Hilbert and Peano Space filling Curves and Beyond:
From Squares and Cubes to Surfaces (bidimensional Manifolds) and tridimensional Manifolds






Jean-François COLONNA
[Contact me]

www.lactamme.polytechnique.fr

CMAP (Centre de Mathématiques APpliquées) UMR CNRS 7641, École polytechnique, Institut Polytechnique de Paris, CNRS, France

[Site Map, Help and Search [Plan du Site, Aide et Recherche]]
[The Y2K Bug [Le bug de l'an 2000]]
[Real Numbers don't exist in Computers and Floating Point Computations aren't safe. [Les Nombres Réels n'existent dans les Ordinateurs et les Calculs Flottants ne sont pas sûrs.]]
[Please, visit A Virtual Machine for Exploring Space-Time and Beyond, the place where you can find thousands of pictures and animations between Art and Science]
(CMAP28 WWW site: this page was created on 01/03/2023 and last updated on 11/16/2023 16:51:18 -CET-)



[en français/in french]


Contents:



1-The bidimensional Peano Surjection:

Giuseppe Peano defined the following surjection :
                    [0,1] --> [0,1]x[0,1]
Let's T being a real number defined using the base 3 :
                    T    = 0.A1A2A3...  [0,1] with Ai  {0,1,2}
Let's X(T) and Y(T) being two real functions of T defined as :
                    X(T) = 0.B1B2B3...  [0,1] with Bi  {0,1,2}
                    Y(T) = 0.C1C2C3...  [0,1] with Ci  {0,1,2}
with :
                    Bn = A2n-1 if A2+A4+...+A2n-0 is even
                    Bn = 2-A2n-1 otherwise

                    Cn = A2n if A1+A3+...+A2n-1 is even
                    Cn = 2-A2n otherwise


These two functions X(T) and Y(T) are the coordinates of a point P(T) inside the [0,1]x[0,1] square. The displayed "curve" -as little spheres- is the trajectory of P(T) when T varies from 0 (lower left corner) to 1-epsilon (upper right corner).


Here are the four first bidimensional Peano curves with an increasing number of digits {2,4,6,8}:

[See the used color set to display the parameter T]



2-The tridimensional Peano Surjection:

A tridimensional surjection can be defined :
                    [0,1] --> [0,1]x[0,1]x[0,1]
as a generalization of the bidimensional one.

Let's T being a real number defined using the base 3 :
                    T    = 0.A1A2A3...  [0,1] with Ai  {0,1,2}
Let's X(T), Y(T) and Z(T) being three real functions of T defined as:
                    X(T) = 0.B1B2B3...  [0,1] with Bi  {0,1,2}
                    Y(T) = 0.C1C2C3...  [0,1] with Ci  {0,1,2}
                    Y(T) = 0.D1D2D3...  [0,1] with Di  {0,1,2}
with :
                    Bn = A3n-2 if A3+A6+...+A3n-0 is even
                    Bn = 2-A3n-2 otherwise

                    Cn = A3n-1 if A2+A5+...+A3n-1 is even
                    Cn = 2-A3n-1 otherwise

                    Dn = A3n if A1+A4+...+A3n-2 is even
                    Dn = 2-A3n otherwise


These three functions X(T), Y(T) and Z(T) are the coordinates of a point P(T) inside the [0,1]x[0,1]x[0,1] cube. The displayed "curve" is the trajectory of P(T) -displayed as little spheres- when T varies from 0 (lower left corner) to 1-epsilon (upper right corner).


Here are the three first tridimensional Peano curves with an increasing number of digits {3,6,9}:

[See the used color set to display the parameter T]



3-The bidimensional Hilbert Curves:

Let's C1(T) being a parametric curve defined by means of 2 real functions of T (T [0,1]) X1(T) [0,1] and Y1(T) [0,1] such as :
                    X1(T=0)=0 Y1(T=0)=0 (lower left corner)
                    X1(T=1)=1 Y1(T=1)=0 (lower right corner)


Then one defines a sequence of curves Ci(T) (i >= 1) as follows :
                    Ci(T) = {Xi(T),Yi(T)}  [0,1]x[0,1] --> Ci+1(T) = {Xi+1(T),Yi+1(T)}  [0,1]x[0,1]

                    if T  [0,1/4[:
                              Xi+1(T) =   Yi(4T-0)
                              Yi+1(T) =   Xi(4T-0)
                                                   Transformation 1
                    if T  [1/4,2/4[:
                              Xi+1(T) =   Xi(4T-1)
                              Yi+1(T) = 1+Yi(4T-1)
                                                   Transformation 2
                    if T  [2/4,3/4[:
                              Xi+1(T) = 1+Xi(4T-2)
                              Yi+1(T) = 1+Yi(4T-2)
                                                   Transformation 3
                    if T  [3/4,1]:
                              Xi+1(T) = 2-Yi(4T-3)
                              Yi+1(T) = 1-Xi(4T-3)
                                                   Transformation 4

Please note that 4=2d where d=2 is the space dimension.



