Light Up is NP-complete

Brandon McPhail
February 28, 2005

(X Y)

0 0
w
a 1 b
0 0 a’ b’ 0 0
X x’ 2 z 2 y’ Y
0 0 0 0 0 0

X Y

Figure 1: An OR/XNOR gate for our encoding of logic circuits as a Light

Up puzzle.

Abstract
Light Up is one of many paper-and-pencil puzzles recently popular
in Japan (and now elsewhere). The question, “Is this Light Up puzzle
solvable?” turns out to be very hard to answer in general. We devise
a polynomial-time reduction from Circuit-SAT to Light Up to prove
that Light Up is NP-complete. We introduce Light Up and illustrate
how Boolean circuits can be encoded as Light Up puzzles. These
Light Up circuits are constructed from a collection arrangeable circuit
gadgets, like the one in Fig. 1.

1 Introduction
Light Up is one of many popular pencil-and-paper puzzles originating in
Japan. Most, if not all, paper-and-pencil puzzles consist of a grid of cells, a

1
list of rules, and some initially specified cells. The objective is to complete
the puzzle by shading in sections of the grid in compliance with the rules.
Yato[23] provides the following list of popular paper-and-pencil puzzles:

• Nurikabe[13] *

• Nonogram[19] (or Paint-by-Numbers) *

• Slither Link[22] *

• Cross Sum[17] (or Kakkuro) *

• Number Place[17] (or Sudoku) *

• Heyawake

Those marked with a * are known to be NP-complete. According to Ueda

and Nagao[19], Nonogram was the first paper-and-pencil puzzle to be proved
NP-complete, although various NP-completeness results have been found
for paper-and-pencil puzzles since then[23]. By constructing a polynomial-
time reduction from Circuit-SAT to Light Up, we present here a new NP-
completeness result for paper-and-pencil puzzles.

2 How to play Light Up

Perhaps the best way to learn how to play Light Up is to visit the website of
the puzzle’s creator, Nikoli, at:
[Link]
Many more puzzles can be printed out or played online at:
[Link]

3 The decision problem

Light Up puzzles seem hard to solve. We characterize the decision problem
for Light Up as, “Given a Light Up puzzle, does a solution exist?”
The complexity class P comprises of decision problems we can answer
in reasonable (polynomial ) time. Those decision problems we can verify in

2
reasonable (polynomial ) time belong to the complexity class NP. A problem
is NP-hard if any other problem in NP can be reduced to it. A problem is
NP-complete if it is in NP and is NP-hard.
Light Up is easily seen to be in NP. Given a Light Up puzzle and a
placement of lights, we can quickly determine whether each of the rules has
been satisfied.

4 Theorem: Light Up is NP-hard

Any problem in NP can be reduced in time polynomial in the size of the inputs
to the problem of satisfying Boolean circuits (also known as Circuit-SAT).
By demonstrating a polynomial-time reduction of Circuit-SAT to Light Up,
it follows that any problem in NP can be reduced to Light Up, that is, Light
Up is NP-hard.
We present a proof by construction, similar to the techniques used by
Kaye[11], Friedman[6], and Moore and Robson[14]. Our goal is to model
the properties of a Boolean circuit using only the rules of Light Up. Given
a Boolean circuit, we will construct a corresponding circuit on a Light Up
board in time polynomial in the number of squares in the puzzle grid.

1 1

X x’ 1 x x’ 1 x x’ 1 x x’ 1 x x’ 1 x X

0 0 0 0
X x’ 1 x x’ 1 x X
0 0 0 0

Figure 2: The wire tile (top) allows only two possible places for a light. We
can string together wire tiles to propagate a Boolean signal (middle). Note
that we can also “stretch” a wire tile by inserting arbitrarily long regions
surrounded by zeros (bottom).

3
5 Proof by construction
First, we construct a grid large enough to contain our circuit. Unless other-
wise specified, all cells in our grid will be black. As a Boolean circuit consists
of distinct gadgets, so too will our Light Up puzzles be divided into separate
parts. We refer to each part or gadget as a tile.

5.1 Wire construction

All of our circuits will be constructed on a simple tiling of our gadgets that
is consistent with the rules of Light Up.
The wire tile has one of two possible states. If the x cells are assigned
lights, we say the state is true. If instead the x’ cells are assigned lights, we
say the state is false. We can string these tiles together to propagate this
truth assignment. Note that lining our wires up with the other gadgets in
our grid-based circuit is easy, since, as demonstrated in Fig. 3, we can stretch
and even bend the wires as we see fit.

