This article is translated from a Chinese article on my Zhihu account. The original article was posted at 2019-11-03 20:02 +0800.

Some notations:

  • aa indexes particles. j,k,lj,k,l are indices of vector components. q,s,t\mathbf q,\mathbf s,\mathbf t are arbitrary vectors.
  • I\mathbf I is the unit matrix. Matrix q⊗s≔qsT\mathbf q\otimes\mathbf s\coloneqq\mathbf q\mathbf s^\mathrm T is the outer product of q\mathbf q and s\mathbf s, and q⊗2≔q⊗q\mathbf q^{\otimes 2}\coloneqq\mathbf q\otimes\mathbf q.
  • A rigid body consists of some particles with distances between each other fixed. μa\mu_a is the mass of a particle; ra\mathbf r_a is the position of a particle relative to the center of the rigid body (∑aμara=0\sum_a\mu_a\mathbf r_a=\mathbf 0); and va\mathbf v_a is the velocity of a particle.
  • The total mass of the rigid body is m=∑aμam=\sum_a\mu_a; the velocity of its mass center is V\mathbf V; and its angular velocity is Ω\mathbf\Omega. θa\theta_a is the angle between ra\mathbf r_a and Ω\mathbf\Omega.
  • m≔mI\mathbf m\coloneqq m\mathbf I, J≔∑aμa(ra2I+ra⊗2)\mathbf J\coloneqq\sum_a\mu_a\left(\mathbf r_a^2\mathbf I+\mathbf r_a^{\otimes 2}\right).

Talk about homogeneous scalar fields of degree 2. Homogeneous scalar fields of degree 2 are function ff that satisfies f ⁣(λq)=λ2f ⁣(q).f\!\left(\lambda\mathbf q\right)=\lambda^2f\!\left(\mathbf q\right). It has the form f ⁣(q)=12∑j,kAj,kqjqk,f\!\left(\mathbf q\right)=\frac12\sum_{j,k}A_{j,k}q_jq_k, where the coefficient matrix A\mathbf A is a symmetric matrix (this is why there is a factor of 12\frac12 in front).

Actually, such a function can be written in matrix multiplication as f ⁣(q)=A2q⋅qf\!\left(\mathbf q\right)=\frac{\mathbf A}2\mathbf q\cdot\mathbf q (note that it is not the same as A2q2\frac{\mathbf A}2\mathbf q^2, but is (A2q)⋅q\left(\frac{\mathbf A}{2}\mathbf q\right)\cdot\mathbf q instead).

Similarly, homogeneous vector field with degree 1 has the property g ⁣(λq)=λg ⁣(q),\mathbf g\!\left(\lambda\mathbf q\right)=\lambda\mathbf g\!\left(\mathbf q\right), and they have the form g ⁣(q)=Aq\mathbf g\!\left(\mathbf q\right)=\mathbf A\mathbf q (now the coefficient matrix A\mathbf A is not necessarily symmetric).

Introduce outer product. If the matrix A≔q⊗s\mathbf A\coloneqq\mathbf q\otimes\mathbf s, then Aj,k=qjsk.A_{j,k}=q_js_k. Specially, if A≔q⊗2\mathbf A\coloneqq\mathbf q^{\otimes 2}, then Aj,k=qjqk.A_{j,k}=q_jq_k. The perfect square expansion of outer product is (q+s)⊗2=q⊗2+q⊗s+s⊗q+s⊗2.\left(\mathbf q+\mathbf s\right)^{\otimes 2}= \mathbf q^{\otimes 2}+\mathbf q\otimes\mathbf s+\mathbf s\otimes\mathbf q+\mathbf s^{\otimes 2}.

First, according to geometric relations, we can prove a theorem va=V+Ω×ra.\mathbf v_a=\mathbf V+\mathbf\Omega\times\mathbf r_a. Then, we can discuss the kinetic energy T≔∑aμa2va2T\coloneqq\sum_a\frac{\mu_a}2\mathbf v_a^2 of the rigid body. We will discover that the kinetic energy of the rigid body can be represented as the sum of a homogeneous function of degree 2 w.r.t. V\mathbf V and a homogeneous function of degree 2 w.r.t. Ω\mathbf\Omega, i.e. T=Ttra+Trot,T=T_{\mathrm{tra}}+T_{\mathrm{rot}}, where Ttra≔m2V⋅V,T_{\mathrm{tra}}\coloneqq\frac{\mathbf m}2\mathbf V\cdot\mathbf V, Trot≔J2Ω⋅Ω.T_{\mathrm{rot}}\coloneqq\frac{\mathbf J}2\mathbf\Omega\cdot\mathbf\Omega.

