Namely, when does the polynomial $z^2 + a z + b = 0$ with $a,b \in \mathbb{C}$ have both roots in the left half complex?
As the sum of the roots $z_1 + z_2 = -a$, it follows that if $\Re a < 0$ then one of $\Re z_1$ or $\Re z_2$ must be positive. So a necessary condition is $\Re a > 0$. What is a sufficient condition? (Please provide a proof for it)
Edit: We can modify the answer below for the "more general case". A necessary and sufficient condition for the roots of $$ A z^2 + B z + C = 0$$ with $A>0$, $B,C \in \mathbb{C}$ to lie on the left complex half plane, i.e. $\Re z_i < 0$, is to have $$\sqrt{2} \Re(B) > \sqrt{\Re(B^2 - 4AC) + \vert B^2 - 4AC \vert}.$$ If we let $B = B_1 + i B_2$, $C = C_1 + i C_2$, with $B_i, C_i \in \mathbb{R}$ then the above condition can also be written as $$ B_1 > 0, \qquad B_1^2 C_1 + B_1 B_2 C_2 - A C_2^2 > 0. $$
If
$\mu_1, \mu_2 \in \Bbb C \tag 1$
are the roots of
$z^2 + az + b = 0, \; a, b \in \Bbb C, \tag 2$
then they satisfy the quadratic formula,
$\mu_1, \mu_2 = \dfrac{-a \pm \sqrt{a^2 - 4b}}{2}, \tag 3$
provided we take some care in the interpretation of $\sqrt{\cdot}$; since in general $a^2 - 4b \in \Bbb C \setminus \Bbb R$, we must carefully specify just what we intend
$\theta = \arg{\sqrt{a^2 - 4b}} \in [0, 2\pi) \tag 4$
to be; I think the most generally accepted convention, assuming $a^2 - 4b \ne 0$, is to write
$a^2 - 4b = \vert a^2 - 4b \vert e^{i \arg(a^2 - 4b)}, \tag 5$
and then choose (or, define)
$\sqrt{a^2 - 4b} = \sqrt{\vert a^2 - 4b \vert} e^{i \arg(a^4 - 4b) / 2}; \tag 6$
we check:
$(\sqrt{\vert a^2 - 4b \vert} e^{i \arg(a^2 - 4b) / 2})^2 = (\sqrt{ \vert a^2 - 4b \vert })^2 (e^{i \arg(a^4 - 4b) / 2})^2$ $= \vert a^2 - 4b \vert e^{2i\arg((a^2 - 4b) / 2)} = \vert a^2 - 4b \vert e^{i\arg(a^2 - 4b)} = a^2 - 4b; \tag 7$
of course with (6) in place, we also see that
$(-\sqrt{a^2 - 4b})^2 = a^2 - 4b \tag 8$
as well; and of course, (3) is a necessary for $\mu_1$, $\mu_2$ to satisfy (2), since it follows from completing the square: from (2),
$z^2 + az = -b, \tag 9$
$\left (z + \dfrac{a}{2} \right )^2 = z^2 + az + \dfrac{a^2}{4} = \dfrac{a^2}{4} - b = \dfrac{1}{4}(a^2 - 4b), \tag{10}$
whence
$z + \dfrac{a}{2} = \pm \dfrac{1}{2}\sqrt{a^2 - 4b} \iff \mu_1, \mu_2 = z = \dfrac{-a \pm \sqrt{a^2 - 4b}}{2}. \tag{11}$
It should be noted here that
$a^2 - 4b = 0 \Longrightarrow \mu_1 = \mu_2 = -\dfrac{1}{2}a, \tag{11.5}$
that is, we obtain a double real root. Bearing this in mind as having resolved the case $a^2 - 4b = 0$, we proceed under our hypothesis $a^2 - 4b \ne 0$.