See a special C1(T) curve in order to understand the geometrical meaning of the 4 transformations and of their order .


Here are the five first bidimensional Hilbert curves with an increasing number of iterations :

[See the used color set to display the parameter T]


Here are some examples of Hilbert-like bidimensional curves using different generating curves :










4-The tridimensional Hilbert Curves:

Let's C1(T) being a parametric curve defined by means of 3 real functions of T (T [0,1]) X1(T) [0,1], Y1(T) [0,1] and Z1(T) [0,1] such as :
                    X1(T=0)=0 Y1(T=0)=0 Z1(T=0)=0 (lower left foreground corner)
                    X1(T=1)=1 Y1(T=1)=0 Z1(T=1)=0 (lower right foreground corner)


Then one defines a sequence of curves Ci(T) (i >= 1) as follows :
                    Ci(T) = {Xi(T),Yi(T),Zi(T)}  [0,1]x[0,1]x[0,1] --> Ci+1(T) = {Xi+1(T),Yi+1(T),Zi+1(T)}  [0,1]x[0,1]x[0,1]

                    if T  [0,1/8[:
                              Xi+1(T) =   Xi(8T-0)
                              Yi+1(T) =   Zi(8T-0)
                              Zi+1(T) =   Yi(8T-0)
                                                   Transformation 1
                    if T  [1/8,2/8[:
                              Xi+1(T) =   Zi(8T-1)
                              Yi+1(T) = 1+Yi(8T-1)
                              Zi+1(T) =   Xi(8T-1)
                                                   Transformation 2
                    if T  [2/8,3/8[:
                              Xi+1(T) = 1+Xi(8T-2)
                              Yi+1(T) = 1+Yi(8T-2)
                              Zi+1(T) =   Zi(8T-2)
                                                   Transformation 3
                    if T  [3/8,4/8[:
                              Xi+1(T) = 1+Zi(8T-3)
                              Yi+1(T) = 1-Xi(8T-3)
                              Zi+1(T) = 1-Yi(8T-3)
                                                   Transformation 4
                    if T  [4/8,5/8[:
                              Xi+1(T) = 2-Zi(8T-4)
                              Yi+1(T) = 1-Xi(8T-4)
                              Zi+1(T) = 1+Yi(8T-4)
                                                   Transformation 5
                    if T  [5/8,6/8[:
                              Xi+1(T) = 1+Xi(8T-5)
                              Yi+1(T) = 1+Yi(8T-5)
                              Zi+1(T) = 1+Zi(8T-5)
                                                   Transformation 6
                    if T  [6/8,7/8[:
                              Xi+1(T) = 1-Zi(8T-6)
                              Yi+1(T) = 1+Yi(8T-6)
                              Zi+1(T) = 2-Xi(8T-6)
                                                   Transformation 7
                    if T  [7/8,1]:
                              Xi+1(T) =   Xi(8T-7)
                              Yi+1(T) = 1-Zi(8T-7)
                              Zi+1(T) = 2-Yi(8T-7)
                                                   Transformation 8

Please note that 8=2d where d=3 is the space dimension.



See a special C1(T) curve in order to understand the geometrical meaning of the 8 transformations and of their order .


Here are the four first tridimensional Hilbert curves with an increasing number of iterations :

[See the used color set to display the parameter T]


Here are some examples of Hilbert-like tridimensional curves using different generating curves and in particular "open" knots :









[More information about Peano Curves and Infinite Knots -in english/en anglais-]
[Plus d'informations à propos des Courbes de Peano et des Nœuds Infinis -en français/in french-]



5-Tridimensional Surfaces (Bidimensional Manifolds):

Many surfaces -bidimensional manifolds- in a tridimensional space can be defined using a set of three equations:
                    X = Fx(u,v)
                    Y = Fy(u,v)
                    Z = Fz(u,v)
with:
                    u  [Umin,Umax]
                    v  [Vmin,Vmax]
For example:
                    Fx(u,v) = R.sin(u).cos(v)
                    Fy(u,v) = R.sin(u).sin(v)
                    Fz(u,v) = R.cos(u)
with:
                     u  [0,pi]
                     v  [0,2.pi]
defines a sphere with R as the radius and the origin of the coordinates as the center.


