We say that a triangle \(T\) tiles the polygon \(\mathcal A \) if \(\mathcal A \) can be decomposed into finitely many non-overlapping triangles similar to \(T\). In fact, it is in general very difficult to determine the set of similar triangles which tile a given polygon. Previous papers [14, 6] have looked at tiling polygons with similar triangles. In such case one starts with a collection of triangles and are allowed to scale the triangles when placing them. Recently, Laczkovich [5] has looked at tiling polygons, particularly regular triangles, squares, rectangles, and regular \(n\)-gons with \(n\) large enough, with congruent triangles. Clearly, the number of distinct non-similar triangles \(T\) such that parallelogram \(\mathcal P \) can be dissected into finitely many triangles similar to \(T\) is infinite. In this paper we look at tiling parallelograms with similar right triangles.

Let us recall some definitions from [4]. Let \(\mathcal A \) be a polygon with vertices \(V_{1}, \ldots , V_{N}\), and suppose that \(\mathcal A \) is decomposed into non-overlapping similar triangles \(\triangle _{1}, \ldots , \triangle _{t}\) of angles \(\alpha , \beta , \gamma \). Let \(V_{1}, \ldots , V_{m} (m\ge N)\) be an enumeration of the vertices of the triangles \(\triangle _{1}, \ldots , \triangle _{t}\). For each \(i=1, \ldots , m\) we denote by \(p_{i}\) (resp. \(q_{i}\) and \(r_{i}\)) the number of those triangles \(\triangle _{j}\) whose angle at the vertex \(V_{i}\) is \(\alpha \) (resp. \(\beta \) and \(\gamma \)). If \(i\le N\) and the angle of \(\mathcal A \) at the vertex \(V_{i}\) is \(\delta _{i}\), then

$$\begin{aligned} p_{i}\alpha +q_{i}\beta +r_{i}\gamma =\delta _{i}. \end{aligned}$$
(1)

If \(i>N\) then we have either

$$\begin{aligned} p_{i}\alpha +q_{i}\beta +r_{i}\gamma =2\pi \end{aligned}$$
(2)

or

$$\begin{aligned} p_{i}\alpha +q_{i}\beta +r_{i}\gamma =\pi . \end{aligned}$$
(3)

Namely, (2) holds if \(V_{i}\) is in the interior of \(\mathcal A \) and whenever \(V_{i}\) is on the boundary of a triangle \(\triangle _{j}\) then necessarily \(V_{i}\) is a vertex of \(\triangle _{j}\). In the other cases (3) holds. It is clear that the coefficients \(p_{i}, q_{i}, r_{i}\) must satisfy

$$\begin{aligned} \sum \limits _{i=1}^{m} p_{i}=\sum \limits _{i=1}^{m} q_{i}=\sum \limits _{i=1}^{m} r_{i}=t. \end{aligned}$$
(4)

The tiling will be called regular if one of the following statements is true:

  • \(p_{i}=q_{i}\) for every \(i=1,\ldots , m\);

  • \(p_{i}=r_{i}\) for every \(i=1,\ldots , m\);

  • \(q_{i}=r_{i}\) for every \(i=1,\ldots , m\).

Otherwise the tiling is called irregular.

Let \(\mathcal P (\delta )\) be a parallelogram with acute angle \(\delta \). We first consider the regular tilings of parallelograms with similar right triangles and get the following theorem.

FormalPara Theorem 1

If parallelogram \(\mathcal P (\delta )\) has a regular tiling with similar right triangles of angles \((\alpha , \beta , \pi /2)\), then \((\alpha , \beta )=(\delta , \pi /2-\delta )\).

FormalPara Proof

Suppose that \(\mathcal P (\delta )\) can be regularly tiled with similar right triangles of angles \((\alpha , \beta , \pi /2)\). If at each vertex of the tiling the number of angles \(\alpha \) is the same as that of \(\beta \), then for the angle \(\delta \) at a vertex of \(\mathcal P (\delta )\) we let \(\delta =p(\alpha +\beta )+r\cdot \pi /2\), where \(p, r\) are integers. But \(0<\delta <\pi /2\), which is impossible. Therefore by symmetry we may assume that at each vertex of the tiling the number of angles \(\beta \) is the same as that of right angles \(\pi /2\). In the following we shall consider Eqs. (1)–(3).

