\(\text{ }\) Polynomials and their roots. Recall that by the fundamental theorem of algebra, any degree \(k\) polynomial with complex coefficients has exactly \(k\) roots (possibly complex), counted with multiplicity. That is, for any \(a_{0},\ldots,a_{k-1}\in\mathbb{C}\) the polynomial \(p(r)\) from ([eq:ch5_ho_lin_aut_char_poly]) can be written as \[p(r)=(r-r_{1})(r-r_{2})\ldots(r-r_{k}),\tag{5.95}\] where \(r_{1},\ldots,r_{k}\in\mathbb{C}\) are its roots, and where a root \(r\) has multiplicity \(m\) if exactly \(r=r_{i}\) for exactly \(m\) indices \(i.\) Alternatively letting \(r_{1},\ldots,r_{k'}\) where \(k'\le k\) be the distinct roots, and letting \(m_{i}\) be the multiplicity of \(r_{i},\) so that necessarily \(m_{1}+\ldots+m_{k'}=k,\) we can write the polynomial as \[p(r)=\prod_{l=1}^{k'}(r-r_{l})^{m_{l}}.\tag{5.96}\]

This is in contrast to the situation for polynomials over the reals: a degree \(k\) polynomial with real coefficients can have less than \(k\) roots, even counting multiplicity. For instance the polynomial \(x^{2}+1\) has no real roots, and can’t be factored over the reals, but has complex roots \(\pm i\) and complex factorization \[x^{2}+1=(x-i)(x+i).\] The polynomial \(x^{4}+x^{3}+x+1\) has exactly one real root \(x=-1,\) of multiplicity \(2,\) and can be factored over the reals as \[x^{4}+x^{3}+x+1=(x+1)^{2}(x^{2}-x+1),\] and in addition has complex roots \(x=e^{\frac{\pi}{3}i},x=e^{-\frac{\pi}{3}i}\) (the roots of the second quadratic factor), so over \(\mathbb{C}\) it can be factored as \[x^{4}+x^{3}+x+1=(x+1)^{2}(x-e^{\frac{\pi}{3}i})(x-e^{-\frac{\pi}{3}i})=(x+1)(x+1)(x-e^{\frac{\pi}{3}i})(x-e^{-\frac{\pi}{3}i}).\]

\(\text{ }\)Solutions from roots. The first step in solving a higher order linear autonomous ODE is finding the roots of the characteristic polynomial.

Example 5.27(Continuation of Example 5.24). The ODE ([eq:ch5_ho_lin_aut_first_ex]) has characteristic polynomial \[p(r)=r^{10}-1.\] Clearly \(r=1\) is a root. The complete set of roots is the tenth roots of unity: \(r_{l}=e^{\frac{l-1}{10}2\pi i},1\le l\le10.\) Thus for each \(l\) \[y(t)=e^{r_{l}t}\] is a solution of ([eq:ch5_ho_lin_aut_first_ex]), and furthermore since the ODE is homogeneous \[y(t)=\sum_{l=1}^{10}\alpha_{l}e^{r_{l}t}\] is a solution for any \(\alpha_{1},\ldots,\alpha_{10}\in\mathbb{C}.\)
Remark 5.28

Factoring a polynomial with arbitrary coefficients is difficult (for degrees \(3,4\)) or even impossible (for insoluble polynomials of degree \(\ge5\)). In exercises meant to be solved with pen and paper it is likely that there is a “nice” factorization, which can be found for instance by guessing simple roots like \(\pm1,\pm2,\pm3,\ldots.\). Or the problem can be solved without knowing the roots explicitly .

\(\text{ }\) IVPs/constraints. A first order linear autonomous homogeneous ODEs has a one parameter set of solutions, and if you impose a constraint of the form \(y(t_{0})=y_{0}\) there is a unique solution (Proposition 3.13, Proposition 3.15, Section 3.2). A second order such equation has a two parameter family of solutions, and if you impose only one constraint \(y(t_{0})=y_{0}\) there are infinitely many solutions, but with two constraints \(y(t_{0})=y_{0},y'(t_{0})=y_{1}\) there is a unique solution (Proposition 5.20, Proposition 5.22, Section 5.1). Essentially, if an ODE has a \(k\) parameter set of solutions, then generally solutions become unique with \(k\) constraints. For a \(k\)-th order ODE ([eq:ch5_ho_lin_aut_ODE_homo_def]) whose characteristic polynomial has distinct roots \(r_{l},\) linear combinations of the roots \(y(t)=e^{r_{l}t},1\le l\le k\) are a \(k\)-parameter family of solutions, and indeed with the \(k\) constraints \(y(t_{0})=y_{0},y'(t_{0})=y_{1},\ldots,y^{(k-1)}(t_{0})=y_{k-1}\) it has a unique solution, for any \(y_{0},\ldots,y_{k-1}\in\mathbb{C}\) (this will be proven below).
Example 5.29 Continuation of Example 5.24 and Example 5.27. Let us solve the IVP \[y^{(10)}-y=0,\quad\quad y^{(l)}(0)=\begin{cases} 1 & \text{\,if }l\text{ even},\\ 0 & \text{ if }l\text{\,odd}, \end{cases},\quad\quad0\le l\le9.\tag{5.98}\]