The proof is not difficult: T≔∑aμa2va2=∑aμa2(V+Ω×ra)2=∑aμa2(V2+2V⋅(Ω×ra)+(Ω×ra)2)=12V2∑aμa⏟m+V⋅(Ω×∑aμara⏟0)+12∑aμa(Ω×ra)2=Ttra+12∑aμa(Ω×ra)2,\begin{align*} T&\coloneqq\sum_a\frac{\mu_a}2\mathbf v_a^2\\ &=\sum_a\frac{\mu_a}2\left(\mathbf V+\mathbf\Omega\times\mathbf r_a\right)^2\\ &=\sum_a\frac{\mu_a}2\left( \mathbf V^2+2\mathbf V\cdot\left(\mathbf\Omega\times\mathbf r_a\right) +\left(\mathbf\Omega\times\mathbf r_a\right)^2 \right)\\ &=\frac12\mathbf V^2\underbrace{\sum_a\mu_a}_m +\mathbf V\cdot\left(\mathbf\Omega\times\underbrace{\sum_a\mu_a\mathbf r_a}_{\mathbf 0}\right) +\frac12\sum_a\mu_a\left(\mathbf\Omega\times\mathbf r_a\right)^2\\ &=T_{\mathrm{tra}}+\frac12\sum_a\mu_a\left(\mathbf\Omega\times\mathbf r_a\right)^2, \end{align*} while =  (Ω×ra)2=Ω2ra2sin⁡2θa=Ω2ra2−Ω2ra2cos⁡2θa=Ω2ra2−(Ω⋅ra)2=(∑jra,j2)(∑jΩj2)−(∑jra,jΩj)(∑jra,jΩj)=∑j,lra,l2Ωj2−∑j,kra,jra,kΩjΩk=∑j,lra,l2Ωj∑kIj,kΩk−∑j,kra,jra,kΩjΩk=∑j,k(∑lra,l2)⏟ra2Ij,kΩjΩk−∑j,kra,jra,kΩjΩk=∑j,k(ra2Ij,k−ra,jra,k)ΩjΩk=(ra2I−ra⊗2)Ω⋅Ω.\begin{align*} &\phantom{=}~\,\left(\mathbf\Omega\times\mathbf r_a\right)^2\\ &=\mathbf\Omega^2\mathbf r_a^2\sin^2\theta_a\\ &=\mathbf\Omega^2\mathbf r_a^2-\mathbf\Omega^2\mathbf r_a^2\cos^2\theta_a\\ &=\mathbf\Omega^2\mathbf r_a^2-\left(\mathbf\Omega\cdot\mathbf r_a\right)^2\\ &=\left(\sum_jr_{a,j}^2\right)\left(\sum_j\Omega_j^2\right) -\left(\sum_jr_{a,j}\Omega_j\right)\left(\sum_jr_{a,j}\Omega_j\right)\\ &=\sum_{j,l}r_{a,l}^2\Omega_j^2-\sum_{j,k}r_{a,j}r_{a,k}\Omega_j\Omega_k\\ &=\sum_{j,l}r_{a,l}^2\Omega_j\sum_kI_{j,k}\Omega_k -\sum_{j,k}r_{a,j}r_{a,k}\Omega_j\Omega_k\\ &=\sum_{j,k}\underbrace{\left(\sum_{l}r_{a,l}^2\right)}_{\mathbf r_a^2} I_{j,k}\Omega_j\Omega_k -\sum_{j,k}r_{a,j}r_{a,k}\Omega_j\Omega_k\\ &=\sum_{j,k}\left(\mathbf r_a^2I_{j,k}-r_{a,j}r_{a,k}\right)\Omega_j\Omega_k\\ &=\left(\mathbf r_a^2\mathbf I-\mathbf r_a^{\otimes2}\right)\mathbf\Omega\cdot\mathbf\Omega. \end{align*} Therefore, T=Ttra+12∑aμa(ra2I−ra⊗2)Ω⋅Ω=Ttra+Trot.\begin{align*} T&=T_{\mathrm{tra}}+\frac12\sum_a\mu_a\left(\mathbf r_a^2\mathbf I-\mathbf r_a^{\otimes2}\right)\mathbf\Omega\cdot\mathbf\Omega\\ &=T_{\mathrm{tra}}+T_{\mathrm{rot}}. \end{align*}

Then, we may discuss the momentum of the rigid body P≔∑aμava.\mathbf P\coloneqq\sum_a\mu_a\mathbf v_a. We will discover that it is a homogeneous function of degree 1 w.r.t. V\mathbf V, i.e. P=mV.\mathbf P=\mathbf m\mathbf V.