With (6) in place we may write a precise expression for the real parts of the roots of (2) via (3): we have
$a = \Re(a) + i \Im(a), \tag{12}$
and
$\sqrt{a^2 - 4b} = \sqrt{\vert a^2 - 4b \vert} e^{i \arg(a^2 - 4b) / 2}$ $= \sqrt{\vert a^2 - 4b \vert} \cos \dfrac{ \arg (a^2 - 4b)}{2} + i \sqrt{\vert a^2 - 4b \vert} \sin \dfrac{ \arg (a^2 - 4b)}{2}; \tag{13}$
it now follows from (3), (11) that the real parts of $\mu_1$ and $\mu_2$ are
$\Re(\mu_1), \; \Re(\mu_2) = \dfrac{1}{2}\left (-\Re(a) \pm \sqrt{\vert a^2 - 4b \vert} \cos \dfrac{ \arg (a^2 - 4b)}{2} \right ); \tag{14}$
we can in principle, from (14), determine the relationship(s) $a$ and $b$ must satisfy in order that
$\Re(\mu_1), \; \Re(\mu_2) < 0; \tag{15}$
indeed,
$\Re(\mu_1), \Re(\mu_2) = \dfrac{1}{2}\left (-\Re(a) \pm \sqrt{\vert a^2 - 4b \vert} \cos \dfrac{ \arg (a^2 - 4b)}{2} \right ) < 0$ $\iff -\Re(a) + \left \vert \sqrt{\vert a^2 - 4b \vert} \cos \dfrac{ \arg (a^2 - 4b)}{2} \right \vert < 0, \tag{16}$
which logical equivalence ($\Longleftrightarrow$) binds since
$\pm \sqrt{\vert a^2 - 4b \vert} \cos \dfrac{ \arg (a^2 - 4b)}{2} = \left \vert \sqrt{\vert a^2 - 4b \vert} \cos \dfrac{ \arg (a^2 - 4b)}{2} \right \vert \tag{17}$
for at least one choice of $+/-$ in the "$\pm$" sign.
(16) is in fact completely determinative of (15), since it affirms that the largest of the two $\Re(\mu_1)$, $\Re(\mu_2)$ is negative.
A few words on the special case with both $a, b \in \Bbb R$, that is,
$a = \Re(a), b = \Re(b) \in \Bbb R, \tag{18}$
we have
$\arg(a^2 - 4b) = 0, \pi \Longleftrightarrow \dfrac{ \arg (a^2 - 4b)}{2} = 0, \dfrac{\pi}{2} \Longrightarrow \cos \dfrac{ \arg (a^2 - 4b)}{2} = 1, 0; \tag{19}$
with $\cos (\arg(a^2 - 4b) / 2) = 1$ we find that
$ \left \vert \sqrt{\vert a^2 - 4b \vert} \cos \arg \dfrac{(a^2 - 4b)}{2} \right \vert = \sqrt{\vert a^2 - 4b} \vert, \tag{20}$
and so, via (14), (18)
$\mu_1 = \Re(\mu_1), \; \mu_2 = \Re(\mu_2) = \dfrac{1}{2}\left (-a \pm \sqrt{\vert a^2 - 4b \vert} \right ); \tag{21}$
in such a situation we thus see that both roots $\mu_1$ and $\mu_2$ have negative real part when
$a = \vert a \vert > \sqrt{\vert a^2 - 4b \vert}, \tag{22}$
that is, when
$0 < a, \; 0 < b < \dfrac{a^2}{2}; \tag{23}$
on the other hand, with $\cos (\arg(a^2 - 4b) / 2) = 0$, it is even easier to realize that we must have, again with the aid of (14),
$\Re(\mu_1) = \Re(\mu_2) = -\dfrac{1}{2} \Re(a), \tag{24}$
and $\mu_1, \mu_2 = \bar \mu_1$ are a complex conjugate pair; this is consistent with (13), which in the present instance reverts to
$\sqrt{a^2 - 4b} = i \sqrt{\vert a^2 - 4b \vert}, \tag{24}$
and then (11) yields
$\mu_1, \mu_2 = \dfrac{-a \pm i\sqrt{\vert a^2 - 4b \vert}}{2}. \tag{25}$