Letting \(r_{n}=e^{2\pi i\frac{n-1}{10}},1\le n\le10\) denote the roots of the characteristic polynomial we know that \[y(t)=\sum_{n=1}^{10}\alpha_{n}e^{tr_{n}}\] is a solution for any \(\alpha_{1},\ldots,\alpha_{10}\in\mathbb{C}.\) The derivatives of this solution are given by \[y^{(l)}(t)=\sum_{n=1}^{10}\alpha_{n}r_{n}^{l}e^{tr_{n}},\quad0\le l\le9,\] and for \(t=0\) \[y^{(l)}(0)=\sum_{n=1}^{10}\alpha_{n}r_{n}^{l},\quad0\le l\le9.\] Thus to find a solution that satisfies the constraints we need to solve the system of ten linear equations \[\sum_{l=0}^{9}\alpha_{n}r_{n}^{l}=y_{l},0\le l\le9,\] where \[y_{l}=\begin{cases} 1 & \text{\,if }l\text{ even},\\ 0 & \text{ if }l\text{\,odd}, \end{cases},\quad\quad0\le l\le9.\] Solving a system of ten linear equations in general requires inverting a \(10\times10\) matrix - a non-trivial task - or using the Gaussian elimination algorithm, which will take a while with pen and paper for \(10\) rows. For these specific \(r_{n}\) and \(y_{l}\) we can exploit a special property to arrive at a solution faster. Write the system as a matrix-vector equation: \[M\alpha=y_{*},\tag{5.99}\] where \(\alpha=(\alpha_{1},\ldots,\alpha_{n})\in\mathbb{C}^{10},y_{*}=(1,0,1,0,1,0,1,0,1,0)\in\mathbb{C}^{10}\) and \[M=\left(\begin{matrix}1 & 1 & \ldots & 1 & \ldots & 1 & 1\\ r_{1} & r_{2} & \ldots & r_{6} & \ldots & r_{9} & r_{10}\\ r_{1}^{2} & r_{2}^{2} & \ldots & r_{6}^{2} & \ldots & r_{9}^{2} & r_{10}^{2}\\ & & \ldots & \ldots & \ldots\\ & & \ldots & \ldots & \ldots\\ r_{1}^{8} & r_{2}^{8} & \ldots & r_{6}^{8} & \ldots & r_{9}^{8} & r_{10}^{8}\\ r_{1}^{9} & r_{2}^{9} & \ldots & r_{6}^{9} & \ldots & r_{9}^{9} & r_{10}^{9} \end{matrix}\right)=\left(\begin{matrix}1 & 1 & \ldots & 1 & \ldots & 1 & 1\\ 1 & r_{2} & \ldots & -1 & \ldots & r_{9} & r_{10}\\ 1 & r_{2}^{2} & \ldots & 1 & \ldots & r_{9}^{2} & r_{10}^{2}\\ & & \ldots & \ldots & \ldots\\ & & \ldots & \ldots & \ldots\\ 1 & r_{2}^{8} & \ldots & 1 & \ldots & r_{9}^{8} & r_{10}^{8}\\ 1 & r_{2}^{9} & \ldots & -1 & \ldots & r_{9}^{9} & r_{10}^{9} \end{matrix}\right),\] whereby the first column is \(M_{1\cdot}=(1,\ldots,1)\) since \(r_{1}=e^{\frac{1-1}{10}2\pi i}=e^{0}=1\) and the \(6\)-th column is \(M_{6\cdot}=(1,-1,1,\ldots,-1)\) since \(r_{6}=e^{\frac{6-1}{10}2\pi i}=-1.\) Our target vector \(y_{*}\) is a linear combination of \(M_{1\cdot},M_{6\cdot},\) more precisely \[y_{*}=\frac{M_{1\cdot}+M_{6\cdot}}{2}.\] Therefore \(\alpha_{1}=\alpha_{6}=\frac{1}{2}\) and \(\alpha_{i}=0\) for other \(i\) solves ([eq:ch5_ho_lin_aut_example_IVP_matrix_eq]). Thus \[y(t)=\alpha_{1}e^{r_{1}t}+\alpha_{6}e^{r_{5}t}=\frac{e^{t}+e^{-t}}{2}\tag{5.100}\] solves the constrained ODE ([eq:ch5_ho_lin_aut_first_example_IVP]).

We double-check our calculation by computing from ([eq:ch5_ho_lin_aut_first_example_IVP_final_sol]) \[y^{(l)}(t)=\frac{e^{t}+(-1)^{l}e^{-t}}{2},\quad l\ge0,\tag{5.101}\] and plugging in to the ODE in ([eq:ch5_ho_lin_aut_first_example_IVP]), obtaining \[y^{(10)}(t)-y(t)=\frac{e^{t}+(-1)^{10}e^{-t}}{2}-\frac{e^{t}+e^{-t}}{2}=\frac{e^{t}+e^{-t}}{2}-\frac{e^{t}+e^{-t}}{2}=0\quad\forall t\checkmark,\] and checking the constraints from ([eq:ch5_ho_lin_aut_first_example_IVP]) by noting from ([eq:ch5_ho_lin_aut_first_example_IVP_final_sol_derivs]) that \[y^{(l)}(0)=\frac{1+(-1)^{l}}{2}=\begin{cases} 1 & \text{ if }l\text{ even,}\\ 0 & \text{ if }l\text{\,odd}. \end{cases}\checkmark\]

Remark 5.30

Even though we went through complex valued computations, we ended up with a real valued answer, a phenomenon we already saw in the second order case with complex roots, and which makes perfect sense since the ODE has real coefficients and the r.h.s. of each constraint is real. Below we will prove that in this situation there is always a unique solution, and this solution is real valued.

Exercise 5.31

Like for the case \(k=2,\) it turns out that when the characteristic polynomial has no repeated roots the linear combinations of the \(k\) solutions \(t\to e^{rt},\) for each root \(r,\) is the complete set of solutions (the proof will be given below)

\(\text{ }\) Repeated roots. Let us investigate the case of a repeated root. Then the functions \(t\to e^{rt}\) yield less than \(k\) solutions. Assume that \(r_{*}\) is a repeated root. By analogy to the second order case it makes sense to test if \[y(t)=te^{r_{*}t}\tag{5.103}\] is a solution.

To compute the derivatives we use a general formula for the \(l\)-th derivative of a product \(f(t)g(t)\) of differentiable functions \(f(t),g(t).\) The case \(l=1\) is just the usual product rule: \[\frac{d}{dt}\left\{ f(t)g(t)\right\} =f'(t)g(t)+f(t)g'(t).\tag{5.104}\] We can compute the second \(l=2\) and third \(l=3\) derivatives by iterating the \(l=1\) formula: \[\begin{array}{ccl} \frac{d^{2}}{dt^{2}}\left\{ f(t)g(t)\right\} & = & f''(t)g(t)+2f'(t)g'(t)+f(t)g''(t)\\ \frac{d^{3}}{dt^{3}}\left\{ f(t)g(t)\right\} & = & f'''(t)g(t)+3f''(t)g'(t)+3f'(t)g''(t)+f(t)g'''(t). \end{array}\] The general formula (which can be proved e.g. by induction) for the \(l\)-th derivative is \[\frac{d^{l}}{dt^{l}}\left\{ f(t)g(t)\right\} =\sum_{i=0}^{l}{l \choose i}f^{(i)}(t)g^{(l-i)}(t),\tag{5.105}\] (note the analogy to the binomial formula). All the formulas ([eq:ch5_ho_lin_aut_product_rule])-([eq:ch5_ho_lin_aut_iterated_product_rule]) hold both for real-valued functions \(f,g:I\to\mathbb{R}\) and more generally for complex-valued functions \(f,g:\mathbb{R}\to\mathbb{C}.\)

