8
$\begingroup$

Suppose I have 3 plots a, b and c, where

a = Plot[x, {x, 0, 1}, Frame -> True, 
   FrameTicks -> {{All, None}, {All, None}}, PlotRangePadding -> None];
b = Plot[-x, {x, 1, 2}, Frame -> True, 
   FrameTicks -> {{None, All}, {All, None}}, PlotRangePadding -> None];
c = Plot[-2 + 3 x, {x, 2, 2.5}, Frame -> True, 
   FrameTicks -> {{None, All}, {All, None}}, PlotRangePadding -> None,
    Frame -> True, FrameTicks -> All];

Now I want to combine them into one, exactly as what this figure depicts. chirs That is, the final result looks like this, on which the lines connected to each other:

this

I tried to use this plotGrid function here:

plotGrid[{{a, b, c}}, 500, 300, ImagePadding -> 40]

However, the function is intentionally written for even width figures. What I want to do is different width, proportional to each plot's x ranges, i.e., width of a$:$b$:$c=$1:1:0.5$. I have also tried other ways like this:

c = Plot[-2 + 3 x, {x, 2, 2.5}, Frame -> True, 
   FrameTicks -> {{All, All}, {All, All}}, PlotRangePadding -> None, 
   Frame -> True, FrameTicks -> All, AspectRatio -> 2];
Row[Show[#, ImagePadding -> {{0, 0}, {20, 20}}] & /@ {a, b, c}]

It works but I need to adjust the figure manually, how can I make it automatically?

=======

If the above question is solved, what if I change the code to

a = Plot[-x, {x, 0, 1}, Frame -> True, 
  FrameTicks -> {{All, None}, {All, None}}, PlotRangePadding -> None]
b = Plot[x, {x, 1, 2.5}, Frame -> True, 
  FrameTicks -> {{None, All}, {All, None}}, PlotRangePadding -> None]
c = Plot[-2.5 + 3 x, {x, 2, 2.5}, Frame -> True, 
  PlotRangePadding -> None, ScalingFunctions -> {"Reverse", Identity}]

Actually this is the result I want.

Improve this question
$\endgroup$
1
  • $\begingroup$ Show[a, b, c, PlotRange -> All, GridLines -> {{1, 2}, None}, GridLinesStyle -> Dashed]? $\endgroup$
    – kglr
    Mar 26, 2017 at 0:19

1 Answer 1

Reset to default
11
$\begingroup$

Since you give an example of a bandstructure, I am going to provide the code I have used to generate them, instead of directly answering the question you asked. The code is part of a package at the end, but I will walk through the reasoning for the functions, first. My apologies if this is somewhat rambling, it was culled from a larger document.

Preliminaries

The goal is to create a plotting function that accepts a list of points, and labels, if desired, and displays a function, $f$, along the path connecting those points. By necessity, that entails crafting a Piecewise function, $g$, that we compose with the function to be plotted, $f\circ{g}$. Most of the support functionality is aimed at crafting that.

Support functions

There are five support functions: getVariables, multiDimComposition, makeFunction, arcLength, and paramPath.

getVariables

The built-in function Variables is specifically geared towards polynomials, so it cannot extract the variables from more "exotic" functions like

In[22]:= Sin[x y^2] // Variables
(*Out[22]= {Sin[x y^2]}*)

But, getVariables is able to extract the independent variables from most nested structures, e.g.

In[9]:= getVariables @ {Exp[f[x]], Sin[x y^2]}
(*Out[9]= {x, y}*)

In[10]:= getVariables[{Exp[f[x]], Sin[x y^2]}, Hold]
(*Out[10]= {Hold[x], Hold[y]}*)

Note, getVariables is intentionally not Listable, so that the above expression can be treated as a single function. As per usual, Map can be used, if this behavior is not desirable.

Multidimensional Composition

The built-in Composition cannot handle compositions, $f\circ{g}$, where $f:\mathbf{R}^M\to\mathbf{R}^N$ and $g:\mathbf{R}^N\to\mathbf{R}$. A simple example is

In[11]:= Clear[f, g]
f[x_, y_] := Sin[2 \[Pi] x y^2]
g[s_] := {s, s^3}
Composition[f, g][s]
(*Out[14]= f[{s, s^3}]*)

So, I created multiDimComposition which can

In[15]:= multiDimComposition[f, g][s]
(*Out[15]= Sin[2 \[Pi] s^7]*)