We first consider Eq. (1) and rewrite it as \(p\alpha +r(\beta +\pi /2)=\delta \) or \(\pi -\delta \), where \(p\) and \(r\) are integers. If \(p\alpha +r(\beta +\pi /2)=\delta \), then by \(0<\delta <\pi /2\) we get \(\delta =p\alpha \). This implies that only the angle \(\alpha \) occurs at the vertex with angle \(\delta \). If \(p\alpha +r(\beta +\pi /2)=\pi -\delta \), then by \(\pi /2<\pi -\delta <\pi \) we have \(r=0\) or 1. When \(r=1\), we obtain \(p=0\), and thus \(\pi -\delta =\beta +\pi /2\). So \(\delta =\alpha \) and \((\alpha , \beta )=(\delta , \pi /2-\delta )\), the result is true. When \(r=0\), we have \(\pi -\delta =p\alpha \). Therefore only the angle \(\alpha \) occurs at the vertex with angle \(\pi -\delta \).

Secondly, we consider Eq. (2) and rewrite it as \(p\alpha +r(\beta +\pi /2)=2\pi \). Clearly, \(r\le 3\). If \(r=0\), then only \(\alpha \) occurs. If \(r=1\), then each of \(\beta \) and \(\pi /2\) occurs once and \(\alpha \) occurs \(p\) times with \(p>1\). If \(r=2\), then \(p=2\) and each of \(\alpha , \beta , \pi /2\) occurs twice. If \(r=3\), then \(p=0\) and \(\beta =\pi /6\), and thus \((\alpha , \beta )=(\pi /3, \pi /6)\). By \(\delta =p\alpha \), we get \(p=1\) and \((\alpha , \beta )=(\delta , \pi /2-\delta )\), the result is true.

Thirdly, we consider Eq. (3) and denote it by \(p\alpha +r(\beta +\pi /2)=\pi \). Clearly, \(r\le 1\). If \(r=1\), then \(p=1\), and thus each of the angles \(\alpha , \beta , \pi /2\) occurs once at the vertices with angle \(\pi \). If \(r=0\), then \(\pi =p\alpha \).

From the discussions above we conclude that if \((\alpha , \beta )\ne (\delta , \pi /2-\delta )\) then at the vertices with angles \(\pi \) or \(2\pi \) the number of angle \(\alpha \) occurs is not less than that of \(\beta \) and at vertices with angles \(\delta \) or \(\pi -\delta \) only angle \(\alpha \) occurs, which contradicts Eq. (4), and the proof is complete. \(\square \)

Figure 1 shows that the right triangle \((\delta , \pi /2-\delta , \pi /2)\) can regularly tile the parallelogram \(\mathcal P (\delta )\).

Fig. 1
figure 1

Regular tiling

In order to consider the irregular tiling of parallelograms with similar right triangles, we introduce the following lemma.

FormalPara Lemma 2

The parallelogram \(\mathcal P (3\pi /10)\) cannot be tiled with similar right triangles of angles \((\pi /10, 2\pi /5, \pi /2)\).

FormalPara Proof

Suppose that \(\mathcal{P }(3\pi /10)\) can be tiled with similar right triangles of angles \((\pi /10, 2\pi /5, \pi /2)\). Then by Theorem 2 of [2], we may assume that the coordinates of each vertex of \(\mathcal{P }(3\pi /10)\) and of the triangles of the tiling belong to the field \(\mathbb{Q }(\text{ cot }(\pi /10))\). Now, by Lemma 8 of [2], we have \(\mathbb{Q }(\text{ cot }(\pi /10))\subset \mathbb{Q }(\zeta )\), where \(\zeta =e^{(\pi /20)i}\). Also, since 3 is prime to 20, there is an automorphism \(\phi \) of \(\mathbb{Q }(\zeta )\) such that

$$\begin{aligned} \phi \big (\text{ cot }{\frac{a}{10}}\pi \big )=\text{ cot }{\frac{3a}{10}}\pi \end{aligned}$$

for every \(a\) not divisible by 10. Consider the map

$$\begin{aligned} \Phi (x, y)=(\phi (x), \phi (y)) \quad (x, y\in \mathbb{Q }(\zeta )). \end{aligned}$$