Using ([eq:ch5_ho_lin_aut_iterated_product_rule]) with \(f(t)=t\) (so \(f'(t)=1,f''(t)=0,f'''(t)=0,\ldots\)) and \(g(t)=e^{r_{*}t}\) (so \(g^{(l)}(t)=r_{*}^{l}e^{r_{*}t}\)) we can compute the \(l\)-th derivative of \(y(t)\) in ([eq:ch5_ho_lin_aut_t_exp_t_ansatz]) as \[y^{(l)}(t)=tr_{*}^{l}e^{r_{*}t}+lr_{*}^{l-1}e^{r_{*}t},\quad l\ge0.\] Thus ([eq:ch5_ho_lin_aut_t_exp_t_ansatz]) satisfies \[y^{(k)}+\sum_{l=0}^{k-1}a_{l}y^{(l)}=tr_{*}^{k}e^{r_{*}t}+kr_{*}^{k-1}e^{r_{*}t}+\sum_{l=0}^{k-1}a_{l}\left\{ tr_{*}^{l}e^{r_{*}t}+lr_{*}^{l-1}e^{r_{*}t}\right\} .\tag{5.106}\] For convenience we let \(a_{k}=1,\) so we can write the general homogeneous ODE ([eq:ch5_ho_lin_aut_ODE_homo_def]) as \[\sum_{l=0}^{k}a_{l}y^{(l)}=0,\tag{5.107}\] the characteristic polynomial ([eq:ch5_ho_lin_aut_char_poly]) as \[p(r)=\sum_{l=0}^{k}a_{k}r^{l}.\tag{5.108}\] Then ([eq:ch5_ho_lin_aut_t_exp_t_ansatz_plugged_in]) equals \[\sum_{l=0}^{k}a_{l}y^{(l)}=\sum_{l=0}^{k}a_{l}\left\{ tr_{*}^{l}e^{r_{*}t}+lr_{*}^{l-1}e^{r_{*}t}\right\} .\] We can rearrange the r.h.s. as \[te^{r_{*}t}\sum_{l=0}^{k}a_{l}r_{*}^{l}+e^{rt}\sum_{l=0}^{k}a_{l}lr_{*}^{l-1}.\tag{5.109}\] Note that the first of sum on the r.h.s. is simply \(p(r_{*}),\) cf. ([eq:ch5_ho_lin_aut_char_poly_with_ak]). Inspecting the second sum, we see that it coincides with the derivative \(p'(r_{*}),\) since \[p'(r)=\sum_{l=0}^{k}a_{k}\frac{d}{dr}\left\{ r^{l}\right\} =\sum_{l=0}^{k}a_{k}lr^{l-1}!\] Thus for \(y\) as in ([eq:ch5_ho_lin_aut_t_exp_t_ansatz]) \[\sum_{l=0}^{k}a_{l}y^{(l)}=te^{r_{*}t}p(r_{*})+e^{r_{*}t}p'(r_{*}).\tag{5.110}\] Since \(r\) is a solution of the characteristic equation it holds that \(p(r_{*})=0,\) so the first term vanishes. What about \(p'(r_{*})\)? Recall from ([eq:ch5_ho_lin_aut_FTOA_with_mults]) from that if \(r_{*}\) has multiplicity \(m\) then \[p(r)=(r-r_{*})^{m}q(r)\tag{5.111}\] for the polynomial \(q(r)=\prod_{s\text{ root}:s\ne r_{*}}(r-s).\) Thus \[p'(r)=m(r-r_{*})^{m-1}q(r)+(r-r_{*})^{m}q'(r),\] and if \(m\ge2\) then both terms on the r.h.s. vanish if \(r=r_{*},\) showing that \[p'(r_{*})=0\text{ if }r_{*}\text{\,is a repeated root}.\tag{5.112}\] Thus it follows from ([eq:ch5_ho_lin_aut_repeated_root_rhs_in_terms_of_p_pprime]) that \[\sum_{l=0}^{k}a_{l}y^{(l)}=te^{r_{*}t}\cdot0+e^{r_{*}t}\cdot0=0,\] so \(y(t)=te^{r_{*}t}\) is indeed a solution of ([eq:ch5_ho_lin_aut_ODE_homo_def_with_ak]) if \(r_{*}\) is a repeated root, just as in the second order case.

\(\text{ }\) Repeated roots with higher multiplicity. What if there are repeated roots of higher multiplicity? It turns out that then \(t^{2}e^{rt},t^{3}e^{rt},\ldots\) can be solutions. Indeed consider for \(n\in\{0,1,2,\ldots\}\) \[y(t)=t^{n}el^{rt}.\tag{5.113}\] Let us plug this in to ([eq:ch5_ho_lin_aut_ODE_homo_def_with_ak]). Since the computation is fairly long we encapsulate it in the following lemma, which generalizes ([eq:ch5_ho_lin_aut_repeated_root_rhs_in_terms_of_p_pprime]) to higher powers of \(t.\)

Lemma 5.32

Let \(I\subset\mathbb{R}.\) Let \(k\ge1,a_{0},\ldots,a_{k}\in\mathbb{C}\) and let \(p(r)=\sum_{l=0}^{k}a_{l}r^{l}\) be a polynomial. Furthermore let \(r\in\mathbb{C},n\in\{0,1,2,\ldots\},\) and \(y(t)\) be as in ([eq:ch5_ho_lin_aut_repeated_root_sol_higher_t_power]). Then \[\sum_{l=0}^{k}a_{l}y^{(l)}(t)=\sum_{i=0}^{n}{n \choose i}t^{n-i}p^{(i)}(r),\quad\quad t\in I.\tag{5.114}\]

