See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/273573138

Rectangles in Rectangles

Article  in  Mathematics Magazine · June 2000

DOI: 10.2307/2691523

CITATIONS READS

4 923

1 author:

John E. Wetzel
University of Illinois, Urbana-Champaign
83 PUBLICATIONS   446 CITATIONS   

SEE PROFILE

All content following this page was uploaded by John E. Wetzel on 07 December 2017.

The user has requested enhancement of the downloaded file.

Rectangles in Rectangles
Author(s): John E. Wetzel
Source: Mathematics Magazine, Vol. 73, No. 3 (Jun., 2000), pp. 204-211
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2691523 .
Accessed: 26/10/2011 11:03

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to
Mathematics Magazine.

http://www.jstor.org
 N OTE S

Rectangles in Rectangles
JOHN E. WETZEL
University of Illinois at Urbana-Champaign
Urbana, IL61801-2975

Introduction Wh-endoes one rectangle fit into another? In 1956, in the Monthly,
L. Ford asked for a necessary and sufficient condition for a p X q rug to fit on an
a x b floor. Necessar-yconditions are easy to find: If a p x q rectangle fits into an
a x b rectangle (FIGURE 1), then
pq < ab (the area condition)
p2 + q2 < a2 + b2 (the diametercondition)
p + q < a + b (the perimeter condition)
min { p, q} < min {a, b} (the thickness condition).
But none of these necessary conditions is sufficient. For example, a rectangle with
sides 9 and 4 does not fit into a rectangle with sides 8 and 6, but all four necessary
conditions are satisfied.

a
FIGURE 1

Sufficient conditions seem scarcer. One can show that if the thickness condition
min { p, q} < min {a, b} is satisfied and if either of the conditions
maxfp,q}<min{ca,b} andp+q?< Va2+b2
holds, then a p x q rectangle fits into an a x b rectangle. But neither of these
sufficient conditions is necessary; an 88 X 13 rectangle fits into an 81 X 59 rectangle
(see below), but both of these two conditions are false.
In answer to Ford's question, W. Carver [3] gave the following mysterious-looking
necessary and sufficient condition: a p X q rectangle (p 2 q) fits into an a x b
rectangle (a 2 b) if and only if
p <a and q b
or
2pqa+(p -q2) p2?q2-a2
p>a and b? p2+q2

204
MATHEMATICS MAGAZINE VOL. 73, NO. 3, JUNE 2000 205
Carver'singenious elementary argument is entirely geometric. In this note we give a
straightforwarddevelopment of Carver's conditions that is quite different from the
argument he suggested, and we prove:
THEOREM 1. (Carver [3]) Suppose arna x b rectangle T is given, with the notation
arranged so that a ? b. Then a p x q rectangle R with p ? q fits into T if and only if
(a) p <?a and q < b, or
+ a
(b) p > a, q < b, and ( p+qb -cb >2.

In Section 4 we use this result to find the smallest rectangle of given shape that can
accommodate first every rectangle of perimeter two, and then every rectangle of
diameter one. In Section 5 we pose some related unsolved problems of undetermined
difficulty.

2. Rectangles in rectangles In 1964 H. Steinhaus [16] asked for a necessaiy and

sufficient condition on the six sides for one triangle to fit into another. Nearly thirty
years later, K. A. Post [12] gave a set of 18 inequalities relating the sides of the
triangles that are necessary and sufficient in the sense that if one of the inequalities is
correct, the first triangle fits in the second; and if the first triangle fits in the second,
then one (at least) of the inequalities is correct. Post's argument relies
on a fitting lemma for triangles that is of interest in its own right: if one triangle fits in
another, then it fits in such a way that two of its vertices lie on the same side of the
containing triangle.
Our argument for rectangles also relies on a fitting lemma: The largest rectangle
similarto a given rectangle that lies in a target rectangle T has its vertices on the sides
of T.
LEMMA 2. (Mok-Kong Shen)' Suppose rectangles T and R( are given. Then there is
a. largest rectangle R similar to R( that fits in T, anrd R fits in T with all four of its
vertices on the sides of T.

