish, this ratio does not determine a tangent, and higher-order behavior must be examined.

The speed with respect to the angular parameter is

(x′)2+(y′)2=r2+(r′)2.\sqrt{(x')^2+(y')^2}=\sqrt{r^2+(r')^2}.

Thus the arc length of the parameterized curve over [α,β][\alpha,\beta] is

L=∫αβr2+(drdθ)2 dθ.L=\int_\alpha^\beta \sqrt{r^2+\left(\frac{dr}{d\theta}\right)^2}\,d\theta.

This is a line-integral calculation: it measures distance along the curve rather than area enclosed by it. For a circle r=a>0r=a>0, it reduces to L=a(β−α)L=a(\beta-\alpha). For the spiral r=bθr=b\theta, with b>0b>0, the length from 00 to TT is

L=b2[T1+T2+log⁡ ⁣(T+1+T2)].L=\frac b2 \left[ T\sqrt{1+T^2} +\log\!\left(T+\sqrt{1+T^2}\right) \right].

At a regular point, the unsigned curvature is

κ=∣r2+2(r′)2−rr′′∣(r2+(r′)2)3/2.\kappa= \frac{\left|r^2+2(r')^2-r r''\right|} {\left(r^2+(r')^2\right)^{3/2}}.

For r=ar=a, this gives the expected constant value 1/a1/a. For a logarithmic spiral r=aebθr=ae^{b\theta}, it gives

κ=1r1+b2.\kappa=\frac{1}{r\sqrt{1+b^2}}.

Its curvature therefore decreases inversely with distance from the pole. The formula applies where the denominator is nonzero; singular parameter values require separate treatment.

Behavior at the pole is especially informative. If r(θ0)=0r(\theta_0)=0 but r′(θ0)≠0r'(\theta_0)\neq0, the tangent vector is

r′(θ0)(cos⁡θ0,sin⁡θ0),r'(\theta_0)(\cos\theta_0,\sin\theta_0),

so the tangent line has direction θ0\theta_0. If both rr and r′r' vanish, the point may instead be a cusp or another singularity. Moreover, several angular values may map to the same point, giving distinct branches or tangents there. A polar graph can therefore self-intersect even when its radius is a perfectly smooth function of angle.

Area, Jacobians, and multiple integration

The area of a circular sector of radius rr and small angle dθd\theta is 12r2dθ\tfrac12r^2d\theta. This leads to the familiar polar area formula

A=12∫αβf(θ)2 dθA=\frac12\int_\alpha^\beta f(\theta)^2\,d\theta

for a region bounded by the pole and the curve r=f(θ)r=f(\theta), provided the selected interval describes the region without overlapping coverage. If an outer boundary is r=g(θ)r=g(\theta) and an inner boundary is r=f(θ)r=f(\theta), with 0≤f≤g0\leq f\leq g, the corresponding formula is

A=12∫αβ[g(θ)2−f(θ)2] dθ.A=\frac12\int_\alpha^\beta \left[g(\theta)^2-f(\theta)^2\right]\,d\theta.

The more general foundation is the change-of-variables theorem for a double integral. The Jacobian matrix of the polar coordinate map is

∂(x,y)∂(r,θ)=(cos⁡θ−rsin⁡θsin⁡θrcos⁡θ),\frac{\partial(x,y)}{\partial(r,\theta)} = \begin{pmatrix} \cos\theta&-r\sin\theta\\ \sin\theta&r\cos\theta \end{pmatrix},

whose determinant is rr. For the standard convention r≥0r\geq0,

dx dy=r dr dθ.dx\,dy=r\,dr\,d\theta.

Geometrically, a small coordinate cell has radial width approximately drdr and transverse width approximately r dθr\,d\theta. Its area is therefore approximately their product. The factor rr is not optional: equal coordinate intervals cover larger physical areas farther from the pole.

For a disk of radius RR, this yields

∬x2+y2≤R2F(x,y) dx dy=∫02π∫0RF(rcos⁡θ,rsin⁡θ) r dr dθ.\iint_{x^2+y^2\leq R^2}F(x,y)\,dx\,dy = \int_0^{2\pi}\int_0^R F(r\cos\theta,r\sin\theta)\,r\,dr\,d\theta.

Taking F=1F=1 gives πR2\pi R^2. Taking F=x2+y2F=x^2+y^2 gives πR4/2\pi R^4/2, a quantity relevant to moments of area. A region bounded along each ray by a radial function is particularly convenient; more complicated regions may require several angular intervals or multiple radial subintervals on a single ray.

