Hilbert spaces are friendly to intuition because they look like Euclidean space with infinitely many coordinates. The danger is that Euclidean intuition keeps working just long enough to create confidence, and then fails silently in exactly the places where the subject becomes powerful.
Below are common mistakes that appear in homework, seminar talks, and research reading. Each one comes with a correction pattern you can reuse.
Smart TV Pick55-inch 4K Fire TVINSIGNIA 55-inch Class F50 Series LED 4K UHD Smart Fire TV
INSIGNIA 55-inch Class F50 Series LED 4K UHD Smart Fire TV
A general-audience television pick for entertainment pages, living-room guides, streaming roundups, and practical smart-TV recommendations.
- 55-inch 4K UHD display
- HDR10 support
- Built-in Fire TV platform
- Alexa voice remote
- HDMI eARC and DTS Virtual:X support
Why it stands out
- General-audience television recommendation
- Easy fit for streaming and living-room pages
- Combines 4K TV and smart platform in one pick
Things to know
- TV pricing and stock can change often
- Platform preferences vary by buyer
Treating norm convergence and weak convergence as interchangeable
The mistake: assuming that if $x_n\rightharpoonup x$ weakly, then $\|x_n-x\|\to 0$.
What’s true: strong convergence implies weak convergence, but not conversely. The standard basis $(e_n)$ in $\ell^2$ converges weakly \to $0$ because $\langle e_n, y\rangle\to 0$ for every $y\in\ell^2$, yet $\|e_n\|=1$ for all $n$.
How to avoid it:
- When you extract subsequences using boundedness, you typically get weak convergence, not strong.
- To upgrade weak to strong, you need extra structure: compactness, monotonicity, strict convexity arguments, or norm convergence in addition to weak convergence.
A reliable upgrade lemma in Hilbert spaces is: if $x_n\rightharpoonup x$ and $\|x_n\|\to\|x\|$, then $x_n\to x$ strongly. The proof is a one-line polarization computation:
and weak convergence handles the inner product term.
Assuming every bounded operator has an eigenvector
The mistake: treating “spectrum” as if it were “eigenvalues.”
What’s true: bounded operators may have no eigenvectors. Even many normal operators have purely continuous spectrum in common models. The right invariant is the spectrum $\sigma(T)$, defined by non-invertibility of $T-\lambda I$, not by existence of eigenvectors.
How to avoid it:
- If the problem asks about eigenvectors, check the operator class. Compact self-adjoint operators are the friendly case where eigenvectors do form an orthonormal basis.
- If compactness is absent, learn to speak the language of resolvents, spectral measures, and functional calculus for normal operators.
A practical reading trick is to note which statements are phrased in terms of $T-\lambda I$ being invertible. Those are spectral statements, not eigenvector statements.
Forgetting that closedness is the difference between “subspace” and “Hilbert subspace”
The mistake: using orthogonal projection onto an arbitrary linear subspace $M\subset H$.
What’s true: the projection theorem requires $M$ \to be closed. If $M$ is not closed, a best approximation may not exist in $M$, even if $M$ is dense.
How to avoid it:
- Whenever projection is mentioned, ask immediately: is the subspace closed?
- When given a dense subspace, treat it as a “core” for computations, not as a target for minimization.
A clean rule is: closed subspaces behave like coordinate planes. Dense non-closed subspaces behave like “almost everything” but without a stable nearest point.
Confusing “orthonormal” with “orthogonal” with “independent”
The mistake: assuming any orthogonal set acts like a basis, or assuming linear independence is enough for expansions.
What’s true: orthonormal sets control coefficients through Bessel’s inequality:
That inequality is the engine behind convergence of Fourier series and stability of expansions. Orthogonality without normalization loses immediate coefficient control. Linear independence alone gives essentially no quantitative control on coefficients.
How to avoid it:
- Normalize orthogonal families whenever possible.
- When working with a non-orthonormal basis, expect condition numbers and instability to appear.
- Use Gram–Schmidt to replace a finite independent set with an orthonormal one; this is not just aesthetic, it changes what estimates you can prove.
Dropping conjugates in complex inner products
The mistake: treating the inner product as bilinear on complex Hilbert spaces.
What’s true: complex inner products are linear in one variable and conjugate-linear in the other (depending on convention). Dropping conjugates changes adjoints, normality, and positivity in subtle ways.
How to avoid it:
- When you compute $\langle ax, y\rangle$ or $\langle x, ay\rangle$, pause and apply the convention consciously.
- When defining an adjoint, always verify the identity $\langle Tx, y\rangle = \langle x, T^*y\rangle$ on a dense set first.
A quick sanity check is positivity: $\langle x,x\rangle$ must be real and nonnegative. If your formula can produce a complex value on $x=x$, something is wrong.
Treating pointwise evaluation as a continuous functional in $L^2$
The mistake: writing things like “let $f(0)$” for $f\in L^2([0,1])$ as if it were well-defined and continuous.
What’s true: an $L^2$ function is an equivalence class, and values at a point are not meaningful invariants. Even for nice representatives, pointwise evaluation is not continuous with respect to the $L^2$ norm.
How to avoid it:
- Use integrals and inner products as your primary observables in $L^2$.
- If you need pointwise information, move \to a Sobolev space where evaluation is continuous under the right regularity assumptions, or work with continuous representatives where the norm controls pointwise behavior.
This is a common place where geometric intuition must be disciplined: “small in $L^2$” means small on average, not uniformly small.