The proof is easier: P≔∑aμava=∑aμa(V+Ω×ra)=(∑aμa)⏟mV+Ω×∑aμara⏟0=mV.\begin{align*} \mathbf P&\coloneqq\sum_a\mu_a\mathbf v_a\\ &=\sum_a\mu_a\left(\mathbf V+\mathbf\Omega\times\mathbf r_a\right)\\ &=\underbrace{\left(\sum_a\mu_a\right)}_m\mathbf V +\mathbf\Omega\times\underbrace{\sum_a\mu_a\mathbf r_a}_{\mathbf 0}\\ &=\mathbf m\mathbf V. \end{align*}

Finally, we can discuss the angular momentum of the rigid body M≔∑ara×(μava).\mathbf M\coloneqq\sum_a\mathbf r_a\times\left(\mu_a\mathbf v_a\right). We will discover that it is a homogeneous function of degree 1 w.r.t. Ω\mathbf\Omega, i.e. M=JΩ.\mathbf M=\mathbf J\mathbf\Omega.

The proof is also easy: M≔∑ara×(μava)=∑aμara×(V+Ω×ra)=(∑aμara)⏟0×V+∑aμara×(Ω×ra)=∑aμa(ra2Ω−ra⋅Ωra)\begin{align*} \mathbf M&\coloneqq\sum_a\mathbf r_a\times\left(\mu_a\mathbf v_a\right)\\ &=\sum_a\mu_a\mathbf r_a\times\left(\mathbf V+\mathbf\Omega\times\mathbf r_a\right)\\ &=\underbrace{\left(\sum_a\mu_a\mathbf r_a\right)}_{\mathbf 0}\times\mathbf V +\sum_a\mu_a\mathbf r_a\times\left(\mathbf\Omega\times\mathbf r_a\right)\\ &=\sum_a\mu_a\left(\mathbf r_a^2\mathbf\Omega-\mathbf r_a\cdot\mathbf\Omega\mathbf r_a\right) \end{align*} (note the order of computation in the notation q⋅st\mathbf q\cdot\mathbf s\mathbf t: it is (q⋅s)t\left(\mathbf q\cdot\mathbf s\right)\mathbf t). Therefore, Mj=∑aμara2Ωj⏟∑kIj,kΩk−∑aμa(∑kra,kΩk)ra,k=∑aμa∑k(ra2Ij,k−ra,jra,k)Ωk=∑k∑aμa(ra2Ij,k−ra,jra,k)Ωk=∑kJj,kΩk,\begin{align*} M_j&=\sum_a\mu_a\mathbf r_a^2\underbrace{\Omega_j}_{\sum_kI_{j,k}\Omega_k} -\sum_a\mu_a\left(\sum_kr_{a,k}\Omega_k\right)r_{a,k}\\ &=\sum_a\mu_a\sum_k\left(\mathbf r_a^2I_{j,k}-r_{a,j}r_{a,k}\right)\Omega_k\\ &=\sum_k\sum_a\mu_a\left(\mathbf r_a^2I_{j,k}-r_{a,j}r_{a,k}\right)\Omega_k\\ &=\sum_kJ_{j,k}\Omega_k, \end{align*} i.e. M=JΩ.\mathbf M=\mathbf J\mathbf\Omega.

We discover that, it seems that the kinetic energy is the sum of “translational kinetic energy” and “rotational kinetic energy”, while the momentum only contains “translational momentum”, and the angular momentum only contains “rotational angular momentum”.

We may discover that J\mathbf J seems to represent some inertia related to rotation of the rigid body, so we call it the inertia matrix (or, in tensor language, the inertia tensor). By this means we introduced an important physical quantity. Now we study some of its properties.