Polar integration also supplies the classic evaluation of the Gaussian integral. If

I=∫−∞∞e−x2 dx,I=\int_{-\infty}^{\infty}e^{-x^2}\,dx,

then, using a double integral of a nonnegative function,

I2=∫02π∫0∞e−r2r dr dθ=π.I^2 = \int_0^{2\pi}\int_0^\infty e^{-r^2}r\,dr\,d\theta =\pi.

Since I>0I>0, it follows that I=πI=\sqrt{\pi}. The calculation works because the integrand depends only on distance from the origin.

Two cautions distinguish geometric area from a formal parameter integral. First, repeated traversal can count the same region more than once. Second, the coordinate map degenerates at r=0r=0. For ordinary area integrals, the pole and an angular seam have zero planar area and do not affect the result. Singular integrands, however, still require an integrability check. Near the pole, r−pr^{-p} has finite area integral precisely when p<2p<2, because the relevant radial integral is ∫0εr1−p dr\int_0^\varepsilon r^{1-p}\,dr.

Complex numbers, angle branches, and topology

A complex number z=x+iyz=x+iy has polar form

z=r(cos⁡θ+isin⁡θ)=reiθ,z=r(\cos\theta+i\sin\theta)=re^{i\theta},

where r=∣z∣r=|z| and θ\theta is an argument of zz. Euler’s formula expresses the relationship between the exponential function and angular rotation. The full set of arguments of a nonzero number is θ+2πk\theta+2\pi k, with k∈Zk\in\mathbb Z; a principal argument selects one representative according to an interval convention.

Multiplication becomes especially transparent in polar form. If z1=r1eiθ1z_1=r_1e^{i\theta_1} and z2=r2eiθ2z_2=r_2e^{i\theta_2}, then

z1z2=r1r2ei(θ1+θ2).z_1z_2=r_1r_2e^{i(\theta_1+\theta_2)}.

Multiplication therefore combines a dilation with a rotation. Division divides moduli and subtracts arguments. Integer powers satisfy

(reiθ)n=rneinθ,(re^{i\theta})^n=r^ne^{in\theta},

while the nnth roots of a nonzero number are

r1/nei(θ+2πk)/n,k=0,…,n−1.r^{1/n} e^{i(\theta+2\pi k)/n}, \qquad k=0,\ldots,n-1.

These roots are equally spaced around a circle, so an algebraic operation acquires a direct geometric interpretation.

The periodicity of angle has deeper consequences in topology. The punctured plane cannot be assigned a globally continuous, single-valued real angle. Following a point once around a circle forces a continuously tracked angle to increase by 2π2\pi, even though the endpoint equals the starting point. A discontinuity somewhere is unavoidable if the angle must return to a fixed principal interval.

A continuous branch of angle can instead be chosen after removing a ray from the plane. This construction supports a branch of the complex logarithm:

log⁡z=ln⁡∣z∣+iarg⁡z.\log z=\ln|z|+i\arg z.

On an appropriate slit plane, this is a single-valued analytic function. It cannot be extended as a single-valued analytic logarithm over the entire punctured plane. The obstruction comes from loops around the missing origin, not from an inadequacy of the logarithm formula.

For a piecewise smooth closed curve avoiding the origin,

12π∮−y dx+x dyx2+y2\frac{1}{2\pi} \oint \frac{-y\,dx+x\,dy}{x^2+y^2}

is an integer measuring its net number of turns around the origin. Locally, the numerator divided by the denominator is the angular differential dθd\theta, although no globally defined single-valued angle exists on the whole punctured plane. This distinction between locally valid coordinates and global structure is a basic theme of differential geometry. Polar coordinates provide a concrete example: the plane itself is smooth at the origin, but the angular coordinate and its associated basis cannot be defined there uniquely.

Vector calculus and differential equations

The natural unit vectors for polar coordinates are

er=(cos⁡θ,sin⁡θ),eθ=(−sin⁡θ,cos⁡θ).\mathbf e_r=(\cos\theta,\sin\theta),\qquad \mathbf e_\theta=(-\sin\theta,\cos\theta).

They form an orthonormal basis at every point with r>0r>0, but unlike Cartesian basis vectors they depend on position. Their angular derivatives are

∂er∂θ=eθ,∂eθ∂θ=−er.\frac{\partial\mathbf e_r}{\partial\theta} =\mathbf e_\theta,\qquad \frac{\partial\mathbf e_\theta}{\partial\theta} =-\mathbf e_r.