Or, more interestingly

GraphicsRow[{Show[
   DensityPlot[f[x, y], {x, -1, 1}, {y, -1, 1}], 
   ParametricPlot[g[s], {s, -1, 1}, PlotStyle -> Black]
   ], Plot[multiDimComposition[f, g][s], {s, -1, 1}]}]

enter image description here

A more useful application is changing variables, for instance

In[9]:= Clear[f, g]
f[x_, y_, z_] := Exp[Sqrt[x^2 + y^2 + z^2]]
g[r_, t_, f_] := {r Sin[t] Cos[f], r Sin[t] Sin[f], r Cos[t]}
multiDimComposition[f, g][\[Rho], \[Theta], \[Phi]] // 
 Simplify[#, \[Rho] > 0] &
(*Out[12]= E^\[Rho]*)

As written, multiDimComposition has a flaw, as illustrated in the following:

In[13]:= multiDimComposition[f, {s, s^2}][s]
(*Out[13]= f[s]*)

So, it requires the use of functions, not expressions.

makeFunction

The function makeFunction takes an expression an turns it into a Function, e.g.

In[112]:= makeFunction[x^2]
(*Out[112]= Function[{x}, x^2]*)

In[113]:= Through @ (makeFunction /@ {x^2, Sin[x y^2], x + I y})[3, 4]
(*Out[113]= {9, Sin[48], 3 + 4 I}*)

By default, makeFunction lists the variables in the order they are encountered, but, for completeness, this can be overridden by supplying them in the second argument.

In[114]:= makeFunction[x^2,  {y, x}]
(*Out[114]= Function[{y, x}, x^2]*)

Interlude

At this point, there are enough support functions to create plotPath, and here are a few examples of its use at this stage:

plotPath[{Sin[2 \[Pi] x y^2], Cos[2 \[Pi] x y^2], Exp[x + y]}, 
 {s, s^3}, {s, -1, 1}]

enter image description here

GraphicsRow[{
  ContourPlot[Sin[x + y^2], {x, -3, 3}, {y, -2, 2}, 
   Epilog -> {Thickness[Medium], Circle[{0, 0}]}],
  plotPath[
   Sin[x + y^2], {Cos[\[Theta]], Sin[\[Theta]]}, {\[Theta], 0, 
    2 \[Pi]}]}]

enter image description here 

GraphicsRow[{
  Show[
   ContourPlot[Sin[x + y^2], {x, -3, 3}, {y, -2, 2}],
   Plot[x^2, {x, -3, 3}, 
    PlotStyle -> Directive[Thickness[Medium], Black]],
   PlotRange -> {-2, 2}
   ],
  plotPath[Sin[x + y^2], {x, x^2}, {x, -3, 3}]}]

enter image description here

But, that is unwieldy, and does not quite allow us to make a bandstructure. We need two additional functions.

arcLength and paramPath

Since writing this, there has been an ArcLength function added, but it only works along a known parameterization, and this application needs a way to calculate the length along line segments connected by known points, e.g.

In[139]:= arcLength[{{0, 0}, {1, 0}}]
arcLength[{{0, 0}, {1, 0}, {1, 1}}]
arcLength[{{0, 0}, {1, 0}, {1, 1}, {0, 0}}]
(*Out[139]= 1
Out[140]= 2
Out[141]= 2 + Sqrt[2] *)

Then, we can combine that with a function that will parameterize such a path, and we can do some interesting things.

{path, length} = {paramPath[#][s], 
     arcLength[#]} &@{{{0, 0, 0}, "\[CapitalGamma]"}, {{1, 0, 0}, 
      "X"}, {{1, 1, 0}, "M"}, {{0, 0, 0}, 
      "\[CapitalGamma]"}, {{1, 1, 1}, "R"}, {{1, 0, 0}, 
      "X"}, {{1, 1, 0}, "M"}, {{1, 1, 1}, "R"}}[[All, 1]];

ParametricPlot3D[path, {s, 0, length}]

enter image description here

This is still a bit unwieldy, though:

\[CurlyEpsilon][kx_, ky_] := - 2 (Cos[\[Pi] kx] + Cos[\[Pi] ky])
plotPath[\[CurlyEpsilon][kx, ky], 
 Evaluate[paramPath[{{0, 0}, {1, 0}, {1, 1}, {0, 0}}][s]], {s, 0, 
  arcLength[{{0, 0}, {1, 0}, {1, 1}, {0, 0}}]}, Frame -> True]

enter image description here

So, we need to add a little syntactic sugar, as shown in the examples, below.

