A p-orbital, defined by the quantum number \(l=1\), is a dumbbell-shaped region of space around the nucleus where an electron is most likely to be found. The letter ”p” is of spectroscopic origin, standing for ‘principal’.

One-electron p-orbital wavefunctions can be expressed using the total wavefunction of a hydrogenic atom:
$$\psi=R_{n,l}(r)Y_l^{\vert m_l\vert}(\theta,\phi)$$
where
\(R_{n,l}(r)=N_{n,l}\frac{1}{r}\biggr(\frac{r}{na_0}\biggr)^{l+1}e^{-\frac{r}{na_0}}L_{n-l-1}^{2l+1}\biggr(\frac{2r}{na_0}\biggr)\), the radial wavefunction, is the radial component of \(\psi\).
\(Y_l^{\vert m_l\vert}(\theta,\phi)=N_l^{\vert m_l\vert}P_l^{\vert m_l\vert}(cos\theta)e^{im_l\phi}\), the spherical harmonics, is the angular component of \(\psi\).
\(L_{n-l-1}^{2l+1}=\sum_{k=0}^{n-l-1}(-1)^k\frac{(n+l)!}{k!(k+2l+1)!(n-l-1-k)!}\biggr(\frac{2r}{na_0}\biggr)^k\) are the associated Laguerre polynomials.
\(P_l^{\vert m_l\vert}(cos\theta)=\frac{1}{2^ll!}(1-cos^2\theta)^{\frac{\vert m_l\vert}{2}}\frac{d^{l+\vert m_l\vert}}{dcos\theta^{l+\vert m_l\vert}}(cos^2\theta-1)^l\)are the associated Legendre polynomials.
\(N_{n,l}=\sqrt{\frac{(n-l-1)!}{2n(n+l)!}\frac{2^{2l+3}}{na_0}}\) is the normalisation constant of the radial wavefunction.
\(N_l^{\vert m_l\vert}=\frac{1}{\sqrt{2\pi}}\biggr[\frac{2l+1}{2}\frac{(l-\vert m_l\vert)!}{(l+\vert m_l\vert)!}\biggr]^{\frac{1}{2}}\) is the normalisation constant of the spherical harmonics.
\(n\) is the principal quantum number, where \(n=1,2,\cdots\)
\(l\) is the orbital angular momentum quantum number (also known as azimuthal quantum number), where \(l=0,1,2,\cdots\) and \(l\leq n-1\).
\(m_l\) is the magnetic quantum number, where \(-l\leq m_l\leq l\).
The principal quantum number, \(n\), is also called a shell. Since \(-1\leq m_l\leq 1\) when \(l=1\), there are three p-orbitals that are characterised by the set of quantum numbers \(\{n,l,m_l\}\) of \(\{n,1,0\}\), \(\{n,1,1\}\) and \(\{n,1,-1\}\) in each shell for \(n\geq 2\). Each of the three p-orbitals has an angular node. Consider the set \(\{2,1,0\}\). Substituting these values into the explicit formula for \(R_{n,l}(r)Y_l^{\vert m_l\vert}(\theta,\phi)\) yields:
$$R_{2,1}(r)Y_1^{0}(\theta,\phi)=rcos\theta f(r)$$
where \(f(r)=\frac{1}{4\sqrt{2\pi}}\sqrt{\biggr(\frac{1}{a_0}\biggr)^5}e^{-\frac{r}{2a_0}}\).
Converting from spherical coordinates to Cartesian coordinates, where \(z=rcos\theta\), gives
$$\psi_{2p_z}=zf(r)\;\;\;\;\;\;\;\;474$$
Eq474 is known as the \(2p_z\) orbital. When \(z=0\), the wavefunction is zero everywhere in the \(xy\)-plane, which is known as the nodal plane.