Proof. Using the iterated product rule ([eq:ch5_ho_lin_aut_iterated_product_rule]) with \(f(t)=t^{n}\) (so that \(f^{(i)}=0\) if \(i>n\) and \(f^{(i)}=n(n-1)\ldots(n-i+1)t^{n-i}\) otherwise) and \(g(t)=e^{rt}\) we obtain the formula \[y^{(l)}(t)=\sum_{i=0}^{\max(l,n)}{l \choose i}n(n-1)\ldots(n-i+1)t^{n-i}r^{l-i}e^{rt},\quad\quad l\ge0,\tag{5.115}\] for the \(l\)-th derivative. Define \(C_{l,i}=l(l-1)\ldots(l-k+1).\) With this notation the constant can be rewritten as \[{l \choose i}n(n-1)\ldots(n-i+1)=\frac{l(l-1)\ldots(l-i+1)}{i!}n(n-1)\ldots(n-i+1)={n \choose i}C_{l,i}.\] Then ([eq:ch5_ho_lin_aut_repeated_root_sol_higher_t_power_derivs_first_step]) becomes \[y^{(l)}(t)=e^{rt}\sum_{i=0}^{\max(l,n)}{n \choose i}C_{l,i}t^{n-i}r^{l-i},\quad\quad l\ge0,\] Plugging this into the r.h.s. of ([eq:ch5_ho_lin_aut_repeated_root_sol_higher_t_power_derivs]) gives \[\sum_{l=0}^{k}a_{l}y^{(l)}(t)=e^{rt}\sum_{l=0}^{k}a_{l}\sum_{i=0}^{\max(l,n)}{n \choose i}C_{l,i}t^{n-i}r^{l-i}.\tag{5.116}\] To facilitate switching the order of summation, write the sums as \[\sum_{l=0}^{k}\sum_{i=0}^{n}1_{\left\{ i\le l\right\} }a_{l}{n \choose i}C_{l,i}t^{n-i}r^{l-i}.\] Now the order of the sums can be switched to rewrite the last expression as \[\begin{array}{ccl} \sum_{i=0}^{n}\sum_{l=0}^{k}1_{\left\{ i\le l\right\} }a_{l}{n \choose i}C_{l,i}t^{n-i}r^{l-i} & = & \sum_{i=0}^{n}\sum_{l=i}^{k}a_{l}{n \choose i}C_{l,i}t^{n-i}r^{l-i}\\ & = & \sum_{i=0}^{n}{n \choose i}t^{n-i}\sum_{l=i}^{k}a_{l}C_{l,i}r^{l-i}. \end{array}\] We have arrived at \[\sum_{l=0}^{k}a_{l}y^{(l)}(t)=e^{rt}\sum_{i=0}^{n}t^{n-i}\sum_{l=i}^{k}a_{l}{n \choose i}C_{l,i}r^{l-i}.\tag{5.117}\] The claim now follows since \[p^{(i)}(r)=\sum_{l=i}^{k}a_{l}l(l-1)(l-2)\ldots(l-i+1)r^{l-i}=\sum_{l=i}^{k}a_{l}C_{l,i}r^{l-i}.\]

To check if \(y\) as in ([eq:ch5_ho_lin_aut_repeated_root_sol_higher_t_power]) is a solution to ([eq:ch5_ho_lin_aut_ODE_homo_def_with_ak]) we need to evaluate the r.h.s. of this expression, i.e. we need to compute the derivatives \(p^{(i)}(r).\) But it is a basic property of polynomials that if a root \(r\) has multiplicity \(m,\) then \(p^{(i)}(r)=0\) for all \(i<m,\) as the following generalization of the formula ([eq:ch5_ho_lin_aut_ODE_homo_p_prime_of_rep_r]) proves.

Lemma 5.33

Let \(h(r)\) be a non-zero complex-valued polynomial of degree \(k\) in \(r,\) and \(s\) a root of multiplicity \(m\in\{1,2,\ldots,k\}.\) Then \(h^{(i)}(s)=0\) for all \(i<m.\)

Proof. Recall from ([eq:ch5_ho_lin_aut_repeated_root_rhs_in_terms_of_p_factored_out_root_of_mult_m]), if \(s\) is repeated root with multiplicity \(m\) then \[h(r)=(r-s)^{m}g(r)\] for a polynomial \(g(r).\) Using the iterated product rule ([eq:ch5_ho_lin_aut_iterated_product_rule]) with \(f(r)=(r-s)^{m}\) we obtain that the \(i\)-th derivative of \(h(r)\) is given by \[h^{(i)}(r)=\sum_{j=0}^{i}{i \choose j}f^{(j)}(r)g^{(i-j)}(r).\tag{5.118}\] Since \[f^{(j)}(r)=m(m-1)\ldots(m-j+1)(r-s)^{m-j}\text{\,for }j\le m,\] in particular \[f^{(j)}(s)=0\text{ for }j<m.\] Thus the claim follows since all the terms in ([eq:ch5_ho_lin_aut_homo_deriv_repeated_root_step_1]) vanish.

Combining this lemma and the previous one, we see that if \(r\) is a root of the characteristic polynomial ([eq:ch5_ho_lin_aut_char_poly]) of multiplicity \(m\in\{1,2,\ldots,\}\) then \[e^{rt},te^{rt},\ldots,t^{m-1}e^{rt}\] are \(m\) distinct solutions of the corresponding homogeneous ODE ([eq:ch5_ho_lin_aut_ODE_homo_def]). Note that this means the for any characteristic polynomial of degree \(k,\) regardless of the multiplicity of its roots, there are exactly \(k\) distinct solutions of the form \(t^{l}e^{rt}.\) Furthermore, by taking linear combinations of these one sees that for any complex polynomial \(q(t)\) of degree at most \(m-1\) \[y(t)=q(t)e^{rt}\] is a solution of ([eq:ch5_ho_lin_aut_ODE_homo_def]).