Examples

Single s-orbital with nearest neighbor hopping

plotPath[-2 ( Cos[\[Pi] kx] + Cos[\[Pi] ky] ), {{{0, 0}, 
   "\[CapitalGamma]"}, {{1, 0}, "M"}, {{1, 1}, "X"}, {{0, 0}, 
   "\[CapitalGamma]"}},
 Frame -> True, 
 FrameTicks -> {{#, #} & @ 
    Thread[{Range[-4, 4, 2], Range[-2, 2] "t"}], {Automatic, 
    Automatic}}
 ]

s-orbitals

P-orbitals with nearest neighbor hopping

Another example is p-orbitals also on a square lattice. This has two parameters, p\[Sigma] and p\[Pi], representing the two types of bonds. Note, that this function is multivalued.

orbitals = 
  2 {p\[Sigma] Cos[\[Pi] kx] + p\[Pi] Cos[\[Pi] ky], 
    p\[Pi] Cos[\[Pi] kx] + p\[Sigma] Cos[\[Pi] ky], 
    p\[Pi] (Cos[\[Pi] kx] +  Cos[\[Pi] ky])};
(* Setting -3 p\[Pi] \[Equal] p\[Sigma] \[Equal] 1 for convenience *)
\
plotPath[Evaluate[% /. {p\[Pi] -> -1/3, p\[Sigma] -> 1}], {{{0, 0}, 
   "\[CapitalGamma]"}, {{1, 0}, "M"}, {{1, 1}, "X"}, {{0, 0}, 
   "\[CapitalGamma]"}},
 Frame -> True]

p-orbitals

One primary observation from these multi-orbital Hamiltonians is the splitting of the orbitals as k changes, and this is directly related to how the local symmetry is changing with respect to k. The points where the $\pi$-orbitals cross the $\sigma$-orbital are likely accidental degeneracies as they have different group representations.

D-orbitals with nearest neighbor hopping

Or, d-orbitals on the same lattice. Note, this explicitly requires solving for the eigenvalues.

plotPath[
 Evaluate[
  Eigenvalues[{{1/
       2 (dd\[Delta] + 3 dd\[Sigma]) (Cos[kx \[Pi]] + Cos[ky \[Pi]]), 
      0, 0, 0, 
      1/2 Sqrt[
       3] (dd\[Delta] - dd\[Sigma]) (Cos[kx \[Pi]] - Cos[ky \[Pi]])},
     {0, 2 dd\[Pi] (Cos[kx \[Pi]] + Cos[ky \[Pi]]), 0, 0, 0},
     {0, 0, 2 (dd\[Pi] Cos[kx \[Pi]] + dd\[Delta] Cos[ky \[Pi]]), 0, 
      0},
     {0, 0, 0, 2 (dd\[Delta] Cos[kx \[Pi]] + dd\[Pi] Cos[ky \[Pi]]), 
      0},
     {1/2 Sqrt[
       3] (dd\[Delta] - dd\[Sigma]) (Cos[kx \[Pi]] - Cos[ky \[Pi]]), 
      0, 0, 0, 
      1/2 (3 dd\[Delta] + dd\[Sigma]) (Cos[kx \[Pi]] + 
         Cos[ky \[Pi]])}} /. {dd\[Sigma] -> 1, dd\[Pi] -> -1/2, 
     dd\[Delta] -> 1/3}]
  ],
 {{{0, 0}, "\[CapitalGamma]"}, {{1, 0}, "M"}, {{1, 1}, "X"}, {{0, 0}, 
   "\[CapitalGamma]"}},
 Frame -> True]

d--orbitals

Package

BeginPackage["PlotPath`"];
getVariables;
multiDimComposition;
makeFunction;
variableList;
arcLength;
paramPath;
plotPath;

Begin["`Private`"];

Clear[getVariables]
SetAttributes[getVariables, HoldFirst];
getVariables[expr_, f_:Identity,  
  Optional[excludedContexts:{__String},{"System`"}]]:=
Cases[Unevaluated[expr], 
  a_Symbol/;!(MemberQ[excludedContexts, Context[a]] || MemberQ[Attributes[a], Locked | ReadProtected]) :> f[a], 
  {0, Infinity}
]//DeleteDuplicates

Clear[multiDimComposition]
multiDimComposition[flst__]:=
 With[{fcns = Reverse@List[flst]},Fold[#2[ Sequence @@ #1 ]&, First[fcns][##], Rest[fcns]]&]

Clear[makeFunction];
SetAttributes[makeFunction, HoldAll];

(* This first form allows pure functions to be used *)
makeFunction[afcn_Function, _.]:= afcn
makeFunction[fexpr_] := makeFunction[fexpr, Automatic]
makeFunction[fexpr_, vars:{__Symbol}|Automatic]:= 
Module[{ivars = Hold[vars]},
ivars = If[ivars===Hold[Automatic],
             (* GetVariables returns {Hold[x_] ..} we want Hold[{x_ ..}] *)
                 Distribute[Sort[getVariables[fexpr, Hold]], Hold],
                 ivars
           ];
Function @@ Join[ivars, Hold[fexpr]]
]