X x’ 1 x x’ 1 x x’ 1 x x’ 1 x x’ 1 x x’ 1
x
x’
0 0 0 0 0 0 1
1 x’ x 1 x
x 0 0 0 0 0 0 x’ 0 0
x’ 0 0 0 0
1 0 0 0 0
x 0 0 0 0
x’ x x’
1 1 x’ x 1 x’ x 1
x

Figure 3: By introducing a set of corner tiles, we can bend our wires to take
more interesting and useful paths.

4
 0 0 0 0 0 0 0 0
1 x X 1 x X
x’ 0 0 0 0 x’ 0 0 0 0
0 0 0 0 1 x 0 0 0 0 0 0 0 0 1 x
X x’ 2 x’ X X x’ 2 x’ x
0 0 0 0 1 x 0 0 0 0 0 0 0 0 1 x
x’ 0 0 0 0 x’ 0 0 0 0
1 x X 1 x X
0 0 0 0 0 0 0 0

x’
0 0 0 0 1 x 0 0 0 0
X x’ 2 x’ X
0 0 0 0 1 x 0 0 0 0
x’

Figure 4: The branch/NOT gate (top left) allows us to split a wire into a
signal and its complement. We can modify the tile in Fig. 4 to produce a
NOT gate (bottom) and a branch gate (top right).

0 0 0 0
0
0 0 0 0

Figure 5: We can fix terminate both ends of this wire to form a complete
Boolean circuit. The 0 on the right forces us to place a light bulb at the
beginning of the wire on the left; we have found a satisfying truth assignment
for this circuit.

5.2 Assigning values

A Boolean circuit is satisfiable if and only if our Light Up puzzle has a
solution. To ensure this bijection, we fix the final output wire to true. The
input wires may take on different values, and if we find any placement of
lights for them that results in a solution, this will correspond to a satisfying
truth assignment for our circuit.
Fig. 5 shows a complete Boolean circuit corresponding the Boolean ex-
pression x consisting only of a single literal. By capping the right end of the
wire with a 0, we’ve actually forced the final output to be true.1 The only
satisfying assignment, of course, it to set x := true, which in our Light Up
puzzle means placing a light bulb in the left end of the wire.
1
If we had capped the wire instead with a 1, we would have forced the final output of
the wire to be false.

5
5.3 Branch and NOT gates
We’d like to split our wires to allow the output of one gadget to form the
input for multiple other gadgets. The gadget in Fig. 4 splits our wire into
three wires, but the signal is flipped in the middle outgoing wire. We’ll call
this our branch/NOT gate. If we cap the middle outgoing wire of the
branch/NOT gate, we get a branch gate. If we instead cap the top and
bottom outgoing wires of the branch/NOT gate, we are left with just a
NOT gate.
X Y

X Y 0 0 0 0
(X Y) x’ z’ y’
1 x x’ 2 z 2 y’ y 1
0 0 0 0 a’ b’
0 0 0 0 w’ 0 0 a 1 b
w 0 0 w 0 0 w
a 1 b 0 0 a 1 b 0 0 1 w’ 1
0 0 a’ b’ 0 0 x’ a’ b’ y’ w’ w 2 w w’
X x’ 2 z 2 y’ Y 1 x x’ 2 z 2 y’ y 1 w’
0 0 0 0 0 0 0 0 0 0

X Y X Y X Y

Figure 6: Given inputs x and y, the OR/XNOR gate (left) outputs both the
exlusive NOR and inclusive OR of the inputs. Capping one of these outputs
produces an OR gate (middle) or an XOR gate (right).

5.4 The OR gate

The OR/XNOR gate, like the branch/NOT gate, is a two-for-one logic
gate. Given two input wires (we’ll label them x and y), the OR/XNOR gate
outputs both the exclusive NOR ¬(x ⊕ y) = (x ∧ y) ∨ (¬x ∧ ¬y) and the
inclusive OR x ∨ y. We can of course cap one of these two outgoing wires to
transform the OR/XNOR gate into just an XNOR gate or just an inclusive
OR gate.

6
 X Y

Y X
Figure 7: From XOR gates like the one in Fig. 6, we can construct a wire
crossing.

5.5 Important details

Our circuit lies in a plane, so we need to either describe explicitly how wires
should cross without interacting or show that wire crossing aren’t necessary.
As it turns out, we already have sufficient gadgetry to build a wire crossing.
From an XNOR gate and a NOT gate, we can build an XOR gate. We can
then use 3 branch and 3 XOR gates to allow two wires to cross.