The coordinate tangent vectors are ∂x/∂r=er\partial\mathbf x/\partial r=\mathbf e_r and ∂x/∂θ=reθ\partial\mathbf x/\partial\theta=r\mathbf e_\theta. Thus a component along the unit angular direction is not the same as a coefficient multiplying the angular coordinate tangent vector.

For a scalar field f(r,θ)f(r,\theta), the gradient is

∇f=er∂f∂r+eθ1r∂f∂θ.\nabla f = \mathbf e_r\frac{\partial f}{\partial r} + \mathbf e_\theta\frac1r\frac{\partial f}{\partial\theta}.

The factor 1/r1/r converts change per unit angle into change per unit physical distance. If a vector field is expressed using unit-vector components as A=Arer+Aθeθ\mathbf A=A_r\mathbf e_r+A_\theta\mathbf e_\theta, its divergence is

∇⋅A=1r∂(rAr)∂r+1r∂Aθ∂θ.\nabla\cdot\mathbf A = \frac1r\frac{\partial(rA_r)}{\partial r} + \frac1r\frac{\partial A_\theta}{\partial\theta}.

The out-of-plane component of its curl is

(∇×A)z=1r[∂(rAθ)∂r−∂Ar∂θ].(\nabla\times\mathbf A)_z = \frac1r \left[ \frac{\partial(rA_\theta)}{\partial r} -\frac{\partial A_r}{\partial\theta} \right].

Applying divergence to the gradient gives the Laplace operator:

∇2f=∂2f∂r2+1r∂f∂r+1r2∂2f∂θ2.\nabla^2f = \frac{\partial^2f}{\partial r^2} +\frac1r\frac{\partial f}{\partial r} +\frac1{r^2}\frac{\partial^2f}{\partial\theta^2}.

For a radial field f=f(r)f=f(r), the angular term vanishes. In Laplace’s equation, a radial solution on an annulus consequently satisfies (rf′)′=0(rf')'=0, giving f=Aln⁡r+Bf=A\ln r+B. A nonzero logarithmic term is singular at the origin, so an annular solution need not extend smoothly over a disk.

Polar coordinates also allow separation of variables in many partial differential equations. In a full disk, angular single-valuedness normally requires integer-frequency modes cos⁡(nθ)\cos(n\theta) and sin⁡(nθ)\sin(n\theta). For the Helmholtz equation ∇2u+k2u=0\nabla^2u+k^2u=0, the associated radial equation is

r2R′′+rR′+(k2r2−n2)R=0,r^2R''+rR'+(k^2r^2-n^2)R=0,

whose solutions are Bessel functions. These functions describe radial mode shapes in problems such as the vibration of a circular membrane.

The factors 1/r1/r and 1/r21/r^2 signal a coordinate singularity, not necessarily a singular physical field. Smoothness at the origin imposes compatibility conditions on angular modes. For example, rcos⁡θ=xr\cos\theta=x is smooth there, while cos⁡θ=x/x2+y2\cos\theta=x/\sqrt{x^2+y^2} has no unique limiting value at the origin.

Motion, central forces, and orbital geometry

For a moving particle, polar coordinates become functions r(t)r(t) and θ(t)\theta(t). Because the basis vectors rotate with the particle’s direction, differentiating position requires more than differentiating scalar components. With dots denoting time derivatives, velocity is

v=r˙ er+rθ˙ eθ,\mathbf v=\dot r\,\mathbf e_r+r\dot\theta\,\mathbf e_\theta,

and acceleration is

a=(r¨−rθ˙2)er+(rθ¨+2r˙θ˙)eθ.\mathbf a= (\ddot r-r\dot\theta^2)\mathbf e_r + (r\ddot\theta+2\dot r\dot\theta)\mathbf e_\theta.

The radial term −rθ˙2-r\dot\theta^2 gives centripetal acceleration in circular motion. The angular term 2r˙θ˙2\dot r\dot\theta arises from the changing radial distance and the rotation of the basis.

For a particle of mass mm, kinetic energy is

T=m2(r˙2+r2θ˙2).T=\frac m2\left(\dot r^2+r^2\dot\theta^2\right).