Note that J\mathbf J is a real symmetric matrix, so it has real eigenvaluese Jpri\mathbf J_{\mathrm{pri}} (this is a vector containing all the eigenvalues as its components), and its (normalized) eigenvectors epri\mathbf e_{\mathrm{pri}} (this is a matrix, with its columns being the eigenvectors) orthogonal to each other. These eigenvalues are called the principal moments of inertia, and the axes parallel to the eigenvectors and passing through the center of mass are called the principal axes of inertia.

J\mathbf J is diagonalizable, i.e. epri−1Jepri=diag⁡Jpri.\mathbf e_{\mathrm{pri}}^{-1}\mathbf J\mathbf e_{\mathrm{pri}} =\operatorname{\mathbf{diag}}\mathbf J_{\mathrm{pri}}. Because the eigenvectors of J\mathbf J are orthogonal to each other, the matrix epri\mathbf e_{\mathrm{pri}} is an orthogonal matrix, i.e. epriT=epri,j,\mathbf e_{\mathrm{pri}}^{\mathrm T}=\mathbf e_{\mathrm{pri},j}, so qpri=(qTepri)T=epriTq=epri−1q,\begin{align*} \mathbf q_{\mathrm{pri}}&=\left(\mathbf q^\mathrm T\mathbf e_{\mathrm{pri}}\right)^\mathrm T\\ &=\mathbf e_{\mathrm{pri}}^\mathrm T\mathbf q\\ &=\mathbf e_{\mathrm{pri}}^{-1}\mathbf q, \end{align*} i.e. epriqpri=q.\mathbf e_{\mathrm{pri}}\mathbf q_{\mathrm{pri}}=\mathbf q. We may find that M=JΩ=(epriepri−1)J(epriepri−1)Ω=epri(epri−1Jepri)(epri−1Ω)=epri(diag⁡Jpri)Ωpri.\begin{align*} \mathbf M&=\mathbf J\mathbf\Omega\\ &=\left(\mathbf e_{\mathrm{pri}}\mathbf e_{\mathrm{pri}}^{-1}\right)\mathbf J \left(\mathbf e_{\mathrm{pri}}\mathbf e_{\mathrm{pri}}^{-1}\right)\mathbf\Omega\\ &=\mathbf e_{\mathrm{pri}}\left(\mathbf e_{\mathrm{pri}}^{-1}\mathbf J\mathbf e_{\mathrm{pri}}\right) \left(\mathbf e_{\mathrm{pri}}^{-1}\mathbf\Omega\right)\\ &=\mathbf e_{\mathrm{pri}}\left(\operatorname{\mathbf{diag}}\mathbf J_{\mathrm{pri}}\right)\mathbf\Omega_{\mathrm{pri}}. \end{align*} On the other hand, we have M=epriMpri,\mathbf M=\mathbf e_{\mathrm{pri}}\mathbf M_{\mathrm{pri}}, so Mpri=(diag⁡Jpri)Ωpri,\mathbf M_{\mathrm{pri}}=\left(\operatorname{\mathbf{diag}}\mathbf J_{\mathrm{pri}}\right)\mathbf\Omega_{\mathrm{pri}}, i.e. Mpri,j=Jpri,jΩpri,j.M_{\mathrm{pri},j}=\mathbf J_{\mathrm{pri},j}\Omega_{\mathrm{pri},j}. This means that the component of the angular momentum along some certain principal axis of inertia is the product of the principal inertia along this axis and the component of the angular velocity along this axis. This can simplify the calculation of the angular momentum a lot when the principal inertia can be calculated.

