The answer is depend on the size of your objects and type of substrate.

In general, AFM and SEM are more suitable for such tasks. A sample for TEM must be such thin as possible (usually, about 50 nm or less), it isn't so simple to prepare it from a bulk specimen. 

Good luck!  

Dear Sara,

You will find valuable pieces of information in order to help you to solve your question by checking for the four references :

Thin Films with Applications in Photonics", 2016 ( https://pdfs.semanticscholar.org/ff5b/3643dd877b262a1a2a1200babfb73daa9bd1.pdf )

MicroAnalysis to Determine the Thickness of Thin Films in Materials Science" ( http://cdn.intechopen.com/pdfs/27340.pdf )

Here is an excerpt from the first reference:

"...perhaps the most important consideration to bear in mind is the difference between computationally-derived film thickness values, such as are provided by techniques like UV-VIS Spectrophotometry, XRR and ellipsometry, and thickness values directly correlated to an empirical measurement, as are provided by techniques like STM and AFM. The latter variety are typically contact-based technologies that often offer sub-nanometer depth resolutions and extremely reliable measurements owing to the use of vibration-damping systems to mitigate noise and error; however, they also necessitate transition regions on the samples to provide vertical contrast between substrate and film layers and, as such, are limited either to only those samples that incorporate such regions or those samples that allow for destructive testing conditions.

When such conditions are not met, the former variety of techniques is preferable, typically guaranteeing non-destructive testing and nano-scale resolutions. However, such methods rely on computational modeling or curve fitting to derive film characteristics and, as such, are only as accurate as the margin of error for the models. Consequently, the parameters of the model, such as refractive index and film thickness, may place limitations on the reliability of the measurement, as the error residuals increase. [...] As a result, derived thickness values are subject to computational error, inaccuracies of fit and potentially substantial deviations away from the theoretically-expected model. Moreover, when making use of any such

computational techniques, it is essential that the operator also be aware of the theoretical assumptions upon which the software is operating. Moreover, said assumptions will likely change based on the various types of models and statistical analyses available in the software."

Finally, about topography analysis and roughness characterization in accordance with ISO standards, I suggest you to check for this reference:

• De Oliveira et al., Chapt. 07, "Measurement of the Nanoscale Roughness by Atomic Force Microscopy: Basic Principles and Applications" ( http://cdn.intechopen.com/pdfs/33450.pdf )

Hoping it will be helpfull

Actually it depends on the magnitude of roughness. I used AFM for measuring the surface roughness; however, it has some limitations.

What is the size and morphology of your sample? Please post photos if available.

XRR, X-ray Reflectivity, would be the best option generally for wafer samples.