Proof: The existence of R is a routine consequence of compactness, but for

completeness we include a short argument using sequences. Suppose that Ro has
sides tto and v0. Let A = sup{t > 0: a tuo x tuo rectangle fits in T}, and let R be the
rectangle with sides u = Atto and v = Auo.To show that R fits in T, take a sequence
{Ak} of positive reals that increase to A, and let Rk be the rectangle with sides
Uk = Aktuo and uk = Akvo.For each index k there are points Kk, Lk, Mk, Nk in T so
that the rectangle Kk Lk Mk Nk is congruent to Rk. Passing successively to convergent
subsequences (using the Bolzano-Weierstrasstheorem), we arrange for {Kk}, {LU,
{Mk}, and {Nk} all to converge, say to K, L, M, and N, respectively. Then the
rectangle R = KLMN is congruent to Ro and lies in T.
How is such a maximal rectangle R situated in T? At least two vertices of R must
lie on the sides of T, or a larger rectangle similar to R0 could be found in T. We
examine the possibilities.
If two vertices of R lie on the same side of T, then the sides of R must be parallel
to those of T; and the maximalityforces the other two vertices to lie on the sides of T
as well.

.In response to my May 7, 1998 post to the newsgroup sci. math. research concerning tllis
question, Mok-Kong Shen posted on May 11, 1998 a partial solution that includes a version of this fitting
lemma. Mr. Shlendeclinecl my invitation to be coauthaorof thaisnote.
206 ? MATHEMATICALASSOCIATION OF AMERICA

If two vertices of R lie on opposite sides of T and if at least one of the two other
vertices of R is inside, a small translation of R moves both inside T; and then a
suitable small rotation about its center moves R completely inside T, contrary to the
maximality.So if two vertices of R lie on opposite sides of T, then all four vertices
must lie on T.
If two vertices of R lie on adjacent sides of T and either of the other two vertices is
on a side, we have two opposite vertices of R on the sides of T, and the previous
argument applies. If both the other vertices are inside T, a small translationmoves R
completely inside T, contrary to the maximality.So in every case, all four vertices of
R must lie on T. U
We say that a rectangle R is inscribed in a rectangle T if each (closed) side of T
contains a vertex of R (see FIGURE 2). When a rectangle with sides p and q is
inscribed in a rectangle with sides a and. b, the lengths p, q, a, and b are related by
the quartic equation

q4 - (a2 +b2+ 2p2)q2+ 4abpq - p2(a2+b 2p_ p2) =? (1)

an equation that over the years has appeared frequently in the problem literature. The
earliest references of which I am aware are in 1896, although (1) surely must have
been known many years earlier. In a problem posed in the Monthly, M. Priest asked
for the " . . . length of a piece of carpet that is a yard wide with square ends, that can
be placed diagonally in a room 40 feet long and 30 feet wide, the corners of the carpet
just touching the walls of the room." A solution [14] published the following year by
C. D. Schmitt, et al., gives a trigonometric derivation of (1). Equation (1) also appears
as an example in a discussion of quartic equations in Merriman and Woodward [10,
p. 20], an 1896 mathematics reference book for engineers, with the curious footnote:
"This example is known by civil en ineers as the problem of finding the length of a
strut in a panel of the Howe truss."
The question was posed again in 1914 by C. E. Flanagan. The solution by Otto
Dunkel [4], published in 1920, includes a geometric derivation of (1) in the elegant
symmetric form
/ a+b\2 (a- b\2
2- q - =0
P+q P- q
and, regarding this as an equation in q for given a, b, and p, an elaborate
investigation of the location of the roots.

3. The fitting region Suppose the given target rectangle T has sides of length a
and b, where a 2 b. By the fitting region F = F(T) we mean the set of points {p, q}
so that the rectangle with sides p and q fits in T. We find it convenient not to
assume that p ? q. It will simplify the language to regard a point as a 0 x 0 rectangle
and a line segment of length f> 0 as an fx 0 rectangle. With this agreement we see
that F is a closed subset of the closed quarter-disk of radius Va2 + b2 that is
symmetric in the 450 line.
The fitting region F(T) is star-shaped with respect to the origin, that is to say, if
P e F(T) and if Q lies on the segment OP (where 0 is the origin), then Q e F(T).

2Tbe Howe truss, a liniearassemblage of cross-braced rectanigularstructural"paniels,"was inivenitedby

onie WVilliamHowe (1803-1852) anid patented in 1840 anid 1842. It was the most popular bridge truss
designiin tbe Uniited States in tbe latter lhalfof the 19th cenitury.See [6].
MATHEMATICS MAGAZINE VOL. 73, NO. 3, JUNE 2000 207
More generally, if (p,q)EF(T) and if O r?p and O s q (or if O r?q and
O s ?p), then (r,s)EF(T).
An x X y rectangle R fits in T with its sides parallel to those of T precisely when
x < a andy<borx<band y < a. Writing S,,, = {(x, y):O < x < u, 0 < y < V}, we
conclude that Sab U Sba C F(T). We will find that F(T) is formed from SabU SbaI by
adding two curvilinear triangular tabs T1 and T2 symmetric in the 450 line (see
T1= UVW in FIGURE4).
It remains to analyze the situation in which the largest fitting rectangle R = KLMN
similar to a given rectangle Ro fits in T with exactly one vertex on each side of
T = ABCD. Suppose the vertices are labeled as in FIGURE 2, with AB = CD = a,

