Which lamp is failing in your UV-Vis spectrophotometer

A UV-Vis spectrophotometer has two lamps in it. An HPLC detector usually has one.

That single difference makes a spectrophotometer fault easier to diagnose than most people expect. It is also why ordering the wrong lamp is such a common way to waste a few hundred dollars.

The wavelength where the problem shows up tells you which lamp to buy.

The two sources, and where they change over

Typical published coverage for the two sources. The exact changeover point varies by instrument and on some models it is a setting rather than a fixed value, so check yours before using it to diagnose anything.

The deuterium lamp covers the ultraviolet. The tungsten halogen lamp covers the visible range and beyond. Between roughly 340 and 370 nm they overlap, and the instrument changes from one to the other.

That changeover point is the diagnostic. Everything below it is the deuterium lamp’s work. Everything above it is the tungsten lamp’s.

Read your baseline before you read the catalogue

Run a baseline or a blank scan across the full range your instrument covers, and look at where it misbehaves.

What the scan shows	Which lamp
Noise or drift only at the short-wavelength end, worst below about 250 nm	Deuterium
Absorbance readings low or unstable in the UV, visible region fine	Deuterium
Noise or instability only above about 400 nm	Tungsten halogen
A step or discontinuity right at the changeover, around 340 to 370 nm	Usually one lamp weakening, so the two no longer match at the joint
Noise across the whole range, UV and visible alike	Probably neither. Look at the sample compartment, the cuvette, or a dirty optical surface

The last row is the one that saves money. A lamp fault is wavelength-selective, because each lamp only serves part of the range. A problem that appears everywhere is usually not a lamp at all. The same reasoning applies on an HPLC detector, and before you replace the deuterium lamp lists the things that imitate a dying lamp there.

flowchart TD
    A["Run a baseline scan<br/>across the full range"] --> B{"Where is the<br/>problem?"}
    B -->|"Only below<br/>about 340 nm"| C["Deuterium lamp"]
    B -->|"Only above<br/>about 370 nm"| D["Tungsten halogen lamp"]
    B -->|"Right at the<br/>changeover"| E["One lamp is weakening.<br/>Check hours on both"]
    B -->|"Everywhere,<br/>UV and visible"| F["Not a lamp.<br/>Check cuvette, sample<br/>compartment, optics"]
    C --> G{"Lamp hours<br/>near 2,000?"}
    D --> G
    E --> G
    G -->|"Yes"| H["Replace it"]
    G -->|"No, well under"| I["Run the energy or<br/>intensity test before<br/>you order anything"]

Two free checks first

Read the lamp hours. Every UV-Vis keeps a counter for each lamp separately. A deuterium lamp is generally rated around 2,000 hours. Tungsten halogen lamps are quoted at around 2,000 hours too. In practice they are often replaced every 6 to 12 months, depending on how much the instrument is used. A lamp at 300 hours is rarely the problem.

Turn one lamp off. Many instruments let you switch the deuterium and tungsten sources independently, or have a shutter for each. If you can turn off the lamp you suspect and the fault stays in the range served by the other one, you have learned something for free.

Worth knowing: the hour counter runs whenever the lamp is lit, not whenever you measure something. An instrument left switched on all week uses up lamp life far faster than the sample count suggests. If your deuterium lamp keeps dying early, check whether anyone is turning it off.

What tends to fail, and how

A deuterium lamp fades rather than dies. It loses output slowly, and it loses it at the short-wavelength end first. So the noise appears at 220 nm long before anything looks wrong at 500 nm. At the very end it fails to ignite and the instrument reports a lamp error.

A tungsten halogen lamp behaves more like a bulb. It runs at close to full output and then fails, often suddenly. Gradual decline happens, but the abrupt failure is more common.

So a slow rise in UV noise over months points to the deuterium lamp. A visible-range problem that appeared overnight points to the tungsten lamp.

Under GMP there is more to do after the lamp change

This is the part that gets missed, and it costs more than the lamp.

Replacing either lamp can change the light path and the intensity reaching the detector. So a lamp change is a component replacement, and the instrument has to be shown to still perform before it goes back into use.

USP General Chapter <857>, and Ph. Eur. 2.2.25 in Europe, set out what to verify: wavelength accuracy, photometric accuracy, resolution, stray light and baseline stability.

After a lamp change	Typically verified with
Wavelength accuracy	A certified holmium oxide standard, commonly to a tolerance of ±1 nm
Photometric accuracy	Certified neutral density filters, or acidic potassium dichromate solution, at the top and bottom of your working absorbance range
Stray light and baseline stability	The instrument’s own procedure and your SOP

Your own SOP is what governs here, and it may ask for more or less than this. The point is that somebody has to do it, it takes time, and it should be planned into the job rather than discovered afterwards.

The practical consequence: order the lamp before the old one dies. A planned replacement lets you schedule the requalification. An emergency one puts the instrument out of use until somebody can run the checks.

One more thing worth checking. Benchtop spectrophotometers are often on no service contract at all, even in labs where the chromatography is fully covered. If you are not sure which applies to yours, who pays for the lamp explains what these agreements usually do and do not include.

Before you order

Which lamp, from the baseline scan. Deuterium or tungsten. The changeover wavelength decides it.
The instrument model, exactly. A Shimadzu UV-1800 and a UV-3600 do not take the same lamp, and neither do the different SPECORD models.
The part number from the old lamp, if you can read it. The most reliable input you have.
Whether you need the change to be documented, because that decides how much time the job takes, not just what the part costs.

If you would rather ask

Tell me the instrument model and where on the scan the problem appears, and I will tell you which lamp it is before anything gets quoted. The deuterium, tungsten and hollow cathode lamps are on the spectrophotometer and AAS lamps page.

If it appears everywhere on the scan, I will tell you it is probably not a lamp. There is nothing on this site I can sell you for a dirty cuvette.

Sources

On the lamps and their ranges:

Shimadzu, light sources for spectrophotometers and what types of light sources are used in a typical spectrophotometer.
Deuterium lamps against tungsten lamps, for the working ranges and typical lamp life.
Chemistry LibreTexts, sources of radiation.

On what to verify afterwards:

USP General Chapter <857>, Ultraviolet-Visible Spectroscopy, and Ph. Eur. 2.2.25. Summarised by Mettler Toledo and in this comparison of the two chapters.
Lab Manager, validating photometric accuracy and stray light, for the practice of re-verifying after a component replacement.

Quoted lamp lives are typical published figures rather than a specification for any one lamp. Your instrument manual and the lamp you actually buy both take priority over anything here.