This map defines a collineation that maps every triangle of the tiling into a triangle of angles \((3\pi /10, \pi /5, \pi /2)\). Also, \(\Phi \) does not change the orientation of these triangles (see pp. 290–291 of [2]). Now \(\Phi \) maps \(\mathcal{P }(3\pi /10)\) into a parallelogram of angles \(9\pi /10\) and \(\pi /10\). Therefore, the conjugate tiling induced by \(\Phi \) gives a tiling of parallelogram of angles \(9\pi /10\) and \(\pi /10\) with triangles of angles \((3\pi /10, \pi /5, \pi /2)\). But this is clearly impossible, since the angles of the triangles are greater than \(\pi /10\). This proves the lemma. \(\square \)

FormalPara Theorem 3

If the parallelogram \(\mathcal P (\delta )\) has an irregular tiling with similar right triangles of angles \((\alpha , \beta , \pi /2)\), then \((\alpha , \beta )=(\delta , \pi /2-\delta )\) or \((\alpha , \beta )=(\delta /2, \pi /2-\delta /2)\).

FormalPara Proof

Suppose that \(\mathcal P (\delta )\) has an irregular tiling with similar right triangles of angles \((\alpha , \beta , \pi /2)\). If \(\alpha =\beta \), then \((\alpha , \beta , \pi /2)=(\pi /4, \pi /4, \pi /2)\) and this is the case of \((\delta , \pi /2-\delta , \pi /2)\). Therefore, we may assume that \(\alpha >\beta \), and thus \(\pi /4<\alpha <\pi /2\). Since the tiling is irregular, there is an equation \(p\alpha +q\beta +r\cdot \pi /2=\tau \) with \(p>q\) and \(\tau \in \{\delta , \pi -\delta , \pi , 2\pi \}\), where \(p, q, r\) are integers. That is, \((p-q)\alpha =\tau -(q+r)\cdot \pi /2\). We need to consider the following cases.

Case 1: \(\tau =\delta \). In this case we have \((p-q)\alpha =\delta -(q+r)\cdot \pi /2\). Since \(p>q\) and \(\delta <\pi /2\), we get \(q+r=0\), and thus \(p\alpha =\delta \). Since \(\alpha >\pi /4\), this implies \(p=1\), \(\alpha =\delta \), and \((\alpha , \beta )=(\delta , \pi /2-\delta )\).

Case 2: \(\tau =\pi -\delta \). Suppose \((p-q)\alpha =\pi -\delta -(q+r)\cdot \pi /2\). This implies \(q+r\le 1\). If \(q+r=0\), then \(p\alpha =\pi -\delta \). Since \(\pi /4<\alpha <\pi /2\), we get \(p=2\) or \(p=3\). If \(p=2\), then \(\alpha =\pi /2-\delta /2\) and we have \((\alpha , \beta )=(\pi /2-\delta /2, \delta /2)\).

If \(p=3\), then \(\alpha =(\pi -\delta )/3\) and \(\beta =\pi /6+\delta /3\). Since \(\alpha >\beta \), this implies \(\delta <\pi /4\). Thus the angle \(\delta \) cannot be tiled with angles \(\alpha \). Therefore, we have \(\delta =n\beta \). Thus \(n=1\) or \(n=2\), which give \(\delta =\pi /4\) or \(\delta =\pi \), both are impossible.

If \(q+r=1\), then \((p-q)\alpha =\pi /2-\delta \). Since \(\alpha >\pi /4\), this implies \(\alpha =\pi /2-\delta \) and \((\alpha , \beta )=(\pi /2-\delta , \delta )\).

Case 3: \(\tau =\pi \). In this case we have \((p-q)\alpha =\pi -(q+r)\cdot \pi /2\). This implies \(q+r\le 1\). If \(q+r=0\), then \(p\alpha =\pi \). Since \(\pi /4<\alpha <\pi /2\), we get \(p=3\), \(\alpha =\pi /3\), and \((\alpha , \beta )=(\pi /3, \pi /6)\). If \(\delta =\pi /3\), then \(\delta =\alpha \), and thus \((\alpha , \beta )=(\delta , \pi /2-\delta )\). If \(\delta =n\cdot \pi /6\), then \(n=1\) or \(n=2\). When \(n=2\), this is the above case. When \(n=1\), then \(\delta =\pi /6\), and thus \((\alpha , \beta )=(\pi /2-\delta , \delta )\).

If \(q+r=1\), then \((p-q)\alpha =\pi /2\). From \(\pi /4<\alpha <\pi /2\) we know that this case is impossible.