D M c

A K B
K
FIGURE 2

BC=DA=b, K cAB, LCBC, M ECD, and NCDA. Write x=KL=NM and

y = KN = LM. Taking the directed angle 0 = / BKL as parameter, we see that
x cos 0 + y sin 0 = a,
x sin 0 + y cos 0 = b
(cf. [14]). Solving for x and y gives parametric equations for part of the boundary
curve of the fitting region:
a cos 0 - b sin 0
cos20
b cos 0 - a sin 0
Y cos20
It follows that
a +b 1
y -r2 sin(0+
|X+ ) (
{21 a-b 1 (2)
Cos(0+ 4)

Rotating the coordinate axes through the angle - -/4 using the usual coordinate
transformation x = ( + -q)/ x/, y = ( + q)/ v2, we arrive at the parametric
equations
a2 a-b 1
2 cos(0+ '4r
a+b 1
si= n . , ,
4s
.
208 C MATHEMATICALASSOCIATION OF AMERICA

To eliminate the parameter, we solve for sin(0 + 4) and cos(0 + 4T)and square and
add, obtaining the equation 22 _ C.- d22 = 0, i.e., (since -Y= (x + y)/ 94> 0),
_ clj
d6- 2

where c = '(a + b) and d = '(a - b). Suppose a = b. Then d = 0, and the curve
has two branches symmetric in the aq-axiswith asymptotes ij = c and g = ?d (see
FIGURE 3). Write ij = F ) for the branch with > 0 and ij> 0. Plainly F is a
decreasing function whose graph is concave upwards.

l l
l l
l I

I I

FIGURE3

In termsof (x, y), F can be wiitten (from(2)) in the formf(x, y; a, b) =2, where

A(X, Y;a,b) =(x + y )+ ( X-C )

The curve F is asymptotic to the lines x + y= (az+b)/ 2 and y-x =

? (a - WIF. For (x, y)inthe quadrant x > Iy1,we see that f(x, y;a, b) <2 o
the concave side of F and f(x, y; a, b) > 2 on the side of rF that contains the
asymptotes. The cuirve F meets the x-axis at x = VaT72+ b2- with slope
- (a2 + b2 )/2 ab < -1, and it passes through the point (a, b) with slope alb. Since
F is concave, it meets the line segment x = az,O< y < b at (a7,b) and again at a point
U whose y-coordinate bo is the root of the cubic

y 3+ b y2-3a 2y+ a2b =

in the inteirvalO< y < b. Let U,V, W be the pointswith coordinates(az,bo), (az,O),

(WA-2+b, O), respectively, and let T, be the curvilinear triangular region bounded
by the line segments UV and VVVand the arc of F with endpoints U and W (see
FIGURE4). Let 7T,be the mirrorimuageregionof T, in the 45? line throughthe origin.
We shallsee that the fittingregion F(T) is preciselyS,b U Sb, u T, u T2 .
In the square case az= b the graph F collapses to the line 7}=c, i.e., x + y = azF;
and Ti and T2,become ordinairytriangles.
So finally we come to our result.
THEOREM3. A rectangleR wvith sidlesp andlqfits in az recta7ngle T twithsidlesa andl
b (with a > b) if and only if the point ( p, q) lies in the set Sab U Sba U T, U 7T2.
Proof: If ( p, q) C=Sab,u Sb,a,then p < azand q < b or vice versa,andin eithercase
the rectangyle with sides p, a fits in T with its sides poarallel to those of T. If
MATHEMATICS MAGAZINE VOL. 73, NO. 3, JUNE 2000 209

Y y

p~~~~~~~~
aiX A/

v w ~~~'

0 b a O,0 b
(a) (b)
FIGURE 4

(p, q) c T1, the ray from the origin through (p, q) meets F at a point (x, y) whose
coordinates are the sides of the largest rectangle R' in T similar to a p x q rectangle
R; and R fits in R' and hence in T. If (p, q) c T9, then (q, p) c T1, and the above
argument applies. So if (p, q) is any point in Sab U Sb, U T1 U T2, the rectangle with
sides p and q fits in T.
Conversely, if a rectangle with sides p and q fits into T, the largest similar
rectangle R', with sides x ? p and y 2 q, fits in T in one of only two possible ways.
If two vertices of R' lie on the same side of T, then either x < a and y < b or
vice versa, and (x, y) c Sab U Sba; so (p, q) c S U Sb,. Otherwise, R' is inscribed
in T, (x, y) lies on F, and (p, q) lies on the radius to (x, y) and so lies in
S,b U Sba U T1 U T2. U