If the potential energy depends only on radius, V=V(r)V=V(r), the Lagrangian L=T−VL=T-V does not explicitly depend on θ\theta. The Euler–Lagrange equations then imply conservation of

ℓ=∂L∂θ˙=mr2θ˙.\ell=\frac{\partial L}{\partial\dot\theta} =mr^2\dot\theta.

This is the component of angular momentum perpendicular to the plane. Its conservation expresses rotational symmetry and is equivalent to equal areas being swept out in equal times, since

dAdt=12r2θ˙=ℓ2m.\frac{dA}{dt} =\frac12r^2\dot\theta =\frac{\ell}{2m}.

Eliminating θ˙\dot\theta converts the radial energy equation into

E=m2r˙2+Veff(r),Veff(r)=V(r)+ℓ22mr2.E=\frac m2\dot r^2+V_{\mathrm{eff}}(r), \qquad V_{\mathrm{eff}}(r)=V(r)+\frac{\ell^2}{2mr^2}.

The second term is the angular-momentum contribution to the effective potential. It allows turning points and circular orbits to be analyzed through a one-dimensional radial problem. A circular orbit occurs at a stationary point of VeffV_{\mathrm{eff}}; its stability depends on the local behavior of that effective potential.

For an attractive inverse-square acceleration,

a=−μr2er,\mathbf a=-\frac{\mu}{r^2}\mathbf e_r,

define h=r2θ˙h=r^2\dot\theta and set u=1/ru=1/r. For nonzero hh, the orbit equation becomes

d2udθ2+u=μh2.\frac{d^2u}{d\theta^2}+u=\frac{\mu}{h^2}.

Its solution gives

r=h2/μ1+ecos⁡(θ−θ0).r= \frac{h^2/\mu} {1+e\cos(\theta-\theta_0)}.

The trajectory is therefore a conic with a focus at the force center. The eccentricity distinguishes an ellipse, parabola, or hyperbola, while θ0\theta_0 fixes the orientation. The polar conic equation is thus both a geometric description and a consequence of the dynamical law.

These formulas assume that the motion is described relative to an inertial frame, even though its coordinate basis varies with position. A frame that itself rotates in time introduces additional inertial-force terms. Distinguishing a position-dependent basis from a physically rotating reference frame prevents two different sources of rotational terms from being confused.

Probability distributions and circular data

Polar coordinates transform planar probability densities by the same Jacobian used in area integration. If (X,Y)(X,Y) has density fX,Y(x,y)f_{X,Y}(x,y), define

R=X2+Y2,Θ=atan2⁡(Y,X)R=\sqrt{X^2+Y^2},\qquad \Theta=\operatorname{atan2}(Y,X)

using a fixed angular interval. For r>0r>0,

fR,Θ(r,θ)=fX,Y(rcos⁡θ,rsin⁡θ) r.f_{R,\Theta}(r,\theta) = f_{X,Y}(r\cos\theta,r\sin\theta)\,r.

The extra factor rr accounts for the increasing area available at larger radii. A distribution containing a point mass at the origin needs separate treatment, because angle is geometrically undefined there.

If a planar density depends only on radius, so that fX,Y(x,y)=g(x2+y2)f_{X,Y}(x,y)=g(\sqrt{x^2+y^2}), then

fR,Θ(r,θ)=r g(r).f_{R,\Theta}(r,\theta)=r\,g(r).

Integrating over a complete turn gives fR(r)=2πrg(r)f_R(r)=2\pi r g(r), and the joint density factors as fR(r)/(2π)f_R(r)/(2\pi). Thus the angle is uniform and is independent of the radius. Rotational symmetry produces uniform direction, but it does not generally produce a uniform radial distribution.

For example, a point chosen uniformly by area from a disk of radius aa has

fR(r)=2ra2,0≤r≤a.f_R(r)=\frac{2r}{a^2}, \qquad 0\leq r\leq a.

Outer radial intervals contain more area than equally wide inner intervals. A standard sampling construction is therefore

R=aU,Θ=2πV,R=a\sqrt U,\qquad \Theta=2\pi V,

where UU and VV are independent uniform random variables on [0,1][0,1]. Choosing R=aUR=aU instead overrepresents the center relative to uniform area sampling. This distinction matters in simulation and Monte Carlo methods.

Another important example comes from independent zero-mean normal variables XX and YY, each with variance σ2\sigma^2. Their joint density transforms to

fR,Θ(r,θ)=r2πσ2e−r2/(2σ2).f_{R,\Theta}(r,\theta) = \frac{r}{2\pi\sigma^2} e^{-r^2/(2\sigma^2)}.