Now, we may study the rotational kinetic energy Trot≔J2Ω⋅Ω=12M⋅Ω=12M⋅Ω=12epri(diag⁡Jpri)Ωpri⋅Ω=12(diag⁡Jpri)(epriΩ)⋅Ωpri=12(diag⁡Jpri)Ωpri⋅Ωpri=12∑jJpri,jΩpri,j2.\begin{align*} T_{\mathrm{rot}}&\coloneqq\frac{\mathbf J}2\mathbf\Omega\cdot\mathbf\Omega\\ &=\frac12\mathbf M\cdot\mathbf\Omega\\ &=\frac12\mathbf M\cdot\mathbf\Omega\\ &=\frac12\mathbf e_{\mathrm{pri}}\left(\operatorname{\mathbf{diag}}\mathbf J_{\mathrm{pri}}\right) \mathbf\Omega_{\mathrm{pri}}\cdot\mathbf\Omega\\ &=\frac12\left(\operatorname{\mathbf{diag}}\mathbf J_{\mathrm{pri}}\right) \left(\mathbf e_{\mathrm{pri}}\mathbf\Omega\right)\cdot\mathbf\Omega_{\mathrm{pri}}\\ &=\frac12\left(\operatorname{\mathbf{diag}}\mathbf J_{\mathrm{pri}}\right) \mathbf\Omega_{\mathrm{pri}}\cdot\mathbf\Omega_{\mathrm{pri}}\\ &=\frac12\sum_jJ_{\mathrm{pri},j}\Omega_{\mathrm{pri},j}^2. \end{align*} This also makes the calculation simplified when the principal inertia can be calculated.

Study how J\mathbf J will change if ra\mathbf r_a experience some transformations.

First, consider translation. Let ra′≔ra+d,\mathbf r_a'\coloneqq\mathbf r_a+\mathbf d, J′≔∑aμa(ra′2I−ra′⊗2)\mathbf J'\coloneqq\sum_a\mu_a\left(\mathbf r_a'^2\mathbf I-\mathbf r_a'^{\otimes2}\right) (we may regard J′\mathbf J' as the inertia matrix of the rigid body w.r.t. the point with d\mathbf d distance away from the center of mass, but not (in usual cases) w.r.t. the center of mass), and then we will discover that J\mathbf J can be calculated from J′\mathbf J' by J=J′−m(d2I−d⊗2).\mathbf J=\mathbf J'-m\left(\mathbf d^2\mathbf I-\mathbf d^{\otimes2}\right). This provides us a useful way to calculate J\mathbf J when J′\mathbf J' is easy to calculate while J\mathbf J is not.

The proof is not difficult: J′≔∑aμa(ra′2I−ra′⊗2)=∑aμa((ra+d)2I−(ra+d)⊗2)=∑aμa((ra2+2ra⋅d+d2)I−(ra⊗2+d⊗ra+ra⊗d+d⊗2))=∑aμa(ra2I−ra⊗2)⏟J+(∑aμa)⏟m(d2I−d⊗2)+∑aμa(2ra⋅dI−d⊗ra−ra⊗d)=J+m(d2I−d⊗2)+2(∑aμara)⏟0⋅dI−d⊗(∑aμara)⏟0−(∑aμara)⏟0⊗d=J+m(d2I−d⊗2),\begin{align*} \mathbf J'&\coloneqq\sum_a\mu_a\left(\mathbf r_a'^2\mathbf I-\mathbf r_a'^{\otimes2}\right)\\ &=\sum_a\mu_a\left(\left(\mathbf r_a+\mathbf d\right)^2\mathbf I-\left(\mathbf r_a+\mathbf d\right)^{\otimes2}\right)\\ &=\sum_a\mu_a\left(\left(\mathbf r_a^2+2\mathbf r_a\cdot\mathbf d+\mathbf d^2\right)\mathbf I -\left(\mathbf r_a^{\otimes2}+\mathbf d\otimes\mathbf r_a+\mathbf r_a\otimes\mathbf d+\mathbf d^{\otimes 2}\right)\right)\\ &=\underbrace{\sum_a\mu_a\left(\mathbf r_a^2\mathbf I-\mathbf r_a^{\otimes2}\right)}_\mathbf J +\underbrace{\left(\sum_a\mu_a\right)}_m\left(\mathbf d^2\mathbf I-\mathbf d^{\otimes2}\right) +\sum_a\mu_a\left(2\mathbf r_a\cdot\mathbf d\mathbf I-\mathbf d\otimes\mathbf r_a-\mathbf r_a\otimes\mathbf d\right)\\ &=\mathbf J+m\left(\mathbf d^2\mathbf I-\mathbf d^{\otimes2}\right) +2\underbrace{\left(\sum_a\mu_a\mathbf r_a\right)}_\mathbf 0\cdot\mathbf d\mathbf I -\mathbf d\otimes\underbrace{\left(\sum_a\mu_a\mathbf r_a\right)}_\mathbf 0 -\underbrace{\left(\sum_a\mu_a\mathbf r_a\right)}_\mathbf 0\otimes\mathbf d\\ &=\mathbf J+m\left(\mathbf d^2\mathbf I-\mathbf d^{\otimes2}\right), \end{align*} i.e. J=J′−m(d2I−d⊗2).\mathbf J=\mathbf J'-m\left(\mathbf d^2\mathbf I-\mathbf d^{\otimes2}\right).