It follows, as asserted, that F(T) = Sab U Sb, U T1 U T2. Theorem 1 follows easily: if
p ? q, conditions (a) are equivalent to (p, q) c Sab; otherwise, (p, q) c T1 if and only
if p ? a, q < b, and f(a, b; p, q) ? 2; and these are conditions (b).
The assertion in the Introduction that the conditions

min(p,q} <min(ba,b} and p+q?< a2+b2

are sufficient can be read off FIGURE 4, because if p ? q and a > b, then

(p, q): q < b and p + q < la2+ b}c F(T).

It is worth noting that in the square case, a p x q rectangle fits into a square of side a
if and only if max{p, q} < a or p + q < a (cf. [15]).

4. Two rectangle covering problems Geometric covering problems have

attracted considerable interest in recent years. We have recently described the
smallest equilateral triangle that can cover any triangle of perimeter two (see Wetzel
[18]), and in the same article we reported an easily verified claim by one J. Smith that
the smallest equilateral triangle that can cover every triangle of diameter one (i.e.,
every triangle whose longest side has length one) has side (2cos 10O)/ . Here we
use Theorem 3 to solve two analogous problems for rectangles.
First we determine the smallest rectangle R similar to a given rectangle Ro that can
accommodate every rectangle of perimeter two. Evidently the width of R must be at
210 ? MATHEMATICALASSOCIATION OF AMERICA

least 1/2, and if Ro is long and thin, taking the width of R to be 1/2 suffices.
However, if the sides of Ro are more nearly equal, we need further to arrangefor the
diagonal of R to be at least one. More precisely, we prove the following:
THEOREM 4. An a x b rectangle R0 is given, with a ? b. The smallest rectangle R
similar to Ro that can accommodateevery rectangleof perimetertwo is a/2b x 1/2 if
a ? bF3 and a/d x b/d if a < b3, where d = ja2 + b2 is the diagonal of Ro.
Proof: The dual question is more naturalin the present context: given a rectangle R
with sides of lengths a and b, determine the largest possible { so that R can
accommodate every rectangle of perimeter f. In geometric terms, we seek the largest
/' so that the line segment o(/) with endpoints (/72,0) and (0, //2) lies in the
fitting region F(R).
Since F meets the x-axis at the point d = v/a2+ b2 with slope at most -1, as /
increases from 0, o(/) remains in F(R) until it encounters either the point P(b, b)
or the intercept (d, 0) (see FIGURE4). The segment o(/) reaches the point P before
the point (d, 0) precisely when 2b < d, i.e., precisely when a ? b /3. So when a ? b;
the largest possible f is 4b, and when a < b/7 the largest possible f is 2d.
Dually, the smallest rectangle similar to a given a x b rectangle (with a ? b) that
can accommodate every rectangle of perimeter ( is fa/4b x//4 when a ? bF/ and
fa/2d x>b/2d when a < b V3. The assertion follows when f= 2. 2

COROLLARY5. The rectangle of least area that can accommodateevenj rectangle of

perimeter two has sides 1/2 and V3/2.
As a second example, we determine the smallest rectangle similar to a given
rectangle that can accommodate every rectangle of diameter one.
THEOREM 6. An a x b rectangle Ro is given, with a ? b. The smallest rectangle R
similar to Ro that can accommodate every rectangle of diameter one is al/(b x
1/x/2 if a?b/i and 1 xKb/a if a<bxfZ.
Proof: Direct arguments are easy to give, but a dual argument that parallels the
proof of Theorem 4 is straightforward.For the given rectangle Ro with fitting region
F(Rd), we seek the largest d so that every rectangle with diagonal d fits in Ro. Write
CF( p) for the quarter circular arc x2 + y2 = d2, 0 < x < d. In geometric terms, we
seek the largest d so that ?(d) c F(RO).
It is a calculus exercise to show that F (as a function of x and y) has slope -x/y at
precisely one point, the point (xl, y1) where

X1= A/(a+ V/a2

_b
IlYi= a1(a - Fa2-V2
Sincex2 + y2 = a2, it followsthatF andthe quartercircle4?(a)aretangentat the
the largestradiusd so that
point(xl, yl) givenby (3) (see FIGURE4). Consequently,
(c(d) F(RO)is just min{a,bV2}, depending whetherfor increasingd, 4?(d)
c on
meetsP or F first.
Dually,the smallestrectanglesimilarto a givena x b rectangle(witha 2 b) that
canaccommodate everyrectangleof diameterd is ad/bTi x d/ F2 whena ? bF2
andd x bd/a whena ? bF2. Theassertion followswhend = 1. x