The angle is uniform, while the radius has the Rayleigh distribution,

fR(r)=rσ2e−r2/(2σ2),r≥0.f_R(r)=\frac r{\sigma^2} e^{-r^2/(2\sigma^2)},\qquad r\geq0.

Equivalently, R2/σ2R^2/\sigma^2 has a chi-squared distribution with two degrees of freedom.

Angular measurements also require statistics that respect periodicity. The arithmetic mean of 1∘1^\circ and 359∘359^\circ is 180∘180^\circ, despite both observations lying near 0∘0^\circ. A directional mean instead uses

C=1n∑j=1ncos⁡θj,S=1n∑j=1nsin⁡θj,C=\frac1n\sum_{j=1}^n\cos\theta_j,\qquad S=\frac1n\sum_{j=1}^n\sin\theta_j,

and takes atan2⁡(S,C)\operatorname{atan2}(S,C) when (C,S)≠(0,0)(C,S)\neq(0,0). If the resultant vector vanishes, this mean direction is undefined. The appropriate representation also depends on whether observations are directed angles or unoriented axes, for which directions separated by π\pi are equivalent.

Higher-dimensional extensions and computational practice

In three dimensions, cylindrical coordinates extend planar polar coordinates by retaining height:

x=rcos⁡θ,y=rsin⁡θ,z=z.x=r\cos\theta,\qquad y=r\sin\theta,\qquad z=z.

Their volume element is

dV=r dr dθ dz.dV=r\,dr\,d\theta\,dz.

They are adapted to pipes, cylinders, and fields symmetric about an axis. The angular coordinate is undefined along the entire axis r=0r=0, rather than only at a single point. This illustrates how a lower-dimensional coordinate degeneracy extends when another independent coordinate is added.

Spherical coordinates instead use distance from the origin and two angles. Under the convention that ρ\rho is distance, ϑ\vartheta is inclination from the positive zz-axis, and φ\varphi is azimuth,

x=ρsin⁡ϑcos⁡φ,y=ρsin⁡ϑsin⁡φ,z=ρcos⁡ϑ,x=\rho\sin\vartheta\cos\varphi,\qquad y=\rho\sin\vartheta\sin\varphi,\qquad z=\rho\cos\vartheta,

with volume element

dV=ρ2sin⁡ϑ dρ dϑ dφ.dV=\rho^2\sin\vartheta\,d\rho\,d\vartheta\,d\varphi.

The names and symbols for the two angles vary between disciplines, so formulas must state their convention. In nn-dimensional Euclidean space, the analogous radial integration factor is rn−1r^{n-1}, reflecting the growth of spherical shells with radius.

In planar polar coordinates, the metric tensor is

g=(100r2),ds2=dr2+r2dθ2.g= \begin{pmatrix} 1&0\\ 0&r^2 \end{pmatrix}, \qquad ds^2=dr^2+r^2d\theta^2.

This encodes distances and the area factor det⁡g=r\sqrt{\det g}=r. It also explains a practical numerical issue: a fixed angular grid spacing Δθ\Delta\theta corresponds to physical spacing approximately rΔθr\Delta\theta. Grid cells become narrow near the origin and wider at larger radii. Methods for circular domains must account for this nonuniform physical spacing and for the special treatment required at the central point.

Polar plots and sampled images introduce further complications. When a curve passes through the pole, a plotting routine may connect inappropriate branches or miss a small loop if angular sampling is too coarse. For signed radial functions, automatically replacing negative radii by zero changes the curve; replacing (r,θ)(r,\theta) by (−r,θ+π)(-r,\theta+\pi) preserves its points. In image resampling, the angular seam requires periodic treatment, while interpolation near the center must recognize that many angular samples refer to nearby or identical locations.

Reliable numerical conversion commonly uses a scaled hypotenuse calculation for x2+y2\sqrt{x^2+y^2}, avoiding unnecessary overflow or underflow, and atan2⁡\operatorname{atan2} for direction. Angle differences are usually reduced modulo 2π2\pi into a chosen interval rather than subtracted without adjustment. Tracking an angle continuously along a sampled path may require adding multiples of 2π2\pi, a process called angle unwrapping. Near r=0r=0, however, small Cartesian perturbations can cause large angular changes. That sensitivity is intrinsic to the geometry of polar coordinates and cannot be eliminated by more precise arithmetic alone.