[Umin,Umax]*[Vmin,Vmax] then defined a bidimensional rectangular domain D.
                       v ^
                         |
                    V    |...... ---------------------------
                     max |      |+++++++++++++++++++++++++++|
                         |      |+++++++++++++++++++++++++++|
                         |      |+++++++++++++++++++++++++++|
                         |      |+++++++++++++++++++++++++++|
                         |      |+++++++++++++++++++++++++++|
                         |      |+++++++++++++++++++++++++++|
                         |      |+++++++++++++++++++++++++++|
                         |      |+++++++++++++++++++++++++++|
                         |      |+++++++++++++++++++++++++++|
                    V    |...... ---------------------------
                     min |      :                           :
                         |      :                           :
                         O------------------------------------------------->
                                U                           U              u
                                 min                         max



Let's define a curve that fills the [0,1]x[0,1] square :
                         ^
                         |
                       1 |---------------
                         | +++++   +++++ |
                         | +   +   +   + |
                         | +   +++++   + |
                         | +           + |
                         | +++++   +++++ |
                         |     +   +     |
                         | +++++   +++++ |
                       0 O------------------------>
                         0               1


Obviously one can define a mapping between the [0,1]x[0,1] square and the [Umin,Umax]*[Vmin,Vmax] domain :
                       v ^
                         |
                    V    |...... ---------------------------
                     max |      | ++++++++         ++++++++ |
                         |      | +      +         +      + |
                         |      | +      +         +      + |
                         |      | +      +++++++++++      + |
                         |      | +           C           + |
                         |      | ++++++++         ++++++++ |
                         |      |        +         +        |
                         |      |        +         +        |
                         |      | ++++++++         ++++++++ |
                    V    |...... ---------------------------
                     min |      :                           :
                         |      :                           :
                         O------------------------------------------------->
                                U                           U              u
                                 min                         max
Then it suffices to display only the points {Fx(u,v),Fy(u,v),Fz(u,v)} with {u,v} on the preceding curve C to fill the surface with C....


Here are some examples of this process :


Surface ==> C Curve ==> Surface filling Curve


==> ==>
==> ==>
==> ==>


==> ==>


==> ==>
==> ==>
==> ==>


==> ==>
==> ==>
==> ==>


==> ==>
==> ==>
==> ==>
==> ==>
==> ==>
==> ==>
==> ==>
==> ==>


==> ==>
==> ==>
==> ==>


==> ==>


==> ==>


==> ==>


==> ==>




6-Tridimensional Manifolds:

Many tridimensional manifolds in a tridimensional space can be defined using a set of three equations:
                    X = Fx(u,v,w)
                    Y = Fy(u,v,w)
                    Z = Fz(u,v,w)
with:
                    u  [Umin,Umax]
                    v  [Vmin,Vmax]
                    w  [Wmin,Wmax]
[Umin,Umax]*[Vmin,Vmax]*[Wmin,Wmax] then defined a tridimensional rectangular domain D.


It is obvious to generalize the preceding bidimensional process in the tridimensional space...


Here are some examples of this process :


Tridimensional Manifold ==> C Curve ==> Tridimensional Manifold filling Curve


==> ==>
==> ==>
==> ==>
==> ==>
==> ==>
==> ==>
==> ==>
==> ==>
==> ==>
==> ==>
==> ==>
==> ==>
==> ==>
==> ==>


==> ==>
==> ==>
==> ==>


==> ==>


==> ==>
==> ==>




7-Beyond:

Instead of using space-filling curves in order to fill the {u,v} and {u,v,w} bi- and tridimensional domains, obviously one can use any means available and, for example, the bi- and the tridimensional brownian motions respectively....


Here are some examples of these extended processes :







Copyright © Jean-François COLONNA, 2023-2023.
Copyright © CMAP (Centre de Mathématiques APpliquées) UMR CNRS 7641 / École polytechnique, Institut Polytechnique de Paris, 2023-2023.