The special case of it is called the parallel axis theorem in general physics.

Now, consider rotation. Suppose the rotational matrix R\mathbf R represents the rotation around the unit vector u\mathbf u. Let ra′≔Rra,\mathbf r_a'\coloneqq\mathbf R\mathbf r_a, J′≔∑aμa(ra′2I−ra′⊗2),\mathbf J'\coloneqq\sum_a\mu_a\left(\mathbf r_a'^2\mathbf I-\mathbf r_a'^{\otimes2}\right), and we will discover J′=RJR−1.\mathbf J'=\mathbf R\mathbf J\mathbf R^{-1}.

Now we will prove the conclusion. First, we now that R\mathbf R is an orthogonal matrix, i.e. RT=R−1.\mathbf R^\mathrm T=\mathbf R^{-1}. By this we can get other properties like (Rq)2=q2,\left(\mathbf R\mathbf q\right)^2=\mathbf q^2, (Rq)⊗2=Rq⊗2R−1.\left(\mathbf R\mathbf q\right)^{\otimes2}=\mathbf R\mathbf q^{\otimes2}\mathbf R^{-1}. Therefore, J′≔∑aμa(ra′2I−ra′⊗2)=∑aμa(ra2I−Rra⊗2R−1)=R(∑aμa(ra2I−ra⊗2))R−1=RJR−1.\begin{align*} \mathbf J'&\coloneqq\sum_a\mu_a\left(\mathbf r_a'^2\mathbf I-\mathbf r_a'^{\otimes2}\right)\\ &=\sum_a\mu_a\left(\mathbf r_a^2\mathbf I-\mathbf R\mathbf r_a^{\otimes2}\mathbf R^{-1}\right)\\ &=\mathbf R\left(\sum_a\mu_a\left(\mathbf r_a^2\mathbf I-\mathbf r_a^{\otimes2}\right)\right)\mathbf R^{-1}\\ &=\mathbf R\mathbf J\mathbf R^{-1}. \end{align*}

After getting this conclusion, we may be interested in non trivial (i.e. R≠I\mathbf R\ne\mathbf I) cases of J′=J\mathbf J'=\mathbf J. We will prove that in this case u\mathbf u is a principal axis of the rigid body.

By J′=RJR−1\mathbf J'=\mathbf R\mathbf J\mathbf R^{-1}, we can get J=RJR−1,\mathbf J=\mathbf R\mathbf J\mathbf R^{-1}, i.e. RJ=JR.\mathbf R\mathbf J=\mathbf J\mathbf R. Multiply u\mathbf u on both sides, and we have RJu=JRu,\mathbf R\mathbf J\mathbf u=\mathbf J\mathbf R\mathbf u, i.e. R(Ju)=J(Ru).\mathbf R\left(\mathbf J\mathbf u\right)=\mathbf J\left(\mathbf R\mathbf u\right). By Euler’s rotation theorem, the only real eigenvalue of nontrivial R\mathbf R is 11, and its corresponding eigenvector is u\mathbf u, i.e. Ru=u.\mathbf R\mathbf u=\mathbf u. Substitute, and we have R(Ju)=Ju,\mathbf R\left(\mathbf J\mathbf u\right)=\mathbf J\mathbf u, i.e. Ju\mathbf J\mathbf u is an eigenvector of R\mathbf R.

Because the only independent (real) eigenvector of notrial R\mathbf R is u\mathbf u, Ju\mathbf J\mathbf u is parallel to u\mathbf u, so u\mathbf u is an eigenvector of J\mathbf J, i.e. u\mathbf u is a principal axis of the rigid body.

This leads to an important conclusion: if the rigid body maintains its original inertia matrix after it rotates around an axis passing its center of mass by a nontrivial angle (an angle that is not a multiple of 2π2\pi), then this axis is a principal axis of the rigid body.

By this we can also get another important conclusion: if a rigid body is rotational symmetric around some axis, then this axis is a principal axis of the rigid body. This conclusion simplifies the determination of principal axes of a rigid body with some symmetries.