Assuming “dense” implies “equal” in operator identities
The mistake: proving an identity $Tx=Sx$ on a dense set and concluding $T=S$ without checking continuity.
What’s true: if $T$ and $S$ are bounded operators and $Tx=Sx$ holds on a dense set, then $T=S$. The boundedness is what lets you pass to limits. For unbounded operators, domains matter, and density is not enough.
How to avoid it:
- When using dense-set arguments, explicitly note where boundedness enters.
- If an operator is unbounded, treat its domain as part of its definition. You cannot ignore it.
A good habit is to phrase dense-set proofs as: “prove on finite linear combinations, then extend by continuity.” That phrase is a built-in check that you have a continuity mechanism.
Mixing up self-adjoint, symmetric, and normal
The mistake: assuming that “symmetric” automatically means “self-adjoint,” or that “$TT^*=T^*T$” is always easy to check.
What’s true: for bounded operators on Hilbert spaces, “self-adjoint” is the same as $T=T^*$. For unbounded operators, “symmetric” means $\langle Tx,y\rangle=\langle x,Ty\rangle$ on the domain, and that is weaker than self-adjointness. Normality $TT^*=T^*T$ implies spectral structure but is delicate for unbounded operators.
How to avoid it:
- In bounded settings, compute adjoints honestly and check equality.
- In unbounded settings, check domains and closures. Many pathologies live entirely in domain mismatches.
If your argument uses spectral theorem conclusions, you must be in a setting where the theorem actually applies: typically bounded normal operators, or self-adjoint operators with proper domain theory.
Forgetting the difference between “finite rank,” “compact,” and “bounded”
The mistake: treating compactness as if it were boundedness, or treating finite-rank approximations as automatic.
What’s true: every compact operator is bounded, but many bounded operators are not compact. Compactness is about sending bounded sets to relatively compact sets, or equivalently about having sequences $Tx_n$ with convergent subsequences whenever $(x_n)$ is bounded.
How to avoid it:
- In $\ell^2$, test compactness on the standard basis. If $Te_n$ has no convergent subsequence, $T$ is not compact.
- For diagonal operators, remember the criterion: $D(x_n)=(d_n x_n)$ is compact exactly when $d_n\to 0$.
Compactness is the hypothesis that makes “weak information becomes strong information” possible in many arguments.
Treating orthogonal complements as if they always split the space
The mistake: writing $H = M \oplus M^\perp$ for an arbitrary subspace $M$.
What’s true: you always have $M^\perp$, but the direct sum decomposition $H = \overline{M} \oplus M^\perp$ uses the closure of $M$. If $M$ is not closed, then $M\oplus M^\perp$ is not all of $H$.
How to avoid it:
- Replace $M$ by $\overline{M}$ when forming decompositions.
- If you need a projection onto $M$, you need $M$ closed.
This mistake is a cousin of the projection mistake: both fail for the same reason, and both are fixed by checking closedness.
Interchanging limits, inner products, and operators without justification
The mistake: writing $\langle \lim x_n, y\rangle = \lim \langle x_n, y\rangle$ or $T(\lim x_n)=\lim Tx_n$ without checking which kind of convergence is present and whether the maps involved are continuous for that convergence.
What’s true: inner products are continuous in the norm topology, so strong convergence $x_n\to x$ gives $\langle x_n,y\rangle\to\langle x,y\rangle$. Weak convergence already means $\langle x_n,y\rangle\to\langle x,y\rangle$ for fixed $y$, but it does not control $\langle x_n, y_n\rangle$ when both slots move. Bounded operators are continuous for norm convergence and also preserve weak convergence, but limits can fail if the operator is unbounded or if you silently switch topologies mid-argument.
How to avoid it:
- If both arguments in an inner product vary with $n$, look for a bound like $\|x_n\|\le C$ and a convergence statement that is strong in at least one slot.
- If an operator is applied \to a limit, confirm the operator is bounded, or explicitly restrict \to a domain where the required continuity is valid.
- If you need convergence of $\langle x_n, y_n\rangle$, try to rewrite it using adjoints, projections, or orthogonality so that one slot becomes fixed.
This is a subtle place where Hilbert space geometry helps: orthogonality and Pythagoras often let you trade a moving inner product for a norm identity you can actually control.
A short prevention checklist
When reading or writing an argument in Hilbert spaces, check these items before trusting the conclusion.
- Identify which topology you are using: norm, weak, or weak-*.
- Check closedness whenever projection or decomposition is invoked.
- Confirm boundedness whenever you extend identities from dense sets.
- For complex spaces, track conjugates and adjoint conventions.
- If eigenvectors are mentioned, check whether compactness or a normality hypothesis is present.
- For unbounded operators, treat the domain as part of the object.
Hilbert spaces make many deep arguments possible precisely because they combine linear algebra, geometry, and completeness. Most mistakes happen when one of those features is implicitly assumed without being verified. Once you develop the habit of naming the feature you are using, the subject becomes both cleaner and more reliable.
Books by Drew Higgins
Bible Study / Spiritual Warfare
Ephesians 6 Field Guide: Spiritual Warfare and the Full Armor of God
Spiritual warfare is real—but it was never meant to turn your life into panic, obsession, or…
Prophecy and Its Meaning for Today
New Testament Prophecies and Their Meaning for Today
A focused study of New Testament prophecy and why it still matters for believers now.
Leave a Reply