5.6 An example construction

At this point, we have enough gadgetry to “embed” all possible Boolean
circuits in Light Up puzzles. As an example, a satisfying assignment can be
found for the Boolean expression

¬x ∨ ((x ∧ y) ∨ z)

if and only if the Light Up puzzle in Fig. 5.6 has a solution.

7
 1 1 0 0 0 0
1 2 1 0 0 0 0
0 0 0
0 0 0 0 1 1 0 0
0 0 0 0 0 0 0
0 0 1 1 1 1 0 0
0 0 2 2 0 0
0 0 0 0
0 0 0 0 1 0 0 0 0
0 0 0 0
0 0 1 2 2 1 0 0
0 0 0 0 0 0
1 1 0 0
1 1 2 0 0
0 0
1 1 0 0 1 0 0
2 0
1 2 2 1
0 0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 1 0 0
0 0 0
1 2 2 1
0 0 0 0 0

Figure 8: The Boolean expression ¬x ∨ ((x ∧ y) ∨ z) is satisfiable if and only

if this puzzle has a solution.

References
[1] L. Auslander and S. Parter. On Imbedding Graphs in the Sphere. J.
Math. Mechanics, (10):517–523, 1961.

[2] Therese C. Biedl, Erik D. Demaine, Martin L. Demaine, Rudolf Fleis-

cher, Lars Jacobsen, and J.Ĩan Munro. The Complexity of Clickomania,
July 2000. preprint.

[3] Stephen A. Cook. The complexity of theorem-proving procedures. In

Proceedings of the 3rd Annual ACM Symposium on Theory of Comput-

8
 ing, pages 151–158, New York, New York, 1971. Association for Com-
puting Machinery, ACM Press.

[4] H. de Fraysseix, J. Pach, and R. Pollack. How to draw a planar graph

on a grid. Combinatorica, 1(10):41–51, 1990.

[5] Erik D. Demaine, Robert A. Hearn, and Michael Hoffman. Push-2-F is

PSPACE-Complete, August 2002.

[6] Erich Friedman. Spiral Galaxies Puzzles are NP-complete. Technical

report, Stetson University, 2000.

[7] Michael R. Garey and David S. Johnson. Computers and Intractabil-

ity: A Guide to the Theory of NP-Completeness. W.H. Freeman and
Company, 1979.

[8] John P. Hayes. Digital System Design and Microprocessors. McGraw-

Hill, 1984.

[9] John E. Hopcroft and Jeffrey D. Ullman. Introduction to Automata

Theory, Languages, and Computation. Addison-Wesley, 1979.

[10] Richard M. Karp. Reducibility among combinatorial problems. In Com-

plexity of Computer Computations, pages 85–103, New York, New York,
1972. Plenum Press.

[11] Richard Kaye. Minesweeper is NP-complete. The Mathematical Intelli-

gencer, 22(2):9–15, 2000.

[12] Richard Kaye. Some Minesweeper Configurations. Tech-

nical report, The University of Birmingham, August 2000.
[Link]

[13] Brandon McPhail. The complexity of puzzles: NP-completeness results

for Nurikabe and Minesweeper. Reed College, 2003. Undergraduate
Thesis.

[14] C. Moore and J.M. Robson. Hard Tiling Problems with Simple Tiles.
Discrete & Computational Geometry, 26(4):573–590, 2000.

[15] Christos H. Papadimitriou. Computational Complexity. Addison-Wesley,

1994.

9
[16] Michael Sipser. Introduction to the Theory of Computation. PWS Pub-
lishing Company, 1997.

[17] Seta Takahiro. The Complexities of Puzzles, Cross Sum, and their An-
other Solution Problems (ASP). The University of Tokyo, 2001. Under-
graduate Thesis.

[18] Alan Turing. On computable numbers, with an application to the

Entscheidungsproblem. Proceedings of the London Mathematical So-
ciety, Series 2, 42(230):230–265, 1936.

[19] Nobuhisa Ueda and Tadaaki Nagao. NP-completeness Results for

NONOGRAM via Parsimonious Reductions. Technical report, Tokyo
Institute of Technology, 1996.

[20] L.G. Valiant and V.V. Vazirani. NP Is As Easy As Detecting Unique

Solutions, 1985.

[21] Thomas Ryan Wilson. NP Completeness: Why Some Problems Are

Hard. Reed College, 1995. Undergraduate Thesis.

[22] Takayuki Yato. On the NP-completeness of the Slither Link Puzzle. In

IPSJ SIGNotes ALgorithms, pages 25–32, 2000.

[23] Takayuki Yato. Complexity and Completeness of Finding Another So-

lution and its Application to Puzzles. Master’s thesis, The University
of Tokyo, 2003.

10