Clear[plotPath];
Options[plotPath] = Options[Plot];

plotPath[fcn:Except[_List],args__]:=plotPath[{fcn},args]
plotPath[fcns_List, params_, {s_Symbol, smin_,smax_}, opts:OptionsPattern[]]:=
 With[{pfcn=makeFunction[params], fcnlst = makeFunction/@fcns},
  Plot @@ {
   multiDimComposition[#,pfcn][s]& /@ fcnlst,
   {s,smin,smax},
   FilterRules[{opts},Options[Plot]]
  }
 ]


Clear[arcLength];
arcLength[p1_List, p2_List]/; (Length[p1]==Length[p2]):=Norm[p2 - p1]
arcLength[p:{_List ..}]/; Check[Transpose[p];True, False]:= Plus @@ arcLength @@@ Partition[p,2,1]

Clear[paramPath]
paramPath[p1_List, p2_List][s_]/;
    (Length[p1]>= 2 && Length[p2]>= 2 && Length[p1] == Length[p2]):=
    p1 + s (p2 - p1)/Norm[p2 - p1]


paramPath[p:{_List ..}][s_] /; Check[Transpose[p];True, False] := 
    Block[{ptpairs = Partition[p, 2, 1], conds, paths, dists},
        dists = {0}~Join~Accumulate[arcLength@@@ptpairs];
        conds = dists // Partition[#,2,1]&;
        paths = paramPath[Sequence @@ #[[1]] ][s - #[[2]] ]& /@ 
                Thread[List[ptpairs, Most[dists]] ] // Transpose; 
     (* 
       This creates seperate Piecewise functions, one for the x, y, etc. coords,
       respectively.
     *)                  
     Piecewise[ {#[[1]], #[[2,1]]<= s <= #[[2,2]]}&  /@ Thread[List[#, conds]]]& /@ paths
]


(* Accepts lists of points *)
plotPath[fcn_, pts_List, opts:OptionsPattern[]]:=
Module[{s},plotPath[fcn, paramPath[ pts ][s], {s, 0, arcLength[pts]}, opts]]

(* Accepts points plus labels *)
plotPath[fcn_, pts:{{_List, _String} ..}, opts:OptionsPattern[]]:=
Module[{s, xticks, rls, ticks, xgrid, grid,tname},
 (* generate tick marks/gridlines for the labels*)
 xgrid = {0}~Join~Accumulate[arcLength@@@Partition[pts[[All,1]], 2,1]];
 xticks = Thread[{xgrid,pts[[All,2]]}];
 (* Substitute in tick and grid specifications *)
 tname = If[OptionValue[Frame], FrameTicks,Ticks];
 ticks = OptionValue[tname ];
 ticks = tname -> Which[
   ticks === None (* Don't override this one only *),
    None,
   ticks === Automatic,
    {xticks, Automatic},
   True,
    MapAt[#/.Automatic-> xticks&, ticks, If[OptionValue[Frame], 2, 1]]
 ];
 grid = GridLines -> If[
   OptionValue[GridLines]===Automatic || OptionValue[GridLines]===None,
   {xgrid, None},
   MapAt[#/.Automatic -> xgrid&,OptionValue[GridLines],1]
 ];
 rls = {ticks, grid,FilterRules[List@opts, Except[{Ticks, FrameTicks}]]};
 plotPath[fcn, Evaluate[pts[[All,1]]], Evaluate[rls]]
]
End[(*`Private`*)];
EndPackage[(*PlotPath*)];
Improve this answer
$\endgroup$
6
  • $\begingroup$ Yes, exactly. I am going to plot a band structure. You are very helpful, thanks! When I was using your code to generate my plot, a little bit weird thing happened, could you please have a look at where I was wrong? Here is the code. Thank you! $\endgroup$
    – Ryan
    Mar 26, 2017 at 4:04
  • 1
    $\begingroup$ @nix First, Im surprised it worked plotting multiple functions at once. It was only designed to plot one at a time, so I'll have to make sure it is doing the right thing. Second, you're getting a phase jump in your eigenvalues. That answer should help you fix it. $\endgroup$
    – rcollyer
    Mar 26, 2017 at 11:36
  • $\begingroup$ Thank you. But I still do not figure out how to change my code to avoid the phase jump, the example you gave is about eigenvectors, what he do seems to be normalizing. But my code is to find eigenvalues, which cannot be normalized, or information will be lost. I also look up this but it has to add extra code. Would you mind showing how to do with that please? $\endgroup$
    – Ryan
    Mar 27, 2017 at 3:32
  • 1
    $\begingroup$ @nix I'll look that over tomorrow, and let you know. $\endgroup$
    – rcollyer
    Mar 27, 2017 at 4:20
  • 1
    $\begingroup$ @denis I wish. It's been a long time since I looked at it, but had trouble figuring out how to calculate it. It would have been a nice addition. $\endgroup$
    – rcollyer
    Oct 7, 2020 at 3:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.