COROLLARY 7. Therectangle of leastareathatcanaccommodate evenjrectangle of

diameteronehassides1 and 1/ /2.
MATHEMATICS MAGAZINE VOL. 73, NO. 3, JUNE 2000 211
5. Some further questions Here are a few interesting related problems, of unde-
termined difficulty.
Find necessary and sufficient conditions on a, b, c, s for a triangle with sides a, b, c
to fit in a square with side s; for a square with side s to fit in a triangle with sides
a,b,c.
More generally, find necessary and sufficient conditions on a, b, c, p, q for a
triangle with sides a, b, c to fit in a rectangle with sides p, q; for a rectangle with sides
p, q to fit in a triangle with sides a, b, c
Each of these fitting problems has associated covering problems. How small a
square (or rectangle of prescribed shape) can accommodate every triangle of perime-
ter two? Of diameter one? How small a triangle of prescribed shape can accommodate
every rectangle of perimeter two? Of diameter one?
Find necessary and sufficient conditions on the lengths a, b, c and p, q, r for a box
(i.e., a rectangularparallelepiped) with edges p, q, r to fit in a box with edges a, b, c.
(F.M. Garnett asked this question in 1923, and W.B. Carver [2] supplied a fragmen-
tary answer in 1925.) What are the analogous results for orthotopes in Ed? Again,
there are related covering problems.

REFERENCES

1. Robert Blasch, Solution to Problem 30, Pentagon 10 (1950), 42-43. (Posed by John K. Osborn,
Pentagon 9 (1950), 109.)
2. W. B. Carver, Solution to Problem 3036, Amner.Math. Monthly 32 (1925), 47-49. (Posed by F. M.
Garnett, Amer. Aath. Monthly 30 (1923), 337.)
3. W. B. Carver, Solution to Problem E1225, Amner.Math. Monthly 64 (1957), 114-6. (Posed by L. R.
Ford, Amer. Math. Monthly 63 (1956), 421.)
4. Otto Dunkel, Solution I to Problem 416, Amler. Math. Monthlly27 (1920), 327-30. (Posed by C. E.
Flanagan, Amner.Math. Monthly 21 (1914), 156.)
5. Robert C. Fisher and Alan D. Ziebur, IntegratedAlgebra anidTrigonometry,Prentice-Hall, Englewood
Cliffs, NJ, 1957. (See p. 259.)
6. Howe, William, Britannica Online (accessed June 30, 1998).
7. James F. Hurley, Litton's Problematical Recreations, Van Nostrand Reinhold, New York, NY, 1971.
(See Problem 41, p. 205.)
8. W. W. Johnson, Solution to Problem 473, Amer. Math. Monthly 23 (1916), 173. (Posed by Frank B.
Morris, Amier. Math. Monthly 22 (1915), 267.)
9. Samuel I. Jones, Mathematical Nuttsfor Lovers of Mathematics, rev. ed., S. I. Jones, Nashville, TN,
1936. (See problem 14, p. 178.)
10. Mansfield Merriman and Robert S. Woodward, Higher Mathematics, John Wiley & Sons, New York,
NY, 1896.
11. Leroy F. Meyers, Solution to Problem 244, Crux Ma-th.4 (1978), 19. (Posed by Steven R. Conrad,
Crux Math. 3 (1977), 130.)
12. K. A. Post, Triangle in a triangle: on a problem of Steinhaus, Geom. Dedicata 45 (1993), 115-20.
13. William R. Ranson, One Hundred Curious Mathematical Problems, J. Weston Walch, Portland, ME,
1955. (See pp. 60-63.)
14. Cooper D. Schmitt, W. H. Harvey, and B. F. Finkel, Solution II to Problem 67, Amner.Math. Monthlly
4 (1897), 87-88. (Posed by F. M. Priest, Amer. Math. Monthly 3 (1896), 278.)
15. D. L. Silverman, Solution to Problem E1584, Amer. Math. Monthly 71 (1964), 209-10. (Posed by
D. J. Newman, Amer. Math Monthly 70 (1963), 438.)
16. Hugo Steinhaus, One HuindredProblems in ElemlentaryMathenmatics,Basic Books, New York, NY,
1964.
17. Edmund H. Umberger, Solution to Problem 1275, School Sci. Math. 33 (1933), 678. (Posed by Paul
Veigle, School Sci. Math. 33 (1933), 332.)
18. John E. Wetzel, The smallest equilateral cover for triangles of perimeter two, this MAGAZINE70 (1997),
125-30.

View publication stats