Example 5.34

Let us solve the IVP \[y''''-2y'''+2y'-y=0,\quad\quad y(0)=y'(0)=y''(0)=0,y'''(0)=1.\tag{5.120}\] Its characteristic polynomial is \[p(r)=r^{4}-2r^{2}+2r-1.\] We note that \(p(1)=0\) so \(r=1\) is a root, which we can factor out to obtain \[p(r)=(r-1)(r^{3}-r^{2}-r+1).\] Again \(r=1\) is a root of the second factor \(r^{3}-r^{2}-r+1,\) so we can factor out another \((r-1)\) to obtain \[p(r)=(r-1)^{2}(r^{2}-1).\] The last factor is easily factorized as \(r^{2}-1=(r-1)(r+1),\) so the final factorization giving all the roots is \[p(r)=(r-1)^{3}(r+1).\] The root \(r=1\) has multiplicity \(3,\) so for any \(\alpha_{1},\alpha_{2},\alpha_{3},\alpha_{4}\in\mathbb{C}\) the function \[y(t)=(\alpha_{1}+t\alpha_{2}+t^{2}\alpha_{3})e^{t}+\alpha_{4}e^{-t}\] is a solution of the ODE. This function has derivatives \[\begin{array}{ccl} y'(t) & = & (\alpha_{2}+2t\alpha_{3})e^{t}+(\alpha_{1}+t\alpha_{2}+t^{2}\alpha_{3})e^{t}-\alpha_{4}e^{-t},\\ y''(t) & = & 2\alpha_{3}e^{t}+2(\alpha_{2}+2t\alpha_{3})e^{t}+(\alpha_{1}+t\alpha_{2}+t^{2}\alpha_{3})e^{t}+\alpha_{4}e^{-t},\\ y'''(t) & = & 6\alpha_{3}e^{t}+3(\alpha_{2}+2t\alpha_{3})e^{t}+(\alpha_{1}+t\alpha_{2}+t^{2}\alpha_{3})e^{t}-\alpha_{4}e^{-t}, \end{array}\] so that after plugging in \(t=0\) \[\begin{array}{ccl} y(0) & = & \alpha_{1}+\alpha_{4},\\ y'(0) & = & \alpha_{2}+\alpha_{1}-\alpha_{4},\\ y''(0) & = & 2\alpha_{3}+2\alpha_{2}+\alpha_{1}+\alpha_{4},\\ y'''(0) & = & 6\alpha_{3}+3\alpha_{2}+\alpha_{1}-\alpha_{4}. \end{array}\] To satisfy the constraints we need that \[\begin{array}{ccl} 0 & = & \alpha_{1}+\alpha_{4},\\ 0 & = & \alpha_{2}+\alpha_{1}-\alpha_{4},\\ 0 & = & 2\alpha_{3}+2\alpha_{2}+\alpha_{1}+\alpha_{4},\\ 1 & = & 6\alpha_{3}+3\alpha_{2}+\alpha_{1}-\alpha_{4}. \end{array}\] The first line implies \(\alpha_{4}=-\alpha_{1},\) the second \(\alpha_{2}=-2\alpha_{1}\) and then the third \(\alpha_{3}=2\alpha_{1}.\) Plugging these into the last line yields \(1=8\alpha_{1},\) so a solution is given by \[\alpha_{1}=\frac{1}{8},\alpha_{2}=-\frac{1}{4},\alpha_{3}=\frac{1}{4},\alpha_{4}=-\frac{1}{8}.\] Thus \[y(t)=\left(\frac{1}{8}-\frac{t}{4}+\frac{t^{2}}{4}\right)e^{t}-\frac{1}{8}e^{-t}\] solves ([eq:ch5_ho_lin_aut_homo_rep_root]) (when solving a problem yourself this is where you would double-check by differentiating and plugging in).

Exercise 5.35

We now prove that the solutions explored so far are indeed all solutions to the general \(k\)-th order linear autonomous homogeneous ODE ([eq:ch5_ho_lin_aut_ODE_homo_def]), and that \(k\) constraints make solutions unique. We obtain the same for the non-homogeneous version ([eq:ch5_ho_lin_aut_ODE_def]), since we can move between solutions of the the homogeneous and non-homogeneous equations simply by adding/subtracting the stationary solution \(y(t)=b/a_{0}\) (recall Lemma 5.25 a).
Proposition 5.36

(All solutions of higher order complex linear autonomous ODEs) Let \(I\subset\mathbb{R}\) be a non-empty interval. Let \(k\ge1,\) \(a_{0},a_{1},\ldots,a_{k-1}\in\mathbb{C}.\) Consider the ODE \[x^{(k)}(t)+\sum_{l=0}^{k-1}a_{l}x^{(l)}(t)=0.\tag{5.124}\] Assume that the characteristic polynomial \(p(r)=r^{k}+\sum_{l=0}^{k-1}a_{l}r^{l}\) has \(k'\) distinct roots \(r_{1},\ldots,r_{k'}\in\mathbb{C},\) each of multiplicity \(m_{1},m_{2},\ldots,m_{k'}\in\{1,2,\ldots\},\) where \(m_{1}+\ldots+m_{k'}=k.\) Let \[x_{1}(t),\ldots,x_{k}(t)\text{ denote the functions }e^{r_{1}t},te^{r_{1}t},\ldots,t^{m_{1}-1}e^{r_{1}t},e^{r_{2}t},te^{r_{2}t},\ldots,t^{m_{k'}-1}e^{r_{k'}t}.\]

a) A function \(x\) solves the ODE ([eq:ch5_ho_lin_aut_complex_all_sols_hom_ODE]) iff \[x(t)=\sum_{i=1}^{k}\alpha_{i}x_{i}(t),\tag{5.125}\] for some \(\alpha\in\mathbb{C}^{k}.\)
b) For any \(t_{*}\in I,\) the possibly complex \(k\times k\) matrix \(D\) with entries \[D_{i,l}=x_{i}^{(l)}(t_{*}),\] is non-degenerate.
c) Let \(t_{*}\in I\) and \(x_{*}\in\mathbb{C}^{k}.\) A function \(x\) solves the constrained homogeneous ODE \[x^{(k)}(t)+\sum_{l=0}^{k-1}a_{l}x^{(l)}(t)=0,\quad\quad t\in I,\quad\quad x^{(l)}(t_{*})=x_{*,l+1},0\le l\le k-1,\tag{5.126}\] iff \[x(t)=\sum_{i=1}^{k}\alpha_{i}x_{i}(t),\] where \(\alpha\in\mathbb{C}^{k}\) is the unique solution to \[D\alpha=x_{*},\quad\quad\text{i.e. }\quad\quad\alpha=D^{-1}x_{*}.\]
d) If \(b\in\mathbb{C}\) and \(a_{0}\ne0\) a function \(y\) solves the non-homogeneous autonomous ODE \[y^{(k)}(t)+\sum_{l=0}^{k-1}a_{l}y^{(l)}(t)=b,\quad\quad t\in I,\tag{5.127}\] iff \[y(t)=x(t)+\frac{b}{a_{0}},\quad\quad t\in I,\] for an \(x(t)\) as in part a). If \(t_{*}\in I,y_{*}\in\mathbb{C}\) then a function \(y\) solves the constrained non-homogeneous second order autonomous ODE \[y^{(k)}(t)+\sum_{l=0}^{k-1}a_{l}y^{(l)}(t)=b,\quad\quad t\in I,y^{(l)}(t_{*})=y_{*,l+1},0\le l\le k-1,\] iff \[y(t)=x(t)+\frac{b}{a_{0}},\quad\quad t\in I,\] where \(x(t)\) is the unique function from part c) with \[\alpha=D^{-1}z_{*}\quad\quad\text{ for }\quad\quad z_{*}=y_{*}-\left(\begin{matrix}\frac{b}{a_{0}}\\ 0\\ \ldots\\ 0 \end{matrix}\right).\]