Case 4: \(\tau =2\pi \). In this case we have \((p-q)\alpha =2\pi -(q+r)\cdot \pi /2\). Since \(p>q\) and \(\pi /4<\alpha <\pi /2\), we have \(q+r\le 3\).

If \(q+r=3\), then \((p-q)\alpha =\pi /2\). From \(\pi /4<\alpha <\pi /2\) we know this is impossible.

If \(q+r=2\), then \((p-q)\alpha =\pi \). Since \(\pi /4<\alpha <\pi /2\), we get \(\alpha =\pi /3\), which is the result in Case 3.

If \(q+r=1\), then \((p-q)\alpha =3\pi /2\). This implies \(\alpha =3\pi /8\) or \(\alpha =3\pi /10\).

When \(\alpha =3\pi /8\), then \((\alpha , \beta )=(3\pi /8, \pi /8)\). If \(\delta =3\pi /8\) or \(\delta =\pi /8\), our result is true. Otherwise, we have \(\delta =m\cdot \pi /8\). It is clear that \(m\ne 1, 3\). If \(m=2\), then \(\delta =\pi /4\), and this is the case of \((\alpha , \beta )=(\pi /2-\delta /2, \delta /2)\). If \(m\ge 4\), then \(\delta =m\cdot \pi /8\ge \pi /2\), a contradiction.

When \(\alpha =3\pi /10\), then \((\alpha , \beta )=(3\pi /10, \pi /5)\). If \(\delta =3\pi /10\), or \(\delta =\pi /5\), our result is true. Otherwise, we have \(\delta =n\cdot \pi /5\). It is clear that \(n\ne 1\). If \(n=2\), then \(\delta =2\pi /5\), and this is the case of \((\pi /2-\delta /2, \delta /2)\). If \(n\ge 3\), then \(\delta =n\cdot \pi /5>\pi /2\), a contradiction.

If \(q+r=0\), then \(p\alpha =2\pi \). Since \(\pi /4<\alpha <\pi /2\), we have \(\alpha =2\pi /5\), \(\alpha =\pi /3\), or \(\alpha =2\pi /7\).

When \(\alpha =2\pi /7\), then \((\alpha , \beta )=(2\pi /7, 3\pi /14)\). If \(\delta =2\pi /7\), or \(\delta =3\pi /14\), then the result is true. If \(\delta \ne 2\pi /7\) and \(\delta \ne 3\pi /14\), then by \(0<\delta <\pi /2\) we may assume that \(\delta =m\cdot 3\pi /14\). If \(m=2\), then \(\delta =3\pi /7\), and this is the case of \((\pi /2-\delta /2, \delta /2)\). If \(m\ge 3\), then \(\delta =m\cdot 3\pi /14>\pi /2\), a contradiction.

When \(\alpha =\pi /3\), this is a result of Case 3.

When \(\alpha =2\pi /5\), then \((\alpha , \beta )=(2\pi /5, \pi /10)\). If \(\delta =2\pi /5\) or \(\delta =\pi /10\), the result is true. Otherwise, we have \(\delta =n\cdot \tfrac{\pi }{10}\). If \(n=2\), this is the case of \((\pi /2-\delta /2, \delta /2)\). If \(n=3\), then \(\delta =3\pi /10\), and by Lemma 2 we conclude a contradiction. If \(n\ge 5\), then \(\delta =n\cdot \pi /10\ge \pi /2\), a contradiction. The proof is complete. \(\square \)

Figure 2 shows that the right triangles \((\delta , \pi /2-\delta , \pi /2)\) and \((\delta /2, \pi /2-\delta /2, \pi /2)\) can irregularly tile the parallelogram \(\mathcal P (\delta )\), where the numbers 1, 2, 3 in the left figure denote the angles \(\delta , \pi /2-\delta , \pi /2\), respectively; the numbers 1, 2, 3 in the right figure denote the angles \(\delta /2, \pi /2-\delta /2, \pi /2\), respectively.

Fig. 2
figure 2

Irregular tiling

FormalPara Remark 4

Since the above proofs never used that \(\mathcal P (\delta )\) is a parallelogram, only that two of its angles equal \(\delta \) and two equal \(\pi -\delta \), for the trapezoid \(\mathcal T (\delta )\) with acute angle \(\delta \), we have the same results as that of parallelogram \(\mathcal P (\delta )\).