For the sets \(\{2,1,1\}\) and \(\{2,1,-1\}\), we have \(\small{\psi_{p_1}=\frac{1}{8\sqrt{\pi}}\sqrt{\biggr(\frac{1}{a_0}\biggr)^5}re^{-\frac{r}{2a_0}}(1-cos^2\theta)^{\frac{1}{2}}e^{i\phi}}\) and \(\small{\psi_{p_{-1}}=\frac{1}{8\sqrt{\pi}}\sqrt{\biggr(\frac{1}{a_0}\biggr)^5}re^{-\frac{r}{2a_0}}(1-cos^2\theta)^{\frac{1}{2}}e^{-i\phi}}\), respectively. These two wavefunctions include the factor \(e^{\pm i\phi}\), which may complicate calculations when they undergo symmetry operations. Therefore, we take linear combinations of these wavefunctions to form simpler wavefunctions. The first linear combination is \(\small{\psi_{p_1}+\psi_{p_{-1}}=rsin\theta\cos\phi\frac{1}{4\sqrt{\pi}}\sqrt{\biggr(\frac{1}{a_0}\biggr)^5}e^{-\frac{r}{2a_0}}}\), which normalises to \(\small{rsin\theta\cos\phi\frac{1}{4\sqrt{2\pi}}\sqrt{\biggr(\frac{1}{a_0}\biggr)^5}e^{-\frac{r}{2a_0}}}\), or equivalently in Cartesian coordinates:
$$\psi_{2p_x}=xf(r)\;\;\;\;\;\;\;\;475$$
since \(x=rsin\theta cos\phi\).
The second normalised linear combination is \(\small{\psi_{p_1}-\psi_{p_{-1}}=rsin\theta\sin\phi\frac{1}{4\sqrt{2\pi}}\sqrt{\biggr(\frac{1}{a_0}\biggr)^5}e^{-\frac{r}{2a_0}}}\), or equivalently,
$$\psi_{2p_y}=yf(r)\;\;\;\;\;\;\;\;476$$
since \(y=rsin\theta sin\phi\).
\(\psi_{2p_x}\) and \(\psi_{2p_y}\) , like \(\psi_{2p_z}\) , each have an angular node.

Question
If \(\small{\psi_{p_1}}\) and \(\small{\psi_{p_{-1}}}\) are already normalised, why do we need to normalise a linear combination of them? How do we normalise a linear combination of \(\small{\psi_{p_1}}\) and \(\small{\psi_{p_{-1}}}\)?
Answer
When forming a linear combination of normalised wavefunctions, the result is not necessarily normalised. Consider a general linear combination \(\varphi=a\phi_1+b\phi_2\), where \(a\) and \(b\) are coefficients. The normalisation condition for \(\varphi\) requires that \(\int\vert\varphi\vert^2d\tau=1\), where \(d\tau\) represents the volume element in spherical coordinates. Expanding \(\int\vert a\phi_1+b\phi_2\vert^2d\tau=1\) and noting that \(\phi_1\) and \(\phi_2\) are orthonormal, we have \(\vert a\vert^2+\vert b\vert^2=1\). If \(\vert a\vert^2+\vert b\vert^2\neq 1\), which is the case for the two linear combinations of \(\small{\psi_{p_1}}\) and \(\small{\psi_{p_{-1}}}\), \(\varphi\) will not be normalised.
To normalise \(\small{\psi_{p_1}-\psi_{p_{-1}}}\), we have \(\small{N^2\int\vert\psi_{p_1}-\psi_{p_{-1}}\vert^2d\tau=1}\) or
$$-N^2\frac{1}{16\pi}\biggr(\frac{1}{a_0}\biggr)^5\int_0^{\infty}\int_0^{\pi}\int_0^{2\pi}r^2sin^2\theta sin^2\phi e^{-\frac{r}{a_0}}r^2drsin\theta d\theta d\phi=1$$
Using \(\small{\int_0^{\infty}r^4e^{-\frac{r}{a_0}}dr=24a_0^5}\) (see this article for proof) and some basic trigonometry identities yields
$$-\frac{3N^2}{2\pi}\int_0^{\pi}(1-cos^2\theta)sin\theta d\theta\int_0^{2\pi}\biggr(\frac{1}{2}-\frac{1}{2}cos 2\phi\biggr) d\phi=1$$
Setting \(u=cos\theta\) and \(du=-sin\theta d\theta\) gives \(N=\frac{1}{i\sqrt{2}}\).