Proof. a, c) imply d) by Lemma 5.25 a). We prove a) by induction. The base case \(k=1\) was proved in Proposition 5.19 (and the case \(k=2\) in Proposition 5.20, so this could also serve as a base case). Assume thus that \(k\ge2\) and that a) holds for any ODE of order \(k-1\) or less. Consider the ODE ([eq:ch5_ho_lin_aut_complex_all_sols_hom_ODE]) of order \(k\) with characteristic polynomial \(p(r).\) Write \[p(r)=\tilde{p}(r)(r-r_{k}),\tag{5.128}\] where the polynomial \[\tilde{p}(r)=\prod_{l=1}^{k-1}(r-r_{l})\] is of degree \(k-1.\) Denote the coefficients of \(\tilde{p}(r)\) by \(\tilde{a}_{l},\) i.e. \[\tilde{p}(r)=\sum_{l=0}^{k-1}\tilde{a}_{l}r^{l},\] (where \(\tilde{a}_{k}=1\)). In terms of coefficients the equality ([eq:p_r_p_tilde]) reads as \[\sum_{l=0}^{k}a_{l}r^{l}=\sum_{l=0}^{k-1}\tilde{a}_{l}r^{l}(r-r_{k})\quad\forall r.\tag{5.129}\] The r.h.s. can be rewritten as \[\sum_{l=0}^{k-1}\tilde{a}_{l}r^{l+1}-r_{k}\sum_{l=0}^{k-1}\tilde{a}_{l}r^{l}=\sum_{l=1}^{k}\tilde{a}_{l-1}r^{l}-r_{k}\sum_{l=0}^{k-1}\tilde{a}_{l}r^{l}=\tilde{a}_{k-1}r^{k}+\sum_{l=1}^{k-1}\left(\tilde{a}_{l-1}-r_{k}\tilde{a}_{l}\right)r^{l}-r_{k}\tilde{a}_{0}.\] Since ([eq:p_p_tilde_coeff]) holds for all \(r\) it implies that \[a_{k}=\tilde{a}_{k-1}=1,\quad\quad a_{l}=\tilde{a}_{l-1}-r_{k}\tilde{a}_{l},1\le l\le k-1,\quad\quad a_{0}=-r_{k}\tilde{a}_{0}.\tag{5.130}\] Assume now that \(x\) solves ([eq:ch5_ho_lin_aut_complex_all_sols_hom_ODE]). Then \[\sum_{l=0}^{k}a_{l}x^{(l)}=0.\] Using ([eq:tilde_as]) this can be written as \[\tilde{a}_{k-1}x^{(k)}+\sum_{l=1}^{k-1}\left(\tilde{a}_{l-1}-r_{k}\tilde{a}_{l}\right)x^{(l)}-r_{k}\tilde{a}_{0}x=0.\] We rewrite the l.h.s. as \[\begin{array}{l} \tilde{a}_{k-1}x^{(k)}+\sum_{l=1}^{k-1}\tilde{a}_{l-1}x^{(l)}-r_{k}\sum_{l=1}^{k-1}\tilde{a}_{l}x^{(l)}-r_{k}\tilde{a}_{0}x\\ =\tilde{a}_{k-1}x^{(k)}+\sum_{l=0}^{k-2}\tilde{a}_{l}x^{(l+1)}-r_{k}\sum_{l=1}^{k-1}\tilde{a}_{l}x^{(l)}-r_{k}\tilde{a}_{0}x\\ =\sum_{l=0}^{k-1}\tilde{a}_{l}x^{(l+1)}-r_{k}\sum_{l=0}^{k-1}\tilde{a}_{l}x^{(l)}\\ =\sum_{l=0}^{k-1}\tilde{a}_{l}\left(x^{(l+1)}-r_{k}x^{(l)}\right). \end{array}\] Thus \[\sum_{l=0}^{k-1}\tilde{a}_{l}\left(x^{(l+1)}-r_{k}x^{(l)}\right)=0.\] Multiplying this by \(e^{-rt_{k}}\) gives \[\sum_{l=0}^{k-1}\tilde{a}_{l}\left(e^{-rt_{k}}x^{(l+1)}-e^{-rt_{k}}r_{k}x{}^{(l)}\right)=0.\] The l.h.s. equals \[\sum_{l=0}^{k-1}\tilde{a}_{l}\frac{d}{dt}\left\{ e^{-rt_{k}}x^{(l)}(t)\right\} =\frac{d}{dt}\sum_{l=0}^{k-1}\tilde{a}_{l}e^{-rt_{k}}x^{(l)}(t).\] Therefore \(x\) being a solution implies that \[\frac{d}{dt}\sum_{l=0}^{k-1}\tilde{a}_{l}e^{-rt_{k}}x^{(l)}(t)=0,\quad\quad t\in I.\] This in turn implies that \[\sum_{l=0}^{k-1}\tilde{a}_{l}e^{-rt_{k}}x^{(l)}(t)=c,\quad\quad t\in I,\] or in other words \[\sum_{l=0}^{k-1}\tilde{a}_{l}y^{(l)}(t)=ce^{rt_{k}},\quad\quad t\in I.\tag{5.131}\] Letting \(m\) be the multiplicity of the root \(r_{k},\) so that \(x_{k}(t)=t^{m-1}e^{r_{k}t},\) it follows by Lemma 5.32 that \[\sum_{l=0}^{k-1}\tilde{a}_{l}x_{k}^{(l)}(t)=e^{rt_{k}}\sum_{i=0}^{m-1}{m-1 \choose i}t^{m-1-i}\tilde{p}^{(i)}(r).\] The number \(r_{k}\) is a root of \(\tilde{p}(r)\) of multiplicity \(m-1,\) so \(\tilde{p}^{(i)}(r_{k})=0\) for \(i<m-1,\) but \(\tilde{p}^{(m-1)}(r_{k})\ne0.\) Therefore \[\sum_{l=0}^{k-1}\tilde{a}_{l}x_{k}^{(l)}(t)=e^{rt_{k}}\tilde{p}^{(m-1)}(r),\] and so there is a \(\alpha_{k}\in\mathbb{C},\) such that \[\sum_{l=0}^{k-1}\tilde{a}_{l}\frac{d^{l}}{dt^{l}}\left\{ \alpha_{k}x_{k}(t)\right\} =ce^{rt_{k}}.\] Subtracting this from ([eq:basew]) gives \[\sum_{l=0}^{k-1}\tilde{a}_{l}\left\{ x^{(l)}(t)-\alpha_{k}x_{k}^{(l)}(t)\right\} =0,\quad\quad t\in I.\] Thus the function \(t\to x(t)-c'x_{k}(t)\) solves the ODE \[\sum_{l=0}^{k-1}\tilde{a}_{l}x^{(l)}=0,\quad\quad t\in I.\] By the induction assuming it therefore holds that \[x(t)-c'x_{k}(t)=\sum_{l=1}^{k-1}\alpha_{l}x_{l}(t),\quad\quad t\in I,\] for some \(\alpha_{1},\ldots.\alpha_{k-1}\in\mathbb{C}.\) Thus shows that \(x(t)\) has the form ([eq:ch5_ho_lin_aut_complex_all_sols_hom_ODE_sols]), so we have proved a). b) Assume for contradiction that \(D\) is degenerate. Then there exists \(\gamma=(\gamma_{0},\ldots,\gamma_{k-1}),\) not all zero, such that \[D\gamma=0.\] Define the \(k-1\) degree polynomial \[q(r)=\sum_{l=0}^{k-1}\gamma_{k}r^{l}.\] Then the \(l'\)-th entry of \(D\gamma\) equals \[\sum_{l=0}^{k-1}\gamma_{l}x_{l'}^{(l)}(t_{*}).\] By Lemma 5.32 the entries of \(D\gamma\) are thus \[e^{r_{l'}t}\sum_{i=0}^{n}{n \choose i}t^{n-i}p^{(i)}(r_{l'}),0\le n\le m_{l'}-1,1\le l'\le k'.\tag{5.132}\] If \(D\gamma=0\) then all of these are zero. But then considering the ([eq:asdasd]) for which \(n=0\) gives \[q(r)=0\text{ for all roots }r=r_{l'}.\] Next considering \(n=1\) we get that \[q'(r)=0\text{ for all repeated roots }r=r_{l'}.\] Continuing for \(n=2,\ldots\) we obtain that \[q(r_{l})=q^{(1)}(r_{l})=\ldots=q^{(m_{l}-1)}(r_{l})=0,\text{ for }1\le l\le k',\] which implies that \(r_{1},\ldots,r_{k'}\)are all roots of multiplicity \(m_{1},\ldots,m_{k'}.\) But \(m_{1}+\ldots+m_{k'}=k\) which is a contradiction, since \(q\) cannot have more than \(k-1\) roots.

