# How do I plot a proper streamline plot, including spacings and line endings?

Mathematica includes two nice built-in tools to visualize vector fields, VectorPlot and StreamPlot. The latter is a useful tool, and it plots the streamlines of the given 2D vector field.

A streamline is an integral curve of the vector field: that is, a curve $\vec\alpha(s)$ whose derivative $\frac{d\vec\alpha}{ds}$ is proportional at every point to the vector field $\vec F(\vec\alpha(s)$ at that point. These are very useful tools for understanding the qualitative behaviour of many types of vector fields.

However, in certain cases, these diagrams require a bit more structure. In situations where the divergence $\nabla\cdot\vec F$ of the field matters, and in particular for

• streamlines of a fluid flow, and

• field lines of an electric or magnetic field,

there are two constraints which are followed, by convention, very strictly, and which make available quantitative information about the field from the diagram, by encoding the field strength in the local density of streamlines. (For more on how and why they work, see this physics.SE thread). Specifically:

• Streamlines should never end or begin in any region in which $\nabla\cdot\vec F$ is zero; they should start at point sources or regions where $\nabla\cdot\vec F>0$, and end at point sinks or regions where $\nabla\cdot\vec F<0$.

• The vector field flow between adjacent streamlines, i.e. the integral $$\int_L\vec F\cdot d\vec s^\perp$$ where $L$ joins the two streamlines, should be constant. (There is some interplay when the field is in 3D, where it's the flow across unit "cell" surfaces that join three or more streamlines, or when one is displaying a 2D cross-section of a 3D field, but the principle remains.)

Unfortunately, the built-in StreamPlot function does nothing of the sort, at least out of the box, and I can find no built-in options that will enforce this type of behaviour.

For an example, consider the vector fields $\hat{\mathbf{r}}/r^n$, for $n=1$ and $n=2$. These are the electric fields produced in 3D by a line of charge and a point charge, respectively. When seen as a 2D field, the former has no divergence, and the corresponding streamlines should start at the origin and end at infinity; the latter, on the other hand, has negative divergence as a 2D field and its intensity thins out faster than the first, so the local density of field lines should thin out as $1/r$ away from the origin. If you try to plot them with Mathematica, though, the opposite happens:

Table[
StreamPlot[{x, y}/(x^2 + y^2)^n, {x, -2, 2}, {y, -2, 2}
, ImageSize -> 400, PlotLabel -> "(x,y)/r^{n+1}, n="<>ToString[2 n - 1]
]
, {n, {1, 3/2}}] // GraphicsRow


The field from the line charge, on the left, has approximately constant spacing after a while, but here are streamlines that are created almost at $r=1$. The field on the right is from a point charge, and it shows the opposite of what it should:

• streamlines are created away from the origin, instead of destroyed,
• more streamlines are created than for a line charge,
• and to top it off, the field is not even rotationally symmetric.

### Is there a way to overcome these shortcomings and make StreamPlot, with appropriate options, produce correct versions of these plots? Is there some other way to produce such a diagram properly?

• Somewhat related: (27319) – Mr.Wizard Feb 5 '14 at 18:28
• @Mr.Wizard Yeah, I saw that one, as well as (34867) and (687). The problem on that one is that a tailored method would be needed for each specific vector field, i.e. you need to figure out manually where to seed streamlines. Plus, there's no way to make them start or stop at specified locations. – Emilio Pisanty Feb 5 '14 at 18:37
• That's why I only said "somewhat." Unfortunately I don't have anything better to offer. Creating a function to automatically build a specification for StreamPoints is the only thing I can think of apart from an entirely custom Graphics expression. If you know of an algorithm to create the kind of plot you describe we can probably help you implement it. – Mr.Wizard Feb 5 '14 at 18:42
• If you can find a stream function $\psi$ for your vector field, that is, a scalar field such that $\vec F=(\partial_y\psi,-\partial_x\psi)$, then the contours of $\psi$ at equally spaced values are the streamlines you want. However, this only works if $\vec F$ is divergence-free everywhere. – Rahul Feb 5 '14 at 23:18
• @Sektor That misses the point. LineIntegralConvolutionPlot has lines appearing and disappearing in places where they shouldn't in a field line plot, and it produces asymmetric diagrams for circularly symmetric fields. It is a useful commands for many things, but it does not work for this, unless you can show how to overcome all these difficulties. – Emilio Pisanty Feb 12 '14 at 17:30

To satisfy your requirements, it is necessary for the spacing between streamlines to be inversely proportional to the vector field strength, but unfortunately Mathematica only allows us to specify a constant spacing for a streamline plot. As an ugly workaround we could try rendering lots of plots with different constant spacings and combine them according to the field strength. For example,

strengths = Table[2^i, {i, -4, 4, 1/2}] // N;
bounds = Partition[Flatten@{0, MovingAverage[strengths, 2], ∞}, 2, 1];
StreamPlot[{-1 - x^2 + y, 1 + x - y^2}, {x, -3, 3}, {y, -3, 3},
StreamPoints -> {Automatic, 2/#1},
RegionFunction -> Function[{x, y, vx, vy, n}, First@#2 <= n <= Last@#2]] &,
{strengths, bounds}], PlotRange -> {{-3, 3}, {-3, 3}}]


It's not perfect, but it does okay at indicating the field strength by the density of streamlines. On your zero-divergence example it doesn't quite manage to avoid creating or destroying streamlines, though it makes an effort:

Show[MapThread[
StreamPlot[{x, y}/(x^2 + y^2), {x, -2, 2}, {y, -2, 2},
StreamPoints -> {Automatic, 0.5/#1},
RegionFunction -> Function[{x, y, vx, vy, n}, First@#2 <= n <= Last@#2]] &,
{strengths, bounds}], PlotRange -> {{-2, 2}, {-2, 2}}]


I think that the Wolfram Demonstration Electric Field Lines Due to a Collection of Point Charges (available at in case of link rot) does a pretty good job.

• But it is single plot and the question is how to "synchronize" many. – Kuba Jun 10 '16 at 7:56
• Indeed it does. Unfortunately, it is tied to the specific structure of the field: it seeds a number of streamlines at the individual charges and then sends them off to infinity. It therefore doesn't handle disappearing streamlines well (e.g. the case of equal and opposite charges). – Emilio Pisanty Jun 10 '16 at 10:00
• @EmilioPisanty: Have you tried it? The demonstration works perfectly well in the case of equal and opposite charges. – TheDoctor Jun 11 '16 at 9:35
• @Kuba: I don't understand your comment. Note that you can add (or subtract) new charges. – TheDoctor Jun 11 '16 at 9:43
• @TheDoctor Of coure I've tried it - hence the criticism. It does not work well for equal and opposite charges, see this example - notice the duplicated lines? It's essentially claiming an electric field that's twice as strong between the charges as it should be. – Emilio Pisanty Jun 13 '16 at 13:46