c) This is a direct consequence of a), b).

\(\text{ }\) ODEs with real coefficients. Let us now specialize to the situation where the ODE has real coefficients. A degree \(k\) polynomial \(p(r)\) with real coefficients has \(k\) roots counted with multiplicity, but they may be complex. Complex roots must appear in complex conjugate pairs: \(r,r'\) s.t. \(r'=\overline{r}\) (where \(\overline{r}\) denotes complex conjugation, i.e. \(\overline{a+ib}=a-ib\) for \(a+ib\in\mathbb{C}\)): this is because if \(p(r)\) has real coefficients, then \(\overline{p(r)}=p(\overline{r}),\) so if \(r\) is a root then so is \(\overline{r}.\) Each non-real complex conjugate pair \(r=a+ib,\overline{r}=a-ib,\) can give rise to the functions \(\cos,\sin,\) as we saw in the second order case: if \(\alpha\in\mathbb{R},0\le l\le m-1,a,b\in\mathbb{R}\) and \(r=a+ib\) is root of multiplicity \(m,\) then \[\frac{\alpha}{2}t^{l}e^{rt}+\frac{\alpha}{2}t^{l}e^{\overline{r}}=\alpha t^{l}e^{at}\cos(bt)\quad\quad\frac{\alpha}{2i}t^{l}e^{rt}-\frac{\alpha}{2i}t^{l}e^{-rt}=\alpha t^{l}e^{at}\sin(bt),\] are solutions.

We finish by proving that if the ODE has real coefficients and \(x_{*}\) in Proposition 5.36 c) is real, then the unique solution is also real valued.
Proposition 5.37

(All solutions of higher order real linear autonomous ODEs) Let \(I\subset\mathbb{R}\) be a non-empty interval. Let \(k\ge1,\) \(a_{0},a_{1},\ldots,a_{k-1}\in\mathbb{R}.\) Consider the ODE \[x^{(k)}(t)+\sum_{l=0}^{k-1}a_{l}x^{(l)}(t)=0.\tag{5.134}\] Assume that the characteristic polynomial \(p(r)=r^{k}+\sum_{l=0}^{k-1}a_{l}r^{l}\) has \(k'\) distinct real roots \(r_{1},\ldots,r_{k'}\in\mathbb{R},\) each of multiplicity \(m_{1},m_{2},\ldots,m_{k'}\in\{1,2,\ldots\},\) and \(k''\) complex roots \(r_{k'+1},\ldots,r_{k'+k''}\in\mathbb{C}\) with positive imaginary part with multiplicities \(m_{k'+1},\ldots,m_{k'+k''}\in\{1,2,\ldots\},\) where \(m_{1}+\ldots+m_{k'}+2m_{k'+1}+\ldots+2m_{k'+k''}=k.\) Let \[x_{1}(t),\ldots,x_{k'}(t)\text{ denote the functions }e^{r_{1}t},te^{r_{1}t},\ldots,t^{m_{1}-1}e^{r_{1}t},e^{r_{2}t},te^{r_{2}t},\ldots,t^{m_{k'}-1}e^{r_{k'}t},\] and \[\begin{array}{c} x_{k'+1}(t),\ldots,x_{k+k'}(t)\text{ denote the functions }\\ e^{\mathrm{Re}r_{1}t}\cos(\mathrm{Im}r_{1}t),e^{\mathrm{Re}r_{1}t}\sin(\mathrm{Im}r_{1}t),te^{\mathrm{Re}r_{1}t}\cos(\mathrm{Im}r_{1}t),\ldots,t^{m_{k'+l''}-1}e^{\mathrm{Re}r_{k'+k''}t}\sin(\mathrm{Im}r_{k'+k''}t). \end{array}\]

a) A function \(x\) solves the ODE ([eq:ch5_ho_lin_aut_real_all_sols_hom_ODE]) iff \[x(t)=\sum_{i=1}^{k}\alpha_{i}x_{i}(t),\tag{5.135}\] for some \(\alpha\in\mathbb{R}^{k}.\)

b) For any \(t_{*}\in I,\) the \(k\times k\) real matrix \(D\) with entries \[D_{i,l}=x_{i}^{(l)}(t_{*}),\] is non-degenerate.

c) Let \(t_{*}\in I\) and \(x_{*}\in\mathbb{R}^{k}.\) A function \(x\) solves the constrained homogeneous ODE \[x^{(k)}(t)+\sum_{l=0}^{k-1}a_{l}x^{(l)}(t)=0,\quad\quad t\in I,\quad\quad x^{(l)}(t_{*})=x_{*,l+1},0\le l\le k-1,\tag{5.136}\] iff \[x(t)=\sum_{i=1}^{k}\alpha_{i}x_{i}(t),\tag{5.137}\] where \(\alpha\in\mathbb{R}^{k}\) is the unique solution to \[D\alpha=y_{*},\quad\quad\text{i.e. }\quad\quad\alpha=D^{-1}x_{*}.\]

d) If \(b\in\mathbb{R}\) and \(a_{0}\ne0\) a function \(y\) solves the non-homogeneous autonomous ODE \[y^{(k)}(t)+\sum_{l=0}^{k-1}a_{l}y^{(l)}(t)=b,\quad\quad t\in I,\tag{5.138}\] iff \[y(t)=x(t)+\frac{b}{a_{0}},\quad\quad t\in I,\] for an \(x(t)\) as in part a). If \(t_{*}\in I,y_{*}\in\mathbb{R}^{k}\) then a function \(y\) solves the constrained non-homogeneous second order autonomous ODE \[y^{(k)}(t)+\sum_{l=0}^{k-1}a_{l}y^{(l)}(t)=b,\quad\quad t\in I,y^{(l)}(t_{*})=y_{*,l+1},0\le l\le k-1,\] iff \[y(t)=x(t)+\frac{b}{a_{0}},\quad\quad t\in I,\] where \(x(t)\) is the unique function from part c) with \[\alpha=D^{-1}z_{*}\quad\quad\text{ for }\quad\quad z_{*}=y_{*}-\left(\begin{matrix}\frac{b}{a_{0}}\\ 0\\ \ldots\\ 0 \end{matrix}\right).\]

Proof. It suffices to prove b) and that ([eq:blabla]) solves ([eq:ch5_ho_lin_aut_complex_all_sols_hom_IVP-1]). The uniqueness of Proposition 5.36 in the more general complex case then implies the rest of a) b) c) d). Assume that b) does not hold. Then there are \(\gamma=(\gamma_{0},\ldots,\gamma_{k-1})\in\mathbb{R}^{k},\) not all zero, such that \[D\gamma=0.\] The polynomial \[q(r)=\sum_{l=0}^{k-1}\gamma_{l}r^{l}\] has degree \(k-1\) and at most \(k-1\) roots. By Lemma 5.32 the entries of \(D^{T}\gamma\) corresponding to a real root \(r\) equals \[e^{rt}\sum_{i=0}^{n}{n \choose i}t^{n-i}q^{(i)}(r),0\le n\le m_{l'}-1,1\le l'\le k'.\tag{5.139}\] Thus \(D\gamma=0\) implies that \(q^{(n)}(r)=0\) for \(0\le n\le m-1\) for all the real roots \(r,\) where \(m\) is the multiplicity of \(r.\) The \(2m\) of of \(D\gamma\) corresponding to a complex conjugate root pair \(r,\bar{r}\) pair of multiplicity \(m\) can be written as \[\sum_{l=0}^{k-1}\gamma_{l}\frac{d}{dt^{l}}\left\{ t^{n}\frac{e^{rt}+e^{\bar{r}t}}{2}\right\} \quad\text{ and }\quad\sum_{l}^{k-1}\gamma_{l}\frac{d}{dt^{l}}\left\{ t^{n}\frac{e^{rt}-e^{\bar{r}t}}{2i}\right\} ,\tag{5.140}\] for \(0\le n\le m-1.\) By Lemma 5.32 the left member equals \[\frac{1}{2}\sum_{l=0}^{k-1}\gamma_{l}\frac{d}{dt^{l}}\left\{ t^{n}e^{rt}\right\} +\frac{1}{2}\sum_{l=0}^{k-1}\gamma_{l}\frac{d}{dt^{l}}\left\{ t^{n}e^{\bar{r}t}\right\} =\frac{1}{2}e^{rt}\sum_{i=0}^{n}{n \choose i}t^{n-i}q^{(i)}(r)+\frac{1}{2}e^{\bar{r}t}\sum_{i=0}^{n}{n \choose i}t^{n-i}q^{(i)}(\bar{r}).\] But since \(q\) has real coefficients \[q^{(i)}(\bar{r})=\overline{q^{(i)}(r)}\] so the r.h.s. equals \[\frac{1}{2}\mathrm{Re}\left\{ e^{rt}\sum_{i=0}^{n}{n \choose i}t^{n-i}q^{(i)}(r)\right\} =\frac{1}{2}\mathrm{Re}\left\{ e^{\bar{r}t}\sum_{i=0}^{n}{n \choose i}t^{n-i}q^{(i)}(\bar{r})\right\}\] Similarly the second member of ([eq:two_sols]) equals \[\frac{1}{2}\mathrm{Im}\left\{ e^{rt}\sum_{i=0}^{n}{n \choose i}t^{n-i}q^{(i)}(r)\right\} =\frac{1}{2}\mathrm{Im}\left\{ e^{\bar{r}t}\sum_{i=0}^{n}{n \choose i}t^{n-i}q^{(i)}(\bar{r})\right\} .\] Thus \(D\gamma=0\) implies that \[e^{rt}\sum_{i=0}^{n}{n \choose i}t^{n-i}q^{(i)}(r)=0\] for all \(0\le n\le m-1\) for each complex root of \(r\) of multiplicity \(m.\) But then \(q(r)\) has \(k\) roots counting multiplicity, which is a contradiction.

Home

Contents

Weeks